You are on page 1of 8

1088 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 21, NO.

3, AUGUST 2006

Consideration of Input Parameter Uncertainties in


Load Flow Solution of Three-Phase Unbalanced
Radial Distribution System
Biswarup Das, Member, IEEE

Abstract—In this paper, a technique based on interval arith- suggested in [13]. Theoretical aspects of three-phase distribu-
metic is presented for considering the uncertainties of the input tion load flow solution have been studied in [14]. An adaptive
parameters in the power flow solution of three-phase unbalanced power flow method with improved convergence characteristics
radial distribution systems. The uncertainties in both the load
demand and the feeder parameters have been considered. The has been introduced in [15]. A method for incorporating trans-
results obtained from an interval arithmetic-based power flow former nodal admittance matrices into the backward/forward
solution have been compared with those obtained from repeated sweep algorithm has been described in [16].
load flow simulations. In all the above works, the analyses have been carried out as-
Index Terms—Interval arithmetic, power flow, three-phase dis- suming the input quantities (loads at different buses and feeder
tribution system. parameters) are known and fixed. However, in real-life situa-
tions, the values of these input quantities may contain a signif-
icant amount of uncertainties. These uncertainties might occur
I. INTRODUCTION due to: 1) error in the calculation or measurement of the feeder
parameters and 2) error in the metered, calculated, or forecasted
ECAUSE of its application in many distribution system de-
B cision algorithms such as network planning, volt/var con-
trol, service restoration, feeder reconfiguration, state estima-
values of the demands in the system load buses. In [17], use
of interval arithmetic has been first proposed to incorporate the
uncertainties into the power flow solution of a transmission net-
tion, etc., distribution system power flow analysis is nowadays work. In this work, the authors have used a small five-bus trans-
an integral component of distribution system planning, opera- mission network for illustration.
tion, and control functions. Now, due to various reasons, such Motivated by the work of [17], this paper proposes to apply
as unbalanced consumer loads, presence of single, double, and the interval arithmetic to the power flow algorithm of a three-
three-phase line sections, existence of asymmetrical line sec- phase unbalanced radial power distribution system to account
tions etc., present-day distribution systems are primarily un- for the uncertain input quantities. Basically, in this approach, the
balanced in nature. As a result, for reliable and accurate solu- uncertain input quantities are represented as interval numbers
tions, three-phase load flow study of the distribution systems (instead of fixed numbers), and subsequently, interval arithmetic
is required. To cater to this need, substantial effort has been is used to compute the power flow solution of the distribution
devoted in the literature for developing efficient and accurate system. This paper is organized as follows. In Section II, the
three-phase load flow algorithms for distribution systems. In the fundamental concepts of interval arithmetic used in this paper
early endeavors, a direct solution approach using the impedance are discussed. The algorithm for power flow analysis of the
matrix of the unbalanced network [1] and the Gauss ap- three-phase unbalanced radial distribution system using interval
proach [2] have been suggested. Subsequently, different other analysis is described in Section III. Numerical results obtained
techniques, such as the backward/forward sweep algorithm [3], for different cases of input parameter uncertainties are presented
the three-phase fast decoupled power flow algorithm [4], [5], in Section IV. Lastly, Section V concludes this paper.
the rectangular Newton–Raphson-based method and its fast de-
coupled version [6], the current injection method [7], the phase II. INTERVAL ARITHMETIC
decoupled method [8], etc., have also been developed. Another The following notations have been used throughout the paper
direct approach, which utilizes two matrices developed from the to describe the fundamental concepts of interval arithmetic and
topological characteristics of the distribution system, has been application of it to the three-phase radial distribution system
introduced in [9]. In a series of papers, Chen et al. have de- power flow algorithm.
veloped the models for three-phase co-generators, transformers, A lowercase letter, such as , denotes a scalar (real) number.
and loads necessary for three-phase distribution load flow anal- A lowercase letter with subscript “I,” such as , denotes a real
ysis [10]–[12]. Models for voltage control devices have been interval number.
With these notations, the basic concepts of interval arithmetic
Manuscript received July 5, 2005; revised December 8, 2005. Paper no. are described below.
TPWRS-00399-2005. An interval number is the set of real numbers
The author is with the Department of Electrical Engineering, Indian Institute
of Technology, Roorkee 24 667, India (e-mail: biswafee@iitr.ernet.in). such that . and are known as the lower
Digital Object Identifier 10.1109/TPWRS.2006.876698 limit and upper limit of the interval number, respectively. It is
0885-8950/$20.00 © 2006 IEEE
DAS: CONSIDERATION OF INPUT PARAMETER UNCERTAINTIES IN LOAD FLOW SOLUTION 1089

to be noted that and are scalar (real) numbers individually.


