Professional Documents
Culture Documents
Green Circuits
Distribution Efficiency Case Studies
1023518
NOTE
For further information about EPRI, call the EPRI Customer Assistance Center at 800.313.3774 or
e-mail askepri@epri.com.
Electric Power Research Institute, EPRI, and TOGETHERSHAPING THE FUTURE OF ELECTRICITY
are registered service marks of the Electric Power Research Institute, Inc.
Copyright 2011 Electric Power Research Institute, Inc. All rights reserved.
ACKNOWLEDGMENTS
The following organizations prepared this report:
Electric Power Research Institute (EPRI)
942 Corridor Park Blvd.
Knoxville, TN 37932
Principal Investigators
T. Short
D. Brooks
B. Arritt
W.Sunderman
J. Taylor
M. Rylander
H. Sharma
A. Gaikwad
J. Smith
Principal Investigator
B. Fletcher
University of TexasAustin
24th & Speedway, ENS348
Austin, TX 78712
Principal Investigator
M. Grady
ESB Networks
lectricit de France
FirstEnergy
Hydro Quebec
Kansas City Power & Light
New York Power Authority
Northeast Utilities
Pacific Gas & Electric
This publication is a corporate document that should be cited in the literature in the following
manner:
Green Circuits: Distribution Efficiency Case Studies. EPRI, Palo Alto, CA: 2011. 1023518.
iii
Finally, we would like to acknowledge the guidance and leadership of all the advisors and
members of EPRIs R&D Program on Transmission & Distribution Efficiency for a Lower
Carbon Future (Program 172) for their helpfulness and constructive comments during the last
three years.
iv
ABSTRACT
The Electric Power Research Institute (EPRI) Green Circuits project was a collaborative effort of
22 utilities. The main goal of the project was to evaluate ways to improve distribution efficiency.
Modeling, economic evaluations, and field trials formed the core of the research effort. To
evaluate efficiency improvements, 66 circuit case studies were modeled and fine-tuned, based on
field data. Field trials of voltage optimization were implemented on nine circuits. Detailed
advanced metering infrastructure metering data from two circuits also helped to provide
information on transformers and secondary circuits.
From simulations and field trials, optimizing voltages to the lower end of the American National
Standards Institute range was found to reduce energy consumption from 1% to 3% on most
circuits. The optimal efficiency options for a specific circuit depended on circuit characteristics,
load placement, circuit issues (like excessive unbalance), economic ranking criteria, and
economic assumptions. In some cases, the most economically viable option was just voltage
reduction; however, additional improvements were often economically viable also. In other
cases, the optimal project included voltage reduction, phase balancing, var optimization, and/or
targeted re-conductoring.
Keywords
Efficiency
Distribution planning
Losses
Power distribution
Voltage optimization
EXECUTIVE SUMMARY
The Electric Power Research Institute (EPRI) Green Circuits project was a collaborative effort of
22 utilities. The main goal of the project was to evaluate ways to improve distribution efficiency.
Modeling, economic evaluations, and field trials formed the core of the research effort. To
evaluate efficiency improvements, 66 circuit case studies were modeled and fine-tuned, based on
field data. Field trials of voltage optimization were implemented on nine circuits. Detailed
advanced metering infrastructure (AMI) metering data from two circuits also helped provide
information on transformers and secondaries.
Overall Modeling Results
Sixty-six circuits were modeled as part of the Green Circuit project. Each circuit was modeled in
detail from the substation to each customer meter and analyzed, using the long-term dynamic
distribution system electrical simulation package OpenDSS. Nearly all of the circuit models
were augmented with historical circuit-measurement data that allowed for hourly-resolution
simulation of the operation of the circuit for actual load patterns for each hour in a calendar year
(8760 hours). All loss sources through both daily and seasonal load changes were found with this
high-fidelity representation of the circuits electrical characteristics with the temporal and spatial
diversity of the loads. The results of analyzing the 66 circuits point to the following:
Circuit diversity The set of circuits modeled includes a wide variety of circuit
characteristics, covering common voltages and urban to rural circuits.
Annual energy losses Total distribution feeder annual energy losses, excluding the
substation transformer losses, averaged 3.6% of total consumption for the feeder and ranged
from approximately 1.5% to 9%.
Overall losses Losses exceeding 2.52% were reported for 75% of circuits, and 25% of
circuits had losses exceeding 4.32%.
Primary line losses Line losses averaged 1.4% of total consumption. Circuit length is a
reasonably good predictor of percentage of line losses.
Transformer no-load losses Transformer no-load losses averaged about 1.6% of total
energy consumption and ranged from approximately 0.5% to 3.5%. These losses were the
most consistent across circuits, depending mainly on transformer age and transformer
utilization (connected kVA versus load).
vii
Peak demand losses At peak load, losses averaged 4.8% of peak demand and ranged from
approximately 1.6% to 16.5%. Losses at peak load were predominantly primary line losses.
Peak losses exceeding 3.0% were reported for 75% of circuits, and 25% of circuits had losses
exceeding 5.8%.
Voltage optimization for energy reduction Optimizing circuit voltages to the lower
American National Standards Institute (ANSI) C84.1 range reduced energy consumption on
almost all circuits. Using a consistent voltage-reduction approach of controlling an end-ofline primary bus to 118.5 V, the median reduction in energy consumption across all circuits
was 2.34%, with upper and lower quartiles of 3.13% and 1.69%.
Phase balancing and reactive power improvements The reduction in losses from improved
reactive power support and phase balancing was generally small.
Overall, voltage optimization was the most promising option. It provided the most energy
reduction, and it could provide benefit on most circuits. Efficiency improvements depend on
circuit characteristics, and reconfigurations may change characteristics.
Economic Optimization
To better evaluate the economics of distribution-efficiency projects, six circuits were selected for
extended analysis with the goal of finding the most cost-effective approaches for improving
efficiency. For each of these circuits, voltage reduction was modeled, along with several
efficiency-improvement options such as phase balancing. In many cases, options helped reduce
losses while helping flatten voltage profiles. With flatter voltage profiles, voltage reduction
becomes more effective. The costs and economic benefits of each option were calculated, so that
economically optimal efficiency options could be selected. From these results, the following
lessons were learned:
Economic viability In all six cases, distribution efficiency projects were economically
viable. For each circuit, options were available with benefit-to-cost ratios exceeding 3.4, and
all had options with levelized life-cycle costs of less than $0.03/kWh. The majority of the
energy savings came from voltage reduction.
Lost billing from voltage reduction The economic analysis did not include lost billing from
voltage reduction, with the assumption that utilities would recover this by rate adjustments or
other regulatory recovery mechanisms.
Highest benefit-to-cost ratios Circuits with the heaviest loading had the highest benefit-tocost ratios, including higher load densities, 25- and 35-kV class circuits, and bus-regulated
circuits. Longer rural, more voltage-limited circuits had lower benefit-to-cost ratios. These
guidelines may help utilities target efficiency programs.
Best options The highest-ranking efficiency options for a specific circuit depended on
circuit characteristics, load placement, circuit issues (such as excessive unbalance), economic
ranking criteria, and economic assumptions. In some cases, the most economically viable
option was just voltage reduction; however, additional improvements were often
economically viable also. In another scenario, the optimal project included voltage reduction,
phase balancing, var optimization, and targeted re-conductoring. Table ES-1 shows optimal
options by circuit, based on a strategy of maximizing efficiency as long as the benefit-to-cost
ratio exceeded 2.
viii
Table ES-1
Optimal Options for Each Circuit Based on Maximizing Efficiency
Circuit
Option
Levelized
Cost
/kWh
Benefitto-Cost
Ratio
Energy
Savings
3.5
2.9
3.9%
Phase balancing
0.6
17.0
3.6%
2.4
3.8
2.3%
2.1
4.7
1.3%
Voltage regulators
3.9
2.5
1.9%
Voltage regulators
0.6
16.2
2.6%
Efficiency approach Because each circuit was different, improvement options should be
targeted to the specific circuits needs. If phases are unbalanced, the focus should be on tap
phasing to address the issue. It is important to consider the load placement. Maximum benefit
occurs if voltages are flat and controlled at key load centers. This can affect where capacitors
or regulators are located. It is often beneficial to try efficiency options in the following order:
phase balancing by rephasing taps, simple reconfigurations to balance phases or loading
among sections, capacitor control or placement changes, addition of regulators, and then
targeted re-conductoring. Fixing problem transformers/secondaries is another option.
Secondary line losses were generally low, but they were higher than the generic models used
for most circuits. Findings from the overall simulations showed annual secondary line losses
of 0.31%. In more precise modeling of four secondary systems, average losses were 0.76%.
From estimates of AMI metering data from two circuits (44 secondaries total), average
secondary losses were 0.72% on one circuit and 0.87% on another.
ix
As with other circuits, circuit voltage measurements from metering data ranged on the high
side of the ANSI C84.1 range. On two circuits analyzed, the median customer voltage was
close to 122 V. Even at peak load, voltages were generally high, with median customer
voltages above 121 V on both circuits. This leaves significant room to reduce voltages.
Load decreased by 1.6% to 2.7% for substation voltages reduced by 2.0% to 4.0% (see
Figure ES-1). In terms of a conservation voltage reduction (CVR) factor, the percent change
in load for a 1% change in voltage generally ranges from 0.6 to 0.8.
Reactive power responds even more strongly to voltage reduction than real power. Reactive
power CVR factors exceeded 4 on two circuits with AMI metering that were capable of
measuring reactive power.
A statistical regression approach using measurements from one or more circuits with
comparable load patterns works well to normalize results and more precisely evaluate the
impact of voltage on load.
Voltage reduction is most effective in the summer and least effective in the winter, when
more thermostatically controlled heating load (constant energy) is present.
Average energy reductions for the EPRI tests circuits were similar in range to a previous study
by the Northwest Energy Efficiency Alliance (NEEA). This is significant because the climate
and load patterns in the northwestern United States differ from the EPRI circuits, which were
primarily in the Southeast.
Figure ES-1
Comparison of Energy Reduction and Confidence Intervals for Several Circuits
Future Work
Project results have shown that the distribution system can be a significant resource for reducing
energy consumption, particularly when voltage optimization is combined with other distribution
improvements. Based on these considerations, some of the key industry needs for distribution
efficiency are:
Load models Better models of the response of loads to voltage will help utilities target
voltage-optimization projects and claim financial credits for implementing distribution
efficiency projects. Better models will also allow predictions of future benefits of voltage
optimization.
Field trials The field trials consisted of nine circuits, primarily in the southeastern United
States. More field trials would help utilities in other areas better quantify efficiency gains.
Implementation of different voltage control and volt-var control systems will help determine
how much improvement is possible with more sophisticated systems.
AMI More work is needed to determine how to best use AMI data for efficiency
improvements.
Optimization There may be a need for tools to make it easier for utilities to perform
economic evaluations of efficiency options in order to optimize efficiency.
xii
CONTENTS
xiii
xiv
xv
xvi
Compliance Threshold Assessment for Existing and Improved Cases .................... B-35
Non-Compliance Violations Identification on Circuit Maps for Existing Case ........... B-36
Average Distribution Transformer Utilization Assessment........................................ B-37
Assessment of Customer Voltage Compliance with ANSI Voltage Standard ........... B-37
System Improvement Investment Cost Assumptions ............................................... B-38
Description of Recommended System Improvements.............................................. B-39
Location of Recommended Improvements on Maps ................................................ B-40
Average Voltage Analysis and Loss Impacts............................................................ B-40
Existing System Average Voltage Calculation .......................................................... B-42
Improved System Average Voltage Calculation........................................................ B-42
Economic Factors and Evaluation Assumptions....................................................... B-44
Economic Analysis.................................................................................................... B-45
Economic Data Given ............................................................................................... B-45
Summary Exhibits of Benefits and Costs.................................................................. B-47
Bibliography ....................................................................................................................... B-50
C EXTENDED CASE STUDIES ............................................................................................... C-1
Circuit B................................................................................................................................ C-1
Circuit C ............................................................................................................................... C-8
Circuit D ............................................................................................................................. C-13
Circuit E.............................................................................................................................. C-19
xvii
LIST OF FIGURES
Figure 2-1 General Process for Developing Green Circuits Base-Case Model .........................2-4
Figure 2-2 Representation of a ZIP Load ................................................................................2-7
Figure 2-3 Simplified Representation of an OpenDSS Transformer Model ...............................2-9
Figure 2-4 Temperature Control for Summer Switching (May 15 to September 15)................2-12
Figure 2-5 Temperature Control for Non-Summer Switching (September 15 to May 15)........2-12
Figure 2-6 Example of LTC Operation of the Model Verified Against the Measured Bus
Voltage .............................................................................................................................2-13
Figure 2-7 Example of Estimating Load Usage Based on AMI Data .......................................2-15
Figure 2-8 Example of Load Scaling........................................................................................2-16
Figure 2-9 Example of Normalized Load Shape ......................................................................2-16
Figure 2-10 Example of Verifying Simulated Phase Current to Measured Phase Current ......2-18
Figure 2-11 Example of Verifying Simulated Bus Power to Measured Bus Power ..................2-18
Figure 2-12 Example of Verifying Simulated Voltage to Measured Voltage ............................2-19
Figure 2-13 Circuits by Voltage and Distance from the Substation .........................................2-21
Figure 2-14 Number of Customers per Circuit .........................................................................2-22
Figure 2-15 Circuit Load Factors .............................................................................................2-23
Figure 2-16 Load Densities......................................................................................................2-24
Figure 2-17 Load versus Connected kVA ................................................................................2-25
Figure 2-18 Residential Load as a Percentage of Connected kVA..........................................2-26
Figure 2-19 Unbalance Versus Load Current ..........................................................................2-27
Figure 2-20 Circuit AA One-Line to Illustrate the Cause of the Unbalanced Load Current......2-28
Figure 2-21 Average Power Factors by Circuit ........................................................................2-29
Figure 2-22 Peak Load and Total Connected Capacitance .....................................................2-29
Figure 2-23 Statistical Distributions of the Minimum Primary Design Voltage .........................2-31
Figure 2-24 Minimum Primary Design Voltage Versus Simulated Minimum Primary
Voltage .............................................................................................................................2-31
Figure 2-25 Statistical Distributions of the Maximum Secondary Design Voltage Drop...........2-32
Figure 2-26 Maximum Primary and Secondary Design Voltage Drop for Two Utilities............2-32
Figure 2-27 Percent Losses by Location .................................................................................2-34
Figure 2-28 Circuit Loss Breakdowns in Average Percentage ................................................2-35
Figure 2-29 Circuit Loss Breakdowns in Average kW..............................................................2-36
Figure 2-30 Circuit Losses at Peak Load.................................................................................2-37
xix
xx
xxi
xxii
Figure 5-43 Secondary Voltage Drop Probability Distributions for the Metered Subset of
Circuit A............................................................................................................................5-36
Figure 5-44 Secondary Voltage Drop Probability Distributions for the Metered Subset of
Circuit B............................................................................................................................5-37
Figure 5-45 Peak-Load Secondary Voltage Drop Probability Distributions for the Metered
Subset of Circuit A ...........................................................................................................5-37
Figure 5-46 Peak-Load Secondary Voltage Drop Probability Distributions for the Metered
Subset of Circuit B ...........................................................................................................5-38
Figure 5-47 Voltage Drops Versus Different Factors for the Monitored Subset of Circuit A ....5-39
Figure 5-48 Voltage Drops Versus Different Factors for the Monitored Subset of Circuit B ....5-40
Figure 5-49 Transformer Voltage Drop Probability Distributions for the Metered Subset of
Circuit A............................................................................................................................5-41
Figure 5-50 Transformer Voltage Drop Probability Distributions for the Metered Subset of
Circuit B............................................................................................................................5-41
th
Figure 5-51 99 Percentile Transformer Voltage Drop Probability Distributions for the
Metered Subset of Circuit A .............................................................................................5-42
th
Figure 5-52 99 Percentile Transformer Voltage Drop Probability Distributions for the
Metered Subset of Circuit B .............................................................................................5-42
Figure 5-53 Transformer Voltage Drops Versus Loading for the Monitored Subset of
Circuit A............................................................................................................................5-43
Figure 5-54 Transformer Voltage Drops Versus Loading for the Monitored Subset of
Circuit B............................................................................................................................5-44
Figure 5-55 Transformer and Secondary Voltage Drops Versus Loading for the
Monitored Subset of Circuit A ..........................................................................................5-45
Figure 5-56 Transformer and Secondary Voltage Drops Versus Loading for the
Monitored Subset of Circuit B ..........................................................................................5-45
Figure 6-1 Energy Savings Versus Voltage Reduction for the NEEA Study Pilot Feeders........6-2
Figure 6-2 Change in Reactive Power in the NEEA Study Homes ............................................6-3
Figure 6-3 Impact of Voltage and Torque on an Example Motor ...............................................6-5
Figure 6-4 Tests of a Typical Modern 3-kW Air Conditioner ......................................................6-6
Figure 6-5 Temperature Normalization and Comparable-Circuit Normalization ........................6-8
Figure 6-6 Test Circuits Compared to Comparison Circuits.......................................................6-9
Figure 6-7 Circuit A Comparisons to Other Circuits .................................................................6-11
Figure 6-8 Example Load and Voltage Profile .........................................................................6-12
Figure 6-9 Comparison of Energy Reduction and Confidence Intervals for Several
Circuits .............................................................................................................................6-14
Figure 6-10 Comparison of Energy Reduction and Confidence Intervals for Several
Circuits .............................................................................................................................6-15
Figure 6-11 Monthly Comparisons of Average Power Consumption With and Without
Voltage Reduction............................................................................................................6-16
Figure 6-12 Reduction in Average Power Consumption With Voltage Reduction ...................6-17
Figure 6-13 Percent Reduction in Average Power Consumption With Voltage Reduction ......6-18
xxiii
xxiv
xxv
xxvi
LIST OF TABLES
Table 2-1 5-kV Class Transformer Losses ..............................................................................2-10
Table 2-2 15-kV Class Transformer Losses ............................................................................2-10
Table 2-3 15-kV Class Transformer Losses (19862006) .......................................................2-11
Table 2-4 Distribution Circuit Loss Statistics, Percent .............................................................2-33
Table 2-5 Distribution Circuit Loss Statistics Weighted by Load, Percent ...............................2-34
Table 2-6 Average Energy Savings From Voltage Optimization by Component .....................2-47
Table 3-1 Extended Case Study Circuits ...................................................................................3-1
Table 3-2 Average Marginal Cost for Several Green Circuit Participants ..................................3-4
Table 3-3 Economic Parameters Used to Determine BCR and LC ...........................................3-5
Table 3-4 Construction Cost Estimates .....................................................................................3-6
Table 3-5 Efficiency Project Annual Energy Savings .................................................................3-8
Table 3-6 Efficiency Project Peak Demand Savings..................................................................3-9
Table 3-7 Economic Acceptability of Efficiency Options ..........................................................3-10
Table 3-8 Benefit-to-Cost Ratio for the Best Option in each Category ....................................3-11
Table 3-9 Levelized Cost in /kWh for the Best Option in each Category ...............................3-12
Table 3-10 Acceptable Efficiency Projects Without Considering Savings From Option
Voltage Reduction............................................................................................................3-12
Table 3-11 Effect of Parameter Value Increase on Benefit-Cost Ratio and Levelized Cost ....3-13
Table 3-12 Benefit-Cost Ratio Resulting from Scaling Individual Parameter Values by 2
and 0.5 for Circuit D Option VF ........................................................................................3-14
Table 3-13 Levelized Cost Resulting From Scaling Individual Parameter Values by 2 and
0.5 for Circuit D Option VF ...............................................................................................3-15
Table 3-14 Optimal Options for Each Circuit Based on Economic Criteria ..............................3-18
Table 3-15 Efficiency Projects Tested .....................................................................................3-21
Table 3-16 Annual and Peak Savings for End-Use Load and Losses .....................................3-24
Table 3-17 Economic Analysis of Efficiency Projects ..............................................................3-26
Table 3-18 Efficiency Projects Tested .....................................................................................3-28
Table 3-19 Annual and Peak Savings for End-Use Load and Losses .....................................3-32
Table 3-20 Economic Analysis of Efficiency Projects ..............................................................3-34
Table 4-1 Case Study Feeder Summary ...................................................................................4-3
Table 4-2 Power-Factor Comparison of Circuit 1.......................................................................4-4
Table 4-3 Efficiency Analysis Comparison Summary ................................................................4-4
xxvii
xxviii
Table A-12 Case 1 Re-Conductor Losses at the Peak-Hour and Annual Energy Losses
With Respect to the Base Case ...................................................................................... A-17
Table A-13 Case 2 Re-Conductor Model Losses at the Peak-Hour and Annual Energy
Losses............................................................................................................................. A-18
Table A-14 Case 2 Re-Conductor Losses at the Peak-Hour and Annual Energy Losses
With Respect to the Base Case ...................................................................................... A-18
Table A-15 Ideal Var Model Losses at the Peak-Hour and Annual Energy Losses................ A-19
Table A-16 Ideal Var Losses at the Peak-Hour and Annual Energy Losses With Respect
to the Base Case............................................................................................................. A-20
Table A-17 Capacitor Control Model Losses at the Peak-Hour and Annual Energy
Losses............................................................................................................................. A-20
Table A-18 Capacitor Control at the Peak-Hour and Annual Energy Losses With Respect
to the Base Case............................................................................................................. A-21
Table A-19 Efficiency Analysis Comparison Summary ........................................................... A-22
Table A-20 Voltage Optimization Annual Losses With Respect to Base Cases ..................... A-24
Table A-21 Varying LTC Annual Losses With Respect to Base Cases .................................. A-24
Table B-1 Distribution System Efficiency Factors ..................................................................... B-3
Table B-2 Financial Factors Used with Studies ...................................................................... B-24
Table B-3 Distribution Efficiency Study Summary .................................................................. B-25
Table B-4 Typical System Improvement Installed Costs ........................................................ B-26
Table B-5 Distribution Efficiency Savings ............................................................................... B-26
Table B-6 Distribution Efficiency Economic Evaluation .......................................................... B-27
Table B-7 End-Use VO Factors for Northwest H-1 and C-3 Climate Zones ........................... B-33
Table B-8 Heating and Cooling Zone Classifications.............................................................. B-34
Table B-9 System Investment Unit Costs ............................................................................... B-38
Table B-10 Recommended System Improvements ................................................................ B-39
Table B-11 Financial Factors Used in Study........................................................................... B-45
Table B-12 Economic Summary ............................................................................................. B-47
Table B-13 System Improvement Project Costs ..................................................................... B-48
Table B-14 Energy and Demand Efficiency Savings .............................................................. B-49
Table B-15 Economic Evaluation Detail.................................................................................. B-50
Table C-1 Efficiency Projects Tested........................................................................................ C-3
Table C-2 Annual and Peak Savings for End-Use Load and Losses ....................................... C-5
Table C-3 Economic Analysis of Efficiency Projects................................................................. C-7
Table C-4 Efficiency Projects Tested...................................................................................... C-10
Table C-5 Annual and Peak Savings for End-Use Load and Losses ..................................... C-11
Table C-6 Economic Analysis of Efficiency Projects............................................................... C-12
Table C-7 Efficiency Projects Tested...................................................................................... C-15
Table C-8 Annual and Peak Savings for End-Use Load and Losses ..................................... C-17
Table C-9 Economic Analysis of Efficiency Projects............................................................... C-19
xxix
xxx
1
BACKGROUND AND FINDINGS
Electric power transmission and distribution systems typically have aggregate annual energy
losses ranging from 7% to 10%. In aggregate, such percentage losses across all U.S. transmission
and distribution equate to approximately 300 million MWh based on a U.S. annual generation
total of 4,157 million MWh. That energy total is roughly equivalent to the energy needed to
power approximately 29 to 35 million homes.1 Approximately two thirds of these losses are
incurred at the distribution voltage levels.
Historically, power delivery loss reduction, especially distribution system losses, has often been
a secondary priority because of uncertainties in quantifying loss improvements and the difficulty
in obtaining sufficient return on investment for projects undertaken. Recently, however, an
increased industry and regulatory focus on climate change and energy efficiency has led to a
renewed evaluation of power distribution efficiency.
A clear understanding of the magnitude of T&D losses is the first step in improving system
efficiency. Several recent advancements have made it possible to more readily identify options
for reducing distribution loss and improving overall system efficiency, including:
Improved metering that provides data on end-use patterns and diversity factors.
Improved communication and control capabilities that allow more precise voltage and
reactive power (var) control.
An overall improvement in modeling capabilities that allows for better loss estimation,
targeting of solutions, and ways to test and identify improvements.
While specific utility and circuit characteristics often dictate achievable efficiency levels, the
wide variation in distribution losses reported from one utility to another suggests that some
utilities or some circuits particularly have significant opportunity for more efficient operation.
In addition to reducing losses, electric distribution companies can increase efficiency through
management of end-use customer consumption. Utility voltage control can be used to reduce
energy consumption and peak demand. There is still significant work needed to quantify the
potential gains through voltage reduction across regions and load types. Existing work in this
area may need updating because end-use loads are changing with less use of purely resistive
loads and pure motor loads and more use of fluorescent lights, adjustable-speed drives, and
electronics.
Electric Power Industry 2007: Year in Review, EIA report released Jan. 21, 2009
1-1
Compile credible data to quantify the costs, benefits, and risks of using energy efficiency and
loss mitigation as a part of planning.
Demonstrate real-life examples where options for efficiency improvement have been
implemented and validate realized efficiency gains.
These objectives formed the basis for an EPRI collaborative utility project titled Distribution
Green Circuits, which officially began in April 2008. Since launch, 22 utilities have joined the
collaborative effort. Most of the participating utilities have worked with EPRI staff to identify
four or more specific distribution feeders for which detailed models have been or will be
developed to characterize existing circuit losses and prioritize potential options for efficiency
improvement. Since the completion of this analysis, each utility has evaluated actual field
implementation of one or more efficiency options.
Distribution Efficiency
Although not a high priority, distribution efficiency has always been a consideration of
distribution planners. Most reports on distribution planning include efficiency evaluations.2-4
Another good resource is the NRECA Power Loss Management report.5
Conservation voltage reduction was studied extensively in the 1980s, including work by EPRI.6,7
Voltage reduction has recently had a resurgence of interest as a way to use the distribution
system for energy conservation. The Northwest Energy Efficiency Alliance (NEEA) and their
contractor RW Beck and several utilities evaluated voltage reduction along with other efficiency
8
options in the U.S. Pacific Northwest. In many ways, the Green Circuits project is an extension
of the NEEA study to cover more circuits and more parts of the United States and even European
2
3
4
5
1-2
circuits. In 2010, the U.S. DOE PNNL laboratory release a report where they estimated that
conservation voltage reduction could reduce energy consumption by 3% nationwide.9
Voltage optimization is a term used to optimally apply voltage reduction. The DOE Bonneville
Power Administration (BPA) has implemented a distribution system energy efficiency program
10
to help promote energy. Voltage optimization is a key component of the program. BPA
specifies a simplified verification protocol to quantify program energy savings.11
Transformer efficiency is another area of recent interest. The Oak Ridge National Laboratory
estimates that distribution transformers account for 26% of transmission and distribution losses
and 41% of distribution and subtransmission losses.12 The DOE has instituted minimum
efficiency requirements for distribution transformers.13 Amorphous-core transformers offer the
potential for improved energy savings.14
In 2008, EPRI started a research program specifically to evaluate transmission and distribution
efficiency: Research Program 172: Efficient Transmission and Distribution Systems for a Lower
Carbon Future. As part of Program 172, the following reports were developed:
Distribution System Losses Evaluation, Reduction: Technical and Economic Assessment.
EPRI, Palo Alto, CA: 2008. 1016097.
http://my.epri.com/portal/server.pt?Abstract_id=000000000001016097
Main results: Framework for performing a distribution system loss study.
U.S. Department of Energy Pacific Northwest National Laboratory, Evaluation of Conservation Voltage Reduction
(CVR) on a National Level, 2010.
http://www.pnl.gov/main/publications/external/technical_reports/PNNL-19596.pdf
10
BPA Energy Efficiency Implementation Manual, October 2010.
11
Simplified Voltage Optimization (VO) Measurement and Verification Protocol, May 4, 2010.
http://www.nwcouncil.org/energy/rtf/measures/protocols/ut/VoltageOptimization_Protocol_v1.pdf
12
ORNL-6804/R1, The Feasibility of Replacing or Upgrading Utility Distribution Transformers During Routine
Maintenance, Oak Ridge National Laboratory, U.S. Department of Energy, 1995.
13
Department of Energy, Distribution Transformers Rulemaking, Liquid-Immersed Engineering Analysis Results,
September 2007.
14
H. Ng, R. Hasegawa, A. Lee, and L. Lowdermilk, Amorphous Alloy Core Distribution Transformers,
Proceedings of the IEEE, vol. 79, no. 11, Nov. 1991.
1-3
Distribution System Losses Evaluation: Distribution System Loss Calculation Tool User
Documentation and Overview. EPRI, Palo Alto, CA: 2009. 1017899.
http://my.epri.com/portal/server.pt?Abstract_id=000000000001017899
Main results: Spreadsheet tool to help in performing a distribution system loss study.
Amorphous Metal Transformer: Next Steps. EPRI, Palo Alto, CA: 2009. 1017898.
http://my.epri.com/portal/server.pt?Abstract_id=000000000001017898
Main results: Evaluation of the state of the art of amorphous metal transformers and options
for further research to improve adoption and technologies.
Distribution Efficiency: Modeling, Volt-Var Control, and Economics, EPRI, Palo Alto, CA:
2010. 1020147.
http://my.epri.com/portal/server.pt?Abstract_id=000000000001020147
Main results: Guidelines for modeling for efficiency studies; comparison of volt-var control
approaches; methodology for evaluating costs of losses.
1-4
Losses and voltage drops from AMI data analysis of transformers and secondaries
CVR factors (conservation voltage reduction factors) from field trials at nine circuits
Variation by season
1-5
2
ANALYTICAL FRAMEWORK AND MODELING
RESULTS
Introduction
This chapter shows the collective results from 66 circuits modeled as part of the Green Circuit
project. Each circuit was modeled in detail in OpenDSS based on circuit models and information
provided by each of the circuits respective utility. Almost all of the circuit models were refined
with circuit measurement data, and analysis includes hourly simulations for each hour in the year
(8760 hours) to accurately capture loads and losses through daily and seasonal changes.
The main goals of the modeling were to:
Quantify voltage profiles along the primary, through the transformer, and to customers.
Evaluate options to reduce losses, including phase balancing, reactive power improvements,
re-conductoring, transformer replacements, and circuit reconfigurations.
The 66 circuits studied encompassed many different types of distribution circuits that varied
in design practices, load types, voltage class, voltage-regulation techniques, and var control
practices. In addition, the circuits covered many different geographical locations and urban and
rural environments of varying degrees. The circuit modeling implementation also varied from
circuit to circuit, depending on circuit measurement data provided or not provided.
Circuit variability may have reduced the overall consistency across the circuits studied. EPRI
also attempted to address each circuit uniquely with utilities specific interests in mind. With
this said, even when a standardized approach was taken across all circuits (see the voltage
optimization section below), the uniqueness of each circuit still presented results that were not
all inclusive but required further investigation of individual circuits to explain those differences.
These further investigations provided additional insight and understanding and some general
conclusions can still be made. The circuits were not selected randomly, and EPRI does not claim
that the circuits are representative. The circuits were self-selected by utility participants. Each
utility chose circuits based on their goals. Some utilities chose circuits from different operating
regions. Some utilities chose circuits based on availability of monitoring data or availability of
modeling data. Some utilities chose circuits based on voltage or urban verses rural.
2-1
Voltage optimization for energy reduction Voltage reduction successfully reduces annual
energy consumption by improving end-use efficiency on almost all circuits, to varying
degrees. The median reduction in energy was 2.34%, with upper and lower quartiles of 1.69
and 3.13%.
Reactive power improvements The analysis identified a few circuits that have an
opportunity for significant loss savings. However, most circuits did not see significant
savings from improved reactive power support.
Phase balancing As with var improvements, options for phase balancing tend to be circuitspecific. Phase balancing should take the average unbalance into consideration in addition to
peak conditions. The annual line losses increased on a few circuits when only peak balancing
was considered.
Simplified modeling Traditionally, utilities evaluate losses based on a peak case and use an
estimated loss factor to evaluate annual performance. Comparing detailed simulations to the
simplified approach shows that the simplified approach is relatively accurate for predicting
losses in the majority of simulated results. However, certain factors can skew the relationship
between the estimated loss factor and the circuits peak results.
This chapter is divided into five sections: Analytical Framework, General Characteristics,
Loss Characteristics, Improvement Options, and Comparing Detailed Modeling to PeakCase Models. The Analytical Framework section provides a brief description of the modeling
software used along with background information on modeling approaches. The General
Characteristics section provides a general summary of various characteristics to provide the
reader with an overview of the variations in the circuits analyzed. The Loss Characteristic
section provides information on the amount of losses, breakdown of losses, and a general
characterization of losses for the circuit set. The Improvement Options section summarizes
several improvement options investigated, and the Comparing Detailed Modeling to Peak-Case
Models section compares the simulation results to the results when the average losses are
estimated with a loss-factor adjustment.
Analytical Framework
Distribution models frequently include only the components of the primary distribution system
(the medium-voltage system) up to the service transformer and occasionally only the feeder
three-phase mains. This is because models of a distribution feeder are typically used to analyze
peak-demand power flows to ensure that there is sufficient power-delivery capacity to meet the
peak load demand.
For this project, a more detailed distribution feeder model was used. These models included a
representation of all of the major electrical components that contribute to losses. The Green
Circuits feeder models include the following:
2-2
Substation power transformer. Note that this study was feeder-focused. Therefore, loss
reductions in the substation transformer were not examined. However, an accurate
representation of the substation transformer was modeled.
Secondaries/services.
Capacitor controls.
In addition to representing the full extent of the physical system, annual variation in the load
served from the circuit was also represented. This was accomplished by the following:
Individual customer loads were either assigned based on data provided by the host utility or
allocated to each customer point based on the peak demand value at the head of the circuit
and connected transformer kVA.
Each individual customer load was assigned an hourly-resolution annual load shape that
represented the manner in which that load varied throughout a typical year at the point
where the load was electrically interconnected.
The general process of developing the base case model for a given circuit is shown in Figure 2-1.
The bulk of the electrical connectivity of a given circuit was obtained by converting a preexisting model of the circuit either from the utilitys own commercial analysis package format or
GIS (geographic information system) format. The base network was then augmented with
additional circuit data that is typically not included in GIS or typical peak power-flow-based
models. This information usually includes the circuit-control parameters for such equipment as
load tap-changing transformers (LTCs), regulators, and switched capacitors. Characteristics of
transformer loss and secondary lines are also typically not included in base models but are added
during the base-model development process. Loads were converted to voltage-dependent loads.
Finally, annual load shapes were defined from measured data and were attached to individual
loads in the model. These items are discussed in additional detail in the following subsections.
2-3
Measured
Data
Load Profiles
Loss
Breakdown
and Reduction
Options
Imported
Circuit Data
LTC/Regulator/
Capacitor
Controls
Secondary and
Services
Transformer
Loss Data
Figure 2-1
General Process for Developing Green Circuits Base-Case Model
Once the base-case model was developed, long-term time-varying simulations of the full
electrical model serving all circuit loads through an annual hourly-resolution load cycle were
conducted. Various electrical outputs for the year were collected from the simulation and
compared with historical measured data to validate the model. Quantities such as active and
reactive power flows, short-circuit currents, and voltage at available measurement locations on
the circuit were very useful in validating that the modeled circuit was representative of actual
circuit operation.
Then, an annual simulation for each circuit was used to determine the base case losses that
were incurred on the circuit. The base-case losses were broken down as to the specific sources of
the losses (primary versus secondary, load versus no-load, and so on). Losses were normalized to
either the total annual energy consumption (energy losses) or the peak demand (peak losses).
Once the base-case losses were determined, various options for reducing losses were then
modeled and simulated as modifications to the base-case model. The loss-reduction options
selected to be explored in this comprehensive evaluation include the following:
1. Ideal var improvement Impact of load reactive power and var improvement options was
analyzed by evaluating an idealized var improvement case. In this case, all load power
factors were corrected to unity, and the resulting loss improvement was analyzed. This
analysis gives the maximum loss savings possible from improved reactive power
management.
2. Phase balancing Rearranging loads on each phase of the circuit to reduce residual flows.
2-4
3. Selected re-conductoring Replacing selected conductor sections with larger, lowerresistance conductors.
4. Voltage optimization Maintaining distribution circuit voltages at a lower range within the
allowable ANSI range. This case also reviews approaches to improving voltage profiles.
Modeling Software OpenDSS
The OpenDSS distribution simulation package was the main analysis tool used in the Green
Circuit project. The OpenDSS program is a script-driven, frequency-domain electrical circuit
simulation tool that has specific models designed to accurately represent unbalanced, multi-phase
power distribution systems.
The OpenDSS software was used because the Green Circuits study required modeling capability
exceeding what is typically found in commercial distribution system analysis software. Some of
these capabilities are listed below:
Long-Term Simulations: Many utilities perform annual analysis of electrical losses on their
power system, the objective being to quantify and pinpoint locations where significant
electrical losses are occurring. Typically, the utility computes electrical losses that occur
during peak load periods using power flow solver snapshot capabilities. Annual losses are
then estimated using approximation techniques, such as loss factors. OpenDSS provides a
convenient mechanism to perform formal loss evaluation for the entire year, thus greatly
reducing the inaccuracy associated with the approximate methods.
Input Data: The OpenDSS was designed for a research or consulting environment where
input data might come from a variety of sources. This is certainly the case for the Green
Circuit analysis, where circuit and monitoring data came from many sources. Formats used in
the Green Circuits project include Aspen Oneliner, CYMDIST, FeederAll, various GIS, PTI
HUB, SynerGEE, and WindMil. The OpenDSS scripting language is designed to require
minimal translation from other formats of distribution data and is frequently amended to
accommodate new forms of data.
Some of the advanced features of OpenDSS require data not traditionally used in distribution
system analysis but needed for efficiency studies. Some examples include:
Various curves as a function of time for sequential time-varying simulations: load shapes
(such as AMI P, Q data, and load class), temperature shapes, and so on.
Machine data.
2-5
Component Control: Conventional power flow solvers include facilities for simulating the
operation of voltage regulators and capacitor bank controls. For example, most conventional
tools are able to simulate local bus control and line drop compensation control strategies for
voltage regulators and load tap changers. Similarly, these programs can simulate simple
capacitor bank controllers based on local voltage and reactive power measurements.
However, conventional power flow solvers are usually not adequate in simulating the
operation of more sophisticated volt/var optimization strategies that determine control actions
using combinations of remote measurements and other system-level considerations. This
feature was utilized for the voltage-optimization studies along with many of the volt/var
control schemes utilized by many utilities that are modeled.
Data Management: For the annual simulations, interval load-flow computations are needed.
An interval is typically 15 minutes or 1 hour, resulting in 35,040 or 8,760 computations,
respectively. Because the load and loss information needs to be captured at each of these
intervals, data handling is a challenge. OpenDSS provides the capability to characterize
losses into various categories and saves this information at each simulation interval for the
yearly simulation. In addition to loss information, the OpenDSS program also logs voltage
levels, overloads, voltage exceptions, and other circuit-related information at each simulation
interval. The loss data logged includes load and no-load losses, transformer losses, line
losses, losses based on voltage level, zero-sequence losses, and more. Average, peak, and
minimum voltage levels for primary and secondary level voltage buses are also logged,
which was primarily used for the base and voltage optimization analysis.
S. Lefebvre et al., Measuring the Efficiency of Voltage Reduction at Hydro-Quebec Distribution, IEEE Power
and Energy Society General Meeting, 2008.
2
NEEA 1207, Distribution Efficiency Initiative, Northwest Energy Efficiency Alliance, 2007. Available at
http://rwbeck.com/neea/
2-6
Figure 2-2
Representation of a ZIP Load
Another approach is a model that allows the user to specify how active and reactive power vary
with voltage. This type of model can be adapted for user-defined CVR factors based on empirical
or typical values of different load classes.
The CVR factor is defined as a reduction (or increase) of power or energy based on a voltage
variation. A CVR factor for watts can be defined as a percent reduction in active power (watts)
per 1% reduction in voltage. A CVR factor for vars can be defined as a percent reduction in
reactive power (vars) per 1% reduction in voltage. One modeling approach is to have a load
model that represents the power and var relationship to voltage as exponential equations, where
this model can be adapted with user-defined CVR factors.
This is expressed mathematically as:
PL = PnV CVRwatts
Eq. 2-1
QL = QnV CVRvar
Eq. 2-2
Where CVRwatts and CVRvar are the active and reactive power exponents, respectively. Pn and Qn
are the values of active and reactive powers at the nominal (rated) voltages, respectively. V is the
voltage magnitude at the terminals of the load. PL and QL are the resulting values of real and
reactive powers. As modeled, the loads respond instantly to voltage changes. In reality, the
response of load varies with time, especially for step changes in voltage, as might be done for
voltage reduction used to reduce peak demand. This model does not consider the variation with
time. We chose the CVR factors to reflect the expected active and reactive energy response to
voltage for long-term voltage control.
This voltage-sensitive load model was used for all modeling in OpenDSS, where the watts and
vars both vary with voltage based on an exponential relationship. For these simulations, a CVR
factor of 0.8 was used for watts and a CVR factor of 3.0 was used for vars. These values were
based on research by Kirshner and Giorsetto3 and the Northwest Energy Efficiency Alliance
3
D. Kirshner and P. Giorsetto, Statistical Tests of Energy Savings Due to Voltage Reduction, IEEE Transactions
on Power Apparatus and Systems, vol. PAS-103, no. 6, pp. 120510, June 1984.
2-7
(NEEA), along with their contractor RW Beck and several utilities evaluating voltage reduction
in the U.S. Pacific Northwest 4.
Service Transformer Modeling
All service transformers were included in the modeling. The model service transformer included
both series and shunt loss elements. Figure 2-3 shows a simplified model of the elements
included in the OpenDSS model. Utilities were almost universally unable to provide accurate
data on specific transformer impedances and loss characteristics. Based on lack of specific data,
approximations were normally needed.
If the utility had the transformers modeled in their analysis packages, values for load loss (Rs)
and series impedance (Ls) were typically included in the databases. However, the majority of
circuits did not have the service transformers modeled, and these values had to be obtained from
the utilities supplied test sheets or transformer specifications.
Unspecified transformers were manually added to the model using loss and impedance
information from these test sheets or specifications. Definitions of the losses and impedances
were based on the transformers base kVA rating, number of phases, pad- or pole-mounted
design, and voltage rating. If values were not supplied by the utilities, estimates were taken based
on typical information.
The no-load loss values for determining Rm of the distribution transformer models were seldom
included in the databases supplied by utilities. Therefore, the no-load loss values were obtained
from the utilities supplied test sheets or transformer specifications, or values were assumed
based on typical information.
Table 2-1 through Table 2-3 are examples of some references used to estimate the losses of
transformers when utility-specific information was not available. Table 2-1 and Table 2-2
provide the no-load losses (NLL) and load losses (LL) of variously sized transformers for 5-kV
and 15-kV class transformers, respectively. These values were taken from the GE Distribution
5
6
Data Book and the Westinghouse: Distribution Systems reference books and were applied
uniformly to those circuits with older-vintage transformers. Table 2-3 is summary data applied to
those circuits that have transformers manufactured in the last 25 years. Note that this table does
not include those transformers that would be classified as energy efficient. Table 2-3 was
derived from transformer loss data collected from a utilitys test results. This loss information
was collected on multiple transformers whose manufacture date ranged from 1986 to 2006. This
loss information was averaged across each transformer based on kVA rating. Because circuits
typically contain service transformers with different manufacture dates, the average loss values
shown in Table 2-3 were applied uniformly across all transformers in the model.
5
6
NEEA 1207, Distribution Efficiency Initiative, Northwest Energy Efficiency Alliance, 2007. Available at
http://rwbeck.com/neea/.
General Electric GET-1008L, Distribution Data Book.
Westinghouse Electric Corporation, Distribution Systems, vol. 3, 1965.
2-8
If a circuit contains transformers whose manufacturing dates include both those built in the last
25 years and also older transformers, a typical method would be to apply an average of the loss
values in Table 2-2 and Table 2-3 uniformly across all of the transformers in the circuit model.
