Professional Documents
Culture Documents
ABSTRACT ZonePAC bridges this gap between building operations and the ex-
Heating Ventillation and Air Conditioning (HVAC) systems ac- perience by individual occupants. We do this by connecting zonal
count for nearly 40% of primary energy consumption by commer- monitoring and estimation that incorporates participatory occupant
cial buildings. Yet, these systems by and large operate in an open- sensing and occupant experiential feedback to be incorporated in
loop with the building occupants. While the monitoring and feed- the building scale HVAC system.
back of comfort conditions is by (much smaller) zones, the HVAC ZonePAC estimates the heating, cooling and electrical power
control systems operate on energy metering and monitoring at the consumption of each zone in a Variable Air Volume (VAV) type
scale of entire buildings. ZonePAC attempts to bridge this gap be- system using existing infrastructure sensors installed as part of the
tween metering, monitoring and control by providing an embedded Building Management System (BMS). We provide the estimated
sensing and information management architecture that provides for zone power consumption as feedback to the occupants of the build-
effective participation by the building occupants in zonal HVAC ing over the web and on mobile devices along with other thermal
settings that directly affect the building scale HVAC control sys- comfort related measurements such as temperature and setpoint.
tem. Our results from a deployment of 65 users spread across 51 We have built ZonePAC on top of BuildingDepot [3], an open-
zones in a 145,000 square feet commercial building demonstrate source RESTful web service BMS. We have deployed ZonePAC
the viability and effectiveness of ZonePAC. on a 145,000 sqft university building with 237 zones and over 4700
BACnet data points for a period of 10 days. This paper primarily
provides the results of our data collection and its analysis regard-
Categories and Subject Descriptors ing distribution of energy consumption across zones. We identify
H.3.5 [Online Information Services]: Web-based services; H.5.2 anomalous behavior and provide possible causes behind energy in-
[Information Interfaces and Presentation]: User InterfacesUser- efficiency. Since the ZonePAC system also provides the occupant
centered design; K.4.3 [Organizational Impacts]: Computer-supported with the capability to change local HVAC control settings, we pro-
collaborative work vide data on user experience and the results of such individual con-
trol settings collected for 65 occupants spread across 51 zones.
General Terms
Measurement, Human Factors, Design 2. OUR BUILDING TESTBED
Our building testbed is the Computer Science and Engineering
Keywords (CSE) building at University of California, San Diego (UCSD).
HVAC, Thermostat, Energy Estimation, Variable Air Volume, User Built in 2004, CSE building has 145,000 sqft of floor space with
Interface 466 rooms. The HVAC design of the building follows modern de-
sign practices and employs a VAV system which can be managed
1. INTRODUCTION by a BMS using BACnet protocol.
Figure 1 shows the system design of the centralized part of the
HVAC systems contribute to 39.6% of the primary energy con- HVAC system. UCSD has a Central Utility Plant (CUP) that op-
sumption in commercial buildings [2], and is one of the prime erates gas driven turbines with hot and chilled water by products.
targets for improving building energy efficiency. Several studies Chilled water is used for large scale energy storage that accounts for
have shown that providing relevant energy feedback to the occu- nearly 14% of daily energy use of the entire campus. As a result, we
pants of a building can lead to significant energy savings [8, 22]. do not have chillers, cooling towers or boilers in the building nor-
However, the energy feedback has been limited to electricity con- mally associated with commercial HVAC systems. As the weather
sumption [22]and has been designed for residential buildings [4, 8]. is temperate throughout the year in San Diego, the HVAC system
was designed without any humidity control. Thus, we do not have
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are not any humidifiers or de-humidifiers found in buildings with harsher
made or distributed for profit or commercial advantage and that copies bear outdoor environment. The centralized part of the system supplies
this notice and the full citation on the first page. Copyrights for components cold air and hot water to the VAV boxes for providing conditioned
of this work owned by others than ACM must be honored. Abstracting with air to building spaces.
credit is permitted. To copy otherwise, or republish, to post on servers or to The cold water from CUP, supplied at 44 F, is distributed to
redistribute to lists, requires prior specific permission and/or a fee. Request the rest of the system by four cold water pumps, two of which are
permissions from Permissions@acm.org.