A rational number is represented as an interval number
. The addition, subtraction, multiplication, and division of
two interval numbers and are defined
as follows [18]:

(1)
(2)

(3)
(4)
where
Fig. 1. Typical radial distribution system.

(5)
quently, the complex interval arithmetic has been used to com-
The distance between two interval numbers is defined as [18] pute the power flow solution.
The basic feeder model used in this paper is same as that
depicted in [3, Fig. 1] and hence is not again shown in this paper.
(6) The impedance matrix of a feeder section between nodes and
is given by [3]
A complex number , where “ ” is the complex
operator, is defined as a complex interval number if both its real
and imaginary parts are interval numbers. The complex conju-
gate of , denoted as , is given by . For any two (11)
complex interval numbers and ,
the addition, subtraction, multiplication, and division operations
are defined as [18] If any particular phase of this feeder section does not exist,
then the elements in the corresponding row and column of
(7) this matrix would all be zero. Now, when no uncertainties in
(8) the feeder parameters are involved, all the matrix elements
etc. in (11) are fixed complex quantities. However,
(9) in the presence of uncertainties, each of these elements would
be represented by complex interval quantities.
(10) Fig. 1 illustrates a typical radial distribution system with
where and buses and branches. For the purpose of
. power flow analysis, the voltage of the root node is as-
It is to be noted that the expressions in (7)–(10) can be eval- sumed to be known, and a flat voltage profile (equal to the
uated using the fundamental relationships given in equations voltage of the root node) has been assumed for the initial
(1)–(4). voltages of all the other nodes of the network. Thus, in a
An interval vector is a vector whose elements are all interval per unit system, the voltage of the root node and the ini-
numbers, and the elements of a complex interval vector are all tial voltages of the other nodes have been assumed to be
complex interval numbers. Similarly, the elements of an interval p.u., p.u.,
matrix and a complex interval matrix are interval numbers and and p.u. for phases a, b, and c,
complex interval numbers, respectively. The addition, subtrac- respectively. With these initial voltages, the following steps are
tion, and multiplication operations of two complex interval vec- executed for iterative solution of the distribution system.
tors (or matrices) obey the same corresponding rules for addi- Step 1: At iteration “k,” the three-phase nodal current injec-
tion, subtraction, and multiplication of two complex (non-in- tions at node are calculated as
terval) vectors (or matrices), and the resulting expressions can
be evaluated by using (1)–(4) and (7)–(10).

III. INTERVAL ARITHMETIC-BASED POWER FLOW ANALYSIS


The basic power flow method used in this paper is essen-
tially the backward/forward sweep algorithm described in [3].
However, to account for the uncertainties, the input parameters
(real and reactive power loads at the buses as well as the feeder (12)
parameters) have been treated as interval numbers, and subse-
1090 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 21, NO. 3, AUGUST 2006