Note that the specifications received from utilities and the information contained in Table 2-1
through Table 2-3 were applied to each transformer in the model based on voltage and kVA
rating. The date of manufacture of each particular transformer in the model was typically
unknown. When that was the case, all transformers in the circuit were given the same loss
assumptions. For example, each 15-kVA, single-phase overhead transformer received the same
loss/impedance assumptions. These assumptions may introduce inaccuracies in the overall
transformer losses to various degrees.
The magnetizing current (im) was also modeled to represent the current flow due to generating
the flux in the core. This value was seldom attainable for specific transformers. In most
instances, this value was assumed to be 0.5% of rated current. In general, the modeled
magnetizing current contributed very little to overall losses.
Figure 2-3
Simplified Representation of an OpenDSS Transformer Model
2-9
kVA
NLL (%)
LL (%)
kVA
NLL (%)
LL (%)
0.72
1.78
0.77
2.07
10
0.59
1.21
10
0.57
1.63
15
0.51
1.04
15
0.51
1.57
25
0.44
1.08
25
0.43
1.45
37.5
0.42
0.90
50
0.33
0.89
50
0.35
1.33
75
0.37
0.86
100
0.32
0.87
100
0.32
1.19
Table 2-2
15-kV Class Transformer Losses
GE Distribution Data Book
(Date Unknown) 7.2/12.47Y (kV)
2-10
kVA
NLL (%)
LL (%)
kVA
NLL (%)
LL (%)
0.72
2.04
0.86
2.14
10
0.59
1.25
10
0.67
1.83
15
0.51
1.19
15
0.6
1.7
25
0.44
1.18
25
0.52
1.52
37.5
0.42
1.05
50
0.33
1.01
50
0.43
1.33
75
0.37
0.88
100
0.32
0.88
100
0.37
1.15
Table 2-3
15-kV Class Transformer Losses (19862006)
Average Transformer Losses
(19862006) 1-Phase
kVA
NLL (%)
LL (%)
kVA
NLL (%)
LL (%)
10
0.318
1.153
225
0.157
0.668
15
0.280
1.090
300
0.139
0.664
25
0.230
1.008
500
0.128
0.607
38
0.219
0.831
750
0.109
0.629
50
0.184
0.811
1000
0.097
0.594
75
0.171
0.749
1500
0.111
0.537
100
0.173
0.715
2000
0.096
0.536
167
0.156
0.633
2500
0.079
0.528
Capacitor Modeling
The system capacitors were for the most part included in the utility-provided databases.
However, the capacitor controls had to be manually defined in the OpenDSS model because
control data was seldom provided in the original model. It was important to have the capacitor
controls included to have correct capacitor operation for the annual simulations.
The OpenDSS was able to model the various control types for capacitors analyzed in the project:
current, voltage, kvar, power factor, time, and other more sophisticated volt/var control schemes.
In some cases, multiple control schemes were needed during an annual simulation. Figure 2-4
and Figure 2-5 illustrate a capacitor control that used two different temperature schemes for two
parts of the year. For this particular model, historical temperature measurements needed to be
included in the model to duplicate the capacitor operation
Correct capacitor operation was often critical to match annual reactive power simulation results
with SCADA measurements. Because the customers power factor was typically estimated, the
annual simulation results along with any reactive power measurements could be used to improve
the estimate of the customers power factor.
2-11
Capacitor Switching
100
60
95
50
Current (A)
85
40
80
75
30
70
OFF
20
65
Temperature (F)
90
ON
Capacitor Current
Temperature
60
10
55
0
5400
5450
5500
5550
5600
5650
50
5700
Hour
Figure 2-4
Temperature Control for Summer Switching (May 15 to September 15)
Capacitor Switching
60
80
70
50
40
50
OFF
30
20
40
30
ON
Temperature (F)
Current (A)
60
20
10
0
875
10
925
975
1025
1075
1125
0
1175
Hour
Figure 2-5
Temperature Control for Non-Summer Switching (September 15 to May 15)
2-12
Capacitor Current
Temperature
Voltage-Regulation Modeling
The voltage-regulation equipment such as load tap changers (LTC) and bus/line regulators were
for the most part included in utility-provided databases. For substation bus voltage regulation, it
was important to model the additional feeder loads located on the bus. This was done so that the
bus voltage regulation reacted to the total load through the transformer and not just the feeder
being studied. It is important to point out that the study that EPRI conducted concentrated on
individual feeders; therefore, practices that would reduce substation transformer losses were not
generally examined.
For those circuits that included line-drop compensation (LDC), the R and X values were defined
in the model along with the corresponding PT ratios and CT ratings.
Figure 2-6 is an example of verifying the LTC operation of the model against the measured bus
voltage. In this particular example, the LTC set point varied throughout the year due to an endof-line feedback scheme implemented for voltage optimization.
Substation Bus
126
125
124
123
Voltage
122
121
120
119
118
117
116
0
1000
2000
3000
4000
5000
6000
7000
8000
Hours
Measured VOLT
Simulated VOLT
Figure 2-6
Example of LTC Operation of the Model Verified Against the Measured Bus Voltage
Service-Line Modeling
Few utilities provided customer service lines in their existing feeder models. The majority of
customer service lines and line model parameters were estimated based on information provided
by the utility. The information provided by utilities was typically from their construction
guidelines. The utilities service conductor size was usually based on their requirements to meet
flicker limits, voltage drop, and any field installation limitations. Also, the limit of the maximum
length for each type of service conductor was stated in the construction guidelines. This
information was then used to estimate a typical service conductor size and length to use in the
model.
2-13
Typically the length of the service line used in the model was between 50 to 75% of the defined
maximum length (which varied based on guidance from the utility). The service conductor size
used in the model was estimated based on the transformer rating and the number of customers
connected to a particular transformer. Overhead construction was used on the service conductor
unless the load was served by a pad-mounted transformer.
The residential service line electrical model was based on the physical construction of the
services cable or lines. The neutral conductor was assumed grounded at both ends and reduced
out by Kron reduction. The series impedance and shunt capacitance were represented by a series
impedance and nodal admittance form, respectively, as is typical for all line models.
As with service transformers, accurate secondary information was rare to have. Blanket
assumptions were made on service lines across each of the circuits. The degree to which
assumptions were made depended primarily on the amount of information obtained on the
service conductors themselves and the number of customers served from each transformer. In
some cases, the number of customers connected to each transformer was not identified in the
databases; therefore, the number of service conductors and the size of service conductors had to
be estimated. These assumptions introduced some inaccuracies in the overall service-line losses.
Use of approximate model secondaries is compared to accurate secondary models in more detail
in Chapter 5 for one utility that had accurate secondary impedance and connectivity information
on two circuits.
Load Allocations and Load Variation
The annual variation of feeder load with time was modeled in this project. The implementation
of load allocation and how these loads varied with time differed from circuit to circuit. The
manner in which loads were allocated and how they varied with time depended on available data
for the feeder (or lack of data). The normal procedure for load allocation and load variation
consisted of the following two steps:
Individual customer loads are allocated to each customer point based on the peak demand at
the head of the circuit.
Each individual customer load is assigned an hourly-resolution annual load shape that
represents the manner in which that load varies throughout a year at the head of the circuit.
The majority of circuit models received already contained a load allocation based on the peak
demand. This load allocation was typically done by allocating a portion of the total load to each
spot load (a load connected on the primary level) based on the connected kVA rating of each
transformer. This is much the same way the majority of the circuits peak demand was allocated
in this study; but because each individual customer was typically modeled, the allocation
occurred at the customer level as opposed to the transformer primary level.
Peak-demand load allocation again varied from circuit to circuit based on the available
information. If only the total power was measured at the head of the feeder, the load was
allocated across all three phases, where each phase was allocated one third of the total power. If
data was measured on a per-phase basis, load was allocated to each phase individually. When
feeder data was collected on a per-phase basis, it was typically individual phase currents.
2-14
Therefore, the allocation factors were set to match the peak phase currents for each phase at the
feeder head.
With the few circuits that contained AMI, a typical approach was to use a two-step allocation
procedure. The first step was to estimate a usage profile for each load individually and then
allocate based on the peak demand measured at the head of the feeder. One approach for
determining the usage profile for each load was to take a weekly average of the daily power
usage during the feeders peak week. Figure 2-7 illustrates an example of four selected loads and
their daily usage taken from the AMI data during the feeders peak week. After the average
usage was determined for each load during the peak week, this value was used to assign an
allocation factor to each load individually. This provided each load with a unique allocation.
After this allocation was set, an additional allocation was used to adjust the loads in order to
match the peak demand measured at the head of the feeder. The reasoning for this approach is
provided below.
Figure 2-7
Example of Estimating Load Usage Based on AMI Data
Once the allocation factors were set, an annual load shape was used to provide the load variation
with time. The load shape is essentially an additional scaling factor applied to each load
separately during each interval in the annual simulation. This load-shape scaling is in addition to
the allocation scaling (see Figure 2-8).
Figure 2-9 shows an example of load shape plotted in OpenDSS. Note that typically the load
shape is normalized to its peak value.
2-15
kVA Rating of
Load
(Constant)
*
Time Varying
Load
1 hour increments for 8760 hrs
* Voltage variation
and power factor
impacts final kW
and kvar values
Allocation Factor
based on peak demand
(Constant)
Loadshape
1 hour increments for 8760 hrs
Figure 2-8
Example of Load Scaling
Figure 2-9
Example of Normalized Load Shape
2-16
How each load shape was defined depended on the available data for the feeder. In the absence
of time-variation data, class load shapes were used. Class load shapes were usually categorized
as residential, commercial, and industrial class customers. The class load shapes were usually
provided by the utility. Each load was assigned a load shape that corresponded to its customer
class. The classification of each load was typically defined in the database (the circuit definition
files for the commercial analysis package, or GIS) or was obtained from the utility. The load
shape would scale each load of that load class during the annual simulation.
When time-varying data was available at the head of the feeder, this data was used to vary the
load with time across the annual simulation. If only the total power was collected at the head of
the feeder, this load shape was used across all of the loads (every load varies by the same ratio
throughout the annual simulation). If data was collected on a per-phase fashion, a load shape for
each phase was derived and allocated to the loads of each phase separately (each load connected
to the same phase varies by the same ratio throughout the annual simulation). For three-phase
loads, a separate load shape was assigned based on an average of the three load shapes from each
phase.
Once load allocations and load shapes were assigned, modeled results were compared to the
measured data to verify the model. Adjustments to the scaling were typically made to further
tune the model to improve the agreement between the model results and the measurements.
Figure 2-10 shows a comparison to the simulated and measured current. Because the loads
power factor is typically estimated, it may need to be adjusted to improve the models
representation of the in-the-field operation.
Figure 2-11 compares the simulated and measured bus power. If feeders were attached to the
same bus as the feeder under investigation, load shapes were used to vary the other feeder loads
as well. This was done so that the substations LTCs or regulator operation takes this additional
load into consideration, and a much better bus voltage match is achieved.
The model was also checked against other measured data (vars, power factor, and so on). On
some of the circuits with AMI data, voltage measurements may have also been checked
throughout the feeder. Figure 2-12 shows an example of this. In this example, the raw AMI data
at the customer level was sampled at a 15-minute interval and compared to the voltages
computed in the model using an hourly load shape derived from AMI data. Several circuits also
had downline measurements from reclosers, voltage regulators, capacitor banks, or from other
primary metering installations. These measurement sites also led to better checks for voltages
and load allocations in circuit simulations.
2-17
Current Comparison
350
300
Current (A)
250
200
I1-Model
I1-SCADA
150
100
50
0
22000
22500
23000
23500
24000
Figure 2-10
Example of Verifying Simulated Phase Current to Measured Phase Current
60
50
Power (MW)
40
MWATT - Sim
30
MWATT - SCADA
20
10
0
0
1000
2000
3000
4000
5000
6000
7000
8000
Hours
Figure 2-11
Example of Verifying Simulated Bus Power to Measured Bus Power
2-18
1.03
1.02
Voltage (PU)
1.01
Model Voltage
AMI Voltage
0.99
0.98
0.97
0.96
20:00
8:00
20:00
8:00
20:00
8:00
20:00
8:00
20:00
8:00
20:00
Time
Figure 2-12
Example of Verifying Simulated Voltage to Measured Voltage
Substation SCADA data was also used for load allocation for those circuits that contained AMI
data. There was much difficulty in incorporating AMI data into the circuits load shapes. Some
of these difficulties include:
Unsynchronized data The data collected was not always synchronized between customers.
This skews the relative load shapes between the customers, making it difficult to match the
annual load profile measured at the substation.
Variability in sampling interval With AMI data collections, the sample time of the data
collection varied between 15 minutes and 24 hours. Most AMI supplied only included daily
power usage.
Coverage of AMI data Most of the circuits with AMI data did not have 100% AMI
coverage of customers. Some of the circuits that included 100% AMI coverage included
significant periods of time when metering data was missing or corrupt.
The AMI data was used primarily for comparing model results to measured data. More work is
needed to determine how best to use the customers AMI data for annual load shapes. These
efforts include using the available AMI data in the most efficient matter, which is an important
factor for a study such as this due to the many annual simulations that are required.
General Characteristics
This section includes several graphs that summarize various characteristics of the circuits
modeled in the study. A brief description of each figure is provided below along with some
statistical data. This section is intended to provide the reader with a concise summary of the
various circuits analyzed.
2-19
Please note that each circuit is coded with a two-letter designation that is consistent throughout
the graphs. Each of the circuits was self-selected by the utility. No statistical weighting or
normalization has been done for these circuits.
Figure 2-13 provides a breakdown, by voltage class, of the circuit lengths. Circuit length in this
graph is defined as the distance to the farthest load from the substation. The 5-kV, 15-kV, 25-kV,
and 35-kV class circuits had an average length of 2, 6.7, 16.1, and 9.4 miles, respectively. The
longest feeder was a 69-mile, 25-kV feeder, which was lightly loaded (450 kW at peak). The
shortest feeder was a 15-kV class feeder that was 1.3 miles long. This was a circuit dedicated
primarily to commercial loads.
Figure 2-14 shows a breakdown, by voltage class, of the number of customers on each of the
circuits studied. The 5-kV, 15-kV, 25-kV, and 35-kV circuits had average numbers of customers
of 372, 1355, 1981, and 2951, respectively. The largest number of customers on a single feeder
was 3885 on a 35-kV feeder. The smallest number of customers was 58 on a 15-kV class circuit
that was dedicated primarily to commercial loads.
The load factor for each of the circuits is provided in Figure 2-15. The average load factor was
46%, with a maximum and minimum load factor of 71 and 29.3%, respectively. The circuit with
minimum load factor was comprised of all residential customers, and the feeder with the
maximum load factor circuit was only 27% residential.
The load density for each of the circuits is shown in Figure 2-16. The load density is defined as
the number of customers per total length of primary line. The average load density was 76
customers per circuit mile. The circuit with the highest load density was an urban feeder, and the
circuit with the lowest load density was a long rural feeder, which would be expected.
Figure 2-17 provides the percentage of peak load to the connected kVA. The average percentage
of peak load to connected kVA was 42%, with the maximum being 101%. The circuit with the
101% loading was a 4.16-kV circuit. After the completion of the efficiency study, the utility with
the circuit having 101% loading upgraded the existing transformers with larger transformers.
Figure 2-18 shows the percentage of residential customers for each circuit. The average
percentage of residential customers for all the circuits was 76%.
2-20
5 kV
AU
BW
BU
Circuit
15 kV
BK
AE
CS
AI
AR
BC
AA
BA
BJ
AZ
BS
CO
AP
AG
BZ
CC
AO
BH
AC
BT
CQ
CE
BV
BO
AV
AJ
CD
BR
BP
AD
CM
CU
CN
CG
AB
AY
AT
AW
BB
CJ
CA
CW
CL
BD
AX
BE
BG
CR
AM
25 kV
* CH
CV
BM
AN
AQ
BQ
CT
CK
AL
35 kV
CP
AK
BI
CX
10
15
20
25
2-21
5 kV
BW
AU
BU
Circuit
15 kV
BV
BR
AV
CG
CL
BJ
BK
AW
BA
AE
BG
BS
BZ
CS
AC
AR
AT
CM
CE
BD
CJ
AZ
AX
CO
CD
CN
CC
BY
CR
CQ
AJ
AP
BT
CW
CA
BE
AD
BC
AB
BP
AY
AI
AG
BO
AO
AA
BH
CU
AM
25 kV
CV
BM
BQ
AL
CK
CT
AN
AQ
CH
35 kV
AK
CX
CP
BI
1000
2000
Number of customers
Figure 2-14
Number of Customers per Circuit
2-22
3000
Circuit
BR
CD
BH
CJ
BB
BY
CP
AI
BP
AO
BZ
AJ
BE
AC
AT
AR
BK
AA
BQ
AV
AE
AK
BI
BD
CH
BG
AL
BS
AP
BV
CV
AM
CG
CN
CQ
CM
CE
BJ
AW
AY
AB
AU
BO
BM
CU
CT
CO
AD
CX
CA
AZ
CR
AX
CW
CL
AQ
BU
BC
CK
CS
BW
CC
BT
AN
BA
AG
20
40
60
80
2-23
Circuit
CG
CL
AW
AM
AU
BU
CR
BR
BT
BW
AC
CQ
CD
CJ
AX
BE
AJ
CM
CU
AV
BD
AT
BS
BV
BA
AE
BG
AO
CN
AL
BQ
CE
CK
AK
AY
CX
BM
CT
CA
BJ
CW
BO
AB
BP
CO
BZ
AP
AQ
AN
BI
AR
BK
BY
BH
AD
CC
CP
CS
AZ
AG
CV
AA
AI
BC
CH
100
200
300
2-24
400
Circuit
BU
BY
BW
CN
CL
CW
AO
CD
BT
CA
CJ
AW
AX
CC
AV
AC
CM
BD
AT
AJ
CK
CR
CO
AB
CS
AZ
AM
AK
CE
BQ
CQ
AQ
AU
CX
BM
AL
AN
CG
BC
AD
BA
BS
AG
AA
BK
CP
BJ
AR
BV
BB
BG
BR
BP
AE
CH
AI
BZ
CT
CU
BE
CV
BI
20
40
60
80
100
2-25
Circuit
AG
CT
AN
AR
CS
AC
BV
AD
AZ
CO
BW
BY
BI
BU
AO
CK
AL
BM
BC
BT
CW
AX
BP
AA
CA
CR
AK
AT
CC
AV
CP
CG
AW
CL
BG
AB
CX
AI
AJ
BA
AU
CD
CQ
BD
CM
CE
BS
CU
CJ
BB
CN
AE
AQ
BR
BE
AM
20
40
60
80
2-26
The percent current unbalance verses the average load current for each circuit is shown in
Figure 2-19. The maximum unbalance was on Circuit AA, which was an anomaly. This circuit
had a large unbalance due to the fact that the substation transformer/bus supplied a single-phase
feeder, and the circuit studied did not carry much load on this phase to better balance the total
substation load (see Figure 2-20). The substation transformer was balanced to less than 15%;
however, the feeder that was studied (Feeder 1 in Figure 2-20) was unbalanced as much as 75%
just downstream of the bus. The CH circuit was a very long, lightly loaded circuit and had the
second most percent current unbalance.
AA
60
40
CH
20
0
0
100
200
300
Figure 2-19
Unbalance Versus Load Current
2-27
Phase B 48.4A
Feeder 2
Substation
Feeder 1
(Circuit Studied)
Phase A 79.2A
Phase B 12.8A
Phase C 66.2A
Figure 2-20
Circuit AA One-Line to Illustrate the Cause of the Unbalanced Load Current
Average power factors on circuits varied significantly, as shown in Figure 2-21. Sixty-one
percent of circuits had a power factor above 0.98, while 12% of circuits had less than a 0.9
power factor. Fifty-one percent had leading power factors. A plot of the total reactive power
available from capacitors (Mvar) verses the peak active power load (kW) is provided in
Figure 2-22. As can be seen in the plot, the amount of reactive power from capacitors tended to
be circuit-specific. Many factors likely contributed to differences in capacitor usage (load power
factors, circuit loading, circuit length, voltage level, and voltage regulation practices). The circuit
with the largest amount of reactive power from capacitors, AE, was a circuit that contained no
LTCs or regulators but depended solely on switched capacitors for voltage regulation.
2-28
14
pf = 0.90
pf = 0.98
10
8
6
4
2
1.0
0.5
0.0
0.5
1.0
1.5
0
6
6
0.78
0.9
Leading
0.98
0.98
0.9
0.78
Power factor
0.6
Lagging
Figure 2-21
Average Power Factors by Circuit
12
AE
12
10
8
6
4
2
0
0
10
15
20
Peak load, MW
Figure 2-22
Peak Load and Total Connected Capacitance
2-29
The voltage design practices for some of the circuits studied are provided in Figure 2-23 through
Figure 2-25. The design criteria shown in these graphs were obtained by surveying the utilities of
the circuits studied. Of all the utilities surveyed, approximately 67% responded to various
degrees. As expected, the utilities budgeted voltage drop across the primary and secondaries;
therefore, utilities that allowed more voltage drop on the primary allowed less voltage drop on
secondaries and vice versa. This fact is illustrated in Figure 2-26, where two utilities (A and B in
this example) had the same minimum design voltage, but the voltage drop was budgeted
differently. In this example taken from the survey results, both utilities allowed for a total
voltage drop of 12 V. Utility A allowed for a 9-V drop on the primary, leaving a 3-V drop on the
secondary, but Utility B allowed for a 6-V drop on both the primary and secondary. Note the
maximum designed voltage drop for the secondary included both the distribution transformer and
the secondary service drop.
Figure 2-24 displays the simulated minimum voltage at peak. In this figure, each of the circuits is
color coded to distinguish the individual utilities (the circuits with the same color belong to the
same utility). There are a few circuits that had what would be considered low voltages. These
voltage problems may or may not have existed in the actual circuit, but they did exist in the
peak-load model provided by the utility. As highlighted in this figure, the European circuits had
the lowest voltage. This is due to the fact that their design guidelines allow for lower voltages
compared to the guidelines used in the United States. This is can be seen in Figure 2-25, where
an 8-V drop was allowed for the European circuits on the secondary.
On the primary, the median design minimum for this set of circuits was 117 V. Assuming a
minimum goal of 114 V for the service entrance, this leaves 3 V of drop through the transformer
to the meter. Note that some of the utility circuits were not designed for the lower limit of 114 V.
There was more spread in the amount of secondary drop designed for. Most circuits were
designed for a 3-V drop on the secondary, but there was a large grouping designed for a 5- or
6-V drop. It is also important to point out that the number of circuits analyzed for each utility
varied from as little as one circuit to as many as seven circuits; therefore, the design criteria may
have been biased to those utilities with the most circuits.
2-30
14
1.0
12
8
6
4
2
0
0.8
0.6
0.4
0.2
0.0
115
116
117
118
119
120
116
117
118
119
Figure 2-23
Statistical Distributions of the Minimum Primary Design Voltage
Frequency
10
120
115
110
European Circuits
Figure 2-24
Minimum Primary Design Voltage Versus Simulated Minimum Primary Voltage
2-31
1.0
Frequency
15
10
European
Circuits
0.8
0.6
0.4
0.2
0.0
0
10
10
9
8
7
6
5
4
3
2
1
0
Utility A
Utility B
Figure 2-26
Maximum Primary and Secondary Design Voltage Drop for Two Utilities
2-32
Figure 2-25
Statistical Distributions of the Maximum Secondary Design Voltage Drop
Loss Characteristics
This section is intended to provide the reader with an overall view of the circuits loss results.
See Table 2-4 and Table 2-5 for summaries of percent losses from simulation results. Table 2-5
is based on the same data as Table 2-4, but losses were weighted by circuit results. Because more
heavily loaded circuits (generally higher load densities) had lower relative losses, the weighted
results show slightly smaller percentage losses.
The average and peak losses in both percentage and absolute values are displayed in Figure 2-27
through Figure 2-31. The losses shown in these figures are broken down into primary, secondary,
and service transformers load and no-load losses. Generally, the longer rural feeders tended to
have higher losses compared to the shorter urban feeders. Additional points of interest are as
listed:
Total losses Total distribution losses, not including the substation transformer averaged
3.64% of total consumption. Seventy-five percent of circuits had losses exceeding 2.49%,
and 25% of circuits had losses exceeding 4.35%. The largest category of losses was no-load
losses, which averaged 42% of the total average losses.
Line losses Line losses averaged just under 1.34% of total consumption. Circuit length is a
reasonably good predictor of percentage line losses. Line losses had the most spread among
circuits.
Transformer no-load losses Transformer no-load losses averaged about 1.4% of total
consumption. These losses were the most consistent across circuits, depending mainly on
transformer age and transformer utilization (connected kVA versus load).
Secondary losses Secondary losses averaged about 0.3% of total consumption. For the most
part, these tended to be low.
Demand At peak load, losses average 4.8% of consumption. Of all circuits, 75% had peak
losses exceeding 2.99%, and 25% of circuits had losses exceeding 5.87%. One circuit had
peak losses of 16.5%. At peak, 72% of losses were line losses.
The additional graphs in this section show average energy losses and peak losses in several
different ways.
Table 2-4
Distribution Circuit Loss Statistics, Percent
Quartiles
Average
25%
50%
75%
1.40
0.61
1.04
1.84
0.38
0.24
0.34
0.46
1.59
1.03
1.49
1.89
0.31
0.16
0.27
0.44
Total losses
3.64
2.52
3.09
4.32
2-33
25%
50%
75%
1.34
0.64
1.03
1.83
0.35
0.23
0.35
0.41
1.39
0.89
1.34
1.82
0.31
0.18
0.27
0.44
Total losses
3.35
2.21
2.96
3.91
60
40
Percent of Total
20
60
40
20
0
0
Losses, percent
Figure 2-27
Percent Losses by Location
2-34
Circuit
BK
BY
AA
CS
AZ
CV
AI
CC
AE
CT
BA
BP
CP
AR
CH
BW
BJ
BV
BC
CE
AG
AC
AQ
BZ
BS
BM
AO
AV
CU
BB
AD
BE
BQ
BG
BD
CM
CO
BI
AN
AU
CD
AJ
AT
AP
AM
BR
BH
BO
BU
BT
CQ
AL
CJ
AK
CG
CN
CK
CX
AW
CR
AY
CL
AX
CW
AB
CA
2-35
Circuit
CP
AZ
AK
AE
CS
BH
CC
BQ
BM
CJ
BS
AL
CM
AO
BK
BR
CV
BV
CE
BP
CX
AR
AV
BD
CD
BI
AT
CN
AC
BA
BB
BZ
BG
CK
CO
AQ
BC
AP
BJ
CU
CG
BY
CT
AW
AN
AJ
BE
CQ
CL
AG
BO
AD
BT
AY
AX
AA
CW
AI
CR
BW
CA
AB
AU
AM
BU
CH
100
200
300
Distribution losses, kW
Figure 2-29
Circuit Loss Breakdowns in Average kW
2-36
400
Circuit
BK
CS
AA
AZ
BY
CV
CC
BW
BJ
AI
CP
BC
BP
AC
AR
BA
BV
CT
CE
AE
BM
AG
AO
AQ
BQ
AV
BS
CU
BT
CD
CM
AJ
AP
BZ
CH
CO
BD
AN
BU
AT
AD
BG
BO
BB
CN
BH
BR
BE
CQ
BI
AU
CJ
AK
AM
AW
CW
CG
CR
AX
CX
AL
AY
CA
CL
CK
AB
10
2-37
Circuit
CS
CP
AZ
CC
AK
BM
BQ
AE
BK
CE
CM
CV
AO
BV
BH
BS
AV
BA
CN
AR
CX
CJ
BJ
BP
AL
AC
BC
BD
AT
BR
AQ
BI
CD
CO
AP
AN
CU
AG
AW
BG
BT
CG
CT
CK
BZ
BB
BW
CQ
AJ
CL
BY
BO
CW
AX
AD
BE
AA
CR
AY
CA
AI
AB
AU
BU
AM
CH
200
400
600
800
1000
1200
Figure 2-32 plots the peak distribution loss percentages versus the average distribution
percentage losses. As can be seen in the graph, there was a relatively linear relationship between
the two values across the circuits studied. This topic will be discussed further in the Comparing
Detailed Modeling to Peak-Case Models section below. Figure 2-33 plots the average losses
percentage verses the system voltage. As can be seen in the figure, the 15-kV class circuits had
the greatest range of losses due primarily to the fact that more 15-kV class systems were
analyzed.
2-38
14
12
10
8
6
4
2
2
10
12
Figure 2-32
Peak Versus Average Losses
*
* European Circuits
8
7
*
6
5
*
4
3
2
10
15
20
25
30
35
System voltage
Figure 2-33
Average Losses by System Voltage
2-39
Figure 2-34 plots the average distribution percentage losses versus the load density. The load
density is defined as the number of customers divided by the total length of primary circuit
miles. As can be seen in the plot, the circuits with the higher loss percentage tended to have a
lower load density. This is due to the fact that the longer rural feeders tended to have higher
losses compared to the shorter urban feeders which had higher load densities. This fact is
highlighted by the facts that the circuit with the highest percent distribution losses was a rural
feeder (BK) and the circuit with the highest load density, which was located in an urban
environment (CG), had relatively low losses.
BK
8
7
6
5
4
3
CG
100
200
300
400
Figure 2-34
Losses by Load Density
Figure 2-35 plots the average loss percentage versus the longest distance from the substation.
This length is defined as the circuit farthest load from the substation. Generally, circuit length is
a reasonably good predictor of percentage losses, in particular line losses. The outlying point
identified as Circuit CH in the plot was a unique circumstance where the feeder had a very low
average load. In this case, the circuit with the highest loss percentage was the second longest
feeder, BK circuit.
Figure 2-36 plots the average loss percentage versus the number of customers served. The
sporadic nature of the plot shows that it is difficult to predict losses based on customer count
alone.
2-40
BK
8
7
6
5
4
CH
3
2
10
20
30
40
50
60
70
Figure 2-35
Losses by Circuit Length
8
7
6
5
4
3
2
1000
2000
3000
4000
Number of customers
Figure 2-36
Losses by Number of Customers
2-41
Improvement Options
This section summarizes several improvement options that were investigated. For every circuit,
the following lower-cost improvement options were evaluated: reactive-power optimization,
phase balancing, and voltage optimization. Other options such as re-conductoring, transformer
replacements, and circuit reconfigurations were modeled based on the circuit characteristics and
requests of the utility.
The impact of load reactive power and feeder var improvement options was analyzed by
evaluating an idealized var improvement, with results given in Figure 2-37. With this approach,
the reactive power component of all loads is set to zero, and capacitors are turned off, resulting
in approximately unity power factor. For those circuits where the idealized var case shows that
significant improvements are possible, more realistic capacitor application and control were
investigated. On average, the ideal var improvements reduced the average line losses by 6.8 kW,
or 17%. The circuit with the most room for improvement (Circuit AZ) had too many fixed
capacitors. In this particular case, the circuit power factor was good at peak but operated at an
excessively leading power factor for most of the rest of the year. This increases the average
current flow on the lines.
80
AZ
60
40
20
0
1000
1000
2000
3000
Figure 2-37
Reduction in Line Losses With Ideal Var Improvement
An idealized phase-balancing case was also conducted for each circuit model. Unbalance
increases line losses because the more heavily loaded phase conductors will have much higher
losses because the line losses are a function of the current squared. There will also tend to be
higher residual currents in the neutral, leading to more losses. The idealized phase balancing is
done by re-adjusting the load to allocate it equally across phases. This forces the phases to be
balanced at the substation. This assumes that much of the line losses are in conductors close to
2-42
the substation. Figure 2-38 shows the reduction in line losses versus the amount of residual
current.
If phase balancing appeared to be an option, more realistic balancing options were attempted by
moving single-phase taps. As can be seen in Figure 2-38, some circuits actually got worse,
mainly because the balancing only balances at the substation, and this can create further
unbalance downstream. The CP circuit had the greatest line savings in average energy. In this
circuit, the average and peak unbalance was reduced to 1.5% from a base-case unbalance of
16.5%. The BS circuit had an increase in average line losses due to the fact that balancing the
circuit at peak created higher unbalances during the off peak hours.
15
Reduction in line losses, average kW
CP
10
BS
10
20
30
40
50
60
Figure 2-38
Reduction in Line Losses with Ideal Phase Balancing
Re-conductoring was also considered for several circuits, as shown in Figure 2-39.
Re-conductoring was done on a case-by-case basis. Conductor replacement was determined by
utility requests or decided by EPRI on the basis of targeting conductors with relative high I2R
losses. Figure 2-40 provides an example of an OpenDSS plot used to identify sections with the
highest line-loss densities. The conductor upgrade size was determined by the present utility
inventory or upgrade plans or arbitrarily selected for analysis reasons. Because of this variability
in re-conductoring studies, it is difficult to say much about the results, because each case was
different regarding the amount of re-conductoring done and the conductor sizes involved. With
this said, circuits that tended to be higher in line lossesrural feeders with long conductors
tended to show more improvement from re-conductoring than urban feeders.
2-43
Circuit
CP
AZ
BK
AE
BP
CC
BM
BS
CV
BB
BV
BC
AK
BJ
BA
CX
AQ
AC
CW
AP
CE
CU
CT
AI
BW
AX
AJ
CQ
AB
AW
CL
AO
BT
CO
BU
CA
AA
CR
CG
20
40
60
80
2-44
Figure 2-40
Example of OpenDSS Plot to Identify Sections With the Highest Line Losses
Voltage optimization is another approach applied to all circuits. Voltage is reduced (while still
keeping within limits) to improve end-use efficiency and reduce energy supplied. Most of the
gain is from reduced end-use consumption. The standardized voltage reduction case involves the
following assumptions:
Voltage set point = 118.5 V (this value varied on each circuit due to voltage drop beyond this
monitored point).
The 118.5-V set point was used to standardize the voltage drop across all of the circuits. This
118.5-V set point was a starting point for the analysis. It had to be verified that there were no
service voltages that dropped below 114 V with this setting. If the voltage did fall below 114 V,
the set point was raised (0.5-V increments) until there were no service voltages below 114 V.
Under certain circumstances, the service voltage was allowed to drop below 114 V. The service
drop was allowed to exceed 114 V only if it was allowed by the utility or if the base case peak
snapshot recorded a lower voltage. However, the voltage set point was never set lower than
118.5 V.
2-45
Although voltage feedback from monitored points at the end of a regulated line section is
possible, it is not always easy to implement. More commonly, line-drop compensation would be
used to control regulators. It is expected that line-drop compensation could achieve results close
to those reported with voltage feedback, and this has been verified in some cases where the linedrop compensation was modeled.
Figure 2-41 shows the reduction in load computed when this standardized voltage-optimization
scheme was used. The median reduction in energy was 2.34%, with upper and lower quartiles of
1.69 and 3.13%. Figure 2-42 shows the same information on an average kilowatt basis. While
CVR factors are expected to vary from circuit to circuit, a constant CVR factor was used in this
analysis. Because a constant CVR factor was used, these simulations mainly show how much
room there is to drop voltage across the circuit throughout the course of an annual cycle. Most of
the circuits had significant room to reduce voltage; most regulator controls had relatively high set
points, and use of line-drop compensation was unusual for the circuits in this study. On a circuit
such as CU in Figure 2-41 (the circuit with the least room for voltage improvement), the
voltage profile was already fairly optimized with the use of substation and mid-line regulators
both utilizing load-drop compensation; therefore, a smaller reduction was realized.
The AK circuit in Figure 2-42 shows the greatest reduction in average energy delivered. This
was a 35-kV circuit, which had the highest value of connected kVA for all circuits studied. For
this particular circuit, the base case had a relatively high average voltage profile across the entire
year (123.9 V) and a minimum peak primary voltage of 121.6 V. Because this circuit had
adequate voltage margin at both average and peak and a large amount of connected load, it had
potential to significantly reduce the amount of energy delivered through the implementation of
the standard voltage optimization scheme.
In contrast, the AI circuit in Figure 2-42 shows the smallest reduction in supplied energy
delivered in average kW. This was a 12.5-kV circuit which had one of the lowest value of
connected kVA and percentage of peak load per connected kVA. The base case had an average
primary voltage profile of 122.9 V across the entire year, and the minimum peak primary voltage
was 118 V. The 118-V minimum voltage at peak indicates that there was significant voltage drop
on the circuit. Therefore, the voltage profile on the circuit was not flat during peak loading
conditions. Also, the only voltage-regulation equipment on the circuit was located on the
substation bus, and the three-phase point where the end-of-line feedback was monitored for
voltage control was approximately 9.5 miles upstream of the farthest point on the circuit (see
Figure 2-43). This fact impacted how much the voltage set point can be lowered because the
additional voltage drop after the monitored point must be compensated for in the bus regulation
equipment. This increased the average voltage on the circuit compared to a case where the
voltage could be regulated closer to the minimum voltage point. With all of these factors
considered, this circuit did not have as much potential to reduce the amount of energy delivered
through the implementation of voltage optimization. Additional measures would be needed to
flatten the voltage profile on a circuit like this. Options could include adding line regulators or
adding capacitor banks.
2-46
As noted in Table 2-6, reduced end-use consumption made up 95.6% of overall energy savings
from voltage optimization. Reductions in no-load losses composed 4.1% of total savings, with
minor savings from load losses.
Table 2-6
Average Energy Savings From Voltage Optimization by Component
Voltage Optimization Energy Savings
Breakdown
of Overall Energy
Savings per
Component
Portion of Total
Savings by Category
No-Load Loss
1.6%
5.6%
4.1%
Load Loss
1.8%
0.6%
0.2%
Consumption
96.6%
2.3%
95.6%
2-47
Circuit
BI
CD
AY
BO
CC
AB
CA
AP
AK
BU
CL
BT
CQ
AG
CR
BD
CK
AL
AW
BB
CX
CH
AD
CW
CS
AX
CT
CE
CG
AZ
AT
CM
CJ
AJ
BS
BW
CP
AN
AR
AU
BH
BM
BY
AQ
AM
BV
BE
BG
CO
AA
BA
AI
BQ
BC
BP
BZ
CU
2-48
Circuit
AK
AL
CX
CP
CJ
CD
BI
BH
BD
CC
CK
CM
AZ
CL
AT
AW
AP
BS
BB
CS
CE
CG
BM
CQ
AY
BQ
BO
CW
CA
AB
AR
AX
BT
BV
CR
AJ
AN
AD
BG
CO
AG
CT
AQ
BE
BA
BP
AU
BC
BW
BY
BZ
BU
CU
AM
AA
CH
AI
100
200
300
400
2-49
3-Phase Bus
That Was
Regulated
1-Phase
Lateral
ABC
Low Voltage
Point
Reach of
Bus Voltage
Regulation
SubstationBus Voltage
Regulation
Figure 2-43
Circuit Diagram to Illustrate Reach of Voltage Regulation Monitoring Point
Figure 2-44 shows the impact of the standardized voltage optimization results on peak load. For
peak demand reduction, this standardized approach benefited fewer circuits. At peak load, many
circuits did not have room to reduce voltage. Voltage reduction may still be an option with more
careful regulator control settings and by using capacitors and/or voltage regulators to flatten
voltage profiles. The flatter the voltage profile is, the more potential there is in reducing
demand at peak load. The cases of extremely bad performance likely indicate an existing voltage
problem on the circuit that was corrected by raising voltages (the voltage problem may or may
not exist in real life, but it does exist in the peak-case model provided by the utility).
Figure 2-45 compares the average energy reduction to the peak demand reduction. Again, the
cases where the peak demand increases indicate an existing voltage problem on the circuit that
was corrected by raising voltages. The AI and AK circuits that were discussed previously are
indicated in the plot.
2-50
Circuit
AK
CX
AL
CL
AW
BI
CK
CW
CD
CM
BD
CG
BT
CA
AX
AB
AY
CR
AQ
AP
CQ
BA
AG
BH
AT
BB
BP
BO
AJ
AD
CJ
AN
BW
CS
BY
BE
BZ
CU
BU
AU
CH
AZ
CT
BS
AM
AA
AI
BG
BV
AR
CO
BC
CP
BM
BQ
CE
CC
200
200
400
600
800
2-51
800
AK
600
400
200
0
AI
200
0
100
200
300
400
Figure 2-46 shows a plot of the percent energy savings versus the average primary voltage. In
this figure, 53% of the circuits are represented in the upper right quadrant of the graph, indicating
a trend that the higher the average primary voltage, the higher the energy savings with voltage
optimization implemented. However, there are some circuits showing less energy reduction than
some of the circuits that had lower average voltages. This illustrates the fact that there are many
variables in addition to average voltage that determine the amount of energy savings that are
achievable with voltage optimization. One critical variable is where the load is concentrated on
the circuit. The figure is for the average primary voltage, but the more important voltage is the
customer voltage, which is not necessarily the same as the primary voltage.
Other factors that impact energy savings include the lowest voltage at peak, bus voltage
regulation technology (LTC or regulators), the length of the circuit that continues past the point
where the circuit is being controlled (sometimes a long portion of a single-phase lateral will
extend out beyond the three-phase point that is being monitored, impacting the overall voltage
profile, as shown in Figure 2-43), the use of line regulators and capacitors (flattens out the
voltage profile), and any other item that may impact voltage profile, such as current unbalance
and the use of step-down transformers.
2-52
3.5
3.0
2.5
2.0
1.5
1.0
0.5
118
120
122
124
126
Figure 2-46
Average Primary Voltage Prior to Reduction Versus Reduction in Energy
FLD =
Average load
Peak load
Eq. 2-1
The loss factor (FLS) is defined as the ratio of the average power loss to the peak-load power loss
during a specified period of time.7 Please note that the loss factor does not include the no-load
losses.
American Standard Definitions of Electric Terms, Group 35, Generation, Transmission and Distribution, ASA
C42.35, 1957.
2-53
FLS =
Average losses
Peak losses
Eq. 2-2
8
The loss factor is traditionally estimated by means of an approximate equation that takes the
form:
2
FLS = (1 C ) * FLD + (C ) FLD
Eq. 2-3
By means of determining the constant C, we can estimate the loss factor and therefore the
average power loss during a specified period of time (typically a year). To do this, we must also
know the losses at the peak hour, because by definition the loss factor has the peak-load power
loss as its denominator. This C term is often referred to as the C-factor.
Some common approximations are for determining the loss factor based on load factor are:
(a C-factor of 0.85)
Eq. 2-4
(a C-factor of 0.7)
Eq. 2-5
Figure 2-47 compares the loss factor evaluated using Fls = 0.3Fld + 0.7 Fld 2 with the loss factor
calculated from the detailed simulations. The results are fairly linear but do not line up on a oneto-one basis. The loss factors from the detailed modeling are less than the estimate based on the
load factor.
More recently, Gustafson and Baylor10 proposed the following estimate for the loss factor:
(a C-factor of 0.92)
Eq. 2-6
This model aligns closely to the majority of results in this study, as shown in Figure 2-48. The
circuits that deviated the most from this estimate are highlighted in the graph.
The outliers in Figure 2-48 tend to have characteristics that differ from the way the majority of
circuits were analyzed or contain unique features that altered line losses. These differences are
mostly centered on the way the loads vary throughout the year. If each load was scaled the same
throughout the annual simulation where the average-to-peak current ratio (Iaverage/Ipeak) was
approximately the same across the circuit, the C-factor remained closer to unity (typically 0.9
to 1). The majority of circuits in the Green Circuits project were analyzed in this way where a
peak load allocation was used, and all loads would scale proportionality to match the substations
8
F.H. Buller and C.A.Woodrow: Load Factor Equivalent Hour Values Compared, Electr. World, vol. 92, no. 2,
July 14, 1928, pp. 5960.
9
M. W. Gangel and R. F. Propst, Distribution Transformer Load Characteristics, IEEE Transactions on Power
Apparatus and Systems, vol. 84, 1965, pp. 67184.
10
M. W. Gustafson, J. S. Baylor, and S. S. Mulnix, The Equivalent Hours Loss Factor Revisited, IEEE
Transactions on Power Systems, vol. 3, 1988, pp. 15028.
2-54
SCADA measurement. In addition to load allocations, other factors influenced the C-factor on
other circuits, such as varying distributed generation, fixed loading, and no-load current
contributing significantly to the I2R losses. All of these factors are highlighted in a short
description of the outlying circuits shown in Figure 2-48:
The circuit with the lowest C-factor, Circuit CH, was a very long (69 miles) and lightly
loaded (< 7 A average line current). In this circuit, the line capacitance was a significant
portion of the average line current. This line capacitance caused a large percentage of the line
losses (I2R).