BuildSys 13, November 13-14 2013, Roma, Italy dedicated for supplying water to the Computer Room Air Condi-
Copyright 2013 ACM 978-1-4503-2431-1/13/11 ...$15.00. tioning (CRAC) unit for the server room in CSE. The cold water is
Domestic Water Supply Air Return Air
SUPPLY
COLD WATER FAN RETURN
Room 1 Room 2
PUMP FAN HVAC Zone
*CENTRAL PLANT Zone Temperature
SUPPLIES HOT AND T Thermostat Adjust
Supply Air Return Air COLD WATER
Figure 1: Centralized part of HVAC system showing details of Figure 2: VAV with reheat system used for controlling the tem-
air handler unit - cooling coils, heat exchanger, water pumps, perature of discharge air in each HVAC zone.
supply/return fans, domestic water and CRAC units.
a button on the thermostat. The basement student labs and public
circulation area are an exception to the schedule, and are always set
passed through cooling coils, which cool the mixture of outside and to Occupied mode.
return air to the appropriate setpoint ( 55 F) to provide supply As an experimental testbed, our building is also instrumented to
air to all the zones in the building. The supply air is dispensed to the allow us to measure the total cooling and heating thermal power,
VAV boxes via ductwork using supply fans, and the flow of return and electric power consumption of the entire HVAC system as well
air is facilitated using return fans. The air mixer uses economizers as the lighting, computer room and plug loads subsystems. This
to increase the proportion of outside air if outdoor conditions are allows us to estimate and track zonal power use as discussed next.
favourable for reducing energy usage.
The hot water from CUP, supplied in the form of pressurized
steam at 325 F, passes through heat exchangers for heating up 3. ZONE POWER ESTIMATION
the hot water returned by the VAV boxes. Part of the hot water is The goal of ZonePAC is to provide a real-time estimate of to-
used to heat the domestic water. The hot water from the HVAC heat tal power consumption of individual HVAC zones to the building
exchanger is supplied via pipes to the VAV boxes using hot water occupants. We use the measurements from existing sensors, and
pumps. All the pumps and the fans used in the centralized part of apply first principles to estimate power consumption of an HVAC
the system employ Variable Frequency Drives (VFDs). The CRAC zone which consists of three parts - cooling thermal, heating ther-
units do not use the hot water, and have an electric reheat system mal and electrical. The cooling thermal power is used for convert-
for environment control. ing the warm mixed air to the cold supply air, the heating thermal
Figure 2 shows the HVAC design of the VAV boxes in CSE. power is used for reheating the cold supply air when the temper-
The amount of cold air supplied to each zone is modulated using ature setpoint of the zone is too high to be satisfied by reducing
a damper, and a flow sensor measures the airflow rate. The zonal the supply air to minimum, and the electrical power is used by the
temperature is controlled by modulating the amount of cold air and fans and the pumps used for supplying cold air and hot water to the
by using the hot water coil to reheat the air. The amount of hot zones from the central HVAC equipments.
water used in heating the air is modulated using an electronically
controlled 2-way valve. Every zone has a thermostat which mea- 3.1 Cooling Thermal Power
sures the current temperature, and acts as the feedback for the VAV We estimate cooling thermal power using the heat transfer equa-
control system. Occupants are allowed to change their temperature tion:
setpoints by 1 F using the thermostat dial. Qcooling = C q (Tzone Tsupply ) (1)
The BMS operates the HVAC system on a weekly schedule. On
weekdays, the HVAC system is put to Occupied from 6am to where, = density of air at 20 C, C = specific heat of air, q =
6pm, then changed to Standby mode till 10pm, and switched to rate of airflow, Tzone = zonal air temperature, Tsupply = supply air
Unoccupied for the rest of the night. In the Occupied mode, temperature.