The circuit with the highest C-factor, Circuit BJ, was a circuit that included a distributed
generator located on the feeder, which varied in power output throughout the year. For this
circuit, there was no contribution from the generator during the peak demand hour, which
resulted in a lower loss factor (losses at peak were much greater than average losses) than if
the generator were present at peak.
The other circuits that deviated the most from the 0.92 C-factor were the AE, AB, CL, and
AZ circuits. The AE and AB circuits included a large percentage of industrial and
commercial customers, and loads were allocated based on load class (each load class had its
own load shape, which varied independently relative to the other class load shapes). Other
circuits contained a large percentage of industrial and commercial customers, but they were
not allocated based on load class, and their results were closer to the 0.92 C-factor. The CL
and AZ circuits both had fixed loads on the feeder (some loads did not vary with the annual
load shape).
Figure 2-49 plots the annual losses evaluated by both the detailed model loss-factor and with the
loss-factor estimate using the Gustafson-Baylor equation. The load losses in these examples
include both line losses and transformer load losses. The outliers from Figure 2-48 are
highlighted in this graph. Note that the outliers with the higher line losses deviated more from
the estimate than the outliers with relatively low line losses. With the exception of the BJ circuit,
the outlying circuits simulated annual losses were higher than the predicted losses.
2-55
0.5
0.4
0.3
0.2
0.1
0.1
0.2
0.3
0.4
0.5
0.6
Figure 2-47
Comparing Estimated Loss Factors to Loss Factors From Detailed Modeling
(C-Factor = 0.7)
2-56
0.5
0.4
CH
AE
0.3
AB
CL & AZ
0.2
BJ
0.1
0.1
0.2
0.3
0.4
0.5
0.6
Figure 2-48
Comparing Estimated Loss Factors Using the Gustafson-Baylor Model to Loss Factors
From Detailed Modeling
2-57
6
AZ
AE
BJ
CH
CL
AB
Figure 2-49
Percent Losses From Detailed Modeling and Peak-Case Modeling (C-Factor = 0.92)
It is difficult to determine how the estimated loss factor translates to actual system losses. As the
simulation results indicate, load allocations and load variation have an impact on annual losses.
Also, predicting annual losses based on peak may not provide insight into energy-saving options.
Some energy-saving options such as capacitor control and voltage optimization may not show a
savings at peak but do provide average annual savings.
Conclusions
The modeling for the Green Circuit project included elements not typically represented in
distribution feeder models. Traditionally, feeder models include only the components of the
primary distribution system up to but not including the service transformer. Because of this, the
primary-side model is usually more refined. Therefore, the primary losses, voltage drop, and
voltage profiles are considered to be reasonably accurate.
With the inclusion of the service transformers and service conductors, more approximations are
introduced into the model. The use of typical values versus actual values for customer services
varied from circuit to circuit, depending on the available information.
2-58
Because traditional models typically only analyze peak-demand power flows, the inclusion of
time-varying loads also adds additional variables to the modeling. For the annual simulation,
estimates were made on the way the load was represented. In particular, the power factors for the
loads were estimated based on assumptions from the utility and/or estimated from modeling. The
time variation of the load power factors was also not considered for the most part. Load
representation will continue to improve as more is learned from voltage optimization field trials
and laboratory testing of equipment.
Some work was conducted on incorporating AMI data into some of the models; however, much
work still remains in determining the optimal usage of AMI data.
The set of circuits analyzed was very diverse. Circuits varied based on design approach, voltage
class, circuit lengths, load density, and so on. Although the circuits varied a great deal, certain
general characteristics could still be identified. The circuits that did not follow these general
characteristics provided additional insight.
The traditional approach of predicting losses based on load profile and peak losses matches the
majority of circuits analyzed. However, it is difficult to determine how the estimated loss factor
translates to the actual system losses.
Having the capability to perform chronological annual simulations helped evaluate efficiency
options that cannot be modeled by simply using an estimate from the peak case:
The optimal efficiency improvement approach depended on the circuit. On average, the
re-conductoring and ideal var control resulted in the greatest reduction in losses.
Voltage optimization applied full-time provided the most energy reduction by improving end-use
efficiency as well as reducing no-load losses. This option provided benefit on almost all circuits.
The circuits that showed the most improvements with voltage optimization had sufficient voltage
margin already existing in the feeder. Additional improvement is possible by flattening voltage
profiles by phase balancing, circuit reconfigurations, or with additional voltage regulators or
capacitors.
2-59
3
EXTENDED ANALYSIS AND ECONOMIC
COMPARISONS
Six circuits from the Green Circuit project were selected for extended analyses. The goal of the
extended analyses was to optimize efficiency using circuit enhancements that reduce losses and
improve voltage profiles. Improving voltage profiles allows greater reduction of circuit voltage,
which improves end-use efficiency and reduces overall consumption.
The six circuits represent different topologies, operating conditions, and control strategies.
Table 3-1 provides information on the six circuits.
Table 3-1
Extended Case Study Circuits
Circuit
Circuit A
Circuit B
Circuit C
Circuit D
Circuit E
Circuit F
Primary
Voltage
12.5 kV
34.5 kV
25 kV
12.47 kV
25 kV
13.2 kV
Voltage
Regulation
Substation and
Three on
Feeder
Substation
Substation
Substation
Substation
Substation
Primary
Conductor
Overhead
Overhead
Overhead
Underground
Overhead
Overhead
Total Primary
Circuit Miles
104.8 mi
72.7 mi
43.8 mi
15.4 mi
29.7 mi
84.3 mi
Farthest
Distance from
Sub
10.5 mi
4.6 mi
9.2 mi
2.0 mi
8.5 mi
8.5 mi
Total Reactive
Compensation
3900 kvar
2400 kvar
0 kvar
3600 kvar
2250 kvar
12750 kvar
Peak Demand
12.6 MW
14.6 MW
7.1 MW
5.7 MW
5.7 MW
29.2 MW
Load Factor
0.33
0.38
0.32
0.40
0.39
0.55
Modeled Load
Power Factor
0.98
0.93
0.93
0.92
0.85
0.90
3-1
Analysis Approach
Each model circuit was built in the OpenDSS simulation platform. The OpenDSS tool solves the
8760 hour (annual) circuit response with controls such as voltage regulators and capacitor banks,
adjusting to the hourly load. Losses, load, and voltage were monitored for all hours of the
simulation. The analysis approach presented in this chapter essentially follows that described in
Chapter 2, with the exception that more efficiency options were evaluated in a consistent fashion
to allow better economic comparisons.
Five primary categories of efficiency improvements were analyzed for each circuit in this
extended analysis. The categories and label abbreviations include:
VR Voltage reduction
The order of improvements listed above is generally a good order in priority for circuit
improvements; the benefits tend to accumulate. In the analysis approach, each option was
evaluated separately. Combinations of promising options were also evaluated. Normally,
simulations of combinations resulted in the same answer as linearly adding the energy savings
from each component.
By combining voltage reduction with other efficiency options, losses were reduced while voltage
profiles were flattened. This improved the distribution system efficiency (lower losses) and enduse efficiency (lower end-use consumption).
In addition to the options considered in the case studies in this chapter, efficiency improvements
on a specific feeder could include:
More significant reconfigurations (adding new feeders and redistributing load to new feeders,
for example)
Model verification and improvement may be needed to implement an efficiency study. In the
Green Circuits project, most utilities needed to field check phasing of models along with settings
of capacitor and voltage regulator controls. Many utilities also needed to upgrade their circuit
models.
Longer-term efforts to improve efficiency may involve upgrading monitoring, substation feeder
metering, GIS systems, distribution modeling databases, and communication systems. For
example, changing planning criteria for allowable secondary voltage drop is another long-term
approach to improving voltage profiles to improve efficiency.
3-2
Voltage feedback was used to model voltage reduction. Regulators were controlled to a
monitored remote voltage measurement at the end-of-line three-phase primary bus to 118.5 V
(on a 120-V base). The set point was adjusted for each efficiency option to maintain greater than
118 V on all primary feeder buses during the annual simulation. If the circuit voltage prior to
voltage reduction was at the limitthus restricting the application of voltage reductionminor
circuit modifications were made to bring all voltages within limits.
Phase balancing reduces line losses and evens out voltage drops among phases, which helps
flatten voltage profiles. Phases were balanced by moving a single-phase tap to another phase,
rotating a three-phase tap, converting a single-phase lateral to two-phase, or splitting a singlephase lateral with the isolated segment reconnected at a new location. Other simple
reconfigurations (like changing open tie positions) were also considered.
Reactive power optimization reduces reactive flows while stabilizing voltage profiles. Options
evaluated for reactive power optimization included the removal of capacitor banks, reducing
capacitor bank size, and adding switched reactive power control algorithms. For each case study,
capacitor banks were also optimally applied from scratch. In the optimal capacitor application,
15-kV class circuits used 300- or 900-kvar banks, while the rating doubled for 25-kV class
circuits. New capacitor banks were placed on the circuit based on a light-load case (500 kvar
above minimum reactive demand) and/or a high-load case (500 kvar below maximum reactive
demand). In the light-load case, fixed capacitor banks were placed on the circuit such that the
capacitor bank serves 50% reactive load upstream and 50% downstream. In the high-load case,
switched capacitor banks were placed on the circuit to provide 50% support upstream and 50%
downstream.
Re-conductoring reduces losses and reduces voltage drop along that section of circuit. For each
re-conductoring analysis, a new conductor was chosen with impedance not below the lowest
existing in the circuit. It was assumed that the new conductor replaces the existing conductor
without the need to require new right-of-ways, but pole replacement may be required if the
conductor size significantly increases. The amount of circuit replaced in the re-conductoring
options was limited to approximately 1-, 2-, or 3-mile sections so that comparisons can be made
across the different circuits. However, size and cost will vary for the different circuits.
Additional voltage regulators also help flatten voltage profiles. Because line regulators control
each phase independently, the use of line regulators helps compensate for unequal phase loadings
and can be particularly beneficial on circuits that have significant voltage drop on long singlephase laterals. Voltage regulators were not tested as an option on circuits with relatively flat
voltage profiles.
In addition to calculating efficiency gains, the extended case study analyzed the economic
viability of each efficiency option. The economic analysis incorporated many financial factors
discussed in the next section.
This chapter shows results from extended case studies on six circuits based on using annual
simulations using OpenDSS. See Appendix B for a similar approach that can be implemented
with a traditional load-flow program using peak snapshot cases, including an example for one
circuit.
3-3
Financial Factors
The system upgrade costs associated with each efficiency option were compared to the energy
savings in the economic analysis. For proper comparison over the lifetime of each project, the
net present value (NPV) was calculated. The system upgrade costs included the investment costs
in addition to those for operation and maintenance.
The two metrics used to determine the economic acceptability of each project included the
benefit/cost ratio (BCR) and levelized cost (LC). The benefit-cost was determined with the
project cost, value of energy saved, and value of peak demand reduction over the lifetime of each
project. The levelized cost was determined with the project cost levelized by the energy saved
over the lifetime of each project and does not include potential benefit from total peak demand
reduction. An economically viable project has a benefit-cost value greater than one or a levelized
cost less than or equal to the maximum marginal cost of purchase power.
A mini survey sent to several Green Circuit participants revealed their average marginal cost of
purchase power, as shown in Table 3-2. The maximum marginal cost of purchase power used in
this analysis to determine economically acceptable projects was $0.08/kWh.
Table 3-2
Average Marginal Cost for Several Green Circuit Participants
Participant
Cents/kWh
2.55
7.38
9.60
2.40
3.50
6.00
The economic analysis used the parameters and values listed in Table 3-3 to determine present
worth of the efficiency option costs and savings. The present worth of the option costs depends
on the investment cost, fixed charge rate, annual operation and maintenance expenses, associated
inflation rates, capital and planned equipment life, and the present worth rate. The present worth
of the efficiency option savings were based on annual energy and peak demand reduction, energy
and demand rates, associated inflation rates, planned equipment life, and present worth rate.
Additional explanation can be found in Appendix B.
The economic analysis did not include impacts on reduced kWh sales associated with voltage
reduction. This analysis assumed that billing losses were recovered by adjusted billing strategies
or other rate-recovery mechanisms.
3-4
$0.0736
$49.00
0.160
3.0%
3.0%
4.0%
4.0%
2.0%
6.0%
5.0%
35
15
The general construction cost estimates used for the economic analysis are shown in Table 3-4.
The voltage reduction cost estimate included a single three-phase voltage regulator, voltage
feedback control, and installation. The cost was based on several estimates for individual circuit
control. However, bus-level control or alternate control algorithms may add/reduce the cost
estimate during actual implementation. Tap adjustment cost estimates were derived from several
work order bids and reflect the labor cost for adjusting a single tap. Capacitor costs were derived
from vendor cost tables. All reactive power optimization options were based on the assumption
that new capacitors would be utilized regardless of existing capacitors on each circuit. The
existing capacitors were assumed to have remaining life, and therefore a salvage value was
assigned. The cost of new conductor was also derived from vendor price sheets. For significant
conductor size upgrades, pole replacement was assumed. Re-conductoring labor costs and pole
replacement/labor costs were based on several work order bids.
3-5
$63.6 K
$1.4 K
Underground
$2.8 K
Capacitors
300-kvar Fixed
$3,000.00
600-kvar Fixed
$5,175.75
900-kvar Fixed
$8,000.00
1200-kvar Fixed
$7,597.00
300-kvar Switched
$4,500.00
600-kvar Switched
$12,573.33
900-kvar Switched
$12,000.00
1200-kvar Switched
$23,084.50
Re-conductoring
Conductor
477 kcmil AA
$2,000.80/kft
556 kcmil AA
$2,328.64/kft
795 kcmil AA
$3,481.60/kft
Conductor installation
Voltage regulators
3-6
Single-phase 100 A
$15,000.00
Three-phase 100 A
$22,000.00
Three-phase 219 A
$42,600.00
Three-phase 328 A
$50,650.00
Three-phase 548 A
$62,000.00
3-7
Losses
Consumption
Total Energy
Table 3-5
Efficiency Project Annual Energy Savings
Ckt A
Ckt B
Ckt C
Ckt D
Ckt E
Ckt F
Ckt A
Ckt B
Ckt C
Ckt D
Ckt E
Ckt F
Ckt A
Ckt B
Ckt C
Ckt D
Ckt E
Ckt F
Increased Savings
R
R
PB
PB
R
VAR
REG
PB
R
PB
PB
PB
PB
R
R
REG
VAR
R
PB
R
VAR
R
R
R
R
VAR
R
REG
VAR
VAR
PB
VAR
PB
REG
VAR
VAR
VAR
VAR
R
R
VAR
VAR
VAR
R
REG
VAR
PB
PB
VAR
PB
REG
PB
Base Line
VR
VR
VR
VR
VR
VR
VR
VR
VR
VR
VR
VR
VR
VR
VR
VR
VR
VR
3-8
Phase balancing on the densely loaded Circuit C and Circuit D had the highest potential to
reduce total peak demand via end-use load reduction. Re-conductoring and reactive power
optimization also occurred as the most effective in reducing total peak demand by the combined
influence of reducing peak load and losses. Re-conductoring on all circuits had the potential for
additional peak hour savings.
Table 3-6
Efficiency Project Peak Demand Savings
Losses
Load
Total Demand
Increased Savings
Ckt A
R
PB
VAR
Ckt B
VAR
R
PB
Ckt C
PB
R
VAR*
Ckt D
PB
VAR
R
Ckt E
R
Ckt F
VAR
REG
R
Ckt A
R
PB
VAR
Ckt B
VAR
PB
Ckt C
PB
R*
VAR*
Ckt D
VAR
PB
R
Ckt E
Ckt F
VAR
REG
Ckt A
R
VAR
PB
Ckt B
R
VAR
PB
Ckt C
R
VAR
PB
Ckt D
R
PB
VAR
Ckt E
R
PB
Ckt F
R
VAR
REG
PB
* Increase in total demand, load, or losses compared to the base case.
Base Line
VR
VR
VR*
VR
VR
VR
VR
VR
VR*
VR
VR
VR
VR
VR
VR*
VR
VR
VR
Economic Analysis
Economic Acceptability
The economic viability or acceptability of each efficiency option was strongly correlated to the
consumed energy and peak demand on each circuit. The benefit-cost ratio and levelized cost for
each efficiency option was also significantly influenced by the efficiency gains from voltage
reduction. The voltage reduction option was used as the baseline to compare economic
acceptability of each efficiency option in Table 3-7. All options were economically acceptable
except those indicated with an asterisk. For this set of circuits, levelized costs and benefit-cost
ratio had the same order of economically acceptable options. For three circuits, phase balancing
or reactive power optimization increased acceptability beyond that from voltage reduction alone.
This increase was due to low additional project costs and/or high additional efficiency gains.
3-9
Phase balancing tended to be viable for most circuits primarily due to low cost and because
phase balancing allows additional voltage reduction to save more energy. Reactive power
optimization tended to work for similar reasons, but the cost was slightly higher, so its ranking
was slightly lower. Re-conductoring and additional voltage regulators tended to be closer to
break-even because the capital cost was higher.
Table 3-7
Economic Acceptability of Efficiency Options
Increased Acceptability Base Line
Ckt A
PB
VAR
VR
Ckt B
VR
Benefit
Ckt C
PB
VR
Cost
Ckt D
VAR
PB
VR
Ratio
Ckt E
VR
Ckt F
VR
Ckt A
PB
VAR
VR
Ckt B
VR
Levelized Ckt C
PB
VR
Cost
Ckt D
VAR
PB
VR
Ckt E
VR
Ckt F
VR
* Economically unacceptable project.
Decreased Acceptability
R
VAR
VAR
R*
PB
VAR
R
VAR
VAR
R*
PB
VAR
PB
R*
VAR
PB
REG
REG
PB
R*
VAR
PB
REG
REG
R*
R
R*
R
Figure 3-1 illustrates the levelized cost for all acceptable projects analyzed for each circuit.
Multiple projects were tested under each category. The levelized cost for voltage reduction alone
(base voltage feedback) decreased as peak demand/annual energy increased. Phase balancing,
reactive power optimization, re-conductoring, and additional voltage regulation then slightly
shifted the levelized cost based on additional project cost and energy savings. For the three
circuits with highest peak demand and lowest levelized cost for voltage reduction (Circuit A,
Circuit B, and Circuit F), re-conductoring became an acceptable option. The significant decrease
in levelized cost for the reactive power optimization project on Circuit D was a result of low
project cost due to the salvage value from the removal of three 1200-kvar pad-mounted capacitor
banks used for transmission system support.
Table 3-8 and Table 3-9 show the best benefit-cost ratio and best levelized cost for each category
of options for each circuit.
3-10
Figure 3-1
Acceptable Efficiency Projects With Respect to Levelized Cost
Table 3-8
Benefit-to-Cost Ratio for the Best Option in each Category
Circuit
A
8.1
17.6
3.5
1.8
3.6
29.9
Phase balancing
9.8
17.0
3.9
2.3
3.4
25.2
Var optimization
9.5
17.2
3.3
4.1
3.1
29.0
Reconductoring
4.7
2.4
0.2
0.4
0.7
15.6
3.1
22.4
2.1
5.3
Voltage regulators
Combinations
11.3
10.8
3.8
4.7
3-11
Table 3-9
Levelized Cost in /kWh for the Best Option in each Category
Circuit
A
1.1
0.6
2.3
5.4
2.8
0.3
Phase balancing
0.9
0.6
2.3
4.4
2.9
0.4
Var optimization
0.9
0.6
2.5
2.3
3.0
0.3
Reconductoring
2.0
4.1
47.3
24.1
14.0
0.6
3.2
0.4
4.6
1.7
Voltage regulators
Combinations
0.8
0.9
2.4
2.1
Table 3-10 shows the project categories that are economically acceptable without savings and
cost accrued from voltage reduction option VR. Because voltage reduction was applied in all
options, projects could not be completely decoupled. These projects would be acceptable as a
standalone project after voltage reduction has already been implemented. These projects included
reactive power optimization, phase balancing, additional voltage regulation, and the combination
(C) of phase balancing and reactive power optimization. Reactive power optimization had the
highest additional benefit for circuits under 15 kV due to greater savings from loss reduction. For
circuits greater than 15 kV, the benefit was greatest from phase balancing (with the exception of
Circuit E).
Table 3-10
Acceptable Efficiency Projects Without Considering Savings From Option Voltage
Reduction
Highest
Project
Viability
After
Option
VR
Circuit A
VAR
Circuit B
PB
Circuit C
Lowest
C
PB
PB
VAR
Circuit D
VAR
PB
Circuit E
None
Circuit F
VAR
REG
3-12
The voltage reduction option can be applied to provide demand reduction only at peak or near
peak hours. The peak hour demand reduction of 168 kW equates to benefit-cost ratio = 1 when
assuming an average marginal purchase demand rate of $49/kW/yr and $114K for voltage
reduction net present value ($63.6K initial investment cost). Based on these assumptions, the
voltage reduction option was acceptable only due to peak hour demand reduction for Circuit B
and Circuit F, where the peak hour reduction was 382 kW and 361 kW, respectively. Due to less
than 168 kW peak hour savings on Circuit A, Circuit D, and Circuit E, additional annual energy
savings would be necessary for voltage reduction acceptability. Circuit A would need 78 MWh
annual energy reduction (9% of that possible when annually applying voltage reduction), Circuit
D would need 87 MWh annual energy reduction (49% of that possible when annually applying
voltage reduction), and Circuit E would need 57 MWh annual energy reduction (16% of that
possible when annually applying voltage reduction). The peak demand and losses increased with
voltage reduction on Circuit C due to voltage regulation-up at peak hour.
Peak hour savings are influenced by feeder voltages and load. The feeder voltage for Circuit B
was relatively high and stiff, thus allowing for significant peak hour voltage and demand
reduction. Circuit F had high circuit load and thus allowed for significant demand reduction with
a smaller change in regulated voltage.
Economic Parameter Sensitivity
Levelized cost (LC) by definition has fewer economic factors than benefit-cost ratio because LC
is normalized to energy saved rather than value of energy saved. The benefit-cost ratio (BCR)
incorporates the energy and demand rates to define the monetary value of reduced energy and
peak demand. The sensitivity of these to economic parameters is shown in Table 3-11. The table
expresses the direction of change of benefit-cost ratio and levelized cost when each economic
parameter was increased individually. The value Constant is given if the benefit-cost ratio or
levelized cost was independent of the parameter.
Table 3-11
Effect of Parameter Value Increase on Benefit-Cost Ratio and Levelized Cost
BenefitCost Ratio
Levelized
Cost
Down
Up
Up
Constant
Up
Constant
Down
Up
Down
Up
Up
Constant
Up
Constant
Down
Up
Option Cost
Up
Down*
Down
Constant
Constant
Constant
Up
Down
3-13
The economic parameters were individually varied for the evaluation of Circuit D voltage
reduction option. The parameters were varied by 0.5 and 2 times the default value. The benefitcost ratios with each modified parameter are shown in Table 3-12, and the new levelized costs
are shown in Table 3-13. The benefit-cost ratio and levelized cost are also shown for the default
parameters. The option cost, fixed charge rate, energy saved, and marginal cost of energy
dictated the acceptability of the efficiency option.
Table 3-12
Benefit-Cost Ratio Resulting from Scaling Individual Parameter Values by 2 and 0.5 for
Circuit D Option VF
Benefit-Cost Ratio
3-14
2x
Parameter
1x
Parameter
0.5x
Parameter
0.903
1.806
3.613
3.388
1.016
2.031
1.694
3.388
1.016
2.031
1.694
0.968
3.188
1.747
1.831
2.427
1.587
1.895
1.775
1.594
1.935
2.589
1.465
1.265
2.210
2.323
1.560
Table 3-13
Levelized Cost Resulting From Scaling Individual Parameter Values by 2 and 0.5 for Circuit
D Option VF
Levelized Cost
2x
Parameter
1x
Parameter
0.5x
Parameter
0.108
0.054
0.027
0.027
0.108
0.101
0.031
0.056
0.053
0.061
0.050
0.048
0.058
The benefit-cost ratio was inversely proportional to project cost and fixed charge rate; linearly
proportional to energy/demand saved, energy/demand rates, and planned life; and nonlinearly
proportional to the additional economic parameters. The levelized cost was linearly proportional
to project cost, fixed charge rate, and maintenance cost; inversely proportional to energy saved;
and nonlinearly proportional to the additional economic parameters.
There was a linear relationship between benefit-cost ratio for voltage reduction and peak
demand, as shown in Figure 3-2, for the six circuits, and there was an inverse relationship
between levelized cost for voltage reduction and peak demand, as shown in and Figure 3-3. The
relationship was approximately linear (as shown by the linear curve fit with a slope of 1.2) for
benefit-cost ratio because the potential benefit from energy reduction was in the numerator. The
relationship was approximately inverse for levelized cost (as shown by the power curve fit with
exponent -1.5) because the potential benefit from energy reduction was in the denominator. The
economic acceptability of voltage reduction also increased with average hourly demand, as
shown in Figure 3-4.
These relationships occurred because the potential benefit from voltage reduction depended on
the load demand. The same percent reduction in voltage will have higher potential benefit from
higher initial demand. The benefit-cost ratio and levelized cost were not perfectly linear or
inverse due to the magnitude of voltage reduction achieved on each circuit. Also, bus regulated
substations may be more economical because voltage control on all feeders may be achieved
with a single voltage controller. Feeder-level voltage regulation, however, may be more effective
because the voltage on each feeder will be controlled more effectively. One feeder with a large
voltage drop will not limit voltage reduction on parallel circuits.
3-15
BCR
20
15
10
5
0
5
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
P e ak D e mand (M W)
Figure 3-2
BCR (Linear Curve Fit) Relationship With Peak Demand
0.06
0.05
LC
0.04
0.03
0.02
0.01
0.00
5
6 7
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
P e ak D e mand (M W)
Figure 3-3
LC (Inverse Curve Fit) Relationship With Peak Demand
3-16
0.06
30
0.05
25
BCR
20
0.03
LC
BCR
0.04
15
LC
Linear (BCR)
Power (LC)
0.02
10
0.01
5
0
0.00
2
10
11
12
13
14
15
16
Figure 3-4
BCR (Linear) and LC (Power) Relationship With Average Hourly Demand
Application of Results
There are several ways to use the results of the efficiency/economic analysis to optimize
efficiency. Options can include:
Highest benefit-to-cost ratio Pick the option or combination of options with the highest
overall benefit-to-cost ratio. Combination options that target unique efficiency areas
approximately add linearly with respect to total cost and energy savings.
Incremental benefit-to-cost ratio greater than one Pick an option whose incremental
benefit-to-cost ratio is greater than one. This maximizes return on incremental investment for
project components and must include voltage reduction.
Largest efficiency benefit Pick the option or combination of options that saves the most
energy. Even though this may have a lower benefit-to-cost ratio, it squeezes the most
efficiency out of the system. This is most applicable for a distribution company that can
effectively sell its kilowatt-hour savings to an efficiency group. By extracting the most from
this distribution resource, more is invested in the distribution system.
The most appropriate strategy will depend on the utilitys goals, regulations, incentive programs,
billing, and more. Table 3-14 shows optimal options for each circuit based on two criteria: one
with the lowest levelized cost (and highest benefit-to-cost ratio) and one with the most energy
savings with a benefit-to-cost ratio greater than two. Note that optimal options varied widely
depending on circuit and economic criteria.
3-17
Circuit
A
Levelized
Cost
/kWh
Benefitto-Cost
Ratio
Energy
Savings
Economic Criteria
Option
Lowest levelized
cost
0.8
11.3
3.5%
Maximum efficiency
3.5
2.9
3.9%
Lowest levelized
cost
Voltage reduction
0.6
17.6
3.5%
Maximum efficiency
Phase balancing
0.6
17.0
3.6%
Lowest levelized
cost
Phase balancing
2.3
3.9
2.2%
Maximum efficiency
2.4
3.8
2.3%
Lowest levelized
cost
2.1
4.7
1.3%
Maximum efficiency
2.1
4.7
1.3%
Lowest levelized
cost
Voltage reduction
2.8
3.6
1.8%
Maximum efficiency
Voltage regulators
3.9
2.5
1.9%
Lowest levelized
cost
Voltage reduction
0.3
29.9
2.4%
Maximum efficiency
Voltage regulators
0.6
16.2
2.6%
Substation
Capacitors () Regulators ()
Figure 3-5
Circuit A Map
3-19
3-20
Option Details
Voltage Reduction
VR
Phase Balancing
PB1
PB2
Var Optimization
VAR1
VAR2
+ Remove one 900-kvar fixed and one 900-kvar switched capacitor bank
+ Add one new 900-kvar switched capacitor bank on var control
VAR3
VAR4
Re-conductoring
R1
R2
R3
R4
R5
Combinations
C1
+ PB2
+ VAR1
C2
+ PB2
+ VAR1
+ R1
+ R5
3-21
The total annual energy saved and peak demand reduction for the efficiency options are shown in
Figure 3-7. These savings include reduction in end-use load and losses. The majority of
efficiency option savings were due to reduction in end-use load by voltage reduction, as shown
by option VR. Additional reduction in consumption was also possible by flattening the feeder
voltage profile and applying further voltage reduction. In particular, phase balancing option PB2
allowed for reduction in regulator voltage set point after load was transferred off the heaviest
loaded/voltage constrained phase. The reduction in consumption, however, was limited due to
multiple regulation zones caused by feeder regulators and circuit branching. Buses beyond feeder
regulators and on feeder spurs were not necessarily controlled by the substation voltage
reduction feedback control. The selection of one voltage feedback point was difficult when the
circuit had a lot of branching. The control algorithm could utilize an extra factor to account for
the fact that some other phases on other branches are lower.
Load location is also very important. To manage consumption, the voltage must be reduced at
the location of the load. Voltage drop on the front part of the circuit was not much of an issue
because the load is primarily at the ends of the circuit, and there were regulators that separated
the zones. Because of different regulation zones, line losses and consumption were also
decoupled. Near substation var improvements or re-conductoring reduced losses yet did not
provide significant reduction in consumption. End-of-line balancing allowed lower voltage and
reduced consumption yet did not significantly reduce line losses.
On circuits with a lot of branching, switched capacitor banks beyond taps may not have switched
on sequentially, which left less vars online than expected over some parts of the year. The
smaller 300-kvar capacitor banks were also found to be less effective than the 900-kvar banks.
The combined efficiency improvement from multiple options occurred due to the individual
options ability to target and reduce specific losses and energy use. When this occurred, the
overall efficiency improved approximately linearly. The combined loss reduction from var
optimization and phase balancing in combination option C1 was due to reduction in line losses
from var optimization and reduction in consumption from phase balancing. Combination option
C2which combines phase balancing, var optimization, and re-conductoringled to the highest
overall annual efficiency benefit. The detailed change in energy/demand and load/losses for all
options are given in Table 3-16.
3-22
600
1400
1200
400
1000
800
300
600
200
500
400
100
200
0
0
VR
PB1
R1
OPTION
Total Energy Saved MWh
R2
R3
R4
R5
C1
C2
Figure 3-7
Annual Energy Saved and Peak Demand Reduction
3-23
Table 3-16
Annual and Peak Savings for End-Use Load and Losses
Option
Annual
Consumption
Saved
MWh
(% Base)
Annual
Losses
Saved
MWh
(% Base)
Peak
Demand
Saved
kW
(% Base)
Peak
Losses
Saved
kW
(% Base)
Total Annual
Energy Saved
MWh
(% Base)
Total Peak
Reduction
kW
(% Base)
VR
832.2 (2.4%)
36.8 (1.8%)
33.0 (0.0%)
18.3 (0.2%)
869.0 (2.4%)
51.3 (0.0%)
PB1
849.7 (2.5%)
35.0 (1.7%)
81.0 (0.1%)
20.2 (0.2%)
884.8 (2.4%)
101.2
(0.1%)
PB2
1068.7 (3.1%)
45.6 (2.3%)
106.0 (0.1%)
23.0 (0.3%)
1114.3 (3.1%)
129.0
(0.1%)
VAR1
832.2 (2.4%)
201.5
(10.0%)
24.0 (0.0%)
17.7 (0.2%)
1033.7 (2.8%)
41.7 (0.0%)
VAR2
849.7 (2.5%)
176.1 (8.8%)
43.0 (0.0%)
17.0 (0.2%)
1025.8 (2.8%)
60.0 (0.1%)
VAR3
849.7 (2.5%)
225.1
(11.2%)
81.0 (0.1%)
31.6 (0.3%)
1074.9 (3.0%)
112.6
(0.1%)
VAR4
858.5 (2.5%)
212.0
(10.6%)
65.0 (0.1%)
26.3 (0.3%)
1070.5 (2.9%)
91.3 (0.1%)
R1
841.0 (2.4%)
135.8 (6.8%)
103.0 (0.1%)
93.0 (1.0%)
976.7 (2.7%)
196.0
(0.2%)
R2
841.0 (2.4%)
220.8
(11.0%)
148.0 (0.1%)
165.0
(1.8%)
1061.7 (2.9%)
313.0
(0.3%)
R3
876.0 (2.5%)
310.1
(15.5%)
255.0 (0.3%)
230.0
(2.5%)
1186.1 (3.3%)
485.0
(0.4%)
R4
832.2 (2.4%)
148.9 (7.4%)
55.0 (0.1%)
68.1 (0.7%)
981.1 (2.7%)
123.1
(0.1%)
R5
832.2 (2.4%)
80.6 (4.0%)
50.0 (0.0%)
54.8 (0.6%)
912.8 (2.5%)
104.8
(0.1%)
C1
1077.5 (3.1%)
212.0
(10.6%)
129.0 (0.1%)
23.7 (0.3%)
1289.5 (3.5%)
152.7
(0.1%)
C2
1077.5 (3.1%)
355.7
(17.7%)
235.0 (0.2%)
136.3
(1.5%)
1433.1 (3.9%)
371.3
(0.3%)
The economic benefit is not always in line with the societal benefit from energy reduction.
Figure 3-8 shows the benefit-cost ratio and levelized cost per kWh of re-conductoring to be the
least desirable of all options and economically unacceptable for re-conductoring option R3, even
though option R3 provided significant annual energy and demand reduction. Voltage reduction
contributed substantially to the overall economic acceptability of each option, as shown by
option VR. Phase balancing and var optimization increased acceptability due to additional
savings and low cost of implementation. The most economically acceptable option was
combination option C1, which combines phase balancing, var optimization, and voltage
reduction.
3-24
Table 3-17 shows the benefit-cost ratio, levelized cost, and net present value (NPV) of the option
cost and savings for all efficiency options. The incremental levelized cost and benefit-cost ratio
describe the economic acceptability of the additional project component without cost and savings
from option voltage reduction. This would be the case if the incremental project is analyzed after
voltage reduction has already been implemented.
The savings in all options were primarily influenced by voltage reduction. Although all options
showed a total benefit-cost ratio greater than one (except re-conductoring option R3), the
incremental levelized cost and benefit-cost ratio was acceptable only for phase balancing and var
optimization. Reactive power optimization option VAR1 had an exceptionally high incremental
acceptability due to efficiency improvement and low implementation costs resulting from the
salvage value of removed equipment. All re-conductoring options were incrementally
unacceptable due to higher implementation costs than additional savings.
$0.14
12
$0.12
$0.10
8
$0.08
6
$0.06
4
10
$0.04
2
$0.02
$0.00
VR
PB1
R1
R2
R3
R4
R5
C1
C2
OPTION
Benefit Cost Ratio
Figure 3-8
Economic Acceptability Shown by Benefit-Cost Ratio and Levelized Cost for Efficiency
Projects
3-25
Option
BenefitCost
Ratio
Levelized
Cost
($/kWh)
Total
Efficiency
Option Cost
(NPV $K)
VR
8.1
$0.011
$114
$922
PB1
8.0
$0.012
$122
PB2
9.8
$0.009
VAR1
9.5
VAR2
Incremental
LC After
Option VR
Incremental
BCR After
Option VR
$972
0.040
6.6
$125
$1,225
0.004
27.7
$0.009
$114
$1,083
0.000
1,126.3
8.1
$0.011
$134
$1,088
0.011
8.1
VAR3
7.1
$0.013
$166
$1,173
0.021
4.9
VAR4
6.8
$0.013
$171
$1,154
0.024
4.1
R1
2.3
$0.042
$486
$1,130
0.289
0.6
R2
1.5
$0.068
$866
$1,296
0.327
0.5
R3
0.9
$0.115
$1,624
$1,540
0.399
0.4
R4
3.2
$0.029
$344
$1,085
0.172
0.7
R5
4.7
$0.020
$214
$1,003
0.192
0.8
C1
11.3
$0.008
$125
$1,420
0.002
44.9
C2
2.9
$0.035
$598
$1,715
0.072
1.6
Unbalanced circuit branching and regulation zones limited the potential of voltage reduction.
Phase balancing and reactive power optimization options were incrementally acceptable.
Cost and efficiency of combination options added nearly linearly with the individual options.
Circuit F
Circuit F consists of four separate 13.2-kV overhead feeders that are served off one substation
transformer. All four feeders were modeled in detail as part of the efficiency analysis. There is a
substation load tap changer and 15 capacitor banks on the circuits: twelve with fixed control: two
1800-kvar, one 800-kvar, one 600-kvar, one 450-kvar, two 400-kvar, and five 300-kvar. Three
are on time control: 3600-kvar, 800-kvar, and 600-kvar. The circuit map is shown in Figure 3-9.
3-26
The efficiency options analyzed are described in Table 3-18. The voltage reduction option VR
involved moving four single-phase taps to rebalance load. This adjustment was necessary to
improve bus voltages that would prevent the application of voltage reduction. The peak hour
primary bus voltages prior to the tap adjustment in option VR are shown in Figure 3-10. The
peak hour primary bus voltages after the tap adjustment in option VR are shown in Figure 3-11.
Substation
Capacitors ()
Figure 3-9
Circuit Map
3-27
Option Details
Voltage Reduction
VR
Phase Balance
PB1
Var Optimization
VAR1
VAR2
VAR3
Re-Conductor
R1
R2
R3
Voltage Regulation
REG1
REG2
REG3
REG4
REG5
Combination
3-28
C1
+ VAR3
+ R3
C2
+ VAR3
+ R3
+ REG3
Figure 3-10
Base Case Peak Hour Bus Voltage versus Distance From Substation
Figure 3-11
Option VR Peak Hour Bus Voltage w.r.t Distance From Substation
3-29
Voltage reduction option VR was applied to regulate all four feeders out of the substation based
on remote voltage feedback from one location. This option had the most significant impact on
reducing end-use consumption. The additional efficiency optionsincluding var optimization
and re-conductoringwere effective in reducing line losses; however, as losses decreased,
feeder voltages increased along with end-use consumption. Voltage increase on the individual
feeders that were not monitored by voltage feedback were not adequately adjusted by substation
bus voltage reduction.
In option REG3, additional mid-feeder regulators were installed to reduce voltages on the three
feeders that are not directly controlled by the substation regulator in option VR. This provided
further reduction in end-use consumption. In option REG4, the substation bus regulator was
exchanged for four individual feeder regulators. The additional regulators slightly reduced enduse consumption by decreasing the voltage profiles to the same set point on all four feeders. The
improvement, however, was limited by feeder branching that caused additional regulation zones.
The reduction in consumption was similar to that with one substation regulator due to the
similarities in preexisting voltage conditions on all four circuits.
Prior to the tap adjustments in option VR, the four circuits had very different voltage profiles.
This would limit the performance of one substation bus regulator controlling all four feeders. In
the case without tap adjustments, the individual feeder regulators would be more beneficial to
overall efficiency improvement by allowing voltage reduction to occur on permissible feeders.
The benefit-cost ratio was greater than one for all options. The economic acceptability was
strongly influenced by the improvement from voltage reduction in option VR. Reactive power
optimization option VAR2 provided further savings in total energy and peak demand and was
justified by the incremental benefit-cost ratio greater than one. Option VR3 with mid-feeder
voltage regulators also had an acceptable incremental benefit-cost ratio. All other options were
incrementally unacceptable due to added cost and minimal or decreased savings.
3-30
4000
450
3500
400
350
3000
300
2500
250
2000
200
1500
150
1000
100
500
50
0
VR
R2
OPTION
Total Energy Saved MWh
C2
Figure 3-12
Annual Energy Saved and Peak Demand Reduction
3-31
3-32
Option
Annual
Consumption
Saved
MWh
(% Base)
Annual
Losses
Saved
MWh
(% Base)
Peak
Demand
Saved
kW
(% Base)
Peak
Losses
Saved
kW
(% Base)
Total Annual
Energy Saved
MWh
(% Base)
Total Peak
Reduction
kW
(% Base)
VR
3306.2 (2.4%)
84.6 (2.5%)
349.4 (0.1%)
11.4 (0.1%)
3390.8 (2.4%)
360.8
(0.1%)
PB1
3301.1 (2.4%)
86.6 (2.5%)
348.7 (0.1%)
12.1 (0.2%)
3387.7 (2.4%)
360.8
(0.1%)
VAR1
2901.9 (2.1%)
78.9 (2.3%)
324.8 (0.1%)
-2.2 (0.0%)
2980.9 (2.1%)
322.6
(0.1%)
VAR2
3497.0 (2.5%)
50.1 (1.5%)
414.2 (0.2%)
10.0 (0.1%)
3547.1 (2.5%)
424.2
(0.2%)
VAR3
3376.5 (2.5%)
208.5 (6.1%)
245.1 (0.1%)
21.3 (0.3%)
3585.0 (2.5%)
266.4
(0.1%)
R1
3294.4 (2.4%)
105.3 (3.1%)
343.7 (0.1%)
20.0 (0.3%)
3399.7 (2.4%)
363.6
(0.1%)
R2
3287.0 (2.4%)
119.3 (3.5%)
340.1 (0.1%)
25.5 (0.3%)
3406.3 (2.4%)
365.6
(0.1%)
R3
3282.9 (2.4%)
125.8 (3.7%)
338.2 (0.1%)
28.6 (0.4%)
3408.6 (2.4%)
366.7
(0.1%)
REG1
3294.3 (2.4%)
83.4 (2.4%)
320.7 (0.1%)
9.6 (0.1%)
3377.7 (2.4%)
330.4
(0.1%)
REG2
2657.7 (1.9%)
72.8 (2.1%)
206.7 (0.1%)
9.7 (0.1%)
2730.6 (1.9%)
216.4
(0.1%)
REG3
3548.4 (2.6%)
83.4 (2.4%)
321.8 (0.1%)
8.9 (0.1%)
3631.8 (2.6%)
330.7
(0.1%)
REG4
3309.3 (2.4%)
97.5 (2.8%)
321.3 (0.1%)
13.3 (0.2%)
3406.8 (2.4%)
334.6
(0.1%)
REG5
3387.3 (2.5%)
95.8 (2.8%)
360.3 (0.1%)
10.9 (0.1%)
3483.0 (2.5%)
371.3
(0.1%)
C1
3352.2 (2.4%)
248.6 (7.2%)
234.0 (0.1%)
38.4 (0.5%)
3600.8 (2.6%)
272.3
(0.1%)
C2
3333.8 (2.4%)
246.0 (7.2%)
238.9 (0.1%)
37.3 (0.5%)
3579.8 (2.5%)
276.1
(0.1%)
$0.03
25
$0.02
20
15
$0.01
30
10
$0.00
VR
R2
C2
OPTION
Benefit Cost Ratio
Figure 3-13
Economic Acceptability Shown by Benefit-Cost Ratio and Levelized Cost for Efficiency
Projects
3-33
Option
BenefitCost
Ratio
Levelized
Cost
($/kWh)
Total
Efficiency
Option Cost
(NPV $K)
Total Efficiency
Option Savings
(NPV $K)
VR
29.9
$0.003
$124
$3,705
PB1
25.2
$0.004
$147
VAR1
23.1
$0.004
VAR2
29.0
VAR3
Incremental
LC After
Option VF
Incremental
BCR After
Option VF
$3,702
-0.611
-0.1
$141
$3,261
-0.003
-26.0
$0.003
$135
$3,908
0.006
18.8
10.5
$0.009
$365
$3,839
0.104
0.6
R1
15.6
$0.006
$239
$3,716
1.079
0.1
R2
10.5
$0.009
$354
$3,724
1.236
0.1
R3
7.6
$0.012
$491
$3,728
1.722
0.1
REG1
22.4
$0.004
$164
$3,671
-0.253
-0.9
REG2
14.5
$0.006
$203
$2,933
-0.010
-9.8
REG3
16.2
$0.006
$242
$3,931
0.041
1.9
REG4
6.4
$0.014
$580
$3,704
2.385
0.0
REG5
5.8
$0.016
$661
$3,807
0.488
0.2
C1
5.3
$0.017
$732
$3,860
0.242
0.3
C2
4.5
$0.020
$850
$3,841
0.322
0.2
Unbalanced circuit branching and regulation zones limit the potential of voltage reduction.
Cost and loss efficiency of combination options add near linearly with the individual options.
Reactive power optimization and voltage regulation options are incrementally acceptable.
Substation bus regulation and individual feeder regulators provide similar savings when the
individual feeder load and voltage profiles are similar.
3-34
3-35
Circuits with significant branching and unequal voltage drop on branches benefit from additional
voltage regulators. The branching can reduce the performance of voltage reduction from
difficultly in selecting one remote location for voltage feedback. Additional voltage regulators
create multiple regulation zones; in each zone, potential savings are maximized. This situation
also occurs for substation bus regulation when multiple feeders out of a substation have
drastically different loads, length, and voltage profiles.
When flattening voltage profiles, one of the most important considerations is the location of the
load. If a circuit has clusters of load, to maximize energy savings, it is important to account for
load location. This can impact regulator placement, capacitor placement, how rebalancing is
done, and circuit reconfiguration options. On a circuit with a pocket of load, it may be useful to
apply a voltage regulator downstream of the load center if there is a down-line, voltage-limiting
section. In some cases, its important to control phase balancing and capacitor placement and
control to have uniform voltage drops to the load center.
Economic feasibility depends on the total energy and peak demand saved; however, the
feasibility is also sensitive to economic parameters. In particular, option cost, fixed charge rate,
and energy rate are the most influential. Keep in mind that this analysis did not include lost
revenue from reduced kilowatt-hour billing. Lost revenue is expected to be recovered by billing
adjustments or other regulatory recovery mechanisms. The economic analysis may need to be
adjusted based on a particular utilitys billing adjustments.
The relatively low investment cost of voltage reduction and high energy savings lead to high
benefit-to-cost ratios. With the assumptions used here, all circuits had economically viable
options. In cases studied, phase balancing and reactive power optimization had the most potential
to increase the overall viability based on low additional implementation costs and greater overall
returns. The incremental benefit-to-cost ratio was often higher than one for phase balancing,
reactive power optimization, and additional voltage regulation options.
3-36
4
CASE STUDIES
This chapter provides more technical details and insights from a number of case studies
including var control, transformer efficiency, voltage optimization, and modeling issues.
Case Studies
placement. Again, this is the original configuration of the feeder that the Green Circuits project
was analyzing; hence, this will be the configuration to improve upon. The main issue with this
circuit is too many fixed capacitors relative to the load. At normal or light load, the capacitors
increase losses.
Regulator
Capacitor
300kvar
900kvar
Substation
900kvar
900kvar
900kvar
(Controlled)
900kvar
Figure 4-1
Circuit 1 One-Line Diagram
Table 4-1 summarizes some of the characteristics of the feeder used for this case study. Note that
these characteristics did not change throughout the analysis (that is, they are the same for both
Circuit 1 and 2).
4-2
Case Studies
Table 4-1
Case Study Feeder Summary
Base Characteristics
Circuit
12.47 kV
Residential
97%
24.0
80.8
10.5
Substation Control
Regulator
Average power factor was determined by calculating the average annual power (Pavg) and the average absolute
reactive power (Qavg). Average Power Factor = (Pavg)/(sqrt(Pavg2+Qavg2)).
4-3
Case Studies
The ideal var case had an additional annual savings of 699.6 MWh, and the capacitor control
case had an additional annual savings of 652.7 MWh when compared to the losses of the base
case. Therefore, with an ideal var scenario (load pf = 1), a savings of 699.6 MWh could be
achieved annually; but through the implementation of capacitor control, a savings of 652.7 MWh
was more realistic.
Table 4-2
Power-Factor Comparison of Circuit 1
Model
Phase A
Phase B
Phase C
Circuit 1
(Original Base Model)
0.90
0.86
0.82
Circuit 1 Capacitor
Control (Original Base
Model w/ Capacitor
Control)
0.99
0.99
0.99
Table 4-3
Efficiency Analysis Comparison Summary
Base
GWh Consumption
GWh Losses
Delta Loss (MWh)
Delta Consumption (MWh)
% Loss (Base)
% Consumption (Base)
% Base
37.81
2.72
7.19%
Capacitor
Control
37.51
2.07
652.7
300.2
5.46%
99.2%
24.01%
Ideal var
(pf=1)
37.83
2.02
699.6
-18.5
5.34%
100.0%
25.73%
Note: The consumption of the ideal var case was adjusted to match the base case.
4-4
Case Studies
41.00
GWh Losses
40.50
GWh Consumption
40.00
39.50
2.72
GWh
39.00
2.02
2.07
38.50
38.00
37.50
37.00
37.81
37.83
37.51
36.50
36.00
Base
Capacitor Control
Figure 4-2
Efficiency Comparison Summary Graph
4-5
Case Studies
Regulator
Capacitor
300kvar
900kvar
Substation
1200kvar
(controlled)
900kvar
900kvar
900kvar
(controlled)
900kvar
(controlled)
900kvar
Figure 4-3
Circuit 2 One-Line Diagram
Results
The loss percentage decreased to 5.85% from 7.19% with the improvements implemented on the
feeder and modeled in Circuit 2. The average power factor2 improved from 0.86 to 0.99 at the
head of the feeder. Table 4-4 shows the average power factor for each phase at the head of the
feeder. Again, the consumption of the capacitor-control case was reduced slightly because feeder
voltage was maintained, and the voltage increases did not occur during the light load conditions.
As can be seen in Table 4-5, these results are comparable to the results in Figure 4-4. Circuit 2
had an additional annual savings of 507.3 MWh, and the capacitor-control case of Circuit 1 had
an additional annual savings of 652.7 MWh when compared to the base case of Circuit 1. The
difference was a result of reduced capacitance and the fact that the controlled capacitors were not
as close to the loads as before; however, it was still a significant savings and comparable to the
ideal var case.
2
Average power factor was determined by calculating the average annual power (Pavg) and the average absolute
reactive power (Qavg). Average Power Factor = (Pavg)/(sqrt(Pavg2+Qavg2)).
4-6
Case Studies
Table 4-4
Power-Factor Comparison of Circuit 1 and Circuit 2
Model
Phase A
Phase B
Phase C
Circuit 1
(Original Base Model)
0.90
0.86
0.82
Circuit 2
(Base Model w/ Implemented
Capacitor Control)
0.99
0.99
0.99
Table 4-5
Efficiency Analysis Comparison Summary
Base
(Circuit 1)
GWh Consumption
GWh Losses
Delta Loss (MWh)
Delta Consumption (MWh)
% Loss (Base)
% Consumption (Base)
% Base
37.81
2.72
7.19%
Base
(Circuit 2)
37.58
2.21
507.3
234.6
5.85%
99.4%
18.66%
41.00
GWh Losses
40.50
GWh Consumption
40.00
39.50
GWh
39.00
2.72
2.21
38.50
38.00
37.50
37.00
37.81
37.58
36.50
36.00
Base
(Circuit 1)
Base
(Circuit 2)
Figure 4-4
Efficiency Comparison Summary Graph
4-7
Case Studies
Summary
Figure 4-5 displays the measured data captured at the head of the feeder for the same two-month
time period. Circuit 1 data was collected in 2008, and Circuit 2 data was collected in 2009 after
the capacitor and capacitor control changes were made to the feeder. The average power factor
improved from 0.83 to 0.95, a 14.5% improvement. This is comparable to the simulation results
of Circuit 1 and Circuit 2, which improved from 0.86 to 0.99 (a 15.1% improvement). When the
metered data of both configurations (Circuit 1 and Circuit 2) was compared, the average reactive
power reduced by 1.3 Mvar when the capacitor controls were added.
Note that the average reactive power in Figure 4-5 was computed as the mean of the absolute
value. Because reactive power can be both positive and negative, one does not want the positives
to cancel the negatives when averaging, so the absolute value was used to better reflect the
impact of reactive power. Because the power factor was leading a majority of the time, a
negative sign was assigned to the average value.
As shown in Figure 4-6, the average power usage was relatively constant during this time period:
3.67 MW and 3.55 MW for Circuit 1 and 2, respectively. In summary, the capacitor-control
simulations predicted that significant savings could be achieved. When 2009 field measurements
of the actually implemented capacitor controls (Circuit 2) were compared to the 2008
measurements of the original configuration (Circuit 1), there was a reduction in reactive power
demand and improved system power factor. This improvement in power factor translates to a
reduction in line losses.
Total Reactive Power
1
0.5
-0.5
Circuit 2 (2009)
Mvar
Circuit 2 (-Mean(|x|))
-1
Circuit 1 (2008)
Circuit 1 (-Mean(|x|))
-1.5
-2
-2.5
-3
-3.5
Aug-05
Aug-15
Aug-25
Sep-04
Time
Figure 4-5
Meter Data of Reactive Power
4-8
Sep-14
Sep-24
Oct-04
Case Studies
Total Real Power
9
6
Circuit 2 (2009)
Circuit 2 (Mean(|x|))
5
MW
Circuit 1 (2008)
Circuit 1 (Mean(|x|))
0
Aug-05
Aug-15
Aug-25
Sep-04
Sep-14
Sep-24
Oct-04
Time
Figure 4-6
Meter Data of Real Power
4-9
Case Studies
Table 4-6
Standard Efficiency Levels for Liquid-Immersed Distribution Transformers
The efficiency levels are calculated following the procedure and equations in Section 5 of the
Federal Register (Federal Register / Vol. 71, No. 81 / Thursday, April 27, 2006 / Rules and
Regulations). Please note that there is also a correction to the test procedure found in Federal
Register / Vol. 71, No. 199 / Monday, October 16, 2006 / Rules and Regulations, 60662, which
affects the calculation procedure in Section 5.
One point to take away from the calculation procedure (and Table I.1 above) is that all efficiency
levels are at 50% (0.5 per unit) of the nameplate-rated load. Therefore, when calculating the
efficiency of a transformer from its no-load loss and load-loss percentages, the load loss should
be corrected to 50% loading. Other corrections apply for variations in temperature at which the
loss values were originally measured, as well as for other measurement-dependent quantities that
deviate from the test procedure.
For discussions here, it is assumed that all percent no-load loss and percent load loss values were
measured under laboratory conditions that are consistent with the test procedure indicated in the
DOE efficiency standards.
Comparison of DOE Efficiency Standard Levels to Green Circuits Utility Data
Distribution transformers (pole-top/pad-mounted service transformers) have been modeled as
having a no-load loss value and a load loss value, from a loss perspective in the Green Circuits
collaborative project. The no-load loss represents the losses in the transformer due to stray losses
and core energization losses and other non-load-dependent losses. The no-load losses are fairly
constant over time but do vary with the applied voltage.
4-10
Case Studies
The load loss represents the losses in the transformer due to the loads that the transformer
supplies. Obviously this value varies with the load current required at any given time.
The no-load loss and the load loss values that were applied to the transformers in the Green
Circuits feeders were taken from several data sources.
Two published data sources from which loss values were obtained are the General Electric
Distribution Data book (published date unknown, but likely circa 19701980) and the
Westinghouse Distribution Systems book (published 1959). These data sources provide a few
values for losses based on the transformer nameplate rating and voltage level. These sources
provide two snapshots in time of transformer losses.
For many of the Green Circuits utilities that did not have specific data with regards to number of
transformers by kVA for a given year, the Green Circuits team asked for a rough estimate of
new versus old transformers. It is from the preceding data sources and the rough estimates of
new/old transformers that actual loss values by kVA and transformer primary voltage class were
applied.
Additional sources of transformer-loss data were provided by two utilities that are part of the
Green Circuits collaborative project. One of the utility data sources covers 19862008 vintage
transformers. The other utility data source provides coverage from 1910 to 2008 for distribution
transformers, although very few transformers are from earlier than the 1950s. The total number
of transformers that these sources represent is greater than 100,000.
One question that has arisen is: How do the transformer losses for the new and old
transformers compare to the DOE standards that have taken effect in January of 2010?
In line with the two larger available data sources (provided by the two Green Circuits utilities),
EPRI segmented the old transformers and the newer transformers and calculated what their
losses would be according to the DOE standard measurement/calculation procedure. The plot
resulting from the calculation procedure is shown in Figure 4-7. The older transformers are
shown in yellow in the figure, while the newer transformers are shown in blue.
4-11
Case Studies
99.5
% efficiency
99
New Xfmr
DOE Efficiency
Old Xfmr
98.5
98
97.5
0
50
100
150
200
250
300
kVA
Figure 4-7
Green Circuits Old/New Transformer Efficiencies and DOE Standard-Efficiency
Transformers at Various Nameplate Ratings
For transformer nameplate ratings for which a reasonable number of data points were available
to the study team, the older transformers were found to be less efficient than the DOE standardefficiency transformers. The newer transformers were found to be more efficient than the DOE
standard-efficiency transformers.
Transformer Efficiency Versus Loading
One question that has been asked relates to the loading of a transformer to achieve the highest
efficiency. This should not be confused with the optimal loading given economic or other
constraints because these depend on the utility costs related to:
Annual cost of voltage regulation (providing acceptable voltages to the customer under
normal and abnormal conditions)
Annual cost of fixed charges on the first cost of the installed transformer
4-12
Case Studies
Continuing with the question posed above, the load at which the efficiency is highest can be
determined by the following equation:
%load =
NLL
* 100
LL
Eq. 4-1
Where:
NLL
LL
%load =
That is, the load at which the efficiency is highest is the load at which the no-load losses equal
the load losses. Therefore, if one knows the no-load losses in watts and the load losses as a
percentage of the load, one can derive the loading associated with highest efficiency.
In the ideal case, the maximum efficiency would be 100%. From known loss characteristics of a
transformer, and the percentage of nameplate rating where the efficiency is highest, one can see
the effect on efficiency as varying with load. This can best be illustrated by an example.
Taking a 25-kVA single-phase transformer that is determined to have an efficiency equal to the
DOE standard Table I.1 (98.91% at 50% of nameplate), we can see the variation of efficiency
with per-unit loading of the transformer, as shown in Figure 4-8.
4-13
Case Studies
98.8%
98.6%
98.4%
% efficiency
98.2%
98.0%
97.8%
97.6%
97.4%
97.2%
97.0%
0
0.2
0.4
0.6
0.8
1.2
per-unit loading
Figure 4-8
Transformer Efficiency Versus Loading
4-14
Case Studies
Table 4-7
Circuit Characteristics
Circuit A
Circuit B
System Voltage, kV
12.5
12.5
Number of Customers
1723
721
99
97
22475
10615
79.4
37.8
17.7
6.5
14.8
4.1
336
1650
600
124
124
substation
capacitors
regulators
miles
Thick black lines indicate three-phase sections, with single-phase taps color-coded by phase.
Figure 4-9
Circuit A
4-15
Case Studies
substation
capacitors
regulators
miles
Figure 4-10
Circuit B
Figure 4-11 and Figure 4-12 summarize the results of the yearly and daily loss profiles for
Circuit A.
4-16
Case Studies
Peak
Average
Load kilowatts
6000
5000
4000
3000
Line losses, kW
200
150
100
50
Jun
Sep
Dec
Figure 4-11
Annual Profile of Load and Losses on Circuit A by Week
4-17
Case Studies
Load kilowatts
4000
3500
3000
2500
Line losses, kW
80
70
60
50
40
10
15
20
Figure 4-12
Hourly Profile of Load and Losses on Circuit A
Case Studies
Minimum load
126
126
124
124
Peak load
122
120
118
116
114
122
120
118
116
114
10
15
Distance, miles
10
15
Distance, miles
126
126
124
124
Peak load
122
120
118
116
114
122
120
118
116
114
Distance, miles
Distance, miles
Figure 4-14
Voltage Profile Along Circuit B
Line-drop compensation on the substation feeder regulators was used to reduce voltages. On
Circuit A, compensation settings were used on the line regulators, too. Figure 4-15 shows
voltage profiles with the settings given in the next section. The end-use energy consumption was
predicted to decrease by about 2.5% compared to the base case, but the peak demand actually
increased slightly. At peak, the line-drop compensators needed to boost the voltage above 124 V,
keeping voltages high for the large block of customers near the substation.
4-19
Case Studies
Minimum load
126
126
124
124
Peak load
122
120
118
116
114
122
120
118
116
114
10
15
Distance, miles
10
15
Distance, miles
Figure 4-15
Voltage Profile Along Circuit A
On Circuit B, line-drop compensation worked well for both energy and demand reduction.
Settings were used on the line regulators, too. Figure 4-16 shows voltage profiles with the
settings given in the next section. The end-use energy consumption was predicted to decrease by
about 3% compared to the base case, and the peak demand decreased by about 2%. No capacitors
or other changes were needed on this circuit.
Minimum load
126
126
124
124
Peak load
122
120
118
116
114
120
118
116
114
Distance, miles
Figure 4-16
Voltage Profile Along Circuit B
4-20
122
Distance, miles
Case Studies
Eq. 4-2
Xset = 0
Vset = Vmin pf * Rset * Imin
Eq. 4-3
Where
Imin = the minimum load current / CTrating
Imax = the peak load current / CTrating
pf = load power factor
Both the Imin and Imax are given as a portion of the primary CT rating. From monitoring, these
were estimated as:
Imin = 80 A / 400 A = 0.20 A
3
4
T. A. Short, Electric Power Distribution Handbook. Boca Raton, FL: CRC Press, 2004.
General Electric, Omnitext, 1979. GET-3537B.
4-21
Case Studies
Eq. 4-4
Xset = 0
Vset = Vmin pf * Rset * Imin = 115.44 V
Eq. 4-5
These were rounded to Rset = 13 V and Vset = 115.5 V. With these settings, if the load were to
rise to the full 400 A of the voltage regulator, that would produce a voltage rise of 13 V to give
an output of 128.5 V. At light load, the resulting voltage is 115.5 V + (80 A / 400 A) * 13 V =
118.1 V.
Using the procedure given above, the following settings are suggested for the remaining voltage
regulator sites. All have Xset = 0.
Circuit A: Set of three feeder regulators at four miles out (CT primary rating = 100 A):
Rset = 7 V
Vset = 117 V
Circuit A: Feeder regulators on the middle phase at ten miles out (CT primary rating = 100 A):
Rset = 12 V
Vset = 118 V
Circuit B: Substation feeder regulators:
Rset = 5.6 V
Vset = 117.8 V
For all regulator controllers, the minimum voltage setting of the regulators should be set at 118 V
or 118.5 V to minimize the risk of undervoltages if significant load is dropped during an outage.
Once the regulator-control settings are in place, the utility will have to carefully watch customer
complaints and possibly make voltage measurements. This particular utility was installing AMI,
which can monitor voltages. These line-drop compensation settings are relatively aggressive.
Having AMI metering allows an aggressive approach because customer voltages can be
monitored, and settings can be changed if voltages are different than expected.
4-22
Case Studies
Control of the local bus with a load-tap-changer (LTC), sensing the local bus voltage
Control of the local bus with an LTC, sensing a single remote bus via voltage feedback
Control of the local bus with an LTC, utilizing line-drop compensation (LDC)
Control of the local buses with an LTC and remote (out on the feeders) regulators using
remote voltage feedback for the LTC and the remote regulators
Control of the local buses with feeder head single-phase voltage regulators using remote
voltage feedback
A simple substation and feeder-wide volt-var optimization scheme with an LTC, where the
LTC is sensing a remote bus
The four circuits that form the basis of each case presented here are served by a single substation
transformer, which has a load-tap-changer as the means of regulation. The set of four circuits is
shown in Figure 4-17, with the substation indicated.
The four circuits have greater than 90% residential customers, by customer count. The assumed
power factor for all loads was set at 0.9 lagging. This is lower than would be expected for
residential loads. The substation serves several large industrial customers, which comprise a
large percentage of the total load, so a closer match on reactive power was found with this lower
assumed power factor, across all of the loads. The load factors on the four circuits were found to
be 44%, 63%, 43%, and 47%, respectively. The overall load factor for the substation is
approximately 53%.
4-23
Case Studies
Remote voltage
measurement point
substation
Figure 4-17
One-Line Diagram of the Substation/Feeders Under Study
4-24
Case Studies
Figure 4-18
Voltage Profile Resulting From the Remote Voltage Feedback With the LTC
5
6
4-25
Case Studies
Eq. 4-6
Xset = 0
Vset = Vmin - pf * Rset * Imin
Eq. 4-7
Where
Imin = the minimum load current / CTrating
Imax = the peak load current / CTrating
pf = load power factor
Both the Imin and Imax are given in per-unit of the primary CT rating. From monitoring, these
were estimated as:
Imin = 290 A / 1200 A = 0.242
Imax = 1200 A / 1200 A = 1.0
The load power factor was taken to be 0.9. This gives the following for Rset and Vset:
Rset = (Vmax - Vmin) / pf / (Imax - Imin) = 4.44 V
Eq. 4-8
Xset = 0
Vset = Vmin - pf * Rset * Imin = 120.73
Eq. 4-9
The LTC was simulated under the control of line drop compensation with these parameters. This
resulted in the following voltage profile at the peak hour (see Figure 4-19).
4-26
Case Studies
Figure 4-19
Voltage Profile Resulting From the Line Drop Compensation Approach to Voltage
Optimization
4-27
Case Studies
Figure 4-20
Voltage Profile at Peak Hour for the Case With Multiple Remote (Feeder) Regulators and
the LTC Operating on Remote Voltage Feedback
4-28
Case Studies
2nd regulator
3rd regulator
1st regulator
substation
Figure 4-21
Locations of the Three Remote Feeder Regulators and the LTC Feedback Bus
4-29
Case Studies
Figure 4-22
Voltage Profile Plot for the Case of Three Single-Phase Regulators at the Head of Each of
the Four Circuits
Reduce losses
Coordinate control of voltage regulation equipment and var equipment on the system
Minimize the number of operations of voltage regulation equipment and var equipment on
the system
The LTC was set to regulate a remote three-phase bus, while switching commands were sent to
one or more capacitors. The OpenDSS solved the system under this new configuration. Once the
system was solved, voltages and powers were used to calculate the value of an objective
function. The objective function is defined as:
4-30
Case Studies
where:
weightlosses
losses
weightconsumption
consumption
weightlow_voltages
low_voltage_violation_count
weighthigh_voltages
high_voltage_violation_count
The objective function allows the engineer to chose weighting factors to apply to each of the
components that make up the function, as well as to set the voltage violation criteria (both high
and low) to values of their choosing. Looking at various capacitor-switching scenarios, the
objective function was minimized for each time interval of the simulation.
Figure 4-23 shows the voltage profile for a configuration that minimized the objective function.
This voltage profile was very flat and resulted in no voltage violations for this case. This
simple approach to volt-var optimization shows some of the expected benefits from a system
implementation in-the-field.
4-31
Case Studies
Figure 4-23
Voltage Profile Resulting From a Volt-Var Optimization Scheme
Summary
This chapter describes a few voltage optimization approaches, showing examples for each.
Table 4-8 summarizes results of various approaches to voltage optimization. Key findings are:
An LTC that is controlling one or more feeders and that is performing voltage control of a
single remote bus may result in lower voltages at other locations on one or more feeders,
including adjacent feeders.
Volt-var optimization, using a very simple approach, shows a marked improvement in the
overall voltage profile, while lowering consumption and losses. The utility would be able to
drive the set point lower on regulators/LTCs using volt-var optimization, reaping greater
benefits from voltage reduction.
4-32
Case Studies
Table 4-8
Comparison of Voltage Optimization Approaches
Case
Peak-Case
Average
Primary
Voltage
120-V Scale
Peak-Case
Per-Unit
Consumption
Annual Average
Primary Voltage
120-V Scale
Annual
Per-Unit Average
Consumption
Base LTC
122.0
1.0
124.6
1.0
121.3
0.996
121.2
0.982
122.9
1.01
122.4
1.01
120.1
0.989
119.9
0.974
Feeder Head
Single-Phase
Voltage Regulators
With Remote
Voltage Feedback
119.2
0.987
119.1
0.975
Simple Volt-Var
Optimization
119.9
0.941
4-33
Case Studies
substation
Figure 4-24
Substation With Four Feeders Modeled in Detail
The four feeders are regulated by means of a single substation transformer with a load-tap
changer (LTC). The substation transformer has the characteristics listed in Table 4-9.
Table 4-9
Substation Transformer Characteristics
Characteristic
Value
Base Rating
45,000 kVA
Distribution Voltage
13,200 V (LL)
126 V
Xhl
10.58%
Resistance
0.423%
No-load losses
0.18%
Load shapes for each of the feeders were available on a per-phase basis from measurements
provided by the utility from their SCADA system. This data set had instances where the power
values dropped to zero where there is a gap in data collection, and also instances where the
power values increased to a very high level. For most of these instances, the last valid measured
power value was averaged with next the valid measured power value. Where large gaps
4-34
Case Studies
occurred, the power values of similar days (weekend, weekday) were used to fill in the missing
data points. From calculations, the overall load factor was found to be 53.3%. This appears to be
in line with the large known loads on the system being industrial-class loads.
The load allocations began by utilizing the feeder-head measurements. The data for development
of the electrical model was a GIS data source. It did not have peak kW usage data for each of the
customers, so the load allocation for each customer was determined by the kVA rating of the
service transformer. Where multiple customers were served by a single service transformer, the
loads were divided evenly among those common customers, because more precise data was not
available.
A load model that takes into account variation of consumption with respect to terminal voltage
was utilized in the electrical model in the OpenDSS software. Two inputs to that particular load
model are the variation in watts (CVRwatts) and the variation in vars (CVRvars), with respect to
changes in terminal voltage from a user-defined nominal voltage. The CVRwatts factor used in this
model was 0.7, and the CVRvars factor was set to 3.
Analysis of Cases
The following cases were analyzed for their impacts on the substation transformer only. Note
that the base case with voltage optimization is the case to which all subsequent cases are
referred.
The percent increase or decrease in losses was calculated with respect to the values from the base
with the voltage optimization case. A negative percentage indicates a reduction in losses, while a
positive percentage indicates an increase in losses:
Base
Phase balancing
Var optimization
Re-conductoring
Voltage regulators
4-35
Case Studies
These load losses were sub-divided into peak-hour losses (in kW) and average annual losses
(also in kW). The values that are expressed as percentages are the change from the Base with
Voltage Optimization case for each of the cases that were simulated. For the percentages, a
reduction in losses is indicated by a positive value. An increase in losses is indicated by a
negative value for the percentages in parentheses.
Table 4-10
Substation Transformer Load Losses by Simulation Case
Case Identification
4-36
Base
76.5
24.4
75.8
23.6
+1.86%
+1.38%
Balancing - 1 -> 2 or 3 ph
+1.77%
+1.37%
-6.63%
-5.67%
+2.90%
+2.00%
+1.09%
+3.00%
+1.77%
+1.36%
+1.53%
+1.33%
+1.58%
+1.43%
+1.91%
+1.56%
+1.09%
+3.00%
+1.12%
+3.04%
+1.13%
+3.06%
Case Studies
Summary
The total load was slightly over half of the base nameplate capacity on the substation transformer
analyzed in this section. The load losses on the transformer were correspondingly low.
The currents were reasonably well balanced at the feeder-heads, while further out they were
quite unbalanced, even on several of the three-phase primary lines. As with the effects out on a
feeder, unbalanced currents at a substation transformer will contribute to higher losses.
The sum of the feeder loads reactive power consumption and any capacitors on the circuit
will, obviously, determine the power factor at the substation transformer. As was seen with the
last two var modification cases, an improvement in the overall reactive power profile (a
minimization of the var flow at the substation transformer) had positive effects on the losses at
the substation.
Eq. 4-10
Where,
P = Active power demand (watts)
P0 = Active power demand at rated voltage (watts)
CVR = Active power CVR factor
V = Per-unit bus voltage at each load
The change in active power of each load with respect to CVR is then given as:
P
= P0 (V 1)
CVR
Eq. 4-11
4-37
Case Studies
Therefore, the relationship between CVR and P is approximately linear, assuming that the
voltage is relatively unaffected by the changes in active power demand. This linear relationship
is shown in Figure 4-25 for an example circuits total peak hour demand. Moreover, the variation
in V for different assumed CVR factors is expected to be even less during lower demand periods,
resulting in a more linear relationship between consumption over the full calendar year and CVR
factor, as shown in Figure 4-26. Nonetheless, the linear behavior may be reduced for weak
circuits characterized by tight coupling of voltage and active power or for circuits where the
particular changes cause nonlinear circuit operations such as LTC tap changes.
5360
5340
Demand (kW)
5320
y = -226.71x + 5462.6
5300
5280
5260
5240
5220
0.4
0.5
0.6
0.7
0.8
CVR FactorWatts
Figure 4-25
Example Peak Hour Demand Versus CVR Factor
4-38
0.9
1.1
Case Studies
18.60
Consumption (GWh)
18.55
y = -0.3595x + 18.769
18.50
18.45
18.40
0.4
0.5
0.6
0.7
0.8
0.9
1.1
CVR FactorWatts
Figure 4-26
Example Annual Consumption Versus CVR Factor
4-39
Case Studies
P ow er Factor
0.98
0.96
0.94
0.92
0.9
0
/1
0
11
/1
/1
/1
10
9/
1/
10
10
1/
8/
10
1/
7/
10
6/
1/
10
5/
4/
1/
1/
1/
3/
2/
10
10
10
1/
10
1/
1/
12
/1
/0
0.88
Figure 4-27
Variable Power Factor
Interpreting the end-use reactive load from the annual power factor profile in Figure 4-27
can be misleading. The power factor profile shows values between 89% and 98%, which vary
significantly throughout the year. The corresponding end-use reactive load is shown in
Figure 4-28, which illustrates that the reactive load was somewhat constant most of the year and
about doubled in the summer months. Assuming a constant power factor would overestimate the
reactive load and create more variation throughout the year.
3000
Cons tant P F
V ariable P F
2500
2000
1500
1000
500
0
0
1000
2000
3000
4000
5000
6000
7000
8000
H our of Y e ar
Figure 4-28
Hourly End-Use Reactive Load
Modeled load power factor had a significant influence on reactive load, which in turn had a
significant influence on circuit voltages, consumption, and losses. As the variable power factor
became closer to unity, the reactive load decreased, voltages improved, and end-use consumption
increased as shown by comparing variable power factor and fixed power factor annual energy
base case results in Table 4-11. These results were based on the same modeling assumptions
4-40
Case Studies
described in Chapter 2. For variable power factors, the base case transformer no-load losses
increased, while secondary line losses and transformer load losses decreased due to less reactive
load. The change in base case primary line losses were impacted by change in load, losses, and
reactive load compensation from capacitor banks.
When comparing the voltage reduction cases to the base cases, the change in consumption and
losses was greater for the variable power factor scenario due to the voltage improvement from
less reactive load in the base case. Secondary line losses and transformer load losses slightly
increased with variable power factor because the reactive load was lower, which resulted in less
reduction in reactive load during voltage reduction. Active power load current increased more
than the reduction in reactive load current. The primary line losses decreased more for the
variable power factor scenario primarily due to switched reactive power compensation on the
circuit.
4-41
Case Studies
Table 4-11
Annual Energy Results for the Power Factor Sensitivity Case
Base Case
Fixed Power Factor
MWh
MWh
Demand/Consumption
19636.1
19656.9
Total Losses
347.8
343.4
95.5
96.2
34.0
31.6
52.0
48.7
Xfmr No Load
166.4
166.9
MWh
MWh
Demand/Consumption
19462.2
19462.5
Total Losses
344.9
334.2
97.1
91.8
33.9
31.6
51.9
48.7
Xfmr No Load
162.0
162.0
4-42
MWh
MWh
Demand/Consumption
-173.91
-194.36
Total Losses
-2.92
-9.18
1.61
-4.42
-0.08
0.04
-0.12
0.05
Xfmr No Load
-4.33
-4.85
Case Studies
Some of the key findings from the power factor sensitivity analysis include:
Circuit voltage and losses were sensitive to reactive load and thus power factor. Assuming a
constant annual power factor for loads is a reasonable approximation, but it does not paint a
complete picture of how reactive load varies on the circuit. Accurate reactive load affected
voltages, consumption, losses, and capacitor control algorithms.
With a more accurate variable power factor, voltage reduction saved more energy because
there were fewer vars to cause voltage drop along the primary and into the secondary.
Customer voltages tended to be higher with a flatter profile. This allows further overall
voltage reduction.
During the winter, voltage profiles were flatter because there was less voltage drop from
reactive power. Therefore, the overall voltage profiles can be reduced further. Unfortunately,
this is when voltage reduction is least effective (CVR factors are lower from thermostatically
controlled resistive heat).
4-43
5
TRANSFORMERS AND SECONDARIES
Voltage drop Voltage drop through the transformer and across the secondary system is a
limiting factor for voltage reduction. This voltage drop also affects transformer sizing and
secondary conductor sizing.
Measurements from advanced metering infrastructure (AMI) should allow utilities to better
manage transformer and secondary loadings, improve modeling of distribution systems, and
manage voltage profiles. AMI metering should allow more aggressive voltage reduction because
voltage measurements can be used to manage control settings or in real time for direct control.
In this chapter, transformer and secondary losses and voltage drops are examined from different
angles. Using AMI data from one utility in the Midwestern United States that has detailed
secondary models, EPRI compared detailed secondary modeling with measurements.
Some of the key findings from this analysis are:
AMI can help quantify transformer loading and losses. Such data can improve transformer
sizing.
Secondary line losses were generally low. Findings from the overall simulations (see Chapter
2) showed that annual secondary line losses were 0.31%. In more precise modeling of four
secondary systems analyzed in this chapter, average losses were 0.76%. From estimates of
AMI metering data from two circuits (44 secondaries total), average secondary losses were
0.72% on one circuit and 0.87% on another.
Detailed secondary models backed by AMI data show that simplified secondary models may
slightly underestimate losses. In four secondary systems analyzed in detail, average losses
were 0.76% from a detailed model and 0.62% using a generic model.
Voltage drops based on detailed secondary simulations with AMI measurements were
accurate.
5-1
From metering on transformer secondaries and customers, voltage drops across secondaries
averaged 0.33 V on a set of 269 customers. At peak load, 85% of secondary drops were less
than 2 V. Transformer voltage drops were typically less than 1 V, and peak voltage drops
were between 0.5 and 3 V. Eighty percent of peak transformer voltage drops were less than
3 V. The total voltage drop across the transformer and secondaries averaged 1.0 V, and 1%
were above 4.2 V.
Metering at the transformer can help identify circuit errors and better measure losses and
voltage drop.
As with other circuits, circuit voltages range on the high side of the ANSI C84.1 range. On
two circuits analyzed in this chapter, the median customer voltage was approximately 122 V.
Even at peak load, voltages were generally high, with median customer voltages above 121 V
on both circuits. This leaves significant room to reduce voltages.
5-2
Table 5-1 summarizes loss information for the four secondary circuits modeled in detail along
with measurements of losses. Losses varied significantly from system to system. As expected,
longer secondary systems had the most losses. Table 5-2 shows measured and modeled voltage
drops across the secondary systems. Voltage drops also varied significantly by site.
Table 5-1
Characteristic Circuit Statistics
Circuit
Average Load
(kW)
Average Power
Factor
Measured Line
Losses (%)
(Lagging)
Total
Secondary
Length (Ft)
Modeled
Line Losses
(%)
2.87
0.97
215.1
0.54
0.17
3.49
0.97
265.6
0.67
0.57
4.79
0.96
1156.8
1.26
0.75
11.42
0.92
1165.2
NA
0.98
5-3
Table 5-2
Transformer and Secondary Line Voltage Drop (120-V Base) Statistics
Circuit
Average
Average
Maximum
Maximum
Measured
Modeled
Measured
Modeled
Bus #
73422056
0.22
6003267439
0.15
0.28
0.55
0.60
6003332156
0.28
0.15
1.85
0.89
73422122
2
1.12
0.26
1.04
6003035406
0.98
0.52
3.25
2.90
6003035419
0.73
0.2
1.5
1.13
73422107
0.38
1.42
6003151373
0.82
0.22
2.65
1.56
6003151473
0.7
0.59
3.4
2.74
6003151586
0.80
0.8
4.1
3.52
6003151601
1.33
0.90
5.05
3.93
6003151613
1.16
0.77
3.6
3.25
73422104
0.94
2.49
6003056509
0.50
0.41
2.65
1.68
6003056571
0.48
0.34
2.1
1.47
6003116790
0.39
0.29
1.75
1.22
6003116831
1.10
0.94
3.9
2.82
6003116861
1.28
1.19
3.95
3.4
6003116914
1.84
1.19
3.4
6003116973
1.82
1.19
3.4
6003122066
1.86
1.19
5.9
3.4
6003122103
0.7
1.20
5.95
3.44
6003122020
NA
1.19
NA
3.40
Measured results were generally higher than modeled values for both losses and voltage drop.
One factor that may have contributed to this was load split between the 120-V legs of the service.
The models all assumed balanced loading across the 240-V winding. Loading consisted of 240-V
loads, such as water heaters, and 120-V load on the two 120-V legs. If the loading on the 120-V
legs was unbalanced, line losses and voltage drop were higher than predicted by a balanced
model.
5-4
Circuit #1
The first secondary study circuit consisted of two overhead service drops of different lengths and
conductor sizes, as shown in Figure 5-1, with the wire lengths indicated in feet. The blue
numbers in Figure 5-1 indicate the bus number used to reference the specific location of
measured or modeled circuit values. The hourly kW demand for each customer is provided in
Figure 5-2, along with the total demand measured at the secondary winding of the service
transformer in Figure 5-3. This is from AMI metering at the secondary side of the transformer
(73422056) and at each customer meter (6003332156 and 6003267439). The graphs in
Figure 5-2 are for one calendar year of data with hour = 1 starting at December 1.
Using the available measurements, losses across the secondary lines were computed by
subtracting the summed customer demand from the total power measured at the transformer
secondary (see the results provided in Figure 5-40. Loss estimates through direct use of the
available measurement data is examined later.
Figure 5-1
Secondary Circuit #1 One-Line Diagram
5-5
Figure 5-2
Measured Customer Demands of Circuit #1
Figure 5-3
Measured Demand at the Service Transformer Secondary of Circuit #1
5-6
Figure 5-4
Measured Losses on Circuit #1
Using the measurements and circuit data, a reasonable match was obtained between the model
and measured voltages, as shown in Figure 5-5. The relationship between the measured and
modeled voltage drop across both services is provided in Figure 5-6, with each data point
showing the measured-versus-modeled voltage drop for averages from a one-hour interval. As
shown, the measured voltage drop at 6003267439 (the longer, lightly loaded service drop) was
almost double that modeled, while an almost ideal one-to-one relationship existed between the
measured and modeled values for the 600332156 bus. Additionally, errors from significant-digit
rounding when calculating the voltage drop are clearly visible as stair steps in the measured
voltage drops.
Figure 5-5
Transformer Secondary Bus Voltage of Circuit #1
5-7
6003332156
3
2.5
2
1.5
1
0.5
0
0
0.5
1.5
2.5
Figure 5-6
Measured-Versus-Model Secondary Voltage Drop of Circuit #1
A comparison between modeled and measured line losses in the circuit is given in Figure 5-7. In
this case, the measured line losses tended to be approximately twice the value that was estimated
in the model. However, this difference may have been due to difficulty in directly measuring
losses due to the rounding errors and possible modeling errors.
5-8
0.15
0.1
0.05
0
0
0.05
0.1
0.15
0.2
Figure 5-7
Measured-Versus-Model Line Losses for Circuit #1
Circuit #2
This overhead secondary circuit, shown in Figure 5-8, consisted of two direct service drops of
the same conductor type and roughly the same length. However, the customer demands over the
calendar year (see Figure 5-9) were characterized by seasonal variations as well as magnitude
differences.
Figure 5-8
Secondary Circuit #2 One-Line Diagram
5-9
Figure 5-9
Customer Demand (kW) for Circuit #2
The voltage drop between the transformer secondary and both customer meters as estimated in
the model with the measured data was compared (see Figure 5-10). Overall, the modeled values
matched very well with the voltage drop determined from the measurements. However, the
measurements tended to consistently indicate an additional volt or more of drop across the
service drops than indicated in the model. However, this may have been due to offset error in the
voltage measurements because a 1:1 relationship still existed between changes in the measured
and modeled drops calculated points in time. The comparison between the measured and
modeled line losses also showed a good match between the two, with the trend line having a near
1:1 slope, as shown in Figure 5-11.
5-10
6003035419
5
4
3
2
1
0
0
3
4
Modeled Drop (120V)
0.3
0.35
Figure 5-10
Measured-Versus-Model Secondary Voltage Drop for Circuit #2
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0
0.05
0.1
0.15
0.2
0.25
Figure 5-11
Measured-Versus-Model Line Losses for Circuit #2
Circuit #3
This secondary circuit was characterized by a long common secondary supplying multiple
service drops at various distances along the secondary (see Figure 5-12). Additionally, service
drops ranged from 50 to over 200 feet in length.
5-11
Figure 5-12
Secondary Circuit #3 One-Line Diagram
Comparison of the model and measured voltage drops and losses from the transformer secondary
to each customer meter, shown in Figure 5-13 and Figure 5-14 respectively, showed a relatively
good fit between model and measurement data. This circuit was also characterized by a
significant amount of voltage drop between the transformer and the customer meter as a result of
the length of the secondary lines and conductor sizes.
6003151373
6003151473
6003151586
6003151601
6003151613
8
7
6
5
4
3
2
1
0
0
Figure 5-13
Measured-Versus-Model Secondary Voltage Drop for Circuit #3
5-12
0.5
0.4
0.3
0.2
0.1
0
0
0.1
0.2
0.3
0.4
0.5
0.6
Figure 5-14
Measured-Versus-Model Line Losses for Circuit #3
Circuit #4
The fourth circuit, shown in Figure 5-15, served a total of 10 customers, six of which were
served at a single service point. One of the loads fed by this transformer (6003122020) did not
have associated AMI measurements. In this case, the loads hourly variations were derived based
on the difference between the hourly demand measured at the transformer secondary and the
total measured customer demands. As such, a comparison of the measured and modeled losses
was not feasible. However, the measured and modeled voltage drops, provided in Figure 5-16,
indicated that the model reasonably captured the circuit behavior. Note that the 6003116861 bus
represented the voltage drop for all six of the meters sharing a common interconnection point.
5-13
Figure 5-15
Secondary Circuit #4 One-Line Diagram
5-14
6003116861
6003056509
6003056571
6003116790
8
7
6
5
4
3
2
1
0
0
Figure 5-16
Measured-Versus-Model Secondary Voltage Drop for Circuit #4
5-15
3. Aggregate Profile
The circuit configuration is taken directly from the utility circuit model. However, hourly
demand variations are based on single normalized load shape derived from the substation
demand measurements, and peak demand is allocated based on measured customer
annual kWh. This case represented the effect of not modeling the load diversity at the
individual customer level. This assumption is illustrated in Figure 5-17 by contrasting the
actual customer demand to the equivalent Aggregate derived demand profile.
4. Generic Service
Loads are the same as for the full AMI case; however, the circuit model is simplified by
assuming a direct service drop (100 feet of 1/0 triplex wire) for each customer meter.
This case provided an indication of how much impact not having an accurate
representation of the secondary configuration and conductor sizes may have on the
estimated losses.
For most utilities, secondaries were modeled using an aggregate profile and a generic service
(see Chapter 2).
Aggregate Profile
6003267439
3.5
Demand (kW)
3
2.5
2
1.5
1
0.5
0
1010
1034
1058
1082
1106
1130
1154
1178
Hours
Figure 5-17
Model Comparison of Customer Demand Variation Over a Sample Week
Table 5-3 and Table 5-4 summarize peak and average loss information for each of the
secondaries analyzed. Some of the main findings from the loss results are:
The annual no-load losses were the least affected by choice of assumptions in cases because
the individual circuit and load changes were insufficient to influence the average primary
voltage in the circuit models.
The fixed power factor assumption has a minimal effect on loss estimates. This impact may
have a more significant impact when voltage reduction (CVR) is fully considered.
5-16
The largest deviations in the annual secondary loss estimations occurred on circuits with
extensive secondary lines serving multiple customers.
Secondary peak losses can be significantly underestimated when using a highly coincident
load profile due to lower I2R loss estimates.