a minimum amount of airflow is maintained for ventilation, and In the absence of sensors to directly measure the supply air tem-
the minimum airflow setpoint for each zone is determined based on perature of each zone, we approximate it by the supply air temper-
its maximum capacity. The temperature of the zone is maintained ature as measured by the central air handler unit (AHU) as it exits
within 4 F range, and the exact range is determined by the temper- the cooling coils. This, of course, neglects the temperature loss
ature setpoint set by the BMS as well as the thermostat adjustment due to imperfect insulation and leaks in the air ducts. Similarly,
set by the occupant. In the Standby mode, the airflow is reduced we estimate the return air temperature by the zonal temperature as
to minimal amount, and the temperature range is increased to 8 F, measured by the thermostat in the zone. Finally, the airflow rate is
and in the Unoccupied mode, the temperature range is further in- measured directly by the flow sensor in the VAV box.
creased to 12 F. The HVAC system remains in the Unoccupied For HVAC systems which also provide humidity control, the
mode on weekends and holidays, and if an occupant were to use a power consumption estimate would also have to include the latent
zone during that time, she can express her occupancy by pressing heat transfer. The corresponding sensors measuring the supply air
140
Measured
HVAC Controller
BuildingDepot (BD)
HTTP Service (HCS)
HTTP HTTP
Figure 5: System Architecture of HVAC Meter Figure 6: Screenshot of Web User Interface
vice. The data from BACnet sensors are collected using our BAC- points related to each HVAC zone in CSE, and measurement from
net connector, and the HVAC Meter Service estimates the power all the points are shown in the Details section. The user can also
consumption of each zone as explained in Section 3. The Web see the historical trend of each of the data points by clicking on the
User Interface (WebUI) reads the data from virtual power sensors values being measured.
created by ZonePAC, and presents it to the occupants. The interface
also allows for change in control of HVAC zone settings. Figure 5
4.2.2 Control Settings
shows the software architecture of our system.
There are two types of control provided to the users - change
4.1 BuildingDepot in temperature setpoint, and change in HVAC occupancy status.
BuildingDepot (BD) is a web service based Building Manage- For each of the zones, a common temperature setpoint is set by
ment System (BMS) designed for next generation Smart Building the BMS. The setpoint is typically set to 72 F, and is modified
applications [3]. The sensor data is stored in a timeseries database if the occupants of the zone register a comfort complaint with the
from different types of sensors in the building. The sensor data and building manager. We allow the users of WebUI to change their
metadata is organized in a uniform structure and exposed for ac- temperature setpoint by 3 F from the preset setpoint. The set-
cess via a RESTful API. The API allows the sensors to be accessed points are allowed to be changed once every 10 minutes per zone.
by its context like location or sensor type, provides for creation of A note at the bottom of the controls section informs the users of the
virtual sensors, and allows customized organization of data using neighbouring rooms with which they are sharing their zone. When
sensorgroups. there are conflicting thermal comfort requirements, we expect the
The BACnet connector is a PC connected to the BACnet network occupants to resolve their differences by direct communication.
as a Foreign Device. The connector continuously polls the data As explained in Section 2, there are three types of occupancy
points from BACnet, and posts it to BD. The connector also writes modes supported by the HVAC system in CSE: Occupied, Standby
to the BACnet points as indicated by BuildingDepot applications and Unoccupied. When a user turns OFF the HVAC using We-
with appropriate permissions. bUI, the occupancy mode is changed to Standby during week-
The HVAC Meter Service (HMS) subscribes to the relevant BAC- day (6am - 10pm), and is changed to Unoccupied on nights and
net points needed for power estimation, and BD notifies HMS as weekends. We chose to use Standby mode during weekdays as
new data is posted from BACnet via a notification url. HMS esti- the zone status is likely to be changed if occupants come in to the
mates the power as outlined in Section 3 and posts the computed zone again, and the shallow setback temperature of Standby will
power back to BD as virtual sensor data. HMS also computes re- reduce the thermal discomfort caused to the occupants. During the
lated useful data such as aggregate heating and cooling power, zone weekends, when the HVAC zone is turned ON, the zone status is
power consumption per unit area, average zone temperature, etc. changed to Occupied mode for two hours. The change in HVAC
status is also restricted to once every 10 mins per zone.