The impact of the generic service assumption in the loss estimates will depend on how
representative the equivalent model is of the typical feeder secondaries.
Table 5-3
Secondary Annual Consumption and Losses
Total
Line
Transformer
Load
Transformer
No-Load
Full AMI
25645
549.7
43.1
50.8
455.8
Fixed pf
25648
552.7
44.6
52.4
455.7
Aggregate Profile
26003
527.4
30.2
41.5
455.7
Generic Service
25641
545.6
37.5
52.3
455.7
Full AMI
30620
697.9
176.0
75.6
446.3
Fixed pf
30625
703.3
179.9
77.2
446.2
Aggregate Profile
30288
628.5
122.1
60.2
446.2
Generic Service
30502
580.4
59.1
74.9
446.3
Full AMI
41972
899.6
315.8
137.4
446.4
Fixed pf
41931
858.5
287.6
123.9
447.0
Aggregate Profile
41854
778.0
221.7
109.8
446.5
Generic Service
41718
645.6
63.6
135.6
446.4
Full AMI
100031
2101.6
982.7
668.9
450.0
Fixed pf
99993
2063.6
969.3
643.9
450.4
Aggregate Profile
99103
1943.5
885.0
608.1
450.4
Generic Service
99188
1258.7
151.6
657.0
450.1
Circuit #3
Circuit #2
Circuit #1
Model Case
Circuit #4
Losses (kWh)
Total
Consumption
(kWh)
5-17
Table 5-4
Secondary Peak Demand and Losses
Total
Line
Transformer
Load
Transformer
No-Load
Full AMI
14.71
0.25
0.09
0.11
0.05
Fixed pf
14.72
0.26
0.10
0.11
0.05
Aggregate Profile
5.61
0.08
0.01
0.02
0.05
Generic Service
14.70
0.24
0.08
0.11
0.05
Full AMI
15.14
0.52
0.35
0.12
0.05
Fixed pf
15.16
0.54
0.37
0.12
0.05
Aggregate Profile
6.56
0.12
0.05
0.02
0.05
Generic Service
14.91
0.28
0.12
0.12
0.05
Full AMI
20.86
0.60
0.35
0.21
0.05
Fixed pf
20.85
0.59
0.33
0.20
0.05
Aggregate Profile
9.10
0.18
0.09
0.04
0.05
Generic Service
20.60
0.34
0.08
0.20
0.05
Full AMI
30.83
1.32
0.78
0.49
0.05
Fixed pf
30.79
1.27
0.75
0.48
0.05
Aggregate Profile
21.74
0.64
0.35
0.24
0.05
Generic Service
30.17
0.65
0.13
0.47
0.05
Circuit #3
Circuit #2
Circuit #1
Model Case
Circuit #4
Losses (kW)
Total
Demand
(kW)
Secondary Voltages
Modeling simplifications with respect to the estimation of customer voltage was first evaluated
through simulation of the different modeling approaches on the four study circuits. The
simulated voltage results are presented in Figure 5-18 through Figure 5-25. Examination of the
finding leads to the following observations:
The fixed power factor assumption did not have a significant impact to the voltage estimation
at this level of load aggregation.
As expected, the aggregate profile overestimated the customer voltages during peak loading
times but provided similar average voltage results. Furthermore, the transformer secondary
voltage was also overestimated during the peak by as much as three volts. These results
indicate potential difficulties accurately assessing customer-level circuit behavior utilizing
substation-level data.
5-18
Use of a generic equivalent service impedance resulted in inaccurate estimation of the peak
voltage by as much as four volts and as much as one or two volts on average. In this case, the
assumed secondary impedances resulted in higher estimates overall. Naturally, the modeled
results were closer for the circuits whose configuration was similar to the assumed direct
service drop configuration.
246
Secondary Voltage (Volts)
245
244
243
242
73422107
241
6003151373
6003151473
240
239
238
Full
Fixed pf
Aggregate
Profile
Generic
Service
Figure 5-18
Minimum Non-Coincident Bus Voltages (Circuit #1)
245
Secondary Voltage (Volts)
244
243
242
73422107
241
6003151373
240
6003151473
239
238
Full
Fixed pf
Aggregate
Profile
Generic
Service
Figure 5-19
Average Annual Bus Voltages (Circuit #1)
5-19
242
240
238
73422122
236
6003035406
234
6003035419
232
230
Full
Fixed pf
Aggregate
Profile
Generic
Service
Figure 5-20
Minimum Non-Coincident Bus Voltages (Circuit #2)
245
244
243
242
73422122
241
6003035406
240
6003035419
239
238
Full
Fixed pf
Figure 5-21
Average Annual Bus Voltages (Circuit #2)
5-20
Aggregate
Profile
Generic
Service
244
242
73422107
240
6003151373
238
6003151473
236
6003151586
234
6003151601
232
6003151613
230
Full
Fixed pf
Aggregate
Profile
Generic
Service
Figure 5-22
Minimum Non-Coincident Bus Voltages (Circuit #3)
244
242
73422107
240
6003151373
238
6003151473
236
6003151586
234
6003151601
232
6003151613
230
Full
Fixed pf
Aggregate
Profile
Generic
Service
Figure 5-23
Average Annual Bus Voltages (Circuit #3)
5-21
242
240
238
236
234
232
230
Full
Fixed pf
Aggregate
Profile
Generic
Service
73422104
6003056509
6003056571
6003116790
6003116831
6003116861
6003116914
6003116973
6003122066
6003122103
6003122020
Figure 5-24
Minimum Non-Coincident Bus Voltages (Circuit #4)
245
244
73422104
6003056509
6003056571
6003116790
6003116831
6003116861
6003116914
6003116973
6003122066
6003122103
6003122020
243
242
241
240
239
238
Full
Fixed pf
Aggregate
Profile
Generic
Service
Figure 5-25
Average Annual Bus Voltages (Circuit #4)
5-22
1.0
Frequency
0.8
0.6
0.4
0.2
0.0
0.0
0.5
1.0
1.5
0.0
0.5
1.0
1.5
Figure 5-26
Secondary Line Loss Probability Distributions for the Monitored Subset of Circuit A
5-23
1.0
Frequency
0.8
0.6
0.4
0.2
0.0
0.0
0.5
1.0
1.5
2.0
0.0
0.5
1.0
1.5
2.0
Figure 5-27
Secondary Line Loss Probability Distributions for the Monitored Subset of Circuit B
Secondary line losses are a function of many parameters. Figure 5-28 and Figure 5-29 show
secondary line losses measured as described above versus various secondary parameters. All of
the parameters listed had positive correlations. The most influential variable was total secondary
line length.
5-24
1.5
Secondary line losses, percent
1.5
1.0
0.5
0.0
1.0
0.5
0.0
1
0.5
1.5
2.0
1.5
Secondary line losses, percent
1.0
1.0
0.5
0.0
1.0
0.5
0.0
1
200
400
600
800
1000
1200
Figure 5-28
Secondary Line Losses Versus Different Factors for the Monitored Subset of Circuit A
5-25
2.5
2.0
1.5
1.0
0.5
0.0
2.5
2.0
1.5
1.0
0.5
0.0
0.5
1.5
2.0
2.5
2.5
1.0
2.0
1.5
1.0
0.5
0.0
2.5
2.0
1.5
1.0
0.5
0.0
0.00
0.05
0.10
0.15
0.20
0.25
100
200
300
400
500
600
700
Figure 5-29
Secondary Line Losses Versus Different Factors for the Monitored Subset of Circuit B
5-26
1.0
Density
1.5
1.0
0.5
0.8
0.6
0.4
0.2
0.0
0.0
0.0
0.5
1.0
1.5
2.0
0.0
0.5
Percent losses
1.0
1.5
2.0
1.5
2.0
Percent losses
Figure 5-30
Transformer Load Loss Distributions for the Monitored Subset of Circuit A
1.0
1.0
Density
0.8
0.6
0.4
0.2
0.8
0.6
0.4
0.2
0.0
0.0
0.0
0.5
1.0
Percent losses
1.5
2.0
0.0
0.5
1.0
Percent losses
Figure 5-31
Transformer Load Loss Distributions for the Monitored Subset of Circuit B
Transformer loading correlated well with line losses, as shown in Figure 5-32 and Figure 5-33.
On Circuit A, the 5-kV transformers had higher percentage losses, likely because the transformer
resistances were higher.
5-27
Losses, percent
1.5
voltage
5 kV
14 kV
1.0
transformer size
10 kVA
15 kVA
25 kVA
35 kVA
0.5
20
40
60
80
100
120
140
5-28
Losses, percent
1.5
voltage
5 kV
14 kV
kVA range
(0,10]
(10,20]
(20,30]
(30,40]
(40,50]
1.0
0.5
50
100
150
Table 5-5 summarizes the average transformer load loss estimates along with the secondary line
loss averages discussed earlier.
Table 5-5
Loss Summary for Monitored Secondaries
Circuit A
Circuit B
0.54%
0.76%
0.72%
0.87%
Secondary Voltages
The customer voltages generally ranged between 118 and 126 V on both circuits. Figure 5-34
and Figure 5-35 show density and cumulative probability distributions for the voltages on each
circuit. The median customer voltage for both circuits was near 122 V. These customer voltages
show that there is significant room to lower voltages on these two circuits and still be above the
ANSI C84.1 lower range of 114 V.
5-29
0.30
1.0
0.25
Density
0.20
0.15
0.10
0.05
0.00
0.8
0.6
0.4
0.2
0.0
116
118
120
122
124
126
116
118
Meter voltages, V
120
122
124
126
124
126
Meter voltages, V
Figure 5-34
Meter Voltage Probability Distributions for Circuit A
1.0
0.20
Density
0.15
0.10
0.05
0.00
0.8
0.6
0.4
0.2
0.0
116
118
120
122
124
126
Meter voltages, V
116
118
120
122
Meter voltages, V
Figure 5-35
Meter Voltage Probability Distributions for Circuit B
The main primary voltage for both circuits was 24.9 kV. Both had pockets of 4.8-kV. The
voltage drop from the 24.9-kV primary had less impact than did voltage drop through 4.8-kV
portions and the voltage drop on distribution transformers and secondaries. Figure 5-36 shows
voltage profiles as a function of distance from the substation for both circuits. The voltages
shown are the fifth-percentile values (they are not a snapshot at a given point in time). Colors
red, green, and blue indicated phases A, B, and C. Note that both circuits had downline portions
that have a voltage boost from 24.9/4.8-kV step-down banks.
Figure 5-37 shows a similar voltage profile graph. In this case, transformer meter measurements
are shown along with their connected customer meters.
5-30
Circuit A
Secondary voltages, V
124
122
120
118
116
0
Circuit B
Secondary voltages, V
124
122
120
118
Notes: Colors indicate phasing. Secondary voltages are 5th percentile voltages.
Figure 5-36
Meter Voltage Profiles
5-31
124
Secondary voltages, V
123
122
121
120
119
118
Circuit B
Transformer secondary
Meter
Secondary voltages, V
122
120
118
116
0
Notes: Colors indicate phasing. Secondary voltages are 5th percentile voltages.
Figure 5-37
Secondary Voltage Profiles
5-32
Figure 5-38 and Figure 5-39 show probability distributions for transformer meter voltages along
with the set of meters connected to those transformers. The offset between the distributions was
less than 0.5 V.
1.0
Meter
Transformer secondary
Portion less than the xaxis value
0.30
0.25
Density
0.20
0.15
0.10
0.05
0.00
0.8
0.6
0.4
0.2
0.0
116
118
120
122
124
126
116
118
Secondary voltages, V
120
122
124
126
124
126
Secondary voltages, V
Figure 5-38
Meter and Transformer Secondary Voltage Probability Distributions for Circuit A
1.0
Meter
Transformer secondary
Portion less than the xaxis value
0.4
Density
0.3
0.2
0.1
0.0
0.8
0.6
0.4
0.2
0.0
116
118
120
122
Secondary voltages, V
124
126
116
118
120
122
Secondary voltages, V
Figure 5-39
Meter and Transformer Secondary Voltage Probability Distributions for Circuit B
5-33
Figure 5-40 and Figure 5-41 show probability distributions of meter voltages at the circuits peak
load. Even at peak load, voltages were generally high, with a median customer voltage of 121.2
V on Circuit A and a median of 122.9 V on Circuit B. The lower tail stretched closer to the
114-V lower ANSI limit. At peak, there was less room to lower voltages circuit-wide. Still, there
was room to drop voltages; on Circuit A, 95% of customer voltages were above 117.9 V. With
improvements to certain portions of the circuit, it may be possible for further drop at peak. See
Figure 5-42 for a voltage profile graph at the feeder peak load. The secondary fed by a 4.8-kV
transformer at about 1.5 miles had the lowest cluster of voltages. Other low voltages tended to be
at a few transformers. Note that this was only at one load snapshot. Other customer voltages may
be low at other times.
0.20
1.0
Density
0.15
0.10
0.05
0.8
0.6
0.4
0.2
0.0
0.00
114
116
118
120
122
Secondary voltage, V
124
126
114
116
118
122
Secondary voltage, V
Figure 5-40
Meter Voltage Probability Distributions at Peak Load for Circuit A
5-34
120
124
126
1.0
0.10
0.05
0.8
0.6
0.4
0.2
0.0
0.00
114
116
118
120
122
124
126
114
116
Secondary voltage, V
118
120
122
124
126
Secondary voltage, V
Figure 5-41
Meter Voltage Probability Distributions at Peak Load for Circuit B
126
4.8 kV
24.9 kV
124
Secondary voltages, V
Density
0.15
122
120
118
116
114
0
Figure 5-42
Meter Voltage Profile at Peak Load for Circuit A
5-35
1.0
1.2
Density
1.0
0.8
0.6
0.4
0.2
0.0
0.8
0.6
0.4
0.2
0.0
1
Figure 5-43
Secondary Voltage Drop Probability Distributions for the Metered Subset of Circuit A
5-36
1.0
1.5
Density
1.0
0.5
0.0
0.8
0.6
0.4
0.2
0.0
1
Figure 5-44
Secondary Voltage Drop Probability Distributions for the Metered Subset of Circuit B
For voltage drop, voltages at peak load are of most interest. Figure 5-45 and Figure 5-46 show
statistical distributions of voltage drops measured at each circuits summer peak load. Voltage
drops at peak were generally between 0.5 and 2 V.
1.0
0.6
0.5
Density
0.4
0.3
0.2
0.1
0.8
0.6
0.4
0.2
0.0
0.0
1
Figure 5-45
Peak-Load Secondary Voltage Drop Probability Distributions for the Metered Subset of
Circuit A
5-37
0.6
1.0
0.5
Density
0.4
0.3
0.2
0.1
0.8
0.6
0.4
0.2
0.0
0.0
1
Figure 5-46
Peak-Load Secondary Voltage Drop Probability Distributions for the Metered Subset of
Circuit B
Figure 5-47 and Figure 5-48 show how secondary voltage drops varied by several factors. The
loading and voltage drops for each data point were not coincident. Each data point shows the 95th
percentile voltage drop for one meter as a function of that meters loading and other parameters.
The loading parameters are also 95th percentile loadings for that meter. All of these parameters
had positive correlation. The strongest correlations were with secondary length and the term:
R * kW + X * kvar. Voltage drop on a single conductor is approximately found as:
Vdrop = R I R + X I X
Eq. 5-39
Where,
R, X = line resistance and reactance, ohms
IR, IX = resistive and reactive portions of line current, A
The term R * kW + X * kvar is proportional to R * IR + X * X * IX. This relationship is valid for
current through one conductor. For multiple loads on a branched secondary, the relationship is
more complicated.
5-38
2.5
Voltage drop, 120V base
2.5
2.0
1.5
1.0
0.5
0.0
2.0
1.5
1.0
0.5
0.0
0.0
Load, kW
1.5
2.0
2.5
2.5
Voltage drop, 120V base
1.0
2.5
2.0
1.5
1.0
0.5
0.0
2.0
1.5
1.0
0.5
0.0
0.00
0.10
0.20
0.30
2.5
Voltage drop, 120V base
0.5
2.0
1.5
1.0
0.5
0.0
2.0
1.5
1.0
0.5
0.0
200
400
600
Secondary length, ft
800
0.0
0.2
0.4
0.6
0.8
1.0
R * kW + X * kvar
Figure 5-47
Voltage Drops Versus Different Factors for the Monitored Subset of Circuit A
5-39
2.5
Voltage drop, 120V base
2.5
2.0
1.5
1.0
0.5
0.0
2.0
1.5
1.0
0.5
0.0
0.0
Load, kW
1.0
1.5
2.0
2.5
Voltage drop, 120V base
2.5
2.0
1.5
1.0
0.5
0.0
2.0
1.5
1.0
0.5
0.0
0.00
0.10
0.20
0.30
2.5
Voltage drop, 120V base
0.5
2.0
1.5
1.0
0.5
0.0
2.0
1.5
1.0
0.5
0.0
200
400
600
Secondary length, ft
800
0.0
0.2
0.4
0.6
0.8
1.0
R * kW + X * kvar
Figure 5-48
Voltage Drops Versus Different Factors for the Monitored Subset of Circuit B
5-40
transformer voltage drop was between one and two volts on both circuits. On both circuits, the
spread of voltage drops was relatively wide.
Figure 5-53 and Figure 5-54 show the 99th percentile loading versus the voltage drop at that
loading. These strongly correlated. On Circuit A, the 5-kV transformers had higher impedances
than the 14-kV transformers.
1.0
2.0
Density
1.5
1.0
0.5
0.0
0.8
0.6
0.4
0.2
0.0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Figure 5-49
Transformer Voltage Drop Probability Distributions for the Metered Subset of Circuit A
1.2
1.0
1.0
Density
0.8
0.6
0.4
0.2
0.0
0.8
0.6
0.4
0.2
0.0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Figure 5-50
Transformer Voltage Drop Probability Distributions for the Metered Subset of Circuit B
5-41
1.0
0.6
0.5
Density
0.4
0.3
0.2
0.1
0.8
0.6
0.4
0.2
0.0
0.0
0
Figure 5-51
99th Percentile Transformer Voltage Drop Probability Distributions for the Metered Subset
of Circuit A
1.0
Density
0.3
0.2
0.1
0.8
0.6
0.4
0.2
0.0
0.0
0
Figure 5-52
99th Percentile Transformer Voltage Drop Probability Distributions for the Metered Subset
of Circuit B
5-42
voltage
5 kV
14 kV
transformer size
10 kVA
15 kVA
25 kVA
35 kVA
0
0
20
40
60
80
5-43
5
voltage
5 kV
14 kV
kVA range
(0,10]
(10,20]
(20,30]
(30,40]
(40,50]
0
0
20
40
60
80
100
120
140
5-44
0.7
1.0
0.6
Density
0.5
0.4
0.3
0.2
0.1
0.0
0.8
0.6
0.4
0.2
0.0
1
Figure 5-55
Transformer and Secondary Voltage Drops Versus Loading for the Monitored Subset of
Circuit A
1.0
0.6
0.5
Density
0.4
0.3
0.2
0.1
0.0
0.8
0.6
0.4
0.2
0.0
1
Figure 5-56
Transformer and Secondary Voltage Drops Versus Loading for the Monitored Subset of
Circuit B
5-45
Summary
Based on analysis of secondaries on two circuits with detailed modeling, it appears that the
losses from secondaries were underestimated in the general modeling described in Chapter 2. In
four circuits modeled in detail, simulations showed losses of less than 1% of consumption. In
measurements of secondary line losses on two circuits, secondaries on one circuit averaged
0.87% losses, and secondaries on the other circuit averaged 0.72%.
Metering data showed voltages above 121 V throughout most of the year. That leaves significant
room to drop voltages. Voltage drops across secondaries were typically between 0.5 and 2 V at
peak load. Transformer voltage drops were typically less than 1 V, with peak values typically
ranging between 0.5 and 3 V. In comparing detailed modeling of voltage drops to simplified
modeling, detailed models produced more accurate results, and simplifications tended to
underestimate voltage drops. Secondary and transformer voltage drop is significant in that it can
limit circuit voltage reduction.
These analysis results suggest that advanced metering data can help to quantify transformer
loading and losses. It can be used to help maintain better customer voltages and identify problem
secondaries.
5-46
6
VOLTAGE OPTIMIZATION FIELD TRIAL RESULTS
Load decreased by 1.6% to 2.7% for substation voltages that were reduced by 2.0 to 4.0%. In
terms of a CVR factor (conservation voltage reduction factor), the results for percent change
in load for a 1% change in voltage generally ranged from 0.6 to 0.8.
Reactive power responded even more strongly to voltage reduction than real power. Reactive
power CVR factors exceeded four on two circuits with AMI metering capable of measuring
reactive power.
A statistical regression approach using measurements from one or more circuits with
comparable load patterns worked well to normalize results and more precisely evaluate the
impact of voltage on load.
Voltage reduction was most effective in the summer and least effective in the winter when
more thermostatically controlled heating load (constant energy) was present.
Average energy reductions for the EPRI tests circuits were similar in range to a previous study
by the Northwest Energy Efficiency Alliance (NEEA). This is significant because the climate
and load patterns in the Northwest United States differ from the EPRI circuits, which were
primarily in the Southeast United States.
6-1
Advanced customer metering to evaluate and/or control voltage based on voltages measured
at customers
Fixed and switched capacitors along with voltage regulators to flatten voltage profiles on
circuits to allow more precise voltage control
The impact of voltage on loads is often quantified as a CVR factor, the percent change in load
1
for a 1% change in voltage. Kirshner and Giorsetto analyzed trials of voltage reduction at
several utilities. While results varied significantly, most test circuits had energy savings of
between 0.5 and 1% for each 1% voltage reduction at the substation. Their regression analysis of
the feeders found that residential energy savings were 0.76% for each 1% reduction in voltage,
while commercial and industrial loads had reductions of 0.99% and 0.41% (but, the correlations
between load class and energy reduction were fairly small).
More recently, the Northwest Energy Efficiency Alliance (NEEA) and their contractor RW Beck
and several utilities evaluated voltage reduction in the U.S. Pacific Northwest (see Figure 6-1).2
They evaluated changes at the circuit level and also changes directly to residential customers. In
their evaluation of voltage changes at the circuit level, using temperature-adjusted regressions,
they found an average CVR factor of 0.69 based on a substation voltage change of 2.5%.
Source: NEEA 1207, Distribution Efficiency Initiative, Northwest Energy Efficiency Alliance, 2007.
Figure 6-1
Energy Savings Versus Voltage Reduction for the NEEA Study Pilot Feeders
D. Kirshner and P. Giorsetto, Statistical Tests of Energy Savings Due to Voltage Reduction, IEEE Transactions
on Power Apparatus and Systems, vol. PAS-103, no. 6, June 1984, pp. 120510.
NEEA 1207, Distribution Efficiency Initiative, Northwest Energy Efficiency Alliance, 2007. Available at
http://rwbeck.com/neea/.
6-2
In NEEAs evaluation of 395 residential customers, they estimated a CVR factor of 0.57 based
on a voltage change of 4.3% at the customer. The NEEA study found seasonal differences. In the
customer evaluation, they found a CVR factor in the winter of 0.5, compared to a summer CVR
factor of 0.78.
The NEEA study found even more dramatic changes with reactive power. In their feeder
monitoring study, they found CVRvar factors between 3.0 and 3.5 (vars drop by 3% for every 1%
drop in voltage at the substation). That indicates that a large component of the change is due to
the reduction in magnetizing current in motors and transformers because this excitation current
has a highly nonlinear response to voltage. The change in reactive power was not particularly
sensitive to season. Figure 6-2 shows that the response of reactive power to voltage may not be
linear with voltage.
Source: NEEA 1207, Distribution Efficiency Initiative, Northwest Energy Efficiency Alliance, 2007.
Figure 6-2
Change in Reactive Power in the NEEA Study Homes
Lefebvre et al.3 reported on tests of voltage reduction at Hydro Quebec. They found a strongly
temperature-dependent CVR factor. At 20C, they estimated a CVR factor of 0.55, and at -10C,
the CVR factor dropped to 0.15. Overall, for their mix of loads, they estimated a summer CVR
factor of 0.67, a winter CVR factor of 0.20, and an overall CVR factor of 0.4. Table 6-1 shows
their estimated breakdown by load type.
S. Lefebvre et al., Measuring the Efficiency of Voltage Reduction at Hydro-Quebec Distribution, presented at
IEEE Power and Energy Society General Meeting, 2008.
6-3
Winter CVR
Factor
0.67
0.06
0.67
0.12
Commercial
0.97
0.80
Small industries
0.10
0.10
The load response of a circuit to voltage changes ultimately comes down to the characteristics of
the loads on the circuit. Voltage reduction produces the strongest response in energy with
resistive loadsthe power drawn by a resistive load decreases with the voltage squared. Lighting
and resistive heating loads are the dominant resistive loads; these are not quite ideal resistive
loads. For example, the power on incandescent lights varies as the voltage to the power of about
1.6not quite to the power of 2, but close. Residential and commercial loads have higher
percentages of resistive load. For water heaters and other devices that regulate to a temperature,
reducing voltage does not reduce overall energy usage; the devices just run for longer times. This
is likely why CVR factors are lower during the winter when there is significant thermostatically
controlled resistive heat.
Table 6-2 shows the impact of voltage on the energy consumption of several devices published
by EPRI in 1981. The CVR factors are specified for a voltage change from 120 V to 115 V (or
240 V to 230 V).4 The results for motors in Table 6-2 are based on a mechanical torque of one
per unit. For more lightly loaded motors, the CVR factor is higher as shown for one example in
Figure 6-3. This EPRI study also showed that the reactive-power consumption of motors was
highly dependent on voltage. More work is needed to evaluate the impact of voltage reduction on
more modern end-use equipment. Figure 6-4 shows an example of more recent tests of an air
conditioner tested by PG&E.
Effects of Reduced Voltage on the Operation and Efficiency of Electric Loads, vol. 1, EPRI, Palo Alto, CA: 1981.
EL-2036.
6-4
Table 6-2
Comparison of Voltage Impact on End-Use Equipment
Device
CVR factor
0.24
0.26
0.55
Incandescent lamp
1.51
Fluorescent lamp
0.72
Refrigerator
2.28
0.26
0.02
0.02
0.00
6-2
Figure 6-3
Impact of Voltage and Torque on an Example Motor
6-5
12.0
11.9
11.8
11.7
180
190
200
210
220
230
240
220
230
240
Voltage, V
0.98
Power factor
0.97
0.96
0.95
0.94
0.93
0.92
180
190
200
210
Voltage, V
Figure 6-4
Tests of a Typical Modern 3-kW Air Conditioner
Study Approach
The evaluation of voltage optimization or voltage reduction was performed following the
approach set out by the NEEA study. Voltage was controlled on alternate days as follows:
Reduced-voltage mode Voltages are lowered using voltage regulators and/or LTC controls.
After enough time operating in alternating modes, the variations due to weather and load usage
even out, and the impact of the difference in voltage should be seen.
Primary metering from the substation was available to monitor voltages, real power, and reactive
power, usually at 15-minute intervals. Several of the circuits had wide AMI meter coverage.
6-6
There are a number of different ways to implement a reduced-voltage mode. For this study, all
voltage control was implemented by local regulator or LTC controls. The reduced-voltage mode
was implemented as a change in the voltage set point or a change in the voltage set point along
with the addition of line-drop compensation. On some circuits, line-drop compensation was
needed to prevent voltages from going too low under peak load. No significant measures were
taken to adjust voltage profiles to accommodate reduced voltages; neither capacitors nor voltage
regulators were added to flatten voltage profiles.
Eq. 6-1
where
kW = circuit energy consumption
hdh = heating-degree hours
cdh = cooling-degree hours
0, 1, 2 = regression coefficients
One set of these regressions was done for the normal-mode data, and another was done for the
reduced-mode data. From these sets of regression models, estimates of normalized load under
normal voltage and reduced voltage were found. The minimum covariance determinant (MCD)
regression modeling approach was used to reduce the impact of outliers. The NEEA temperaturenormalization model was largely based on the approach of David Bell, PCS UtiliData.6,7
M. Fels, ed. (1986), Measuring Energy Savings: The Scorekeeping Approach, Special PRISM Issue of Energy
and Buildings, vol. 9, no. 1-2.
D. Bell, Estimation of PCS UtiliData AdaptiVolt System Performance Using Observed Energy Demand
Profiles, 2004.
http://www.pcsutilidata.com/userfiles/file/CVR_Performance_Estimation_2004.pdf
D. Bell, Automated CVR Protocol No. 1, PCS UtiliData, 2003.
6-7
For the voltage optimization trials in this project, project engineers attempted temperature
correction along the lines of the NEEA method, but they found that using a regression approach
using measurements from a similar circuit resulted in a better regression model. Compare the
two regression models in Figure 6-5. Each graph includes plots of daily average power
consumption, one plotted against average daily temperature and another plotted versus the
average power consumption of another circuit with a similar load composition. Each point
represents the average results for one day. Other than a few outliers, the model based on the
2
comparable circuit had a higher coefficient of determination (R ) and a higher degree of
statistical significance based on the term related to the voltage control mode.
Figure 6-5
Temperature Normalization and Comparable-Circuit Normalization
Eq. 6-2
where
MW = average power of the test circuit
MWcomparable = average power measured at a comparable circuit
Vstate = 1 for normal voltage, 0 for reduced voltage
k0, k1, k2 = regression coefficients
Especially for determining prediction intervals, the following adjustments were used to improve
the results from the linear regression models:
6-8
Outlier removal Outliers can significantly skew results. Outliers can occur because of
unusual load patterns (holidays, possibly) or from circuit changes (outages or other feeder
switching on either circuit). We excluded days manually based on the regression residuals
(days with large errors between predicted and measured).
Autocorrelation adjustment Time series data has autocorrelation, meaning that the data is
correlated with itself. One day tends to look like the next. In linear modeling, one must
account for this, especially for determining confidence intervals.
See Figure 6-6 for graphs showing how well several of the circuits compare to their comparable
circuit. Table 6-3 shows the regression formulas for the circuits evaluated. Many of the
regression models improved with the addition of a second comparable circuit.
Analyzing daily data proved to be sufficient for the normalization. Using hourly data did not
improve confidence intervals on regression estimates for circuits in which this was tried.
Circuit B
Circuit C
3
3
Circuit D
10
Circuit E
10
10
Figure 6-6
Test Circuits Compared to Comparison Circuits
6-9
Regression formula
wkend + voltagestate
comparable1 + voltagestate
While most of the regression models were straightforward linear equations with one or two
comparable circuits included, one of the circuits (Circuit B) did not have a single comparable
circuit that matched well. Circuit A had significant commercial load, and its load pattern did not
track well with other circuits that were compared. It was possible to use a combination of circuits
to obtain a reasonably good model. One of the comparison circuits compared well when
accounting for weekdays and weekends (top graph in Figure 6-7), and another circuit compared
well if season was accounted for (bottom graph in Figure 6-7). The best regression for Circuit A
included both of these circuits along with two others plus a weekday/weekend indicator and a
seasonal indicator.
6-10
Weekday
Weekend
10
10
11
Nonwinter
Winter
10
10
12
14
16
Figure 6-7
Circuit A Comparisons to Other Circuits
Figure 6-8 shows an example profile for one of the test circuits along with the scaled estimate
based on the comparable circuit. On this particular circuit, the mode was switched at 2 A.M. As
noted earlier, the voltage-mode effect was small relative to the daily swings, but over time, the
impact of the voltage reduction can be seen.
6-11
Figure 6-8
Example Load and Voltage Profile
6-12
Overall Results
See Table 6-4 for characteristics of the nine test circuits. Most of them were in the Southeast
United States. Most had mainly residential load. Table 6-5Table 6-5 summarizes the results from
the normalized regression model used to estimate the energy reduced along with the confidence
interval. The CVR factors ranged from 0.5 to 0.9.
Table 6-4
Circuit Characteristics
Circuit
Voltage
Number of
Customers
Circuit
Miles
Location
Percent
Residential
Voltage Control
Devices
34.5 kV
3597
105
Southeast US
75
Station LTC
34.5 kV
800
Southeast US
80
Station LTC
12.5 kV
1545
73
Southeast US
97
Station regulator
20.0 kV
1515
90
Europe
N/A
Station LTC
20.0 kV
903
Europe
N/A
Station LTC
12.5 kV
1379
48
Southeast US
96
12.5 kV
721
38
Southeast US
97
Station regulator
23.9 kV
2867
48
Southeast US
94
Station LTC
23.9 kV
2088
37
Southeast US
94
Station LTC
Table 6-5
Voltage Optimization Results From Field Trials
Circuit
MonitorDays
Linear
Model
R2
Voltage
Reduction
Energy Reduction
[95% Confidence Interval]
CVR Factor
729
0.852
3.28%
0.613
729
0.017
3.29%
0.071
664
0.983
2.50%
0.858
379
0.938
2.98%
0.701
379
0.980
2.98%
0.556
345
0.989
3.57%
0.665
519
0.963
3.95%
0.608
430
0.992
2.00%
0.584
430
0.995
2.00%
0.756
6-13
Figure 6-9 shows the energy reduction compared with the voltage reduction based on the results
in Table 6-5. The confidence intervals were relative tight. The two European circuits had the
widest confidence limits. Note that the voltage changes given were based on substation
measurements. The percentage change in customer voltages were likely to differ from the
substation values. Figure 6-10 shows the same data with CVR factors on the Y axis. The CVR
factors stayed approximately the same or decline slightly with a deeper voltage reduction.
Figure 6-9
Comparison of Energy Reduction and Confidence Intervals for Several Circuits
6-14
1.0
CVR factor
0.8
0.6
0.4
0.2
0.0
2.0
2.5
3.0
3.5
4.0
Figure 6-10
Comparison of Energy Reduction and Confidence Intervals for Several Circuits
The one data point that stands out is Circuit B, which did not have a statistically significant
energy reduction. Circuit Aa circuit fed from the same bus as Circuit Bhad statistically
significant energy savings. The load characteristics seem similar. Circuit B had less loading than
its sister circuit, and the instrumentation may not have been as accurate as a result (more load
was expected). Circuit B had some office park commercial load, and Circuit A had more
restaurant type commercial load, but both were still composed of more than 75% residential load.
Overall, the difference between the two circuits remains a mystery.
Comparisons by Month
Figure 6-11 compares monthly average circuit loads under normal voltage control and reduced
voltage. Each panel shows results from one circuit. These monthly results are based on the
following assumptions:
Use the same regression models as the total results shown in the previous section.
From the regression model, estimate the normalized average power (in MW) for the normal
mode and for the reduced-voltage mode.
6-15
Figure 6-12 through Figure 6-14 show different ways to compare the monthly results. Whether
on an absolute or a relative basis, voltage reduction generally had somewhat more impact on load
during the summer on some circuits. Note that this data was noisier, because each regression
model was based on 1/12th of the data.
Normal mode
Reduced voltage mode
Circuit A
Circuit B
1.6
1.4
Circuit C
1.2
7
1.0
6
0.8
Circuit D
4.0
Circuit E
2.0
3.5
Circuit F
4.0
3.5
3.0
3.0
1.5
2.5
2.5
1.0
2.0
2.0
Circuit G
Circuit H
Circuit I
6
9
2.0
1.5
6
2
8 10 12
3
2
8 10 12
Month
Figure 6-11
Monthly Comparisons of Average Power Consumption With and Without Voltage
Reduction
6-16
8 10 12
Circuit A
Circuit B
Circuit C
0.14
0.25
0.05
0.12
0.10
0.20
0.00
0.08
0.15
0.06
0.05
0.10
0.04
Energy difference, MW
Circuit D
Circuit E
Circuit F
0.15
0.04
0.10
0.10
0.02
0.05
0.00
0.00
0.02
0.05
0.00
Circuit G
0.05
0.04
0.03
0.02
0.01
0.00
0.01
2
8 10 12
Circuit H
Circuit I
0.14
0.10
0.12
0.08
0.10
0.06
0.08
0.04
0.06
0.02
0.04
0.00
2
8 10 12
8 10 12
Month
Figure 6-12
Reduction in Average Power Consumption With Voltage Reduction
6-17
Circuit A
4.0
3.5
3.0
2.5
3
2
1.5
Circuit C
4
2.0
Circuit B
Circuit D
Circuit E
Circuit F
Circuit G
Circuit H
1.6
Circuit I
2.0
1.4
2
1.5
1.2
1.0
1.0
1
0.8
0.5
0.6
0.0
0.4
2
8 10 12
8 10 12
Month
Figure 6-13
Percent Reduction in Average Power Consumption With Voltage Reduction
6-18
8 10 12
Circuit A
Circuit B
2
3.0
2.5
Circuit C
1.5
2.0
1.0
1.5
1
1.0
0.5
2
0.5
Circuit D
Circuit F
1.0
1.0
1.5
CVR factor
Circuit E
0.8
1.0
0.5
0.5
0.0
0.0
0.5
0.6
0.4
0.2
Circuit G
Circuit H
0.8
Circuit I
1.0
0.8
0.6
0.8
0.6
0.6
0.2
0.4
0.4
0.0
0.2
0.4
0.2
8 10 12
0.0
2
8 10 12
8 10 12
Month
Figure 6-14
CVR Factor by Month
6-19
Normal mode
Reduced voltage mode
Circuit A
Circuit B
7.3
Circuit C
1.25
3.90
7.2
7.1
3.85
1.20
3.80
7.0
1.15
6.9
3.75
6.8
1.10
3.70
Circuit D
Circuit E
Circuit F
1.40
2.8
2.85
1.35
2.6
2.4
1.30
2.80
1.25
2.75
2.2
1.20
2.0
1.15
2.70
Circuit G
Circuit H
Circuit I
7.95
1.54
4.50
7.90
1.52
4.45
7.85
4.40
1.50
7.80
1.48
7.75
4.35
7.70
4.30
Day of week
Day 1 = Sunday
Figure 6-15
Weekly Comparisons of Average Power Consumption With and Without Voltage Reduction
Hourly Results
Figure 6-16 shows daily load profiles for each test circuit by voltage control mode. These were
analyzed using the same regression normalization approach. Figure 6-17 through Figure 6-19
show other comparisons based on these results. The most striking result is that many of the
circuits showed a pronounced upturn just after the voltage-control mode changes, on an absolute
6-20
and especially on a relative basis. The mode changes at midnight on circuits C through F and at
1 AM on circuits A and B. This was not unexpected because the load from resistive heating loads
(hot water heaters, space heaters, and so on) will change as a function of the voltage squared.
Then with lower (or higher) heat output, the heating load will run longer (or shorter) to keep the
overall energy input the same. No other trends stand out among the noise.
Normal mode
Reduced voltage mode
Circuit A
Circuit B
Circuit C
9
1.4
8
7
6
1.2
4.0
1.1
3.5
1.0
3.0
0.9
4.5
1.3
Circuit D
Circuit E
3.0
Circuit F
3.5
1.5
2.8
3.0
1.4
2.6
2.4
1.3
2.2
1.2
2.5
2.0
Circuit G
Circuit H
10
Circuit I
6
1.8
9
1.6
1.4
1.2
10 15 20 25
4
3
0
10 15 20 25
10 15 20 25
Hour
Figure 6-16
Hourly Comparisons of Average Power Consumption With and Without Voltage Reduction
6-21
Circuit A
Circuit B
Circuit C
0.20
0.12
0.02
0.10
0.15
0.01
0.08
0.00
0.06
0.01
0.10
0.05
0.02
0.04
Energy difference, MW
Circuit D
0.06
Circuit E
Circuit F
0.09
0.04
0.08
0.05
0.04
0.03
0.03
0.07
0.02
0.06
0.05
0.02
0.01
0.04
0.01
Circuit G
0.050
Circuit H
0.12
0.08
0.045
0.040
Circuit I
0.07
0.10
0.035
0.06
0.08
0.030
0.05
0.025
0.04
0.06
0.020
0
10 15 20 25
10 15 20 25
Hour
Figure 6-17
Reduction in Average Power Consumption With Voltage Reduction
6-22
10 15 20 25
Circuit A
Circuit B
Circuit C
3
2.0
5
1
1.5
1.0
2
1
0.5
Circuit D
Circuit E
4
2.5
Circuit F
4.5
4.0
2.0
3.5
3.0
1.5
2.5
1.0
2.0
0.5
1.5
Circuit G
4.0
Circuit H
Circuit I
2.0
2.5
1.5
2.0
3.5
3.0
2.5
1.5
1.0
2.0
1.0
1.5
0
10 15 20 25
10 15 20 25
10 15 20 25
Hour
Figure 6-18
Percent Reduction in Average Power Consumption With Voltage Reduction
6-23
Circuit A
Circuit B
Circuit C
2.5
0.8
0.5
2.0
0.6
1.5
0.0
1.0
0.4
0.5
0.5
0.2
Circuit D
Circuit E
Circuit F
1.6
CVR factor
0.8
0.6
0.4
0.2
1.2
1.4
1.0
1.2
0.8
1.0
0.6
0.8
0.4
0.6
0.2
0.4
Circuit G
Circuit H
Circuit I
1.6
1.2
1.0
0.8
0.6
1.4
1.0
1.2
0.8
1.0
0.8
0.6
0.4
0.6
0.4
0
10 15 20 25
0.4
0
10 15 20 25
10 15 20 25
Hour
Figure 6-19
CVR Factor by Hour of the Day
6-24
3.0
EPRI circuits
NEEA circuits
2.5
2.0
1.5
1.0
0.5
0.0
0
Figure 6-20
Comparison of Voltage Reduction on EPRI Feeders and NEEA Feeders
1.0
EPRI circuits
NEEA circuits
CVR factor
0.8
0.6
0.4
0.2
0.0
2.0
2.5
3.0
3.5
4.0
Figure 6-21
Comparison of CVR Factors on EPRI Feeders and NEEA Feeders
6-25
6-26
Circuit
Voltage
Reduction
Reactive Power
Reduction in kvar
kvar Reduced
per MW of
Load
3.62%
90
13
2.80%
309
90
3.71%
30
11
1.90%
49
16
1.90%
32
24
2.76%
3.0
Normal mode
Reduced voltage mode
2.5
Average Mvar
2.0
1.5
1.0
0.5
0.0
0
10
15
20
Figure 6-22
Normalized Var Comparisons for Circuit A
6-27
116
118
120
122
124
1.0
0.8
0.6
0.4
0.2
0.0
126
114
116
Circuit B
0.30
0.25
0.20
0.15
0.10
0.05
0.00
114
116
118
120
122
124
0.20
0.15
0.10
0.05
0.00
120
122
124
116
0.15
0.10
0.05
0.00
120
122
124
0.20
0.15
0.10
0.05
0.00
122
124
126
122
124
126
0.0
116
118
120
1.0
0.8
0.6
0.4
0.2
0.0
116
118
120
122
124
126
1.0
0.8
0.6
0.4
0.2
0.0
114
Voltage, V
Figure 6-23
Statistical Distribution of Customer Voltages by Circuit
6-28
126
0.2
114
Portion less than the xaxis value
0.25
120
124
0.4
Voltage, V
Circuit I
118
122
0.6
126
0.30
116
120
0.8
Voltage, V
114
118
1.0
114
Portion less than the xaxis value
0.20
118
126
0.0
Voltage, V
0.25
116
124
0.2
126
Circuit H
114
122
0.4
Voltage, V
0.30
126
0.6
114
Portion less than the xaxis value
0.25
118
124
Voltage, V
0.30
116
122
0.8
126
Circuit G
114
120
1.0
Voltage, V
0.35
118
Voltage, V
Portion less than the xaxis value
Voltage, V
116
118
120
Voltage, V
For Circuit A, an AMI dataset of 3744 meters covered 23 months of day-on/day-off operation.
Some of the key features of this data set include:
Voltage was also recorded, but it was a spotty measurement without much precision.
The analysis approach for the AMI data is similar to that for feeder-level analysis. The meter
data was grouped together, and then the energy reduction was estimated based on a regression of
average daily watt readings for the grouping. As with the feeder-level estimates, the customergroup estimates were based on regressions based on comparable circuits. The same regression
model was used for the complete circuit as was used for the customer subsets. One difference is
that outliers were excluded automatically rather than by hand.
Circuit A had considerable commercial load. One of the main questions about voltage reduction
is how effective it is on commercial loads. Figure 6-24 shows a breakdown for four load
classifications for the metered customers on Circuit A.
Figure 6-24 shows clear differences by customer type. Not surprisingly, lighting loads had the
largest reduction in energy. The commercial load on this circuit was mainly shopping, bars, and
restaurants. This type of commercial load appears to have had less reduction in energy than
residential load.
On Circuit A, the average voltage reduction at the substation was 3.28%. Because this was a
relatively short 34.5-kV circuit with little voltage drop, the customer voltage reduction was likely
to be similar to the feeder-level voltage reduction. Based on a 3.28% voltage reduction, the CVR
factors for the each category are shown in Table 6-7.
Table 6-7
Circuit A CVR Factors by Customer Type
Customer Type
Energy
Reduction in
Percent
CVR Factor
(W / V)
Residential
2.06
0.63
Small Commercial
1.22
0.37
Restaurant
1.33
0.41
Lighting
4.65
1.42
6-29
1.5
1.5
1.0
1.0
kW
kW
Residential
0.5
0.5
0.0
0.0
2
10
12
10
15
20
10
15
20
10
15
20
10
15
20
kW
kW
3
2
3
2
0
2
10
12
Restaurant/food service
70
60
60
50
50
40
40
kW
kW
70
30
30
20
20
10
10
0
2
10
12
Lighting
0.5
0.6
0.5
0.3
kW
kW
0.4
0.2
0.4
0.3
0.2
0.1
0.1
0.0
0.0
2
Month
10
12
Hour
The right-most column of graph panels shows the energy reduction for that set of
customers. The red dot marks the estimate of the energy savings with voltage reduction,
and the gray line marks the 95% confidence intervals around the estimate. The left two
columns of graph panels show yearly and hourly customer usage profiles for each
grouping of customers.
Figure 6-24
Energy Usage Profiles by Customer Class With Energy Reduction From Voltage Reduction
for Circuit A
6-30
Statistical clustering algorithms are another way to subdivide the metering data to add insights
into the performance of different types of customers. Figure 6-25 compares the results of one
type of regression clustering approach. This was based on finite mixture models.8,9 These
produced a mixture of regression models. Each grouping of customers had its own regression
model. For the non-residential load on Circuit A, customers split naturally into the lighting and
commercial load as shown in Figure 6-25.
1
0.5
0.4
0.4
kW
kW
0.3
0.2
0.3
0.2
0.1
0.1
0.0
0.0
2
10
12
10
15
20
10
15
20
2
8
6
kW
kW
6
4
0
2
Month
10
12
Hour
Figure 6-25
Clustering of Non-Residential Meters for Circuit A
For the non-lighting group (grouping 2 from Figure 6-25), the same regression clustering
algorithm divided this set into the three groupings shown in Figure 6-26. The load grouping with
the most impact from voltage reduction had customer establishments operating later at night and
had higher load in the winter. This may indicate more lighting load present or more nonthermostatically controlled heating load. There was not much discernable 8 A.M. 5 P.M. office
load on this circuit.
M. Wedel and W.S. DeSarbo, A Mixture Likelihood Approach for Generalized Linear Models. Journal of
Classification, vol. 12, 1995, pp. 2155.
F. Leisch, FlexMix: A General Framework for Finite Mixture Models and Latent Class Regression in R.
Journal of Statistical Software, vol. 11. no. 8, 2004. http://www.jstatsoft.org/v11/i08/.
6-31
1
35
30
25
30
20
25
15
20
kW
kW
20
15
kW
25
15
10
10
10
12
10
10
15
20
10
15
20
10
15
20
2
10
8
8
kW
kW
kW
6
4
2
2
0
0
2
10
12
0
1
3
2.5
2.0
2.0
2.0
1.5
1.0
1.5
kW
kW
kW
1.5
1.0
0.5
0.5
0.0
0.5
0.0
2
Month
10
12
1.0
0.0
1
Day of week
Hour
Figure 6-26
Clustering of Non-Residential Meters for Circuit A After Removing Lighting Load
The performances for the twenty largest loads were also evaluated; the energy reduction from
CVR is shown in Figure 6-27. Because individual loads varied from day to day more than the
groupings, the error bands were much higher. The energy reduction estimates were widely
spread. The same regression model was used for all loads. Tighter error bands might be obtained
with custom regression models for each metered load.
Figure 6-28 shows yearly and daily load profiles for each of these 20 loads. No major patterns
stood out. It was not a surprise that Customer S was near the top in terms of percent reduction
because this was largely lighting load. The rest were more muddled. Customer D had the most
striking 8 A.M. 5 P.M. load pattern, and it ranked low in terms of energy reduction. Customer
E, which was very consistent, also ranked low (with tight confidence intervals), probably
indicating a relatively constant energy load. Customer P was also flat in terms of daily profile,
but it showed much more energy reduction.
6-32
A
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P
Q
R
S
T
2
Figure 6-27
Energy Reduction From Voltage Reduction for the 20 Largest Customers on Circuit A
6-33
30
25
15
10
10
kW
kW
30
40
20
15
30
20
20
10
10
25
20
20
15
15
15
kW
kW
15
kW
20
10
10
5
0
25
20
20
15
15
10
10
5
0
35
40
15
10
10
kW
25
15
50
30
20
20
25
25
30
30
15
40
20
25
kW
10
5
30
kW
10
10
25
20
10
50
25
20
30
20
15
20
14
14
12
12
10
10
20
F
kW
kW
15
10
5
15
10
60
60
50
50
40
40
30
30
20
20
10
10
25
20
20
15
15
10
10
10
30
20
10
25
15
25
kW
kW
15
10
15
20
10
60
70
20
35
50
60
I
kW
40
30
30
20
10
10
80
80
60
60
40
40
20
20
0
2
Month
10
12
25
20
10
15
10
5
0
15
15
10
10
100
100
30
15
50
T
kW
kW
30
20
kW
20
10
40
40
20
10
25
kW
kW
20
15
10
30
25
20
15
20
25
15
20
10
10
kW
kW
10
0
0
10
Hour
15
20
10
0
2
Month
10
12
10
Hour
15
20
Figure 6-28
Profiles and Energy Reduction From Voltage Reduction for the 20 Largest Customers on
Circuit A
6-34
10
Circuit A had two residential rates, a standard rate and an efficiency rate. The efficiency rate
targeted residences that have taken more than normal steps to improve thermal and equipment
efficiency. The efficiency rate class had higher rates during the summer. Within the confidence
intervals given, these sets of customers were similar as shown in Figure 6-29.
1.5
1.5
1.0
1.0
kW
kW
Residential
0.5
0.5
0.0
0.0
2
10
12
10
15
20
10
15
20
2.5
2.0
2.0
1.5
1.5
kW
kW
Residential efficiency
2.5
1.0
1.0
0.5
0.5
0.0
0.0
2
Month
10
12
Hour
Figure 6-29
Energy Reduction From Voltage Reduction for Two Residential Rate Classes on Circuit A
While rate type did not show much difference between residential customers, annual load
profiles showed significant difference. Figure 6-30 compares groups of customers clustered by
normalized monthly usage profiles. Clustering was done with the widely used kmeans
algorithm.10,11 The summer-peaking groupings in the top graph panels had much more energy
reduction than did the winter-peaking groupings in the bottom panels. Low gains for winterpeaking loads were likely from thermostatically controlled resistive heat loads, which have
constant energy usage despite voltage reductions. Heat pumps were common in this service
territory, and these often had resistive heating elements to boost heating on colder days. For
summer-peaking loads, gains from voltage reduction were higher because air conditioners
generally run more efficiently at lower voltage.
Figure 6-31 shows the performance of these customer groupings split out by season. In
Figure 6-31, summer is defined as June, July, and August; winter is defined as December,
January, and February; and shoulder is defined as April, October, and November. All five
customer categories had similar energy reductions of about 2% in the summer. Winter months
showed significant reduction in energy savings for winter peaking loads (presumably with
10
J. MacQueen, Some Methods for Classification and Analysis of Multivariate Observations. Proceedings of the
Fifth Berkeley Symposium on Mathematical Statistics and Probability, eds L. M. Le Cam & J. Neyman, vol. 1,
Berkeley, CA: University of California Press. 1967, pp. 281297.
11
S. P. Lloyd, Least Squares Quantization in PCM. Technical Note, Bell Laboratories. Published in 1982 in IEEE
Transactions on Information Theory, vol. 28, 1957, 1982, pp. 128137.
6-35
considerable thermostatically controlled heating load). Shoulder months tended to have a higher
percentage savings, possibly because there was relatively more lighting load during that time.
1
kW
kW
1.5
1.0
0.5
0.0
1.0
0.5
0.0
10
12
10
15
20
0.0
0.5
1.0
1.5
2.0
2.5
3.0
10
15
20
0.0
0.5
1.0
1.5
2.0
2.5
3.0
10
15
20
0.0
0.5
1.0
1.5
2.0
2.5
3.0
10
15
20
0.0
0.5
1.0
1.5
2.0
2.5
3.0
10
15
20
0.0
0.5
1.0
1.5
2.0
2.5
3.0
1.5
kW
kW
2
2.0
1.5
1.0
0.5
0.0
1.0
0.5
0.0
10
12
3
1.5
1.0
kW
kW
1.5
0.5
1.0
0.5
0.0
0.0
2
10
12
4
2.0
1.5
kW
kW
1.5
1.0
1.0
0.5
0.5
0.0
0.0
2
10
12
5
2
kW
kW
1.0
1
0
0.5
0.0
Month
10
12
Hour
Figure 6-30
Daily Profile Grouping With Energy Reduction From Voltage Reduction for Residential
Customers on Circuit A
6-36
1
kW
1.5
summer
shoulder
winter
1.0
0.5
0.0
2
10
12
2
2.0
summer
shoulder
winter
kW
1.5
1.0
0.5
0.0
2
10
12
3
kW
1.5
summer
shoulder
winter
1.0
0.5
0.0
2
10
12
4
2.0
summer
shoulder
winter
kW
1.5
1.0
0.5
0.0
2
10
12
5
2.0
summer
shoulder
winter
kW
1.5
1.0
0.5
0.0
2
Month
10
12
Figure 6-31
Daily Profile Grouping With Energy Reduction From Voltage Reduction by Season for
Residential Customers on Circuit A
Another AMI data set on Circuit G was also evaluated. On this circuit, AMI meters measured
average voltage, watts, and vars every 15 minutes. Only residential customers had the AMI
metering; commercial meters on this circuit were not yet connected to the AMI system.
Figure 6-32 shows a breakdown of this subset of meters by the two main residential rate
categories: standard and all electric. The all-electric rate showed less reduction in energy and
lower CVR factors. These CVR factors were based on the voltage measured at the meter. The
two graph panels on the right side of Figure 6-32 show annual average watt and var profiles for
these customer classes. The voltage reduction averaged 3.95% at the substation on Circuit G.
The voltage reduction measured 4.0% at customer meters. This AMI dataset had a number of
data points missing. Because of missing data, only meters with at least 200 days of measurement
were included in the analysis.
6-37
kW or kvar
Residential standard
2
1
0
2
Residential standard
10
12
0.2
0.4
0.6
0.8
1.0
CVR factor
kW or kvar
0.0
4
3
kW
kvar
1
0
2
10
12
Month
The left graph shows CVR factors for each grouping of customers by billing rate
category. CVR factors are relative to the voltage measured at the customer meter. The red
dot marks the estimate of CVR factor, and the gray line marks the 95% confidence
intervals around the estimate. The right column of graph panels shows the average real
and reactive power consumed for the given set of customers.
Figure 6-32
Customer Class and Voltage Reduction for AMI Subset G
The left graph shows CVR factors for each grouping of customers by billing rate category. CVR
factors were relative to the voltage measured at the customer meter. The red dot marks the
estimate of CVR factor, and the gray line marks the 95% confidence intervals around the
estimate. The right column of the graph panels shows the average real and reactive power
consumed for the given set of customers.
Figure 6-33 shows a breakdown by load size. There were some differences. Larger loads tended
to have more winter heating and lower CVR factors. Figure 6-34 shows breakdowns by power
factor, which turned out to be a good indicator of energy reduction and CVR factors.
6-38
kW or kvar
1.0
0.5
0.0
2
10
12
10
12
10
12
Medium load
kW or kvar
Smaller load
Medium load
Larger load
Larger load
0.0
0.2
0.4
0.6
0.8
1.0
CVR factor
kW or kvar
4
kW
2
kvar
0
2
Month
Figure 6-33
Power Factor and Voltage Reduction for AMI Subset G
6-39
2.0
1.5
1.0
0.5
0.0
2
10
12
[0.948 0.972)
kW or kvar
pf < 0.948
2
1
0
2
[0.948 0.972)
10
12
[0.972 0.988)
kW or kvar
[0.972 0.988)
pf 0.988
3
2
1
0
0.0
0.2
0.4
0.6
0.8
1.0
10
12
10
12
pf 0.988
kW or kvar
CVR factor
3
2
kW
kvar
1
0
2
Month
Figure 6-34
Power Factor Ranges and Voltage Reduction for AMI Subset G
Circuit H also had AMI data. Similar comparisons were made by customer groupings. Trends
were similar between Circuits G and H. Circuits G and H were both from the same utility, but
they were separated by more than 100 miles. Circuit G was more rural. Figure 6-35 shows that
all-electric customers again had lower CVR factors. Figure 6-36 shows that larger customers had
lower CVR factors. Figure 6-37 shows that power factor was again a good indicator of CVR
factor, but the relationship was not as strong as it was on Circuit G.
6-40
Residential standard
kW or kvar
3
2
1
0
2
Residential standard
10
12
0.0
0.2
0.4
0.6
0.8
1.0
CVR factor
kW or kvar
kvar
0
2
10
12
Month
Figure 6-35
Customer Class and Voltage Reduction for AMI Subset H
6-41
Smaller load
kW or kvar
1.0
0.5
0.0
2
10
12
10
12
10
12
Medium load
Smaller load
kW or kvar
Medium load
Larger load
Larger load
0.0
0.2
0.4
0.6
0.8
1.0
CVR factor
kW
kW or kvar
3
2
kvar
1
0
2
Month
Figure 6-36
Customer Size and Voltage Reduction for AMI Subset H
6-42
3
2
1
0
2
10
12
kW or kvar
[0.949 0.975)
pf < 0.949
2
1
0
2
[0.949 0.975)
10
12
[0.975 0.992)
kW or kvar
[0.975 0.992)
pf 0.992
2
1
0
0.0
0.2
0.4
0.6
0.8
1.0
10
12
10
12
pf 0.992
kW or kvar
CVR factor
2
kW
kvar
0
2
Month
Figure 6-37
Power Factor Ranges and Voltage Reduction for AMI Subset H
Figure 6-38 shows a breakdown of CVR factors from groupings of customers on Circuit H.
Clustering was based on normalized monthly watt and var profiles. The profiles from winter
peaking to summer peaking show the impact on resistive heating and air conditioning load on
CVR factors.
6-43
kW or kvar
1
3
2
1
0
2
10
12
10
12
10
12
10
12
10
12
kW or kvar
2
3.0
2.5
2.0
1.5
1.0
0.5
0.0
1
3
2
kW or kvar
2.5
2.0
1.5
1.0
0.5
0.0
4
4
0.0
0.5
CVR factor
1.0
1.5
kW or kvar
5
2.5
2.0
1.5
1.0
0.5
0.0
kW or kvar
5
3
kW
kvar
1
0
2
Month
Figure 6-38
Clusters by Watt and Var Profiles and Voltage Reduction for AMI Subset H
6-44
2000
1500
1000
500
0
0
2000
4000
6000
Figure 6-39
Daily Reactive Power Averages for AMI Subset G
6-45
1.0
Normal voltage
Reduced voltage
Density
0.002
0.001
0.5
0.0
0.000
200
400
600
800
1000
200
400
600
800
1000
Figure 6-40
Distributions of Daily Reactive Power Averages for AMI Subset G
Table 6-8 shows reactive-power CVR factors for two circuits with quite different levels of
voltage reduction. These CVR factors were based on the voltage change at the customer. The
CVR var factors for both were above 4.
Table 6-8
CVR Var Factors for Circuits G and H
Circuit
Voltage
Reduction
in Percent
Reduction in Reactive
Power in Percent
4.0
17.0
4.3
1.9
9.9
5.3
Figure 6-41 shows CVR var factors for customers on Circuit G as split out by residential rate
category. Differences by rate category were not dramatic. Figure 6-42 shows CVR var factors for
Circuit G by rolling month. The CVR var factor is shown for the three-month period centered on
the month given. While not varying greatly, the CVR var factor was highest during the winter.
6-46
Residential standard
kW or kvar
2
1
0
2
Residential standard
10
12
kW or kvar
3
kW
kvar
1
0
2
10
12
Month
Figure 6-41
CVR Var Factors by Customer Billing Class for AMI Subset G
6
0
2
10
12
Month
Figure 6-42
CVR Var Factors by Rolling Month for AMI Subset G
6-47
Develop a method to predict CVR factors based on load composition. This would allow
better confidence for energy conservation groups or regulators to award funding for voltageoptimization programs.
Quantify energy reductions on different types of commercial and industrial customers. The
Green Circuits study had mostly residential loads.
Predict future changes to CVR factors based on changes in load composition. As load
compositions change from incandescent to LED lighting and as plug-in electric vehicles gain
traction, these may change the effectiveness of voltage reduction.
6-48
7
SUMMARY AND FUTURE WORK
Project results have shown that the distribution system can be a significant resource for reducing
energy consumption. Voltage reduction produces the majority of savings. The median energy
reduction from simulations of 66 circuits was 2.3%, and economic models suggest that efficiency
projects can be economically viable on most circuits. Voltage reduction is most effective when
coupled with efficiency improvements like phase balancing or var optimization. The most
appropriate option for a given circuit depends on economic assumptions, economic ranking
criteria, circuit characteristics, and load composition and location.
Voltage optimization stands out as a key strategy to reduce peak and average energy
consumption. Combining system improvements with voltage reduction can improve utility-side
efficiency and end-use efficiency to lower overall consumption. Billing is a significant roadblock
to widespread implementation of voltage optimization. Reduced end-use consumption lowers
customer kilowatt-hour billing. To offset that lost billing, changes in billing rates or structure are
needed. The industry needs to do more work to investigate options to remove this roadblock.
One priority for future research is better estimation of load response to voltage for different load
mixes. As described in Chapter 6, feeder load response to voltage varied. A better understanding
of that variability, especially with load mix, will allow more precise estimates of energy
reductions from voltage optimization. More precise estimates will help utilities target voltageoptimization projects. More precise estimates of energy savings will help utilities claim financial
credits for implementing distribution efficiency projects. EPRI has a Load Modeling for
Voltage Optimization project currently in early stages that aims to develop the following:
Models of load response to voltage developed from laboratory testing that will cover
common residential and commercial loads. In addition, this testing will quantify performance
changes of equipment with voltage.
Feeder or system-level models of voltage response of load based on load mix, season, and
climate zone.
Field demonstration and evaluation of the accuracy of load models based on voltage
reduction field trials.
More field trials of voltage optimization implementations would help the industry. More data
will help identify issues with implementations. Implementation of different voltage control and
volt-var control systems will help determine how much improvement is possible with more
sophisticated systems.
7-1
Use of more efficient transformers is another area ripe for more investigation. As seen from the
modeling, no-load transformer losses make up a significant portion of distribution losses.
Amorphous-core transformers offer significantly reduced no-load losses. Options for research
that could advance the adoption of amorphous-core transformers include field pilots, targeted
laboratory testing, and improvements in manufacturing processes.
Long-range planning is another area for utility review and for future industry collaboration.
Efficiency needs include guidelines to determine conductor loading, substation sizing and
locations, and maximum feeder lengths and voltage drops. Future review of planning criteria for
transformers and secondaries is also warranted. In addition to optimizing voltage drop on
transformers and secondaries, additional factors come into play, including the adoption of plugin electric vehicles, photovoltaic systems, and other end-use changes. Tools are needed to
optimize transformer and secondary based on cost, voltage drop, and loading.
On modeling, some improvements worth investigating in more detail include:
Load reactive power At the customer level or at the feeder head, most of the circuits did
not have good measurement data on power factor or load reactive power.
Transformer data Only one utility in the study had records on transformer losses (and that
was only for units installed after 1986). Impedance information was also suspect for many
utilities. Without good loss and impedance information, it is difficult to quantify transformer
losses and voltage drops.
Secondary models Few utilities had secondary models. As discussed in Chapter 5, accurate
secondary models can provide a more complete picture of secondary voltage drops and
secondary line losses.
Volt-var control systems Some of the more sophisticated volt-var control systems that are
on the horizon are difficult to accurately model. The main impediment is that most are
proprietary systems.
The OpenDSS distribution modeling platform is an advanced distribution analysis tool that
allows unique modeling. For the Green Circuits project, key features included custom voltagedependent load models, custom load shapes, and annual 8760-hour simulations that allow
analysis of different voltage and var controls. Bringing these capabilities to more industrystandard tools like CYMDIST or WindMil would help utilities better evaluate efficiency
improvement options. Another option is to investigate simplified approaches to modeling based
on existing tools. Another challenge is economic analysis; there may be a need for tools to make
it easier for utilities to perform economic evaluations of efficiency options to optimize
efficiency.
Another challenge is how to best integrate AMI metering data into distribution modeling for
better planning and efficiency analyses. Based on AMI data provided by utilities, issues include:
Handling missing data (either missing meters, missing classes of meters, or missing time
slices)
7-2
Overall, we expect continued adoption of efficiency efforts, and as more utilities test out
different approaches to improve efficiency, the industry will gain knowledge for continued
improvement.
7-3
A
EXAMPLE CIRCUIT ANALYSIS
This appendix provides an abbreviated example of a Green Circuits circuit analysis. The
objective of this example is to provide the reader with an overview of the overall system analysis
approach. This analysis, like all other Green Circuits analyses performed, contains the base-case
analysis along with the standard options for loss savings (voltage optimization using the
standardized approach, re-conductoring, ideal var, realistic var control such as adding var control
to existing/new capacitors, and phase balancing). Some circuit studies included loss savings
approaches in addition to these. These additional studies are usually done at the request of the
utility or recommended by EPRI. Some examples include: higher efficiency transformers,
voltage upgrades, and circuit reconfigurations. This example circuit study is no exception. This
study included a special case where specialized voltage regulation had been implemented for
voltage optimization. The results of this case were analyzed along with the other loss savings
approaches and compared to the original base circuit losses.
This example is broken into four sections. The first section provides some circuit background.
The second section provides the overall model development and base case analysis. The third
section provides the assessments of the various options for loss savings, and the fourth section
provides overall conclusions of the study.
Circuit Background
The circuit analyzed in this example was primarily a rural residential circuit. Figure A-1 shows
the one-line layout of the circuit. The farthest electrical distance from the substation is
approximately 8 miles. This circuit was primarily a 34.5-kV circuit; however, the circuit
contained many step-down transformers, which converted the 34.5 kV to 13.2 kV or 19.92 kV to
7.6 kV. These step-down transformers are shown in the circuit diagram in Figure A-1. The
customer transformer locations are shown in Figure A-2. As can be seen in this figure, the
customer loads were distributed fairly evenly.
The loading for this circuit, which will be discussed in more detail below, was approximately
28 MW at peak with a 48% load factor. Including the additional feeders on the substation bus,
the substation transformer was loaded to 50 MW at peak.
The voltage control on this circuit was an LTC at the substation. The circuit contained 3300 kvar
of fixed capacitance used primarily for power-factor correction. Table A-1summarizes some of
the characteristics of the circuit selected for this example.
A-1
Circuit
34.5 kV / 13.2 kV
Residential
87%
48%
Substation Control
LTC
Circuit Miles
(Total Electrical Length of All Primary
Conductors/Cables)
74.3 miles
13.1 miles
8 miles
Substation
Figure A-1
Circuit One-Line (Power Plot Thicker Lines Indicate Higher Power Flow)
A-2
Substation
Figure A-2
Plot of the Circuits Step-Down Transformers
A-3
Substation
Figure A-3
Plot of the Customers Service Transformers
Import the utilitys analysis software database into the EPRIs analysis software (OpenDSS).
A-4
Imported
Circuit Data
Add Service/
Step-down
Transformers
Measured
Data
Add
Transformer
Loss Data
Load Profiles
Evaluate Base
Case
Add LTC
Controls
Evaluate
Reduction
Options
Figure A-4
Model Development Steps for Example Circuit
Several data sources were needed to develop the base case. Many of the circuit components were
available in the utilitys circuit analysis database, but there were certain key elements that were
not. This additional information was obtained from discussions with the utility, construction
sheets, and utility-supplied spreadsheets.
The original database contained limited information about the substation transformer (kVA
rating and impedance). This information was augmented with the utility-provided information
about LTC and LDC (load drop compensation) settings. For the base case: Vset = 123 V,
bandwidth = 3 V, and the LDC was set to R = 7 V and X = 0.
Feeder primary cables (UG) and conductors (OH) were fully specified by their lengths and
impedances/line-geometries in the database as well as capacitor sizes and location. There were
many step-down transformers in the circuit. Their impedances were specified in a look-up table
within the database. This database also included their load loss and no-load loss values. This
information was added to EPRIs circuit model.
Although the original database did not include service transformers, it did have a representation
of kVA ratings associated with groups of customers; but in this form, it was not suitable for
identifying individual service transformers, nor individual customers. The individual service
transformers were supplied via a spreadsheet provided by the utility along with the loss
information. The spreadsheet contained a column that contained the section identifier that linked
the spreadsheet to the circuits database.
Service conductors were not in the supplied circuits database. The service conductors were
added to the OpenDSS circuit files with information provided by the utility:
Residential customers The conductor in service was 4/0 AL with a 2/0 AL neutral. A
distance of 200 feet was used for all residential customers. The distance of 200 feet was an
estimated value provided by the utility.
Commercial customers A distance of 150 feet was used for all commercial customers. The
distance of 150 feet was an estimated value provided by the utility.
Commercial customers, for 200-to-400 amp services 120/208 volts: the conductor in
service was three 4/0 Al phases with a 4/0 Al neutral.
A-5
Commercial customers, for 400-to-800 amp services 120/208 volts: the conductor in
service was three 500 Al phases with a 4/0 Al neutral.
Commercial customers, for larger services The conductor in service was three 750 Al
phases with a 500 Al neutral, or multiple sets of conductors.
The database did contain groups of customers but did not have individual customers modeled.
The utility provided a spreadsheet that linked each customer to its service transformer; therefore,
each customer was modeled individually. A voltage-sensitive load model was used for all loads,
where the watts and vars both vary with voltage based on a linear relationship. A CVR factor of
0.8 was used for watts, and a CVR factor of 3.0 was used for vars. These CVR values were
consistent across all circuits studied in the Green Circuits project.
A spreadsheet was provided by the utility that contained the AMI peak power values associated
with the 2008 peak for each load. This information provided a load allocation that scaled each
load separately.
The 2010 SCADA data was used for the substation load profile, based on the desire to have more
recent load data. A load shape was developed that showed the time-varying nature of the loads
(at the feeder level). Note: The substation bus also supplied other feeders. A representation of
the other feeder load is included in the model to provide the correct total load profile on the
substation LTC.
Figure A-5 shows a plot of the measured line currents verses the modeled line currents. The line
currents were feeder-specific, so they were used to provide the load shape for the feeder. As
previously noted, the other feeders connected to the substation bus were represented in the
model. Figure A-6 shows a comparison of the modeled bus power to the measured bus power.
This power included the circuit under investigation and all other connected feeders.
Current Comparison
500
450
400
C urrent (A )
350
300
A AMPS - Sim
A AMPS - SCADA
250
200
150
100
50
0
0
1000
2000
3000
4000
Hours
Figure A-5
Modeled Verses Measured Line Currents
A-6
5000
6000
7000
8000
50
Power (MW)
40
MWATT - Sim
30
MWATT - SCADA
20
10
0
0
1000
2000
3000
4000
5000
6000
7000
8000
Hours
Figure A-6
Modeled Verses Measured Bus Power Levels
Base-Case Results
After the implementation of the customer and substation data, the base-case losses were
analyzed. The base case was the circuit as originally configured (without any loss-saving options
implemented). Table A-2 summarizes the results of the yearly and peak-day losses for the circuit.
The peak losses were 2.78%, and the annual losses were 2.19%.
A summary of the base-case modeled peak and annual losses are shown in Figure A-7 and
Figure A-8, respectively. As can be seen in the figures, peak losses were dominated by the
primary lines, and annual losses were dominated by the transformers no-load losses.
Table A-2
Model Base Losses at the Peak-Hour and Annual Energy Losses
Peak Demand
kW
Consumption/Demand
Total Losses
Line Losses
Xfmr Losses
Load Losses
No-Load Losses
Primary Losses
Secondary Losses
%Peak
27,901
776
455
320
648
128
656
120
2.78%
1.63%
1.15%
2.32%
0.46%
2.35%
0.43%
Annual Energy
kWh
% Consumption
118,123,596
2.19%
2,587,358
1,019,896
0.86%
1.33%
1,567,462
1,446,723
1.22%
0.97%
1,140,635
2,324,546
1.97%
262,812
0.22%
Avg
13484.43
295.36
116.43
178.93
165.15
130.21
265.36
30.00
A-7
Transformer Load
25%
Primary Line
44%
Transformer NoLoad
16%
Service Lines
15%
Figure A-7
Base-Case Peak Loss Breakdown
Transformer Load
16%
Primary Line
29%
Service Lines
10%
Transformer NoLoad
45%
Figure A-8
Base Case Annual Loss Breakdown
A-8
Voltage Optimization
To model voltage optimization, the substation LTC base was reduced to 119 V from 123 V, and
remote regulation (remote voltage feedback) was used. The base case included LDC, and this
was disabled for the voltage-optimization case. Figure A-9 illustrates the end of the line bus that
the LTC was controlling. Note: The point where the end-of-line bus was being monitored was
the lowest voltage three-phase bus.
The use of remote voltage feedback helps to limit the voltage levels during light load conditions.
The reduction of the LTC setting and the use of remote voltage feedback maintained an average
primary voltage of 118.9 V. Before the voltage reduction, the average primary voltage on the
feeder was 123.9 V. The minimum primary voltage at peak was reduced to 117.3 V1 from 121.6
V.
Table A-3 shows the results of the voltage-optimization simulation, and
Table A-4 compares the results to the base case. For the annual simulation, the consumption was
reduced by 3,782 MWh, and the loss was reduced by 68.8 MWh. Loss savings came primarily
from no-load loss. In fact, the load loss actually increased due to the slight increase in line
currents.
Table A-3
Voltage Optimization Modeled Losses at the Peak-Hour and Annual Energy
Base Case Consumption
27,901
Peak Demand
kW
Consumption/Demand
Total Losses
Line Losses
Xfmr Losses
Load Losses
No-Load Losses
Primary Losses
Secondary Losses
%Peak
27,097
773
460
313
655
119
652
121
2.85%
1.70%
1.16%
2.42%
0.44%
2.41%
0.45%
118,123,596
Annual Energy
kWh
% Consumption Avg
13052.70
114,341,663
2,518,513
2.20%
287.50
1,038,109
0.91%
118.51
169.00
1,480,404
1.29%
167.61
1,468,302
1.28%
1,050,211
0.92%
119.89
2,253,322
1.97%
257.23
265,191
0.23%
30.27
Table A-4
Voltage Optimization Modeled Losses at the Peak-Hour and Annual Energy Losses With
Respect to the Base Case
Change in consumption
Change in demand
change in total losses
change in peak losses
Change in load loss
Change in no-load loss
1
-3,782
-804
-68,845
-2
21,579
-90,424
MWh
kW
kWh
kW
kWh
kWh
The minimum voltage of 117 V was selected so a comparison could be done against the LTC Set-Point
Variation case.
A-9
ABC
LTC
Figure A-9
Voltage Optimization Remote Regulation
A-10
The reduction of the substation bus voltage reduced the average primary voltage to 120.1 V from
123.9 V. The minimum primary voltage at peak was reduced to 117.2 V from 121.6 V.
Table A-5 shows the results of the in-the-field voltage-optimization simulation, and Table A-6
compares the results to the base case. For the annual simulation, the consumption was reduced
by 2,769 MWh, and the loss was reduced by 56.7 MWh. As in the voltage-optimization case
before, loss savings came primarily from reduced no-load losses.
Substation Bus
126
125
124
123
Voltage
122
121
120
119
118
117
116
0
1000
2000
3000
4000
5000
6000
7000
8000
Hours
Measured VOLT
Simulated VOLT
A-11
126
Voltage
124
122
120
118
116
0
1000
2000
3000
4000
5000
6000
7000
8000
Hours
Measured VOLT
Simulated VOLT
A-12
%Peak
27,074
772
460
312
654
118
651
121
2.85%
1.70%
1.15%
2.41%
0.44%
2.41%
0.45%
118,123,596
Annual Energy
kWh
% Consumption Avg
13168.35
115,354,721
288.88
2,530,632
2.19%
117.47
1,029,001
0.89%
171.42
1,501,631
1.30%
166.46
1,458,163
1.26%
122.43
1,072,469
0.93%
258.69
2,266,157
1.96%
30.19
264,475
0.23%
Change in consumption
Change in demand
change in total losses
change in peak losses
Change in load loss
Change in no-load loss
-2,769
-827
-56,726
-4
11,440
-68,166
MWh
kW
kWh
kW
kWh
kWh
The in-the-field voltage regulation had slightly less annual loss and consumption reduction
compared to the voltage-optimization case. Table A-7 shows the results of the voltageoptimization case with respect to the in-the-field LTC set-point variation.
The voltage-optimization case used voltage feedback from monitoring points at the end of the
regulated line section, and the LTC bandwidth was set to +/- 1 V. Figure A-12 displays the bus
voltage results of the voltage-optimization case simulation along with the measured bus voltage.
As can be seen in the figure, the voltage-optimization case had a slightly lower voltage profile
and a tighter bandwidth. Table A-8 summarizes the primary voltage results for both cases along
with the base case. As can be seen in the table, the minimum primary voltage at peak was
approximately the same for both cases; however, the average annual voltage was lower for the
voltage-optimization case.
A-13
Voltage
122
121
120
119
118
117
116
0
1000
2000
3000
4000
5000
6000
7000
8000
Hours
Measured VOLT
Simulated CVR
Figure A-12
Substation Bus Voltage Comparison With Measured and Voltage-Optimization Model
Results
Table A-7
Voltage-Optimization Losses at the Peak-Hour and Annual Energy Losses With Respect to
the In-the-Field LTC Set-Point Variation Case
Change in consumption
Change in demand
change in total losses
change in peak losses
Change in load loss
Change in no-load loss
A-14
Base
Voltage-Optimization
Case
Peak Average
Primary Volts
122.8
118.4
118.5
121.6
117.2
117.3
124.3
119.8
119.9
Average
Primary Volts
123.9
120.1
118.9
Phase Balancing
Based on the load allocations used, the circuit had an annual current unbalance of 2.4% at the
substation; therefore, this circuit was relatively balanced across the annual simulation.
Improvements were made to improve the average unbalanced to 1.95%. The peak current
unbalance was improved to 0.65% from 2%. Figure A-13 shows the balanced case phase currents
at peak.
Table A-9 shows the results of the phase balancing simulation. Table A-10 compares the results
to the base case. The annual losses were reduced by 6.7 MWh, with the majority of savings
coming from the primary line losses.
Note: The unbalanced calculation was based on the ANSI/NEMA Standard MG1-1993 definition.
Table A-9
Phase Balance Case Modeled Losses at the Peak-Hour and Annual Energy Losses
Base Case Consumption
27,901
Peak Demand
kW
Consumption/Demand
Total Losses
Line Losses
Xfmr Losses
Load Losses
No-Load Losses
Primary Losses
Secondary Losses
%Peak
27,875
773
453
319
645
128
653
119
2.77%
1.63%
1.15%
2.31%
0.46%
2.34%
0.43%
118,123,596
Annual Energy
kWh
% Consumption Avg
13469.93
117,996,589
2,580,686
2.19%
294.60
1,015,662
0.86%
115.94
178.66
1.33%
1,565,025
164.41
1.22%
1,440,195
1,140,491
0.97%
130.19
2,318,842
1.97%
264.71
261,844
0.22%
29.89
A-15
Change in consumption
Change in demand
change in total losses
change in peak losses
Change in load loss
Change in no-load loss
Phase A 461A
Phase B 464A
Phase C 459A
Figure A-13
Phase-Balance Case Peak Currents
A-16
-127
-26
-6,672
-3
-6,528
-144
MWh
kW
kWh
kW
kWh
kWh
Re-Conductoring
A loss reduction approach could be to re-conductor the circuit. The conductors that were targeted
for re-conductoring were those line sections with the highest losses. The sections with the
highest losses were identified by using a feature in the analysis tool. Figure A-14 shows a circuit
plot of the line sections with the highest losses (areas highlighted in red).
The re-conductoring was done in two phases: Case 1 and Case 2. Case 1 had higher losses than
Case 2.The Case 1 re-conductoring simulation replaced the overhead three-phase 336 AAC with
477 AAC. This resulted in approximately 1 mile of re-conductoring. The re-conductored section
is highlighted in red in Figure A-15 (labeled Case 1). Case 1 reduced the annual losses by
100.3 MWh. Table A-11 shows the results of the re-conductor simulation, and Table A-12
compares the results to the base case.
Case 2 re-conductoring simulation replaced the overhead three-phase 336 AAC and 477 AAC
with 795 AAC. This resulted in approximately 5 miles of re-conductoring. The re-conductored
section is highlighted in red in Figure A-15 (labeled Case 2). (Note: Only the overhead section
was re-conductored. The highlighted section in Figure A-14 Case 2 that was not reconductored was UG.) Case 2 reduced the annual losses by 257 MWh. Table A-13 shows the
results of the re-conductor simulation, and Table A-14 compares the results to the base case.
Table A-11
Case 1 Re-Conductor Model Losses at the Peak-Hour and Annual Energy Losses
Base Case Consumption
27,901
Peak Demand
kW
Consumption/Demand
Total Losses
Line Losses
Xfmr Losses
Load Losses
No-Load Losses
Primary Losses
Secondary Losses
%Peak
27,937
731
410
321
603
128
528
120
2.62%
1.47%
1.15%
2.16%
0.46%
1.89%
0.43%
118,123,596
Annual Energy
kWh
% Consumption
118,200,592
2.10%
2,487,058
0.78%
917,995
1.33%
1,569,063
1,344,752
1.14%
1,142,307
0.97%
2,224,297
1.88%
262,762
0.22%
Table A-12
Case 1 Re-Conductor Losses at the Peak-Hour and Annual Energy Losses With Respect
to the Base Case
Change in consumption
Change in demand
change in total losses
change in peak losses
Change in load loss
Change in no-load loss
77
36
-100,300
-45
-101,971
1,672
MWh
kW
kWh
kW
kWh
kWh
A-17
27,901
Peak Demand
kW
Consumption/Demand
Total Losses
Line Losses
Xfmr Losses
Load Losses
No-Load Losses
Primary Losses
Secondary Losses
%Peak
27,995
661
340
321
533
129
459
120
2.36%
1.22%
1.15%
1.90%
0.46%
1.64%
0.43%
118,123,596
Annual Energy
kWh
% Consumption
118,315,795
1.97%
2,330,453
758,917
0.64%
1,571,536
1.33%
1,185,568
1.00%
1,144,885
0.97%
2,067,766
1.75%
262,687
0.22%
Table A-14
Case 2 Re-Conductor Losses at the Peak-Hour and Annual Energy Losses With Respect
to the Base Case
Change in consumption
Change in demand
change in total losses
change in peak losses
Change in load loss
Change in no-load loss
Case 1
Figure A-14
Identified Line Sections With the Highest Losses
A-18
192
94
-256,905
-114
-261,155
4,250
MWh
kW
kWh
kW
kWh
kWh
Case 2
Case 2
Case 1
Figure A-15
Conductors That Were Re-Conductored
%Peak
27,864
760
449
311
633
127
566
115
2.72%
1.61%
1.11%
2.26%
0.45%
2.02%
0.41%
Annual Energy
kWh
% Consumption
118,109,441
2,538,992
2.15%
993,414
0.84%
1,545,579
1.31%
1,398,157
1.18%
1,140,835
0.96%
2,288,574
1.93%
250,418
0.21%
A-19
Change in consumption
Change in demand
change in total losses
change in peak losses
Change in load loss
Change in no-load loss
-14
-36
-48,366
-15
-48,566
200
MWh
kW
kWh
kW
kWh
kWh
Capacitor Control
Capacitors with var control were studied for the circuit as another approach to reduce losses.
This is considered to be a more realistic var control simulation than the ideal var case. At the
time of modeling, the capacitors were all fixed in the circuit. The capacitor placement is shown
in Figure A-16.
The 1200-kvar capacitor banks had var control added, which turned on at 600 kvar and turned
off at 900 kvar. The 900-kvar capacitor banks had var control added, which turned on at 450
kvar and turned off at 675 kvar. These settings were set to optimize the var control for loss
savings. This capacitor control improves the average power factor from 0.98 to 0.998. The
annual losses were reduced by 19 kWh. The capacitor control case does not have as much
savings in the transformer load losses and service line loss compared to the ideal var case. This
would suggest that power factor correction capacitors closer to the loads would improve loss
savings.
Table A-17 shows the results of the capacitor control simulation, and Table A-18 compares the
results to the base case.
Table A-17
Capacitor Control Model Losses at the Peak-Hour and Annual Energy Losses
Peak Demand
kW
Consumption/Demand
Total Losses
Line Losses
Xfmr Losses
Load Losses
No-Load Losses
Primary Losses
Secondary Losses
A-20
%Peak
27,962
777
456
321
649
128
657
120
2.78%
1.63%
1.15%
2.32%
0.46%
2.35%
0.43%
Annual Energy
kWh
% Consumption
117,891,677
2,568,560
2.18%
1,007,079
0.85%
1,561,481
1.32%
1,434,092
1.22%
1,134,467
0.96%
2,305,616
1.96%
262,944
0.22%
Change in consumption
Change in demand
change in total losses
change in peak losses
Change in load loss
Change in no-load loss
-232
62
-18,798
2
-12,631
-6,168
MWh
kW
kWh
kW
kWh
kWh
Figure A-16
Capacitor Placement
A-21
Summary
Table A-19 and Figure A-17 summarize the results of all the cases and compare the results to the
base case.
Table A-19
Efficiency Analysis Comparison Summary
Base
In-theField LTC
Set-Point
Variation
Voltage
Optimization
Cap
Control
Phase
Balance
ReConductor
Case 1
ReConductor
Case 2
Ideal
Var
GWh
Consumption
118.1
115.4
114.3
117.9
118.0
118.2
118.3
118.1
GWh Losses
2.59
2.53
2.52
2.57
2.58
2.49
2.33
2.54
Delta Loss
(MWh)
56.7
68.8
18.8
6.7
100.3
256.9
48.4
Delta
Consumption
(MWh)
2768.9
3781.9
231.9
127.0
-77.0
-192.2
14.2
A-22
122.0
GWh Consumption
GWh Losses
120.0
2.59
2.57
2.58
2.49
2.33
2.54
117.9
118.0
118.2
118.3
118.1
GWh
118.0
2.53
116.0
2.52
114.0
118.1
115.4
114.3
112.0
as
e
C
ec
on
du
ct
or
ar
Id
ea
lV
1
as
e
R
ec
on
du
ct
or
C
Ba
la
nc
e
l
Co
nt
ro
ap
C
Ph
as
e
iz
at
io
n
pt
im
O
Vo
l ta
ge
Ac
tu
al
Vo
lt a
ge
Ba
se
eg
ul
at
io
n
110.0
Figure A-17
Efficiency Comparison Summary Graph
As can be seen in the figures, the analysis shows the greatest improvement in losses with reconducting followed by voltage optimization. Voltage optimization may be the most economic
loss savings, especially when compared to re-conductoring. Table A-20 provides a summary of
the voltage optimization annual simulation results compared to the base case.
This circuit provided a special case where advanced voltage regulation was already implemented
for voltage optimization. This voltage regulation was imitated in the circuit model to determine
the loss savings.
Table A-21 provides a summary of the in-the-field voltage regulation simulation results
compared to the base case. Figure A-18 displays the bus voltage results of the in-the-field
voltage regulation simulation along with the measured bus voltage.
As can be seen in the tables, the in-the-field voltage regulation had slightly lower annual loss
and consumption reduction compared to the voltage-optimization case. The voltage-optimization
case used voltage feedback from monitoring points at the end of the regulated line section, and
the LTC bandwidth was set to +/- 1 V. Figure A-19 displays the annual bus voltage results of the
voltage-optimization case simulation along with the measured bus voltage. As can be seen in the
A-23
figure, the voltage-optimization case had a slightly lower voltage profile and a tighter bandwidth.