4.2 Web User Interface The HVAC Controller Service (HCS) relays the commands pro-
We implement an interactive webapp on top of BD which reads vided by WebUI to the corresponding BACnet points. The HCS
sensor data from both BACnet and ZonePAC. Interested occupants was designed such that the control service could be made unavail-
register their email address, and WebUI administrators provide per- able to the users if needed without affecting the feedback services
mission to access the sensor information after manual verification. provided by WebUI.
Access control among users is enforced by BD, and users are only
provided information about zones to which they have physical ac- 4.2.3 Energy Saving Suggestions
cess. In order to encourage the users of ZonePAC to save energy, we
include a suggestion box which shows personalized energy saving
4.2.1 Feedback Information recommendations. For example, if the VAV is cooling a zone ex-
The webapp has been designed for access from both desktop and cessively for over an hour, a suggestion is provided to increase the
mobile browsers. Figure 6 shows a snapshot of the desktop ver- setpoint by 0.5 F to save energy. The suggestion box also provides
sion. The most pertinent information such as the room temperature tips for using the WebUI to better control the HVAC system, and
as measured by the thermostat, and the energy consumption as es- encourages the user to learn further about the sensor information
timated by ZonePAC are displayed prominently. Users can provide provided. To provide a comparison with other zones, we display the
feedback on their thermal comfort on a scale of -3 to +3, compli- average temperature and average zone power normalized by zone
ant with ASHRAE Standard 55 [1]. There are 17 different BACnet area as shown in Figure 6.
150
0 0 0
0 100 200 150 200 150 200
Zones Sorted by Power Zones Sorted by Power Zones Sorted by Power
T from Setpoint
Thermal Feedback
30 30 Increase Setpoint
10
Cold 20
Decrease Setpoint OFF
20
HVAC Status
5
10
10 0
0
0 5
10
10 10
20
Tue-Thu Fri Mon Tue Wed Thu Fri Mon Tue Wed Thu Fri 15 Mon Tue Wed Thu Fri
Day of Week Day of Week Day of Week
Figure 9: Feedback and Control using Web User Interface: Figure 9(a) shows the distribution of the thermal feedback received for
10 days of deployment, Figure 9(b) shows the change in setpoints selected by the users for one week of control, and Figure 9(c) shows
the HVAC status commands sent by the users for the control week.
analysis can be done using established simulation engines such as
2.5 Visualization Only
EnergyPlus [7] and DOE-2 [12]. Building models are built using
Total Energy (MWh)
the simulation tools, and the HVAC system is tuned based on the
2.0 w/Control results of such analyses. The methodology is followed for both
w/Control,Suggestion design of new systems [5], as well as existing buildings [6].
1.5 Continuous commissioning [20] and automated Fault Diagnosis
and Detection (FDD) [16] have been proposed for monitoring of
1.0 HVAC systems using sensors and BMS. Mills et al. [20] report 16%
median energy savings in existing buildings due to commissioning,
0.5 and the savings were accrued due to faults corrected in all parts
of HVAC system [21]. FDD methods have also advanced over the
0.0 Fri Mon Tue Wed Thu Fri
years from practical decision based rules [23], system models [18]
Day of Week to data driven approaches [9]. However, the commissioning and
energy information systems developed are designed for domain ex-
perts, and no feedback is provided to the building occupants. More-
Figure 11: Average energy consumption of zones which were over, energy wastage due to behavorial faults such as anomalous
part of the feedback experiment thermostat settings remain unchecked. ZonePAC provides visibil-
buildings which do not use a centralized plant and use humidity ity into energy consumption of each zone, and the opportunity to
control equipments. From the general principles presented in this detect behavioral faults using modern FDD methods. By providing
paper, it is possible to estimate zone power consumption in differ- feedback directly to occupants, ZonePAC also provides the oppor-
ent types of HVAC systems using the measurement from installed tunity for the behavorial faults to be self-corrected.