This voltage regulation approach may be more ideal than what can be actually implemented in
the field. Therefore, it is uncertain how much the in-the-field voltage regulation could actually
be improved upon to come closer to the results of the voltage-optimization case. With that said,
the present field implementation of the voltage optimization did provide 56.7 MWh of energy
savings.
Table A-20
Voltage Optimization Annual Losses With Respect to Base Cases
Circuit
Annual
Loss
Reduction
(MWh)
Annual
Consumption
Reduction
(GWh)
Annual
Consumption
Reduction
(%)
Annual
Loss
Reduction
(%)
No-Load
Loss
Reduction
Load Loss
Reduction
05410
-68.8
-3.78
-3.2%
-2.66%
-7.93%
1.49%
Table A-21
Varying LTC Annual Losses With Respect to Base Cases
A-24
Circuit
Annual
Loss
Reduction
(MWh)
Annual
Consumption
Reduction
(GWh)
Annual
Consumption
Reduction
(%)
Annual
Loss
Reduction
(%)
No-Load
Loss
Reduction
Load Loss
Reduction
05410
-56.7
-2.77
-2.34%
-2.19%
-5.98%
0.79%
Voltage
122
121
120
119
118
117
116
0
1000
2000
3000
4000
5000
6000
7000
8000
Hours
Measured VOLT
Simulated VOLT
Figure A-18
In-the-Field Voltage Regulation Simulation Results Compared to the Measured Voltage
Substation Bus
126
125
124
123
Voltage
122
121
120
119
118
117
116
0
1000
2000
3000
4000
5000
6000
7000
8000
Hours
Measured VOLT
Simulated CVR
Figure A-19
Voltage Optimization Simulation Results Compared to the Measured Voltage
A-25
B
ALTERNATE METHODOLOGY FOR EFFICIENCY
ANALYSIS
The efficiency modeling results in Chapters 2 and 3 were based on annual 8760-hour simulations
and voltage-dependent load models to accurately predict impacts of efficiency improvements.
This appendix provides an alternative analysis methodology that uses only peak snapshot powerflow solutions. The economic approach described here is similar to the approach described in
Chapter 3.
Distribution systems should be designed to maximize system efficiency by minimizing feeder
losses, construction costs, and customer average voltages while complying with operational
constraints of service reliability, minimum voltages, equipment loading, and safety. The optimal
distribution system design is determined by the minimization of the overall system costs. The
objective cost function includes the cost of equipment installations, O&M, cost of energy and
demand charges, system no-load (core) losses, load (coil and line) losses, and impact of lower
average voltages on consumers.
The objective cost function depends on the load density, forecasted load growth, economic
factors, and system constraints. The demand for electricity must be met in an optimal way by
identifying size and location of distribution system requirements (secondary conductors,
distribution transformers, primary feeder lines, capacitors, regulators, and number of feeders per
substation), taking into consideration all voltage control settings and operation guidelines.
To solve the problem of optimal distribution system planning, the planning engineer will need to
rely on expanded system metering, computer-aided mapping and system modeling, and loadflow applications. For efficiency studies, the engineer will need to determine the load magnitude
and its geographic location, as well as the hourly and annual load shapes for feeders and
substations. Ideally, the distribution substation and feeder metering should include annual hourly
kW and kvar feeder loads, as well as feeder hourly phase amps. The final cost-effective solution
is often determined after several iterative trial solutions. Each trial solution is judged for amount
of energy and demand savings, total benefit-cost ratio, and annual levelized cost of savings.
The distribution system is particularly important to an electrical utility because of its close
proximity to the ultimate customer and because it has the highest investment cost per customer.
Because the distribution system is closest to the customer, its failures affect customer service
more directly than failures on the transmission and generation system. As a result of increasing
marginal cost of energy, equipment, and laborand to ensure adequate system performance and
cost-effective operationsdistribution system planning must be capable of accurately
identifying system cost impacts for improved average system voltages, customer voltage
impacts, and system reliability.
B-1
The issues raised at each level are the optimal size and location of distribution system elements
and the control methods required. The distribution system assessment must include transformer
impedances, distribution insulation levels, marginal cost of energy, and demand charges.
Distribution planning determines cost-effective additions and/or modifications to the distribution
network, including:
The distribution efficiency planning goal is to minimize the cost of the distribution system
feeders and laterals, including the system energy losses and customer consumption. This section
provides an introduction to system factors and input data; modeling, mapping, and system
elements; metering data; system study process steps; and example distribution system efficiency
study.
Table B-1
Distribution System Efficiency Factors
Variable
Description
VFdr (V)
BCR
CFN (pu)
INeutral (A)
IUnbal (%)
kvarMax-Lag (kvar)
kvarMax-Lead (kvar)
kVATrans
kWFdr
10
kWhFdr
11
kWVCZ
12
LCLC
13
LDFCust (pu)
14
LDFFdr (pu)
15
LSFFdr (pu)
16
TransfkVA-Util (%)
17
VDrop-Sec (%)
18
VDrop-VCZ (%)
19
VOf (pu)
20
VPri-EOL (V)
21
VPri-Reg (V)
22
VRise (V)
23
VSet (V)
24
kWLoss-Existing
25
kWLoss-Improved
In addition to the distribution system efficiency factors, a distribution system efficiency analysis
considers economic factors and targets, distribution design and operational limitations, and
system improvement unit costs. A summary of each of the system efficiency factors and
considerations is given below.
B-3
1 Voltage variance between feeders VFdr (V) This factor provides a guide to identify feeder
average voltage issues that can prevent the simplified application of voltage optimization.
Factors applied to the entire substation voltage control zone. A system that lends itself to
simplified average voltage calculations generally has voltage differences between feeder
maximum voltage drops less than 2.5 V amperes (on a 120-V base).
2 Benefit-cost ratio BCR The BCR factor is given as the ratio of the present value of project
benefits to present value of project costs. The BCR is applied as minimum economic
performance target. The distribution system efficiency energy saved is maximized so that the
BCR is greater than 1.0. Some utilities require BCR factors greater than 1.0.
3 Coincident factors CFN (pu) The coincident factor (CF) is determined at each level in the
distribution system and is a function of the total system coincidence and the number of similar
loads. The average maximum diversified demand (AMDD) for any level is determined by
multiplying the average maximum demand (AMD) for customers or components by CF. The
coincident factor formulation used to determine the impact of customer or component loads on
the next higher level in the distribution system where N is the number of parallel components is
given below. This formulation assumes that the total system coincidence is 0.5.
1
CFN = CFTOTAL 1 +
N
Eq. B-1
4 Feeder neutral conductor amperes INeutral (A) This factor provides a guide to identify
inefficient feeder unbalance issues that lead to poor efficiency, higher system losses, and higher
feeder voltage drops. An efficient system will generally have feeder maximum neutral amperes
of less than 50 A measured over a one-hour period during peak conditions and is somewhat
higher for 12.5-kV systems and lower for 24.9-kV systems.
5 Phase unbalance IUnbal (%) This factor provides a guide to identify inefficient feeders with
high unbalance. An efficient system will generally have maximum phase unbalance of less than
20% during peak conditions, averaged over a one-hour period during peak conditions.
6 Compensated lagging var limit kvarMax-Lag (kvar) This factor provides a guide to identify
inefficient feeders with poor var compensation. An efficient system with adequate var
compensation will generally have annual vars of less than 500 kvar for any one-hour period.
For example, a system with +600 kvar during peak var loading conditions is judged inefficient.
7 Compensated leading var limit kvarMax-Lead (kvar) This factor provides a guide to identify
inefficient feeders with poor var compensation. An efficient system with adequate var
compensation will generally have annual vars of less than 500 kvar for any one-hour period.
For example, a system with -1300 kvar during minimum var loading conditions is judged
inefficient.
8 Feeder connected distribution transformer kVA kVATrans Distribution transformer connected
kVA per feeder is used with average voltage calculations to determine net change in no-load
losses.
B-4
9 Feeder annual peak kW load kWFdr Feeder peak kW load is used for load-flow calculations
and for average voltage calculations.
10 Feeder annual kWh load kWhFdr This factor provides the total energy delivered for each
feeder based on annual metered data.
11 VCZ annual peak kW load kWVCZ VCZ peak kW load is used for average voltage
calculations.
12 lifecycle levelized cost LCLC The LCLC factor is given as the maximum annual Lifecycle
Levelized Cost (LCLC) of energy saved. The LCLC is applied as minimum economic
performance target. The distribution system efficiency energy saved is maximized so that the
LCLC is less than or equal to the marginal cost of purchased power. Some utilities require LCLC
factors less than marginal costs. See RUS Bulletin 1724D-104 for additional details and sample
data. The general formulation for LCLC is given by:
I t + M t + Ft
t =1
n
LCLC =
n
t =1
(1 + r ) t
Et
Eq. B-2
(1 + r ) t
Where,
LCLC = Average lifetime levelized cost of energy saved
It
Mt
Ft
Et
= Interest rate
inf
= Inflation rate
= d inf
B-5
13 Customer annual load factor LDFCust (pu) This factor is determined from customer
information collected for customer residential, commercial, and industrial classes served by the
distribution system and identifies the annual demand as a function of usage for each class.
Knowing typical customer consumption and annual load factor helps in understanding the type
of heating and air conditioning load to be served across the distribution system. High kWh in
winter indicates electric heat usage, and high kWh in summer indicates AC. A low load factor (<
50%) generally indicates high electric heat usage. Data collected at the customer via AMI and
AMR applications can be used to provide sample data to determine diversified kW and kvar
hourly load shapes.
14 Feeder annual load factors LDFFdr (pu) The annual feeder load factor is determined from
substation power transformer and feeder on-site installed metering. The LDF is used to estimate
average primary voltage change and annual feeder line load loss factors. Load factors can be
prepared for each level of the distribution system and are the ratio of energy delivered and the
average maximum diversified demand. LDF is calculated at the feeder level with annual kWh
and peak kW:
LDF Fdr =
Annual _ kwh
8760 kW PEAK
Eq. B-3
15 Feeder annual loss factors LSFFdr (pu) The feeder annual loss factor is determined from
substation power transformer and feeder on-site installed metering. From the feeder annual LDF,
an estimate of the annual load loss factor can be made. The LSF is 1.0 for no-load loss. LSF is
used to determine line energy losses as a function of load factor by using a common loss
formulation used by RUS.
LSF Fdr = 0.85 (LDF ) + 0.15 LDF
2
Eq. B-4
Eq. B-5
16 Distribution transformer loading utilization TransfkVA-Util (%) This factor provides a guide to
identify inefficient distribution transformer utilization. The higher the utilization factor, the
higher the voltage drop, which limits the ability to lower the primary voltage and average
voltages to the customer. Typical values range from 70% to 130%.
17 Secondary design voltage drop VDrop-Sec (%) The utilitys typical distribution secondary design
(distribution transformer, secondary conductors, and service wires) has a voltage drop across the
secondary system and is highest during customer peak loading periods. If distribution system
modeling includes secondary modeling, the voltage drop is the absolute difference between the
voltage on the high side of the distribution transformer minus the voltage at the lowest-voltage
customer. Load factors and coincident factors are applied to determine the maximum diversified
voltage drop. In addition, the maximum average secondary voltage drop over one-hour period is
determined by applying a diversity factor of 0.80. It is unlikely that the maximum voltage drop
will be sustained for one hour occurring at the same time as the system peak low primary
voltage. The minimum primary voltage VDrop-EOL (V) is then the minimum ANSI C84.1 Range A
voltage of 114 V plus the maximum average secondary voltage drop (volts). A typical value for
B-6
secondary average coincident voltage drop is 4 V (3.3%), yielding a minimum primary voltage
of 119 V.
18 Voltage control zone maximum voltage drop VDrop-VCZ (%) This factor provides a guide to
identify voltage control zones with poor efficiency. A voltage control zone is the selection of
feeders and feeder sections where the primary voltage is controlled by a substation LTC or line
regulator. It is assumed for efficiency distribution systems that switched capacitors are switched
on vars and not for time, amps, or volts. An efficient system will generally have voltage control
zone annual maximum voltage drops less than 4 V (3.3%) for any one-hour period.
19 Residential voltage optimization factors VOf (pu) This factor represents a measure of the
per-unit annual energy saved for each per-unit average annual voltage reduction. This factor is
used to estimate the end-use energy savings from applying lower average voltages to the end-use
customer. This factor is derived from feeder or system customer load types and the NWPCC
th
1
Voltage Optimization Simplified M&V Protocol dated May 4 , 2010. The average percentage of
residential customers with electric heat or electric heat pumps and the average percentage of
residential customers with AC along with heating and cooling climate zone are used as inputs for
the protocol to select the appropriate VOf. The data was specifically derived for the northwest
states of Montana, Washington, Oregon, and Idaho. These voltage-optimization factors VOf are
applied to end-use load consumption, unlike the CVR factor, which is applied at the substation
feeder total consumption. VOf values typically can range from 0.30 to 0.70.
20 Primary lowest voltage at end-of-line VPri-EOL (V) This factor provides a guide to identify the
minimum allowed voltage at the lowest voltage point for feeders and Voltage Control Zones.
The minimum primary voltage is then the minimum ANSI C84.1 Range A voltage of 114 V plus
regulator bandwidth (BW) plus maximum average secondary voltage drop VDrop-Sec (%). A
typical value for minimum primary voltage is 119 V, assuming a maximum secondary
coincidental voltage drop of 4 V (3.3%) plus the minimum ANSI Range A voltage of 114 V.
When performing load-flow assessments, the lowest primary voltage at the end-of-line is used to
determine the average voltage of a distribution system.
21 Primary voltage at line voltage regulators VDrop-Reg (V) This variable is determined from loadflow assessments and is the primary line voltage on the high side of existing and proposed line
regulator locations. This value is used to determine the average voltage of a distribution system.
22 Primary voltage rise at voltage regulators VDrop-Rise (V) This variable is the voltage rise
determined for a voltage control zone at its voltage regulation source. For existing systems, it is
either simulated using load-flow models or measured from field data. For proposed systems, the
VDrop-Rise (V) is calculated and is generally equal to the maximum voltage control zone voltage
drop VDrop-vcz (%).
23 Feeder or VCZ source regulator voltage setting (no load) VSet (V) This variable is the
voltage setting determined for a voltage control zone at its voltage regulation source. It is the
output voltage at the regulator terminal at no load.
Simplified Voltage Optimization (VO) Measurement and Verification Protocol, May 4, 2010.
http://www.nwcouncil.org/energy/rtf/measures/protocols/ut/VoltageOptimization_Protocol_v1.pdf
B-7
24 Existing system line loss voltage rise at voltage regulators kWLoss-Existing This variable is the
existing peak line loss of the feeder or VCZ for the existing case. It is determined by load-flow
simulations for peak load conditions.
24 Improved system line loss voltage rise at voltage regulators kWLoss-Improved This variable is the
improved peak line loss of the feeder or VCZ for the improved case. It is determined by loadflow simulations for peak load conditions.
Economic factors These factors are used to calculate benefits and costs of a specific alternative
design. Factors may vary greatly among utilities. Financial factors include:
BCR
CLIFE
ELIFE
FCR
IRD
IRE
IRI
IROM
LCLC
MPDR
MPER
OM
PWRE
PWRI
Distribution system design standards An electric distribution system must comply with ANSI
C84.1 Voltage Standard for Range A. This standard provides a maximum and minimum onehour average voltage for the customer service entrance of voltage at 114 V to 126 V measured on
B-8
a 120-V base. The voltage may be infrequently higher or lower than these values for intervals
less than one hour, but the average voltage for each hour interval must comply with this standard.
In addition to industry standards, utilities often provide operating limits, such as maximum load
guidelines for conductors, voltage regulator set points, bandwidths, and/or minimum primary
voltages. For conductors, it is important to determine maximum normal rating versus maximum
emergency rating. Typical maximum (normal) capacity rating for conductors is 70% of
emergency ratings. In some cases, it may be advisable to alter the guidelines and/or suggest
modifications to aid overall system efficiency for the long term (such as lowering conductor
allowable loading percentage, reducing maximum allowable voltage drop percentage, and
reducing maximum allowable transformer and secondary voltage drop percentage).
Distribution System equipment installation unit costs An electric distribution system is
composed of primary lines, shunt capacitors (switched and fixed), distribution transformer
replacements, line voltage regulators, re-configuring costs, switches, metering systems, and
alternative voltage control methods. Line costs should include a variety of configurations and
upgrades (such as single-phase to three-phase). Costs should include materials, labor, and
overheads.
lines and equipment location from the mapping system. For distribution system studies, the
distribution system topology maps need to include the following information:
1. Lines (overhead and underground)
a. Location
b. Distances
c. Conductor sizes (types)
d. Phasing
2. Distribution transformers
a. Location
b. kVA rating
c. Phase connection
3. Shunt capacitors
a. Location
b. Kvar rating
c. Voltage rating
d. Phase connection
e. Switched or fixed control
4. Substation and line voltage regulators
a. Location
b. Phase connection
c. kVA rating
5. Switches
a. Location
b. Ampere capacity
c. Normal open/close status
System Load-Flow Model
A distribution feeder provides service to unbalanced three-phase, two-phase, and single-phase
loads by way of a variety of line element characteristics (capacity and phase configurations). To
electrically analyze distribution systems for precision, it is necessary to model all three phases of
the feeder accurately. For efficiency studies, it is recommended that all elements of the
B-10
distribution system be modeled, including remote lateral single-phase lines. With distribution
system studies, system load flow models are prepared to geographically represent distribution
equipment data, using data obtained from the utilitys mapping files (such as shp files or dxf
files).
If the GIS system uses AutoCAD or ArcView representation, the distribution system load-flow
model can be created by overlaying the electrical model on to the ArcView background shp
files or ArcView background dxf files.
The load-flow model graphical display should closely resemble the system GIS map. The loadflow models are calibrated by allocating load metered at the feeder source (at substation location)
to each distribution transformer by phase along each feeder. The allocation of loads to the
distribution transformer includes both kW and kvar. In general, distribution and secondary
models are not needed except to assess maximum secondary voltage drop violations. In some
situations, it is necessary to include more accurate load modeling (spot loads) for large
commercial and industrial facilities (greater than 2500 kW) on the low side of the distribution
transformer. This heavy commercial and industrial hourly load information is generally available
from the utilitys billing information.
For most electric utility distribution analysis applications, load-flow applications are capable of
performing simulations for one point in time (such as at peak kVA, minimum kvar, or maximum
kW). Some electrical load-flow analysis applications include the ability to model annual 8760hour simulations, which can provide greater detail for load factors, system losses, and average
voltages. These 8760-hour simulations require annual metering profile data for kW and kvar.
The analyses in the main body of this report used 8760-hour simulations. The main purpose of
this appendix section is to describe a similar approach using snapshot load-flow solutions.
For distribution system efficiency studies, the electrical load-flow simulations should be capable
of determining the following:
1. Feeder primary kWFDR and kvar loading at peak load.
2. Primary voltage drop for each voltage control zone at peak load VDrop-VCZ (%).
3. Voltage rise at source terminals for each voltage control zone at peak load VRise (V).
4. Primary equipment percent loading of maximum normal capacity at peak load.
5. Location of lowest primary voltage for each feeder and voltage control zone VPri-EOL (V).
6. Feeder phase amp unbalance at peak load IUnbal (%).
7. Feeder neutral current at peak load INeutral (A).
8. Feeder voltage variance (comparison of feeder lowest voltages) at peak load VFdr (V).
9. Impact of location and size of fixed and switched shunt capacitor banks.
10. Impact of location and size of line-volt regulators.
11. System energy primary line losses (kWh and kvarh).
B-11
B-12
B-13
(bandwidths, R and X settings, and voltage set points). All assessments are referenced on a 120V base.
The load-flow analysis of the Existing Case model will help derive base-case parameters that
describe system performance used for comparison with other trial Improved Case efficiency
scenarios. The load-flow modeling and simulations should be capable of providing the following
information needed for efficiency analyses:
1. Location of line-volt regulators
2. Location of shunt fixed and switched capacitor banks
3. Load allocation for feeder distribution transformers (by distributed or spot kVA)
4. Feeder and voltage control zone maximum kW loading kWVCZ
5. Voltage regulator LTC terminal voltage maximum rise VRise (V)
6. Primary voltage control zone maximum voltage drop VDrop-vcz (%)
7. Location of voltage control zone end-of-line voltage VPri-EOL (V) at peak load
8. High-side voltage on existing/improved regulator locations VPri-Reg (V) at peak load
9. Voltage variance between feeders at peak conditions VFdr
10. System line energy losses (kWh and kvarh)
11. System line demand loss (kW and kvar)
12. Equipment (lines and regulators) percent loading of maximum normal capacity (%)
13. Feeder maximum annual phase amp unbalance IUnbal (%)
14. Feeder maximum annual neutral current INeutral (A)
15. Feeder annual lowest power factor (leading or lagging)
16. Feeder maximum compensated lagging vars kvarMax-Lag (kvar)
17. Feeder maximum compensated leading vars kvarMax-Lead (kvar)
18. Average distribution transformer loading TransfkVA-Util (%)
Step 3: On the Existing Case model, apply existing voltage regulation source voltages VSet and
VRise (V) at each substation LTC/regulator and existing line regulator location for maximum and
minimum kVA conditions (with Existing VCZ voltage settings). Apply regulator LDC control
settings for Vsetting, R%, X%, CT rating, and PT ratio as required.
Step 4: On the Existing Case model, perform a load flow simulation and assess power flow,
voltages, phase balance, kvar flows, and system loss (with Existing VCZ voltage settings) for
peak loading conditions. Determine line voltage drops VDrop-vcz (%), load balance IUnbal (%), neutral
currents INeutral (A), location and values of minimum primary voltage VPri-EOL (V) and VPri-Reg (V),
voltage variance VFdr , and peak kW line loss.
B-15
Step 5: Identify efficiency threshold compliance for voltage drops VDrop-vcz (%), load balance IUnbal
(%), neutral currents INeutral (A), minimum primary voltage VPri-EOL (V), VFdr , and maximum
compensated lagging vars kvarMax-Lag (kvar). The results of the load flow are screened and
compared to efficiency threshold factors to identify possible areas for improvement. Document
the systems compliance and noncompliance with thresholds.
Step 6: Modify the feeder with system improvements in order of priority to improve distribution
efficiency for VDrop-vcz (%), load balance IUnbal (%), neutral currents INeutral (A), minimum primary
voltage VPri-EOL (V), VFdr , and maximum compensated lagging vars kvarMax-Lag (kvar). The priorities
of system improvements are given below.
Step 7: On the Improved Case model, perform a load-flow simulation and assess power flow,
voltages, phase balance, kvar flows, and system loss (with Existing VCZ voltage settings) for
peak loading conditions. Determine line voltage drops VDrop-vcz (%), load balance IUnbal (%), neutral
currents INeutral (A), location and values of minimum primary voltage VPri-EOL (V) and VPri-Reg (V),
voltage variance VFdr , and peak kW line loss.
Step 8: Same as Step 5 Identify efficiency threshold compliance for voltage drops VDrop-vcz (%),
load balance IUnbal (%), neutral currents INeutral (A), minimum primary voltage VPri-EOL (V), VFdr , and
maximum compensated lagging vars kvarMax-Lag (kvar).
Step 9: Repeat Steps 6, 7, and 8 until efficiency thresholds are met or exceeded.
Step 10: Identify system net change in kW peak line losses between the Existing Case and the
final Improved Case. Identify the investment cost of system improvements.
Step 11: Perform average voltage calculations and no-load loss assessment (with Existing VCZ
voltage settings) on Improved Case model.
Step 12: Perform average voltage analysis calculations and no-load loss assessments (with
Improved Case voltage settings) on Improved Case model.
Step 13: Perform economic analysis and prepare exhibits for costs and benefits of Improved
Case system. If benefit-cost ratio is less than the BCR target or the lifecycle levelized cost
exceeds the LCLC target, modify by repeating Steps 613. If BCR exceeds the target or LCLC
is less than its target, you may wish to repeat Steps 613 to search for additional cost-effective
energy savings. The distribution efficiency solution is an iterative process that requires
evaluation of several trials before achieving the optimal solution with the highest amount of
energy-saving potential while meeting the BCR and LCLC targets.
Distribution System Efficiency Thresholds
After the Existing Case has been analyzed, the next step is to identify an alternate system design
to be assessed. This proposed design is called the Improved Case option. Several feasible
Improved Case options or trials will be prepared by identifying additions and/or modifications to
B-16
the Existing Case model with the Existing voltage control settings applied. The threshold
assessment is performed in Steps 5 and 8.
The objective is for the Improved Case option to be a feasible solution that meets efficiency
thresholds with Existing voltage control settings. By improving the power factor (reducing var
flows), the system losses will be reduced. Likewise, by rephasing and/or re-conductoring, the
voltage drop and losses will be reduced. The existing voltage control settings for substation LTC
and existing line voltage regulators should be applied. Multiple feeders leaving the same
substation controlled by the same LTC are grouped and analyzed together.
The universal efficiency threshold guidelines are given below (listed by system factor reference
number):
1.
Maximum voltage variance between feeders VFdr (V) < 2.5 V on 120-V base
4.
5.
6.
Maximum compensated lagging var limit kvarMax-Lag (kvar) < 500 kvar
7.
Minimum compensated leading var limit kvarMax-Lead (kvar) > -500 kvar
18.
Maximum voltage control zone maximum voltage drop VDrop-VCZ (%) < 3.3%
20.
Minimum lowest voltage at end-of-line (V)VPri-EOL (V) = 114V + VDrop-Sec (%) 120V
--
--
Maximum lateral single- or two-phase design voltage drop criteria < 1.5%
In some cases, when the improved case does not comply with minimum operating voltage
thresholds with existing voltage settings, regulators may be installed to improve the voltage. If
the voltage control zone maximum voltage drop VDrop-VCZ (%) is exceeded in the Improved Case, a
proposed regulator can be installed set on a neutral tap. If the primary lowest voltage at end-ofline VPri-EOL (V) is violated with existing voltage settings in the Improved Case, the proposed
regulator is set with a raised tap just enough to meet the VPri-EOL minimum threshold. Place the
regulator on the fixed tap position to mimic, as much as possible, the Existing Case voltage
control environment while still meeting thresholds.
System Improvement Priorities
The objective of adding system improvements for the Improved Case is to reduce primary
voltage drop, reduce line losses, and to enable a lower voltage set point. Typical improvements
include the following measures (listed in order of typical priority):
1. Improve substation and feeder metering Identify metering improvements needed at
substations for power transformer and feeders, end-of-line primary voltages, and at load side
of line voltage regulators. Data collected is hourly three-phase kW/ kvar, single-phase hour
amps at substations at line regulators. Primary end-of-line (at lowest voltage location)
metering collects hourly voltage.
B-17
2. Reconfigure (by switching) Reconfigure feeder by switching line sections from one feeder
to another to offload feeder (by opening and closing tie locations). This will reduce total
system line losses and primary voltage drops.
3. Reconfigure (by tap changes) Relocate feeder sections from one phase to another phase to
balance phase amps (by relocating phase tap connections). This will reduce total system line
losses and primary voltage drops.
4. Add or modify capacitors Add or modify fixed and switched capacitor to achieve optimal
hourly var compensation throughout the year. If correctly applied, var corrections should
reduce line losses and primary voltage drops.
5. Add phase upgrades Add OH and UG phase upgrades (one-phase to two-phase, one-phase
to three-phase, two-phase to three-phase) to balance load and reduce voltage drop. This will
reduce total system line losses and primary voltage drops.
6. Re-conductor line sections Replace heavily loaded conductors above 80% of normal
maximum ratings (for example, > 70% of maximum emergency rating) with larger-capacity
conductors. This will reduce total system line losses and primary voltage drops.
7. Replace distribution transformer/secondary systems Identify distribution secondary
systems where secondary voltage drops VDrop-Sec (%) exceed design targets and where
secondary service voltages are less than 114 V at peak loading conditions. If secondary low
voltages occur with the Existing Case, the cost of the replacement and/or modifications
should not be included in the total cost of improvements. However, if low voltages are
expected with reduced voltage Improved Case, the cost to replace and/or modify secondary
systems should be included as part of the Improved Case. This will enable a lower system
voltage set point and will reduce the average system voltage.
8. Add line voltage regulators Add line regulators to reduce primary voltage drop.
9. Add new parallel feeders This will reduce conductor loading, system losses, and primary
voltage drops.
10. Upgrade feeder to higher primary voltage class This will reduce system line losses,
conductor loading, and primary voltage drop.
Distribution System Kvar Analysis
The Improved Case capacitor placement is based on the feeder annual historical var profiles. The
historical var profiles help determine the minimum and maximum feeder vars needed for
adequate hourly var compensations. The following is a suggested method to determine the
amount of fixed and switched capacitors needed and where they are to be located on the feeder.
1. Review annual kvar load profiles (uncompensated) and determine annual average, maximum,
and minimum var loads for the feeder. Avoid using very extreme isolated events observed for
meter errors, abnormal loading conditions due to switching, and widely varying peak kvar
conditions. Determine average maximum and average minimum kvar conditions for a period
of at least one week.
B-18
2. Avoid large cap banks greater than 600 kvar, and always place compensating kvar near load
centers. If correctly applied, var corrections should reduce line losses and primary voltage
drops.
3. Determine the maximum and minimum amount of capacitors to meet thresholds. The amount
of fixed kvar should result in a compensated kvar flow of less than the compensated lagging
var limit (kvar) target (such as -500 kvar) at average minimum uncompensated kvar flow
conditions. The amount of switched plus fixed kvar should result in a compensated kvar flow
of less than the compensated leading var limit target (such as +500 kvar) at average
maximum uncompensated kvar flow conditions. Switched capacitors should be var
controlled and placed in main feeder sections that have a range of var flows through the year.
4. Capacitor locations are determined at the average maximum uncompensated var conditions
with the appropriate kW load applied. The capacitors are distributed over the feeder, starting
at the end, using the B method by alternating locations of fixed and switched capacitors.
5. Capacitor switched controls:
a. For switched capacitor applications with efficient systems, apply var controls only with
time delay maximum at feeder source locations and minimum at end-of-feeder locations.
b. Other control methods can be applied if the net result is a maximum leading or lagging
kvar that is less than compensation targets at the feeder source for every hour of the year.
c. Switched capacitors with kvar controls: Turn ON with source lagging kvar load greater
than 50% of bank rating, and turn OFF with source leading kvar load greater than 75% of
bank rating.
Average Voltage Impact and Analysis
There are three loss elements addressed here: (1) end-use energy saved as a result of reduced
average voltage, (2) energy saved from distribution feeder line loss reduction, and (3) energy
saved in distribution transformer non-load loss as a result of reduced average voltage.
1. Once thresholds are met by the Improved Case, the total system average primary voltage is
calculated for the Existing voltage settings and compared with the proposed Improved
voltage settings. The net end-use energy saved is a function of feeder annual kWh load
kWhFdr, residential voltage optimization factors VOf, and per-unit change in average voltage
V. This calculation is performed for the Improved Case in Step 11 (with Existing voltage
settings) and in Step 12 (with proposed Improved voltage settings). The voltage optimization
factor VOf is determined from tables.
2. The net change in system line loss is the difference between peak kW line loss in the Existing
Case in Step 4 and the Improved Case in Step 10.
3. The change in distribution transformer no-load loss is the difference between the Improved
Case in Step 11 (with Existing voltage settings) and Step 12 (with proposed Improved
voltage settings). It can be assumed that the distribution transformer average no-load loss is
3.0 watt per kVA. What is of interest is the difference in loss. The no-load losses vary as the
square of the change in average voltage V.
B-19
The change in average annual voltage for electric distribution systems for the Existing Case and
the Improved Case is assessed for each voltage control zone (VCZ), based on voltage regulation
factors determined from peak load assessments. The following data is required to perform the
average voltage calculations (listed in order of system efficiency factor reference number):
1. Feeder connected distribution transformer kVATrans
2. Feeder annual peak kW load kWFdr
3. Feeder annual kWh load kWhFdr
4. VCZ annual peak kW load kWVCZ
5. Feeder annual load factors LDFFdr (pu)
6. Feeder annual loss factors LSFFdr (pu)
7. Voltage control zone maximum voltage drop VDrop-VCZ (%)
8. Residential voltage optimization factors VOf (pu)
9. Primary lowest voltage at the end of the line VPri-EOL (V)
10. Primary voltage at line voltage regulators (at peak and zero load) VPri-Reg (V)
11. Primary voltage rise at voltage regulators VRise (V)
12. Feeder source regulator voltage setting (at no load) VSet-Fdr (V)
Distribution systems designed for efficient operationswith voltage control zone maximum
voltage drops VDrop-VCZ (%) less than 3.3% and var flows compensated to near unity for every hour
of the yearallow the use of simplified average voltage calculation methods. With efficient
systems, each VCZ annual average voltage can be approximated as a function of the feeder
annual load factor LDFFdr (pu), VCZ maximum voltage drop VDrop-VCZ (%), primary voltage rise at
the VCZ voltage regulator VRise (V), and VCZ regulator/LTC voltage set point. The total system
average voltage is the average of all VCZ average voltages weighted by VCZ annual peak kW
load kWVCZ. The net per-unit change in average voltage V is used to estimate potential end-use
energy savings from voltage reduction.
The proposed Improved Case is evaluated for both Existing voltage settings and proposed
Improved voltage settings. The Existing voltage control may include new regulators placed on a
fixed tap position to mimic, as much as possible, the Existing voltage control environment while
still meeting thresholds.
The proposed Improved regulator/LTC voltage controls are revised by incorporating LDC and/or
automatic feedback applications with regulator/LTC voltage set points positioned as low as
possible to maintain acceptable minimum primary voltage limitations (such as 119 V). The
proposed Improved voltage settings for all VCZ regulator/LTC are such that the maximum VCZ
voltage rise is equal to the maximum VCZ voltage drop.
B-20
The per-unit voltage difference V between the Improved Case (with Existing voltage settings)
and the Improved Case (with proposed Improved voltage settings) is used with the voltage
optimization factor VOf , the feeder annual kWh load kWhFdr delivered, and weighted per-unit
change in system primary average voltage V to calculate the total end-use energy saved EVO
from lowering the average annual voltage. VOf is the voltage optimization factor derived from
2
the NWPCC Simplified VO M&V Protocol.
The general voltage control zone formulation for average primary voltage is given by the
following, where VSet-VCR is the VCZ regulator/LTC set point voltage in volts (120-V base), VDropis the VCZ maximum annual voltage drop in %, VRise is the VCZ regulator maximum annual
VCZ
voltage rise in Volts (120-V base), and LDFFdr is the feeder annual load factor. The average
voltages are weighted according to peak kW demand for each VCZ. The average volts V* on a
120-V base is:
V Drop VCZ 120V
Eq. B-6
VRise can be determined for the Existing VCZ voltage settings from actual metered voltage data
observed for the VCZ regulator/LTC for peak loading conditions. However, if near 100% var
compensation has been applied to the Improved Case, VRise can be determined for the Existing
settings by assuming zero var flow at peak load conditions. For LDC controls with R and X
settings equal to zero, VRise is zero. For cases where the existing R and X settings are not zero, R
and X represent maximum real voltage rise on a 120-V base for VCZ loading equal to the control
CT amp rating, and the var flow compensation assumed near 100% at peak load conditions, VRise
is calculated on a 120-V base using the following relation:
VRise
kWVCZ
kV 3
LL
= R
CT
AmpRating
Eq. B-7
Where,
kWVCZ = maximum kW loading for the voltage control zone regulator/LTC
kVLL = primary line to line kV
CTAmpRating = VCZ control primary current transformer amp rating (500 amps)
In cases where a new VCZ line voltage regulator is applied and placed on a neutral tap position
to mimic Existing voltage settings, the voltage rise VRise at the regulator is the difference between
the voltage at the regulator location for maximum and zero load conditions. The voltages at the
regulator location are determined from load-flow simulations for maximum and zero load
2
Simplified Voltage Optimization (VO) Measurement and Verification Protocol, May 4, 2010.
http://www.nwcouncil.org/energy/rtf/measures/protocols/ut/VoltageOptimization_Protocol_v1.pdf
B-21
conditions. In cases where the source voltage rise VRise upstream of the line voltage regulator is
zero (such as R and X = 0), VRise for the line voltage regulator is negative. The voltage rise at the
regulator location is represented as:
Eq. B-8
Where,
VPri-Reg (Max_Load) = Voltage on 120-V base at primary regulator location at maximum load conditions
VPri-Reg (zero_Load) = Voltage on 120-V base at primary regulator location at zero load conditions
The optimal VRise established for the Improved VCZ voltage LDC control settings assumes R > 0
and X = 0, with var flow compensation near 100% at peak load conditions.3 VRise is calculated on
a 120-V base using the following relation:
VRise = VDrop VCZ 120V
Eq. B-9
Where,
VDrop-VCZ = maximum VCZ primary voltage drop in percent
The change in end-use kWh energy consumption EVO is given by the following relation:
EVO = V VO f kWh Fdr
Eq. B-10
V* is the difference between average voltages V* for Existing Case control setting and V* for
Improved Case control setting applied to the Improved Case model.
The assumed annual load factor for end-use energy saved LDFVO is approximately 0.60 based on
field trials. The annual peak demand savings for reduced voltage is:
kWVO =
EVO
8760 h LDFVO
Eq. B-11
The change in line loss energy ELL is determined from the change in peak lines loss kWLineLoss
between the Existing Case and Improved Case models and the feeder annual loss factor LSFFdr
(pu). The line loss energy saved is determined by the following relation:
Eq. B-12
A reasonable estimate of the existing no-load loss can be calculated for the Improved Case by
assuming 3 watts per connected kVA for each feeder. The no-load loss savings is a function of
3
B-22
the feeder connected distribution transformer kVATrans and the per-unit change in average voltage
V. The no-load loss energy kWh saved is determined by the following relation:
E NL = NoLoadLoss BASE _ CASE
1 2
1
1 + V
Eq. B-13
Where,
NoLoadLoss BASE _ CASE =
kVATrans 8760 3W
1000
Eq. B-14
And,
kW NL =
E NL 1000
8760 h
Eq. B-15
The total annual energy kWh saved EANNUAL is the sum of the savings to the customer EVO, ELL
savings in load loss, and ENL savings in no-load loss as follows:
Eq. B-16
kWANNUAL is the sum of the VO demand reduction kWVO, savings in load loss demand kWLL,
and savings in no-load loss demand kWNL as follows:
Eq. B-17
Financial Factors
After system improvements are made and energy savings are calculated, an economic analysis is
performed using financial input factors and capital cost of system improvements. Analysis
should be performed for each feeder and for the overall system (substations and feeders).
Example financial and economic input factors used for efficiency studies are given in Table B-2.
The avoided marginal cost of purchase power is estimated for the base year, with energy cost
inflation rates applied per year thereafter. Economic analysis should include annual fixed charges
for investment and annual O&M expenses. Inflation rates and present worth rates are provided
for energy and investment costs. If the equipment investment life is longer than the expected
energy-saving life, the remaining life of capital investment can be assumed salvaged. The
economic analysis is performed using principles described in D.G. Newnan, et al.4 and IEEE 91
5
EHO 345-9-PWR.
4
5
D.G. Newnan, T.G. Eschenbach, and J.P. Lavelle, Engineering Economic Analysis, Ninth Edition, 2004.
IEEE 91 EHO 345-9-PWR, Engineering Economic Analysis: Overview and Current Applications, 1991
B-23
Costs should include the net change in annual O&M expense as a percentage of the capital
installed cost and the annual capital fixed charges. Fixed charge rate (FCR) represents the annual
fixed cost (not escalated) of investment over the life of the project. The annual cost is calculated
as the product of the FCR and capital installed cost. A typical fixed charge rate is given in Table
B-2 and includes capital recovery rate of returns, sinking funds, insurance, and taxes.
The distribution system efficiency BCR and LCLC analysis is performed using the energy and
cost differences between the Existing Case and the Improved Case. The results of the economic
analysis are compared to the permissible BCR and LCLC targets to determined economic
feasibility.
Table B-2
Financial Factors Used with Studies
BCR
CLIFE
ELIFE
FCR
IRD
IRE
IRI
IROM
LCLC
MPDR
MPER
OM
PWRE
PWRI
Table B-4 shows the estimated installation costs for each substation for the represented
improvement plan.
Table B-5 describes the efficiency savings for each substation and total project. Table B-6
provides a total project economic evaluation of benefits and costs.
Table B-3
Distribution Efficiency Study Summary
General Information
Total Customers Served (#)
Substation Annual Peak MW
Total Annual Energy Consumed (MWh/yr)
Reduction in Annual Energy (%)
Average Customer Voltage Change (%)
Total SI&VO Installed Cost
Utility Energy Savings Potential
Line Loss Saved (MWh/yr)
No-Load Loss Saved (MWhyr)
End-Use VO Energy Saved (MWh/yr)
Total Energy Savings (MWh/yr)
Customer Average Energy Reduction (kWh/yr)
Benefit Cost Projections
Regional Total Resource Cost ($/kWh)
Regional TRC Benefit Cost Ratio
Utility Levelized Cost per kWh Saved
Utility Benefit Cost Ratio
Net Utility Annual Savings ($/yr)
Overall
Sub 1
Sub 2
Sub 3
Sub 4
11,432
65.14
284,349
1.15%
2.63%
$1,120,000
1,607
12.75
55,733
0.99%
2.53%
$86,000
3,325
15.60
68,191
1.16%
2.89%
$91,000
3,935
19.35
84,583
1.38%
2.56%
$742,000
2,566
17.35
75,841
1.01%
2.57%
$201,000
292.9
208.8
2771.8
3273.5
242
-4.7
34.1
521.0
550.4
324
2.7
57.5
728.8
789.0
219
292.4
75.2
801.1
1168.7
204
2.5
42.0
720.9
765.4
281
$0.028
4.10
$0.012
2.68
($81,614)
$0.013
9.10
$0.005
5.87
($18,159)
$0.009
12.30
$0.004
7.95
($27,427)
$0.051
2.20
$0.034
0.93
$3,377
$0.021
5.40
$0.009
3.49
($21,721)
B-25
Table B-4
Typical System Improvement Installed Costs
Overall Project
VO Project Equipment
Capacitor Installations (300 kVAr)
EOL Volt Metering set
UG Line Phase Upgrade miles ( 1ph to 2ph)
OH Line Reconductoring miles (4/0 to 336ACSR)
UG Line Extension & Reconductor miles (4/0 to 1000AL)
Substation Metering set/fdr and mtr data application
Engineering Applications, GIS, and Fielding
Total SI&VO Installed Cost
Units
9
15
0.5
1.1
0.9
15
1 set
Cost
$45,000
$45,000
$100,000
$165,000
$495,000
$90,000
$180,000
$1,120,000
Sub 1
Units
1
4
0
0
0
4
1
Cost
$5,000
$12,000
$0
$0
$0
$24,000
$45,000
$86,000
Sub 2
Units
2
4
0
0
0
4
1
Cost
$10,000
$12,000
$0
$0
$0
$24,000
$45,000
$91,000
Sub 3
Units
2
3
0
1.1
0.9
3
1
Cost
$10,000
$9,000
$0
$165,000
$495,000
$18,000
$45,000
$742,000
Sub 4
Units
4
4
0.5
0
0
4
1
Cost
$20,000
$12,000
$100,000
$0
$0
$24,000
$45,000
$201,000
Table B-5
Distribution Efficiency Savings
Distribution Line Losses
Existing Base System loss (kW)
Baseline Pre-VO System loss (kW)
Peak Loss Reduction (kW)
Line Loss Saved (MWh/yr)
End-Use Energy Consumption
Average Volts Pre-VO (V)
Average Volts Post-VO (V)
Average Customer Voltage Change (%)
VO Factor (for End-Use) from Calculator
Total Annual Energy Consumed (MWh/yr)
End-Use VO Energy Saved (MWh/yr)
Distribution Transformer No-Load Losses
Distribution Transformer Connected kVA
Total No Load Loss (kW)
No-Load Loss Saved (MWhyr)
Total Utility Energy Saved
Reduction in Annual Energy (%)
Total Energy Savings (MWh/yr)
B-26
Overall
Sub 1
Sub 2
Sub 3
sub 4
610.5
493.8
116.7
292.9
48.4
50.2
-1.9
-4.7
108.1
107.0
1.1
2.7
352.5
236.0
116.5
292.4
101.6
100.6
1.0
2.5
122.97
119.81
0.02635
0.370
284,349
2771.8
122.67
119.64
0.02527
0.370
55,733
521.0
123.42
119.96
0.02888
0.370
68,191
728.8
122.97
119.90
0.02560
0.370
84,583
801.1
122.80
119.72
0.02569
0.370
75,841
720.9
156,511
469.5
208.8
26,671
80.0
34.1
39,506
118.5
57.5
58,014
174.0
75.2
32,320
97.0
42.0
1.15%
3273.5
0.99%
550.4
1.16%
789.0
1.38%
1168.7
1.01%
765.4
Overall
Sub 1
Sub 2
Sub 3
Sub 4
$1,120,000
3273.5
11,432
$86,000
550.4
1,607
$91,000
789.0
3,325
$742,000
1168.7
3,935
$201,000
765.4
2,566
3,569,634
5,097,191
1,233,012
4.10
0.028
600,236
857,096
94,677
9.10
$0.013
860,347
1,228,516
100,182
12.30
$0.009
1,274,417
1,819,780
816,870
2.20
$0.051
834,631
1,191,795
221,281
5.40
$0.021
$784,000
$336,000
$287,824
$623,824
$0.012
$60,200
$25,800
$22,101
$47,901
$0.005
$63,700
$27,300
$23,386
$50,686
$0.004
$292,177
$449,823
$190,683
$640,506
$0.034
$140,700
$60,300
$51,654
$111,954
$0.009
($81,614)
($1,672,501)
2.68
($18,159)
$281,232
5.87
($27,427)
$403,104
7.95
$3,377
($597,110)
0.93
($21,721)
($391,055)
3.49
242
($7.14)
$91.73
324
($11.30)
$145.23
219
($8.25)
$105.98
204
$0.86
-$11.03
281
($8.47)
$108.79
B-27
B-28
The feeder primary voltage drop at peak demand is 2.7%, which is low. Some phase upgrades
and reconfiguring of taps are recommended to reduce this voltage drop to 2.3%. The existing
feeder voltage source regulator set point is 122 V with line drop compensation controls applied.
The minimum primary voltage at peak demand is 119.8 V. After system improvements are
implemented, it is recommended that the voltage set point be revised to 119 V.
The single-phase 500k-VA 14.4-kVLN to 4.8-kVLN distribution transformer has a boost tap at
97.5%. At peak demand, the voltage on the load side of the transformer is 122.7 V on 120-V
base with the feeder source voltage regulator at 123.2 V. For Zero load, the voltage on the load
side of the transformer is 124.7 V, assuming that the feeder source regulator voltage is 122 V.
There are approximately 52 customers served downstream from this step-down transformer. A
regulator is recommended to lower the voltage on this lateral.
The existing distribution feeder primary voltage at peak conditions is estimated at 119.0 V 1.5V
bandwidth. The average secondary voltage drop at peak conditions is 1.9% or 2.3V. Based on the
applied 5710-MW maximum load condition, many distribution transformer/secondary
installations exhibit low service voltages and/or high secondary voltage drops. The distribution
transformer/secondary design voltage drop guideline provided by the utility is 5.0%, or 6.0 V.
The study found 24 distribution transformer installations that are expected to have service
voltages less than 114 V, which violates ANSI voltage standards. There are 418
transformer/secondary installations serving a total of 727 customers. There are 33 (7.9%)
transformer/secondary installations that exceed the utilitys maximum design voltage drop
guideline of 5.0% ,or 6 V. There are 42 (10.0%) of installations that exceed 4.2% (5.0 V) voltage
drop. The maximum transformer/secondary voltage drop is 15.0%. Distribution transformer
replacements are recommended to improve service voltages and reduce the secondary voltage
drops.
Available System Data Sources
Data used in this study was provided by utility representatives. Their assistance was integral to
the results of this study. The distribution feeder data provided included maps, actual meter data,
typical annual load shape for kW and kvar (uncompensated), operational voltage drop guidelines,
and capacitor operation strategy. The customer load profiles were provided for a 30-customer
AMI sample. The utility provided a detailed CYME distribution system model. Included with
the distribution model are the feeder connected kVA, distributed allocated load, fixed kvar
capacitors, line configurations and impedances, load balance, and voltage controls.
Assumptions were made regarding customer load characteristics, VO factors, economic factors,
and system investment unit cost options. Annual energy delivered and peak kW and kvar were
based on the typical annual load shape provided. The utilitys preferred conductor types and
sizes were assumed to be the same as the existing system. No feeder growth models were
assumed.
B-29
600 kvar
600 kvar
600 kvar
450 kvar
Figure B-1
Map of Existing Feeder Layout for Example-Feeder
3P kW Total
6,000.000
5,000.000
4,000.000
3,000.000
2,000.000
1,000.000
1
184
367
550
733
916
1099
1282
1465
1648
1831
2014
2197
2380
2563
2746
2929
3112
3295
3478
3661
3844
4027
4210
4393
4576
4759
4942
5125
5308
5491
5674
5857
6040
6223
6406
6589
6772
6955
7138
7321
7504
7687
7870
8053
8236
8419
8602
0.000
Figure B-2
Annual kW Load Profile for Example Feeder
3P kvar Total
4,000.000
3,500.000
3,000.000
2,500.000
2,000.000
1,500.000
1,000.000
500.000
1
184
367
550
733
916
1099
1282
1465
1648
1831
2014
2197
2380
2563
2746
2929
3112
3295
3478
3661
3844
4027
4210
4393
4576
4759
4942
5125
5308
5491
5674
5857
6040
6223
6406
6589
6772
6955
7138
7321
7504
7687
7870
8053
8236
8419
8602
0.000
Figure B-3
Annual kvar Load Profile for Example Feeder
B-31
The sample 30-customer annual load profile for kW and kvar is shown in Figure B-4. There was
no var compensation assumed. The hourly customer power factor is shown in Figure B-5 and
varies between 88% and 99%. Blue is watts, and red is vars.
Blackman-Hurst 30 customer load sample
SumOfkWh
SumOfkVArh
1600
1400
1200
1000
800
600
400
200
08-Nov-10
19-Sep-10
31-Jul-10
11-Jun-10
22-Apr-10
03-Mar-10
12-Jan-10
-200
23-Nov-09
Figure B-4
Annual 30 Customer kW Load Profile for Example Feeder
pf
1.01
0.99
0.97
0.95
0.93
pf
0.91
0.89
Figure B-5
Annual 30 Customer pf Load Profile for Example Feeder
B-32
08-Nov-10
19-Sep-10
31-Jul-10
11-Jun-10
22-Apr-10
12-Jan-10
03-Mar-10
0.85
23-Nov-09
0.87
10
20
30
40
50
60
70
80
90
100
27
29
32
35
39
42
46
51
55
61
66
10
27
30
33
36
39
43
47
51
56
61
66
20
28
31
34
37
40
44
47
52
56
61
66
30
29
31
34
37
41
44
48
52
56
61
66
40
29
32
35
38
41
45
49
53
57
61
66
50
30
33
36
39
42
46
49
53
57
61
66
60
31
34
37
40
43
46
50
54
57
62
66
70
32
35
38
41
44
47
50
54
58
62
66
80
33
36
39
42
45
48
51
55
58
62
66
90
34
37
40
42
45
49
52
55
59
62
66
100
35
38
41
43
46
49
52
56
59
62
66
Heating and Cooling zone classifications were determined using the zone characteristics in
Table B-8
Simplified Voltage Optimization (VO) Measurement and Verification Protocol, May 4, 2010.
http://www.nwcouncil.org/energy/rtf/measures/protocols/ut/VoltageOptimization_Protocol_v1.pdf
B-33
Table B-8
Heating and Cooling Zone Classifications
Zone 1
Zone 2
Zone 3
Max
6000
7500
>7500
Min
6000
7500
Max
300
600
>600
Min
300
600
B-34
-0.0245
-0.0185
-0.0125
-0.0065
-0.0005
0.0055
0.0115
0.0175
0.0235
0.0295
0.0355
0.0415
0.0475
0.0535
0.0595
0.0655
0.0715
0.0775
0.0835
0.0895
0.0955
0.01
0.009
0.008
0.007
0.006
0.005
0.004
0.003
0.002
0.001
0
Figure B-6
Distribution Transformer/Secondary Max Voltage Drops
Maximum compensated lagging var limit kvarMax-Lag (kvar) < 500 kvar
Minimum compensated leading var limit kvarMax-Lead (kvar) < 500 kvar
Maximum voltage control zone maximum voltage drop VDrop-VCZ (%) < 3.3%
Minimum lowest voltage at end-of-line (V) VPri-EOL (V) > VDrop-Sec + 114V.
Maximum lateral single- or two-phase design voltage drop criteria < 1.5%
B-35
The maximum feeder neutral current INeutral was only 28 Amps. The maximum feeder load phase
balance IUnbal was 13%, which is adequate. The maximum primary voltage drop VDrop-VCZ was 3.2
V, or 2.7%. The maximum primary conductor loading was 61% of normal ratings and 42% of
emergency ratings. The maximum lateral voltage drop was 0.9%.
The existing system is noncompliant with three thresholds:
The compensated lagging kvarMax-Lag was +1200 kvar, which is more than 500-kvar limit.
The compensated leading kvarMax-Lead was -1500 kvar, which is less than -500-kvar limit.
The lowest voltage at end-of-line VPri-EOL was 119.8 V, which is less than 120-V limit.
Figure B-7
Location of Primary Feeder Volts Less Than 120 V at Peak Loads
B-36
125
120
115
110
24 transformers with
service voltages
< 114 V
105
100
0
100
200
300
400
Distribution Transfomrer ID
Figure B-8
Lowest Service Voltage for Each Distribution Transformer/Secondary
B-37
Unit Cost
Line Reconfiguration
Modifications
$9,000
$5,000
$12,000
$66,000
$22,000
$8,000
$27,000
B-38
VO Project Equipment
Units
Cost
$9,000
1.00
$5,000
$12,000
0.6
$66,000
$22,000
$8,000
$27,000
$149,000
B-39
Reconfigure Taps
Bph to Cph
Install 4.8kV
132kVA 1ph Reg
Aph to Cph
Reconfigure Taps
Aph to Cph
Modify 600kvar
with switched var
controls
Relocated Trans
Aph to Bph
Modify 600kvar
with switched var
controls
Figure B-9
Recommended System Improvements
Proposed 4.8-kV line regulator VCZ maximum voltage drop VDrop-VCZ = 2.3%
3. Primary voltage at proposed line voltage regulators (at peak load) VPri-Reg = 122.8 V
4. Primary voltage at proposed line voltage regulators (at no load) VPri-Reg = 119.0 V
5. Primary voltage rise at proposed feeder source regulator VRise = 3.0 V
6. Primary voltage rise at proposed line voltage regulator VRise = 3.0 V
7. Feeder source regulator voltage setting (at no load) VSet-Fdr = 119.0 V
B-41
0.023 120V
The average voltage V2 on a 120V base for the proposed line regulator, which serves kWVCZ =
408kW is:
0.023 120V
*
=
V EXISTING
kW Fdr
0.023 120V
The average voltage V4 on a 120V base for the proposed line regulator, which serves kWVCZ =
408kW is:
0.023 120V
B-42
*
V IMPROVED
=
*
VPU
(V
*
V IMPROVED
120V
(122.04 119.64) = 0.01999 pu
=
120V
*
V PU
=
*
EXISTING
Or,
EVO = 243.6 MWh
The assumed annual load factor for end-use energy saved LDFVO is assumed to be 0.60. The
annual peak demand savings for reduced voltage is:
kWVO =
EVO
243.5
=
= 46.3kW
8760h LDFVO 8760h 0.60
The annual feeder primary line loss saved due to system improvements
--
--
The distribution transformer no-load loss reduction per year due to voltage reduction
B-43
E NL
2
1
= 171529 1
= 6656 kWh
1 + 0.020
Or,
ENL = 6.7 MWh
And,
kW NL =
1.0
CLIFE
35
ELIFE
20
FCR
17.9%
IRD
1.84%
IRE
4.31%
IRI
3.00%
IROM
3.00%
LCLC
$0.090
MPDR
$172.44
MPER
$0.060
OM
2.00%
PWRE
7.17%
PWRI
7.20%
Economic Analysis
A general economic analysis is described with the focus on the objective of maximizing the net
energy savings while meeting permissible LCLC and BCR targets. The analysis estimates the
first-year investment costs, net present value of annual fixed charges and O&M expenses, and
total improvement investment net present value. The benefits are estimated for the net present
value of energy and demand savings. The benefit-cost ratio is calculated and compared with the
BCR target. Likewise, the life cycle levelized cost factor is calculated and compared to the
LCLC target. The formulations used are considered universal. Additional economic evaluation
may need to be performed as needed to better meet the needs of the utility financial guidelines.
Economic Data Given
Peak demand reduction (PDR) = 50.103 kW (sum of VO, line loss, and no-load loss
B-45
demands)
Cost Analysis
First-Year Investment Costs (PCI) = $149,000
Annual Fixed Charges for first year (AFC) = PCI FCR = $26,671
Annual O&M Costs for first year (AOM) = PCI OM = $2,980
Uniform Series PWF
[1 (1 + PWRI ) CLIFE ]
= $337,930
NPVFIxedCh arg es (NFC ) = AOM
PWRI
Uniform Series PWF
(1 + PWRI )ELIFE
AFC
Benefit Analysis:
B-46
NEC
= 1 .1
NIC
PWRI IRI
NIC
ELIFE
1 (1 + PWRI IRI )
= $523 .65
Life Cycle Level Cost Demand Saved ( LCLCd ) =
PDR
CRF
PWRI IRI
NIC
ELIFE
1 (1 + PWRI IRI )
= $0.0933
Life Cycle Level Cost Energy ( LCLCe ) =
TES
The distribution efficiency improvement project is cost-effective, with a BCR > 1.0 and LCLC
$0.090.
Summary Exhibits of Benefits and Costs
Table B-12
Economic Summary
General Information
Total Customers Served (#)
Feeder Annual Peak MW
Total Annual Energy Consumed (MWh/yr)
Average Customer Voltage Change (%)
Reduction in Annual Energy Delivered from Sub (%)
Total SI&VO Installed Cost
Utility Energy Savings Potential
Line Loss Reduction (MWh/y)
No_Load Loss Reduction (MWh/y)
VO Energy Savings (MWh/y)
Total Energy Savings (MWh/y)
Total Energy Savings (MWa)
Total Coincidental Demand Reduction (kW)
Customer Average Energy Reduction (kWh/yr)
Benefit Cost Projections
Utility Levelized Cost per kWh Saved
Utility Benefit Cost Ratio
Net Utility PV Savings ($)
727
5.71
19,655
2.00%
1.30%
$149,000
5.0
6.7
243.6
255.2
0.0291
50.1
335
$0.093
1.1
$17,526
B-47
VO Project Equipment
Line Reconfiguration Modifications
Fixed Capacitor Additions / Modifications
Switched Capacitor Additions / Modifications
Line Phase Ugrade Additions (miles)
Line Voltage 1ph rRegulator Addition
Metering Improvements at Sub and EOL
Distribution Transf and Sec upgrades
Total SI&VO Installed Cost
B-48
Cost
3.0
1.0
2.0
0.6
1.0
1.0
9.0
Cost
$9,000
$5,000
$12,000
$66,000
$22,000
$8,000
$27,000
$149,000
216.5
213.5
3.0
5.0
0.0006
122.0
119.6
2.00%
0.62
19,655
46.3
243.6
0.0278
6,527
19.6
3.88%
0.8
6.7
0.0008
1.30%
335
50.1
255.2
0.0291
B-49
727
50.1
255.2
0.0291
$149,000
$337,930
($59,717)
$39,791
$318,005
$0.0933
$524
1.1
$318,005
$335,531
$17,526
27,036
26,701
335
Bibliography
1. ANSI Electrical Power Systems and Equipment Voltage Ratings (60Hz), ANSI Standard
1991, IEEE Trans. 91 EHO 345-9-PWR, IEEE Service Center 445 Hoes Lane,
P.O.Box1331, Piscataway, NJ 08855-1331.
B-50
6. W.H. Kersting, Distribution System Modeling and Analysis, CRC Press LLC, New York,
2002.
7. Mandal, A. Pahwa, Optimal Selection of Conductors for Distribution Feeders, IEEE Trans.
http://www.nwcouncil.org/energy/rtf/measures/protocols/ut/VoltageOptimization_Protocol_v
1.pdf
11. Rural Utility Services, Bulletins 169-4 and 1724D-101A&B, RUS Electric Programs, U.S.
Formation, and Part II Application, IEEE Trans. Power Systems Journal, Vol 5, May 2007.
16. Department of Energy, Distribution Transformers Rulemaking, Liquid-immersed
rmers_finalrule.html
B-51
C
EXTENDED CASE STUDIES
Circuit B
Circuit B is 34.5-kV, relatively long, and primarily overhead. There are two capacitor banks for
voltage support in addition to the substation voltage regulator. Both capacitor banks are 1200kvar fixed online. The locations of the capacitor banks are shown in Figure C-1. The circuit was
well balanced prior to the Green Circuit efficiency analysis, as shown by the peak hour primary
bus voltage profile in Figure C-2. The efficiency options analyzed are described in
C-1
Table C-1. Additional voltage regulators have not been found applicable to this circuit due to a
stiff 34.5-kV system voltage.
Substation
Voltage (pu)
Figure C-1
Circuit Map Indicating Capacitors () and Regulators ()
C-2
Table C-1
Efficiency Projects Tested
Option Label
Option Details
Phase Balance
PB1
Var Optimization
VAR1
VAR2
VAR3
+ Remove one 1200-kvar fixed capacitor bank (different bank from VO2)
VAR4
VAR5
VAR6
VAR7
VAR8
Re-Conductor
R1
R2
R3
R4
Combinations
C1
+ PB1
+ VAR5
C2
+ PB1
+ VAR5
+ R2
C3
+ VAR5
+ R2
C-3
Conservation voltage reduction with voltage feedback was the primary option to reduce the
annual energy consumption and transformer no-load (core) losses. The total annual energy saved
was relatively constant for all efficiency options due to stiff system voltage and prior feeder
balancing. The high system voltage (and associated lower current) led to lower returns in loss
reduction from var optimization and re-conductoring. Slight improvements in system voltage
from the efficiency options prevented additional savings from conservation voltage reduction.
The highest economic benefit came from option conservation voltage reduction. The benefit-cost
ratio was maintained above one for the majority of options based on the savings from
conservation voltage reduction. The incremental benefit-cost ratio was only greater than one for
further phase balancing. This relatively low-cost project provided slightly higher savings in
return. Re-conductoring had the lowest economic benefit and was unacceptable for one option.
2000
450
1800
400
350
1400
300
1200
250
1000
200
800
150
600
100
400
50
200
0
0
VR
R2
R3
R4
C1
OPTION
Total Energy Saved MWh
Figure C-3
Annual Energy Saved and Peak Demand Reduction
C-4
C2
C3
1600
Table C-2
Annual and Peak Savings for End-Use Load and Losses
Option
Annual
Consumption
Saved
MWh (% Base)
Annual
Losses
Saved
MWh
(% Base)
Peak
Demand
Saved
kW
(% Base)
Peak
Losses
Saved
kW
(% Base)
VR
1625.1 (3.44%)
83.1 (8.01%)
375.2
(0.30%)
PB1
1643.6 (3.48%)
84.5 (8.14%)
VAR1
1597.9 (3.38%)
VAR2
Total Annual
Energy Saved
MWh
(% Base)
Total Peak
Reduction
kW
(% Base)
6.4 (0.34%)
1708.2
(3.54%)
381.6
(0.30%)
375.8
(0.30%)
6.9 (0.37%)
1728.1
(3.58%)
382.7
(0.30%)
84.8 (8.17%)
371.4
(0.29%)
6.4 (0.34%)
1682.7
(3.48%)
377.8
(0.29%)
1601.8 (3.39%)
84.5 (8.14%)
347.9
(0.28%)
3.9 (0.21%)
1686.3
(3.49%)
351.8
(0.27%)
VAR3
1604.3 (3.39%)
84.0 (8.09%)
347.3
(0.27%)
4.1 (0.22%)
1688.2
(3.49%)
351.4
(0.27%)
VAR4
1618.6 (3.42%)
85.6 (8.25%)
373.0
(0.30%)
6.2 (0.33%)
1704.2
(3.53%)
379.2
(0.30%)
VAR5
1611.1 (3.41%)
85.6 (8.25%)
402.5
(0.32%)
7.8 (0.41%)
1696.7
(3.51%)
410.3
(0.32%)
VAR6
1590.6 (3.36%)
85.5 (8.23%)
402.2
(0.32%)
7.8 (0.41%)
1676.1
(3.47%)
410.0
(0.32%)
VAR7
1615.0 (3.42%)
84.7 (8.16%)
362.5
(0.29%)
5.9 (0.31%)
1699.7
(3.52%)
368.4
(0.29%)
VAR8
1629.9 (3.45%)
79.2 (7.63%)
312.1
(0.25%)
5.1 (0.27%)
1709.1
(3.54%)
317.2
(0.25%)
R1
1622.5 (3.43%)
105.4
(10.15%)
363.9
(0.29%)
22.5
(1.19%)
1727.9
(3.58%)
386.4
(0.30%)
R2
1645.1 (3.48%)
113.5
(10.93%)
360.7
(0.29%)
27.0
(1.42%)
1758.6
(3.64%)
387.7
(0.30%)
R3
1624.4 (3.44%)
85.0 (8.18%)
374.2
(0.30%)
7.8 (0.41%)
1709.3
(3.54%)
382.0
(0.30%)
R4
1646.1 (3.48%)
115.4
(11.12%)
359.7
(0.28%)
28.4
(1.50%)
1761.5
(3.65%)
388.1
(0.30%)
C1
1624.3 (3.44%)
86.9 (8.37%)
403.1
(0.32%)
8.4 (0.44%)
1711.2
(3.54%)
411.5
(0.32%)
C2
1643.6 (3.48%)
116.0
(11.18%)
388.9
(0.31%)
28.4
(1.50%)
1759.6
(3.64%)
417.3
(0.33%)
C3
1606.8 (3.40%)
114.0
(10.98%)
388.3
(0.31%)
28.0
(1.48%)
1720.8
(3.56%)
416.3
(0.32%)
C-5
20
18
$0.14
16
$0.10
12
$0.08
10
8
$0.06
6
$0.04
4
$0.02
2
0
$0.00
VR
R2
R3
R4
C1
C2
C3
OPTION
Benefit Cost Ratio
Figure C-4
Economic Acceptability Shown by Benefit-Cost Ratio and Levelized Cost for Efficiency
Projects
C-6
$0.12
14
Table C-3
Economic Analysis of Efficiency Projects
Option
BenefitCost
Ratio
Levelized
Cost
($/kWh)
Total Efficiency
Option Cost
(NPV $K)
Total Efficiency
Option Savings
(NPV $K)
Incremental
LC After
Option VF
Incremental
BCR After
Option VF
VR
17.6
$0.006
$114
$2,002
PB1
17.0
$0.006
$119
$2,023
0.021
4.2
VAR1
10.0
$0.010
$197
$1,974
-0.274
-0.3
VAR2
17.1
$0.006
$114
$1,960
-0.001
-147.6
VAR3
17.2
$0.006
$114
$1,962
-0.001
-141.7
VAR4
14.1
$0.007
$142
$1,997
-0.578
-0.2
VAR5
11.0
$0.009
$183
$2,010
-0.505
0.1
VAR6
8.3
$0.012
$239
$1,989
-0.326
-0.1
VAR7
10.1
$0.010
$196
$1,985
-0.801
-0.2
VAR8
15.3
$0.006
$128
$1,960
1.384
-3.1
R1
2.4
$0.041
$846
$2,026
3.118
0.0
R2
1.0
$0.096
$2,016
$2,058
3.165
0.0
R3
2.0
$0.050
$1,022
$2,004
68.874
0.0
R4
0.7
$0.139
$2,924
$2,061
4.419
0.0
C1
10.8
$0.009
$188
$2,026
2.075
0.3
C2
1.0
$0.099
$2,090
$2,079
3.221
0.0
C3
1.0
$0.101
$2,085
$2,039
13.118
0.0
Voltage reduction provided significant savings for both energy consumption and losses.
Relatively flat voltage profiles limited additional consumption savings for projects other than
voltage reduction.
Cost and loss efficiency of combination options added near linearly with the individual
options.
C-7
Circuit C
Circuit C is 25-kV, relatively long, and primarily overhead, with the greatest load density near
the feeder-end approximately 8 miles from the substation. There is a substation regulator, yet
there is no capacitor banks on the circuit. The circuit map is shown in Figure C-5, and the peak
hour voltage profile is shown in Figure C-6. The efficiency options analyzed are described in
Table C-4.
Substation
Figure C-5
Circuit Map
C-8
Figure C-6
Peak Hour Bus Voltage w.r.t Distance From Substation
C-9
Option Details
Phase Balance
PB1
Var Optimization
VAR1
VAR2
Re-Conductor
R1
R2
R3
Combinations
C1
+ PB1
+ VAR2
C2
+ PB1
+ VAR2
+ R3
C-10
60
450
50
40
350
30
300
20
250
10
200
0
150
400
-10
100
-20
50
0
-30
VR
PB1
VAR1
VAR2
R1
OPTION
Total Energy Saved MWh
R2
R3
C1
C2
Figure C-7
Annual Energy Saved and Peak Demand Reduction
Table C-5
Annual and Peak Savings for End-Use Load and Losses
Option
Annual
Consumption
Saved
MWh
(% Base)
Annual
Losses
Saved
MWh
(% Base)
Peak
Demand
Saved
kW
(% Base)
-24.7
(-0.04%)
Peak
Losses
Saved
kW
(% Base)
-0.3
(-0.01%)
VR
385.4 (1.99%)
21.5 (3.82%)
PB1
424.4 (2.19%)
25.0 (4.44%)
21.9 (0.04%)
1.8 (0.09%)
449.4 (2.25%)
VAR1
389.0 (2.01%)
10.8 (1.91%)
-26.5
(-0.04%)
-12.9
(-0.02%)
-5.8
(-0.01%)
-11.6
(-0.02%)
-14.1
(-0.02%)
15.1
(0.72%)
399.8 (2.01%)
VAR2
385.5 (1.99%)
24.2 (4.29%)
5.1 (0.24%)
409.7 (2.05%)
R1
387.9 (2.00%)
30.1 (5.35%)
8.6 (0.41%)
418.0 (2.10%)
2.8 (0.00%)
R2
388.3 (2.00%)
38.7 (6.86%)
427.0 (2.14%)
5.6 (0.01%)
R3
394.1 (2.03%)
42.8 (7.59%)
436.9 (2.19%)
7.0 (0.01%)
C1
435.6 (2.25%)
27.8 (4.94%)
34.7 (0.06%)
7.0 (0.34%)
463.5 (2.32%)
C2
427.6 (2.21%)
47.8 (8.49%)
20.9 (0.03%)
27.2
(1.31%)
475.5 (2.38%)
17.2
(0.82%)
21.1
(1.01%)
Total Annual
Energy Saved
MWh
(% Base)
406.9 (2.04%)
Total Peak
Reduction
kW
(% Base)
-25.0
(-0.04%)
23.8
(0.04%)
-11.4
(-0.02%)
-7.9
(-0.01%)
41.8
(0.07%)
48.2
(0.08%)
C-11
$1.2
$1.0
$0.8
$0.6
$0.4
1
$0.2
$0.0
VR
PB1
VAR1
VAR2
R1
R2
R3
C1
C2
OPTION
Benefit Cost Ratio
Figure C-8
Economic Acceptability Shown by Benefit-Cost Ratio and Levelized Cost for Efficiency
Projects
Table C-6
Economic Analysis of Efficiency Projects
Total
Efficiency
Option
Savings
(NPV $K)
Option
BenefitCost
Ratio
Levelized
Cost
($/kWh)
Total
Efficiency
Option Cost
(NPV $K)
VR
3.5
$0.023
$114
$398
PB1
3.9
$0.023
$122
VAR1
2.7
$0.031
VAR2
3.3
R1
Incremental
LC After
Option VF
Incremental
BCR After
Option VF
$475
0.015
10.2
$146
$400
-0.375
0.1
$0.025
$123
$413
0.283
1.6
0.2
$0.473
$2,360
$428
16.996
0.0
R2
0.1
$0.929
$4,735
$440
19.314
0.0
R3
0.1
$0.955
$4,982
$451
13.621
0.0
C1
3.8
$0.024
$131
$501
0.025
6.1
C2
0.1
$0.881
$4,999
$518
5.968
* Negative incremental values represent unacceptable additional project component.
0.0
C-12
Voltage reduction provided significant savings for both energy consumption and losses.
Voltage reduction led projects to be economically acceptable except those including reconductoring.
Phase balancing and reactive power optimization options were incrementally acceptable.
Cost and loss efficiency of combination options added near linearly with the individual
options.
Circuit D
The 12.47-kV Circuit D is relatively short and primarily underground with uniform load density
along the primary. At peak hour the feeder voltages slowly decay along the primary, as shown in
Figure C-9. There is a substation voltage regulator and three capacitor banks on the circuit. The
capacitor locations are shown on the circuit map in Figure C-10. The capacitor banks are each
1200-kvar with substation power factor control. The power factor control algorithm switches the
three capacitors banks on sequentially based on distance from the substation to maintain 0.98
leading power factor at the substation transformer. The analyzed efficiency options are described
in Table C-7. Additional voltage regulators have not been found applicable for this circuit due to
the short circuit length.
Figure C-9
Peak Hour Bus Voltage versus Distance From Substation
C-13
Substation
Figure C-10
Circuit Map Indicating Capacitors
C-14
Option Details
Phase Balancing
PB1
PB2
PB3
+ PB1
+ PB2
Var Optimization
VAR1
VAR2
VAR3
VAR4
+ Replace PF control with local kvar control on the three original banks
VAR5
VAR6
Re-Conductoring
R1
Combinations
C1
+ PB3
+ VAR3
C2
+ PB3
+ R1
C3
+ PB3
+ VAR3
+ R1
C4
+ PB3
+ VAR1
C-15
The total annual energy saved and peak demand reduction for the efficiency options are shown in
Figure C-11. Savings were primarily influenced by conservation voltage reduction. The highest
annual and peak savings occurred from the combination of phase balancing, var optimization,
and re-conductoring. The detailed change in energy/demand and load/losses are given in
Table C-8.
Reactive power optimization option VAR1 removed all capacitor banks from the circuit. Without
capacitor banks, the reduction in leading reactive flow reduced losses, and conservation voltage
reduction was capable of reducing end-use consumption. In option VAR2, three small fixed
capacitor banks were optimally placed in the circuit to fully compensate light reactive load. This
significantly reduced annual losses and annual consumption from option conservation voltage
reduction. In option VAR3, an additional switched capacitor bank was optimally placed in
service to compensate full reactive load. The var optimization option VAR3 can potentially
reduce annual and peak losses by driving the annual feeder power factor closer to unity. In
option VAR4, local kvar control replaced the original power factor control. This had the
potential to reduce consumption from option VR. However, the annual losses increased because
the capacitor banks were primarily stuck in service. Local switched control was not practical in
this circuit due to remote location of banks on low load taps. The centrally controlled switching
of capacitor banks was not designed to reduce losses and limited the possible savings through
conservation voltage reduction.
100
350
90
300
250
70
60
200
50
150
40
30
100
20
50
10
0
0
VR
OPTION
Total Energy Saved MWh
Figure C-11
Annual Energy Saved and Peak Demand Reduction
C-16
C1
C2
C3
C4
80
Option
Annual
Consumption
Saved
MWh
(% Base)
Annual
Losses
Saved
MWh
(% Base)
Peak
Demand
Saved
kW
(% Base)
Peak
Losses
Saved
kW
(% Base)
Total Annual
Energy Saved
MWh
(% Base)
Total Peak
Reduction
kW
(% Base)
VR
173.9 (0.9%)
2.9 (0.8%)
37.4 (0.1%)
0.3 (0.0%)
176.8 (0.9%)
37.7 (0.1%)
PB1
207.5 (1.1%)
4.5 (1.3%)
48.9 (0.1%)
1.0 (0.1%)
212.0 (1.1%)
49.8 (0.1%)
PB2
209.4 (1.1%)
6.2 (1.8%)
48.8 (0.1%)
2.0 (0.2%)
215.6 (1.1%)
50.8 (0.1%)
PB3
228.7 (1.2%)
7.5 (2.1%)
60.3 (0.1%)
2.6 (0.2%)
236.2 (1.2%)
62.9 (0.1%)
VAR1
206.1 (1.0%)
4.9 (1.4%)
43.9 (0.1%)
-4.5 (-0.4%)
211.0 (1.1%)
39.3 (0.1%)
VAR2
195.5 (1.0%)
15.3 (4.4%)
57.8 (0.1%)
0.2 (0.0%)
210.8 (1.1%)
57.9 (0.1%)
VAR3
196.2 (1.0%)
16.1 (4.6%)
60.0 (0.1%)
2.6 (0.2%)
212.3 (1.1%)
62.6 (0.1%)
VAR4
202.8 (1.0%)
1.7 (0.5%)
60.4 (0.1%)
0.7 (0.1%)
204.5 (1.0%)
61.1 (0.1%)
VAR5
184.6 (0.9%)
7.8 (2.3%)
44.8 (0.1%)
0.4 (0.0%)
192.4 (1.0%)
45.2 (0.1%)
VAR6
206.8 (1.1%)
3.8 (1.1%)
57.1 (0.1%)
0.5 (0.0%)
210.7 (1.1%)
57.6 (0.1%)
R1
189.6 (1.0%)
25.4 (7.3%)
40.0 (0.1%)
13.3 (1.2%)
214.9 (1.1%)
53.3 (0.1%)
C1
236.6 (1.2%)
20.2 (5.8%)
70.0 (0.1%)
4.7 (0.4%)
256.8 (1.3%)
74.8 (0.1%)
C2
230.6 (1.2%)
30.1 (8.6%)
63.2 (0.1%)
15.2 (1.4%)
260.6 (1.3%)
78.4 (0.2%)
C3
252.0 (1.3%)
39.7 (11.4%)
73.0 (0.1%)
16.8 (1.5%)
291.7 (1.5%)
89.9 (0.2%)
C4
258.8 (1.3%)
9.4 (2.7%)
77.2 (0.2%)
-2.0 (-0.2%)
268.3 (1.3%)
75.2 (0.2%)
The economic acceptability of each option is given in the Figure C-12 and Table C-9. Voltage
feedback with conservation voltage reduction had high economic gains that rolled over into all
efficiency options. Phase balancing had a low cost and produced substantial savings, which is
reflected by higher overall economic acceptability. Reactive power optimization option VAR1
had the highest overall economic acceptability; however, the significant jump in acceptability
was due to the reduction in option cost from the salvage value of the removed pad-mount
capacitor banks.
The incremental levelized cost and benefit-cost ratio were acceptable only for phase balancing
and option VAR1. Re-conductoring and var optimization options (besides VAR1) were
incrementally unacceptable due to high implementation costs.
C-17
5.0
4.5
3.5
$0.20
3.0
$0.15
2.5
2.0
$0.10
1.5
1.0
$0.05
0.5
0.0
$0.00
VR
PB1
PB2
C1
C2
C3
C4
OPTION
Benefit Cost Ratio
Figure C-12
Economic Acceptability Shown by Benefit-Cost Ratio and Levelized Cost for Efficiency
Projects
C-18
$0.25
4.0
Option
BenefitCost
Ratio
Levelized
Cost
($/kWh)
Total
Efficiency
Option Cost
(NPV $K)
VR
1.8
$0.054
$114
$206
PB1
2.1
$0.047
$119
PB2
2.1
$0.046
PB3
2.3
VAR1
Incremental
LC After
Option VF
Incremental
BCR After
Option VF
$250
0.012
8.8
$119
$254
0.011
9.6
$0.044
$124
$284
0.014
7.7
4.1
$0.023
$59
$242
0.000
100.0
VAR2
1.5
$0.066
$166
$254
0.128
0.9
VAR3
1.0
$0.103
$260
$259
0.346
0.4
VAR4
0.7
$0.156
$381
$250
0.810
0.2
VAR5
0.7
$0.148
$341
$227
1.220
0.1
VAR6
0.8
$0.120
$301
$254
0.462
0.3
R1
0.4
$0.241
$618
$256
1.107
0.1
C1
1.2
$0.088
$270
$313
0.164
0.7
C2
0.5
$0.202
$628
$319
0.513
0.2
C3
0.5
$0.222
$774
$359
0.481
0.2
C4
4.7
$0.021
$69
$325
0.000
100.0
Phase balancing and reactive power optimization options were incrementally acceptable.
Cost and efficiency of combination options added near linearly with the individual options.
Circuit E
The 25-kV Circuit E is primarily overhead. There is a substation load drop compensator along
with four capacitor banks on the circuit. Three capacitor are 600-kvar, and the forth is 450-kvar.
All banks are fixed online. The circuit map and capacitor locations are shown in Figure C-13.
The analyzed efficiency options are described in Table C-10.
C-19
The peak hour primary bus voltages are shown in Figure C-14. Due to limiting bus voltages in
the base case, the tap setting on the feeder step-down transformer (shown in circuit plot) was
adjusted to provide further voltage support. The peak hour voltages after the tap adjustment in
the conservation voltage reduction option VR is shown in Figure C-15.
Substation
Figure C-13
Circuit Map Indicating Capacitors () and Step-Down Transformer ()
C-20
Option Details
Phase Balance
PB1
PB2
Var Optimization
VAR1
VAR2
VAR3
VAR4
Re-Conductor
R1
R2
Voltage Regulation
REG1
REG2
REG3
+ 4A
+ 4B
Combinations
C1
+ PB2
+ VAR4
C2
+ PB2
+ VAR4
+ R2
C3
+ PB2
+ VAR4
+ R2
+ VR2
C-21
Voltage (pu)
Voltage (pu)
Figure C-14
Base Case Peak Hour Bus Voltage w.r.t Distance From Substation
C-22
The conservation voltage reduction option in this case study involved adjusting the tap setting of
one single-phase step-down transformer. This adjustment was necessary to prevent isolated low
voltages from occurring during conservation voltage reduction.
All options decreased annual losses compared to option conservation voltage reduction;
however, annual consumption increased for all options other than additional voltage regulation.
Including additional regulators for significant feeder branches as in option VR3 further reduced
annual consumption. The high system voltage allowed greater potential for consumption savings
than loss reduction savings.
The economic benefit of the efficiency options were primarily due to conservation voltage
reduction. The benefit from conservation voltage reduction carried over to all options. The
incremental benefit-cost ratio was unacceptable for all additional options. This was due to high
additional costs and/or decreased savings. The best potential for increased savings was due to
additional voltage regulation on the feeder.
450
120
400
300
80
250
60
200
150
40
100
350
100
20
50
0
0
VR
C2
C3
OPTION
Total Energy Saved MWh
Figure C-16
Annual Energy Saved and Peak Demand Reduction
C-23
C-24
Option
Annual
Consumption
Saved
MWh
(% Base)
Annual
Losses
Saved
MWh
(% Base)
Peak
Demand
Saved
kW
(% Base)
Peak
Losses
Saved
kW
(% Base)
Total Annual
Energy Saved
MWh
(% Base)
Total Peak
Reduction
kW
(% Base)
VR
328.1 (1.74%)
26.9 (3.21%)
81.5 (0.17%)
4.9 (0.19%)
355.0 (1.80%)
86.3
(0.17%)
PB1
326.6 (1.73%)
28.3 (3.39%)
65.3 (0.14%)
5.2 (0.20%)
355.0 (1.80%)
70.5
(0.14%)
PB2
323.7 (1.72%)
31.0 (3.70%)
71.0 (0.15%)
7.2 (0.28%)
354.6 (1.80%)
78.2
(0.16%)
VAR1
289.4 (1.53%)
39.4 (4.71%)
58.8 (0.12%)
-5.1
(-0.20%)
328.8 (1.67%)
53.7
(0.11%)
VAR2
282.0 (1.49%)
41.9 (5.00%)
66.6 (0.14%)
4.0 (0.15%)
323.9 (1.64%)
70.5
(0.14%)
VAR3
274.4 (1.45%)
40.3 (4.81%)
53.1 (0.11%)
-5.9
(-0.23%)
314.7 (1.60%)
47.2
(0.09%)
VAR4
266.8 (1.41%)
42.6 (5.09%)
56.0 (0.12%)
-2.7
(-0.11%)
309.4 (1.57%)
53.3
(0.11%)
R1
292.1 (1.55%)
42.7 (5.11%)
58.0 (0.12%)
15.3
(0.60%)
334.9 (1.70%)
73.3
(0.15%)
R2
303.7 (1.61%)
73.1 (8.73%)
67.8 (0.14%)
34.9
(1.36%)
376.7 (1.91%)
102.8
(0.21%)
REG1
357.5 (1.89%)
28.0 (3.35%)
77.6 (0.16%)
2.7 (0.11%)
385.5 (1.96%)
80.3
(0.16%)
REG2
343.0 (1.82%)
28.5 (3.40%)
81.4 (0.17%)
4.9 (0.19%)
371.5 (1.88%)
86.3
(0.17%)
REG3
365.2 (1.94%)
28.6 (3.42%)
78.0 (0.17%)
2.3 (0.09%)
393.9 (2.00%)
80.4
(0.16%)
C1
315.2 (1.67%)
50.4 (6.02%)
74.7 (0.16%)
1.5 (0.06%)
365.6 (1.86%)
76.2
(0.15%)
C2
271.6 (1.44%)
92.4
(11.03%)
54.2 (0.11%)
32.2
(1.26%)
363.9 (1.85%)
86.4
(0.17%)
C3
288.0 (1.53%)
94.3
(11.26%)
52.9 (0.11%)
32.0
(1.25%)
382.3 (1.94%)
85.0
(0.17%)
$0.30
$0.25
$0.20
$0.15
$0.10
1
$0.05
$0.00
VR
PB1
C2
C3
OPTION
Benefit Cost Ratio
Figure C-17
Economic Acceptability Shown by Benefit-Cost Ratio and Levelized Cost for Efficiency
Projects
C-25
Option
BenefitCost
Ratio
Levelized
Cost
($/kWh)
Total
Efficiency
Option Cost
(NPV $K)
VR
3.6
$0.028
$117
$421
PB1
3.4
$0.029
$122
PB2
2.5
$0.038
VAR1
3.1
VAR2
Incremental
LC After
Option VF
Incremental
BCR After
Option VF
$410
30.309
-2.1
$163
$415
-11.764
-0.1
$0.030
$119
$372
-0.009
-17.6
1.9
$0.053
$204
$378
-0.234
-0.5
VAR3
2.6
$0.036
$135
$353
-0.039
-3.6
VAR4
2.3
$0.042
$153
$352
-0.068
-1.9
R1
0.7
$0.140
$559
$392
-1.843
-0.1
R2
0.5
$0.223
$1,002
$454
3.414
0.0
REG1
2.9
$0.034
$156
$448
0.108
0.7
REG2
3.1
$0.032
$143
$438
0.136
0.6
REG3
2.5
$0.039
$183
$457
0.143
0.5
C1
2.1
$0.046
$200
$425
0.653
0.0
C2
0.4
$0.250
$1,085
$430
9.066
0.0
C3
0.4
$0.244
$1,112
$448
3.051
0.0
Utilizing voltage regulators for additional regulation zones increased the effectiveness of
voltage reduction.
Cost and loss efficiency of combination options added near linearly with the individual
options.
C-26
Access to and use of EPRI Intellectual Property is granted with the spe-
compliance with all applicable U.S. and foreign export laws and regu-
under applicable U.S. and foreign export laws and regulations. In the
event you are uncertain whether you or your company may lawfully
mine whether this access is lawful. Although EPRI may make available
EPRIs principal offices and laboratories are located in Palo Alto, Calif.;
export classification for specific EPRI Intellectual Property, you and your
company acknowledge that this assessment is solely for informational
purposes and not for reliance purposes. You and your company acknowledge that it is still the obligation of you and your company to make
your own assessment of the applicable U.S. export classification and
ensure compliance accordingly. You and your company understand and
acknowledge your obligations to make a prompt report to EPRI and the
appropriate authorities regarding any access to or use of EPRI Intellectual Property hereunder that may be in violation of applicable U.S. or
foreign export laws or regulations.
Program:
Distribution Systems
2011 Electric Power Research Institute (EPRI), Inc. All rights reserved. Electric Power
Research Institute, EPRI, and TOGETHER...SHAPING THE FUTURE OF ELECTRICITY are
registered service marks of the Electric Power Research Institute, Inc.
1023518