sensors and modeling the heat transfer process. Prior work has shown that energy feedback can be effective in
Occupancy based HVAC control has been proposed for signifi- motivating users to save energy [4, 8]. In a energy conservation
cant improvement in energy efficiency of buildings [10, 24]. ZonePAC study, Peterson et al. show that motivated occupants saved 20%
provides insight into a variety of situations in which occupancy in- more energy when given feedback on energy consumption in a
formation would be useful for saving energy (Section 5.1). With college dormitory [22]. Recognizing the importance of feedback,
power consumption information at the zone level, researchers would plug meters have been developed to provide feedback on appliance
be able to design more optimized solutions that would exploit the power consumption [15]. However, to the best of our knowledge,
inefficiencies in current HVAC systems. ZonePAC is the first attempt to provide feedback on HVAC energy
Feedback from occupants using ZonePAC showed that they care consumption to building occupants.
about their energy footprint on the building (Section 5.2). Although Prior work has given web based feedback on HVAC system to
our WebUI provides the information about HVAC and zone power the building occupants. Krioukov et al. [17] build a personalized
consumption in a clear manner, it does not adopt sustainable HCI control system, allowing occupants to view the current status of the
solutions such as use of social network [19] or providing compari- system and change settings. Erickson et al. [11] and Jazizadeh et
son against other zones [22]. We hope ZonePAC acts as a stepping al. [13, 14] gather thermal comfort feedback from occupants, and
stone to develop better feedback interfaces so that occupants are change the HVAC settings to match their thermal needs. Unlike
incentivized to save energy. ZonePAC, none of the systems provide energy feedback to the oc-
Several parameters used for estimating HVAC zone power re- cupants. Erickson et al. [11] do estimate zone energy consumption
quired careful study of building plans. Although the information using heat transfer equation, but do not validate its accuracy and do
from sensors installed in the building were readily available through not account for electrical power consumption. ZonePAC provides
BACnet, the details about the type of VAV box, the size of the air occupants with similar web based HVAC information and includes
ducts and water pipes are not provided in a manner that could be the estimated zone power consumption.
easily used for developing applications such as ZonePAC. Auto-
mated methods to scan the existing building plans and extraction of 8. CONCLUSION
relevant information to form a building model would significantly
We have built ZonePAC, a real-time HVAC zone power estima-
accelerate development of smart building applications.
tion system, built on top of a RESTful web service. We present
the trends in zone energy consumption, and provide insights into
7. RELATED WORK improving the energy efficiency of HVAC system. We find that
HVAC power estimation is well understood, and detailed energy the usage characteristics of a zone such as aberrant thermostat set-
tings and presence of cooling demanding equipment can lead to sensing for end user driven control. In Proc. of the 4th ACM
significant wastage of energy. Further, we designed and deployed Workshop on BuildSys. ACM, 2012.
an interactive webapp which provides HVAC sensor information, [14] F. Jazizadeh, F. M. Marin, and B. Becerik-Gerber. A thermal
zone power consumption and control of local HVAC settings to the preference scale for personalized comfort profile
occupants of the building. We present the data collected from the identification via participatory sensing. Building and
feedback study over a period of 10 days, and show that HVAC en- Environment, 2013.
ergy feedback to the occupants in commercial buildings could be [15] X. Jiang, S. Dawson-Haggerty, P. Dutta, and D. Culler.
used to motivate them to save energy. Design and Implementation of a High-Fidelity AC Metering
Network. In IPSN 2009, pages 253264. IEEE, 2009.
9. ACKNOWLEDGMENTS [16] S. Katipamula and M. R. Brambley. Review Article:
We would like to thank Anna Levitt and Michelle Perez from Methods for Fault Detection, Diagnostics, and Prognostics
the UCSD Facilities Management who have helped us get access to for Building Systems-A Review, Part I. HVAC&R Research,
the building HVAC infrastructure. We would like to thank Thomas 11(1):325, 2005.
Weng for providing access to power measurements and the occu- [17] A. Krioukov, S. Dawson-Haggerty, L. Lee, O. Rehmane, and
pants of Computer Science and Engineering building for partici- D. Culler. A Living Laboratory Study in Personalized
pating in the experiments. We also wish to thank our anonymous Automated Lighting Controls. In Proc. of the 3rd ACM
reviewers for their feedback to help improve the paper. This work Workshop on BuildSys. ACM, 2011.
was supported in part by NSF grant SHF-1018632 and a UC San [18] J. Liang and R. Du. Model-based fault detection and
Diego facilities grant for smart building deployment. diagnosis of HVAC systems using support vector machine
method. International Journal of refrigeration,
10. REFERENCES 30(6):11041114, 2007.
[1] Standard 55-2004, Thermal Environmental Conditions for [19] J. Mankoff, D. Matthews, S. R. Fussell, and M. Johnson.
Human Occupancy, Atlanta: American Society of Heating, Leveraging social networks to motivate individuals to reduce
Refrigerating, and Air-conditioning Engineers(ASHRAE). their ecological footprints. In System Sciences, 2007. HICSS
Inc., USA, 2004. 2007. 40th Annual Hawaii International Conference on,
[2] Buildings Energy Databook(DoE). Energy Efficiency & pages 8787. IEEE, 2007.
Renewable Energy Department, 2011. [20] E. Mills. Building commissioning: A golden opportunity for
[3] Y. Agarwal, R. Gupta, D. Komaki, and T. Weng. reducing energy costs and greenhouse-gas emissions.
BuildingDepot: An Extensible and Distributed Architecture Technical report, Ernest Orlando Lawrence Berkeley
for Building Data Storage, Access and Sharing. In Proc. of National Laboratory, Berkeley, CA (US), 2009.
the 4th ACM Workshop on BuildSys. ACM, 2012. [21] E. Mills and P. Mathew. Monitoring Based Commissioning:
[4] H. Allcott. Social norms and energy conservation. Journal of Benchmarking Analysis of 24 UC/CSU/IOU Projects.
Public Economics, 95(9):10821095, 2011. Technical report, Ernest Orlando Lawrence Berkeley
[5] W. Bernal, M. Behl, T. X. Nghiem, and R. Mangharam. National Laboratory, Berkeley, CA (US), 2009.
Mle+: a tool for integrated design and deployment of energy [22] J. E. Petersen, V. Shunturov, K. Janda, G. Platt, and
efficient building controls. In Proc. of the 4th ACM Workshop K. Weinberger. Dormitory residents reduce electricity
on BuildSys. ACM, 2012. consumption when exposed to real-time visual feedback and
[6] D. E. Claridge. Using simulation models for building incentives. International Journal of Sustainability in Higher
commissioning. In International Conference on Enhanced Education, 8(1):1633, 2007.
Building Operation, Energy Systems Laboratory, Texas A&M [23] J. Schein, S. T. Bushby, N. S. Castro, and J. M. House. A
University, 2004. rule-based fault detection method for air handling units.
[7] D. B. Crawley, L. K. Lawrie, F. C. Winkelmann, W. F. Buhl, Energy and Buildings, 38(12):14851492, 2006.
Y. J. Huang, C. O. Pedersen, R. K. Strand, R. J. Liesen, D. E. [24] J. Zhang, G. Liu, R. Lutes, and M. R. Brambley. Energy
Fisher, M. J. Witte, et al. EnergyPlus: Creating a Savings for Occupancy-Based Control (OBC) of
New-generation Building Energy Simulation Program. Variable-Air-Volume (VAV) Systems. Technical report,
Energy and Buildings, 33(4):319331, 2001. Pacific Northwest National Laboratory (PNNL), Richland,
[8] S. Darby. The effectiveness of feedback on energy WA (US), 2013.
consumption. A Review for DEFRA of the Literature on
Metering, Billing and direct Displays, 486:2006, 2006.
[9] Z. Du, X. Jin, and Y. Yang. Fault diagnosis for temperature,
flow rate and pressure sensors in VAV systems using wavelet
neural network. Applied energy, 86(9):16241631, 2009.
[10] V. Erickson, S. Achleitner, and A. Cerpa. POEM:
Power-Efficient Occupancy-Based Energy Management
System. In IPSN 2011. IEEE, 2013.
[11] V. L. Erickson and A. E. Cerpa. Thermovote: participatory
sensing for efficient building HVAC conditioning. In Proc. of
the 4th ACM Workshop on BuildSys. ACM, 2012.
[12] J. J. Hirsch et al. Doe 2.2 building energy use and cost
analysis program. See http://www. doe2. com, 2003.
[13] F. Jazizadeh and B. Becerik-Gerber. Toward adaptive
comfort management in office buildings using participatory