where, in (12), is the complex interval cur- tion systems that have been obtained after making certain sim-
rent for phase a (b, c) at node corresponding to the th plifications in the original IEEE 13-bus, 34-bus, and 123-bus
iteration, is the constant, pre-specified, radial distribution test feeders [19]. The simplifications adopted
complex interval injected power at phase a (b, c) of node in this paper are essentially the same simplifications followed
is the complex interval voltage in [20]. However, for the sake of completeness, these simplifi-
of phase a (b, c) of node at the th iteration, and cations are mentioned also in this paper as follows.
is the total complex interval shunt admittance 1) The transformers, voltage regulators. and the switches are
connected at phase a (b, c) of node . omitted, and hence, the corresponding feeder sections and
Step 2: At this step, known as the backward sweep, the cur- nodes are deleted.
rents in each branch are calculated starting from the feeder sec- 2) The distributed load along any feeder section is lumped and
tion in the last layer and progressively moving toward the root allocated equally between the two terminal nodes of that
node. Thus, with reference to Fig. 1, at iteration “k,” the com- particular feeder section.
plex interval current at branch “L” is calculated as 3) All the spot loads have been assumed to be constant PQ
load and star connected.
With these three modifications in place, the resulting loading
patterns and feeder parameters in any of these three test sys-
(13)
tems are henceforth termed as “base loading pattern” and “base
feeder parameter,” respectively, of that particular test system.
where, in (13), is the complex interval current For all these three test systems, three different cases have been
for phase a (b, c) flowing through feeder section at the th considered in this paper: 1) uncertainties only in the load pa-
iteration, and is the set of feeder section connected to node rameters, 2) uncertainties only in the feeder parameters, and 3)
. uncertainties both in the load and feeder parameters. However,
Step 3: In this step, known as the forward sweep, starting due to space limitation, the results obtained only in the IEEE
from the root node and progressively moving toward the last 123-bus system are shown in this paper.
layer, the node voltages are updated from the knowledge of the
latest updated voltages of the previous nodes. Thus, with refer- A. Uncertainties in Load Parameters Only
ence to Fig. 1, at iteration “k,” the voltages at the node are In this case, the feeder parameters have been kept fixed at their
calculated from the knowledge of the voltages of the node as corresponding “base feeder parameters,” and uncertainties are
assumed to be present only in the load parameters. As already
mentioned in Section I, the uncertainties in the load parame-
ters have been taken into account by assuming that the loads
are varying over a certain range or interval. For this purpose,
it has been assumed that for each phase of each node, the un-
(14) certainties in its real and reactive loading values are limited to
% variation with respect to the corresponding values at the
where, in (14), is the complex interval “base loading pattern.” Thus, for phase a of node , the real
voltage of phase a (b, c) of node at the th iteration, and and reactive loads are assumed to vary over the intervals of
etc. are the complex interval elements of the KW and KVAR, respec-
impedance matrix of the feeder section . tively, where and are the real and reactive power loads,
Step 4: For each node , the distance between its respectively, of phase a of node at the “base loading pattern.”
three-phase voltages of present iteration and those of pre- Similarly, for the remaining two phases of node and for all
vious iteration, henceforth denoted by , is calculated at the three phases of all the other nodes, the intervals of variation
each iteration “k” by the following procedure. For each of load demands have been decided.
phase, e.g., phase a, the voltages and are com- When the load demands in a system vary within some inter-
plex interval numbers, and hence, they can be expressed as vals, the bus voltages and feeder power flows (for all the three
and . Subsequently, phases) also vary within certain intervals. These intervals have
the distance between and is calculated as been calculated by performing load flow study using the algo-
. Similarly, the quantities rithm described in Section III. For further reference, this algo-
(distance between and and (distance between rithm would henceforth be termed as “interval load flow method
and are also calculated. Finally, is calculated (ILFM).” For this purpose, the voltage error tolerance limit has
as . If been chosen as 0.000001 p.u. in this paper, and the algorithm
(“n” is the number of buses in the system) is less than a specified took nine iterations to converge.
tolerance limit, the load flow is considered to be converged; For comparison purpose, the intervals of variations of bus
otherwise, the algorithm goes back to step 1. voltages and feeder power flows for all the three phases have
also been calculated by repeated power flow simulations
IV. SIMULATION RESULTS AND DISCUSSION (RPFS). In this method, the load demands for all the three
The proposed algorithm described in the previous section has phases at any bus have been fixed at some arbitrary value
been applied to three different three-phase radial test distribu- within their corresponding, pre-specified intervals (the intervals
DAS: CONSIDERATION OF INPUT PARAMETER UNCERTAINTIES IN LOAD FLOW SOLUTION 1091

Fig. 3. Phase-b voltage profile for load uncertainties only. (a) Magnitude.
(b) Angle (degree).

Fig. 2. Phase-a voltage profile for load uncertainties only. (a) Magnitude.
(b) Angle (degree).

already adopted in ILFM). Similarly, the real and reactive load


demands for all the three phases at all the other nodes have
also been specified at some arbitrary values within their corre-
sponding, pre-specified intervals. Thus, a loading pattern other
than the “base loading pattern” has been generated, and normal
load flow solution has been carried out for this loading pattern
by the method of [3]. Similarly, by fixing the load demands at
other arbitrary values within their corresponding pre-specified
intervals, many other loading conditions for the given system
can be generated, and the corresponding load flow solution can
be obtained. Theoretically, by this method, an infinite number
of loading conditions can be generated. As it is not possible to
perform load flow studies for an infinite number of loading pat-
Fig. 4. Phase-c voltage profile for load uncertainties only. (a) Magnitude.
terns, in this paper, a very large number of loading conditions, (b) Angle (degree).
in the range of lacs, has been generated, and the corresponding
load flow solutions have been computed. The minimum and
maximum values of the real and imaginary parts of the complex
ILFM always suggests little wider intervals than RPFS. How-
bus voltages (for all the three phases) can be obtained from the
ever, from these figures, it can be observed that the intervals ob-
results of these 1 million load flow solutions, and these consti-
tained by these two techniques are actually quite close to each
tute the intervals of variations of the bus voltages obtained from
other. In all these results, the differences in the values suggested
RPFS. Similarly, the minimum and maximum values of the real
by these two methods start after the decimal point only. The
and reactive power flow in the feeders can be determined, and
same pattern has also been observed with the interval of varia-
these would constitute the intervals of variations of the feeder
tions of feeder power flows (real and reactive), and hence, these
power flows obtained with RPFS. It is to be noted that, even by
plots are not given in this paper.
increasing the number of operating points further, the intervals
of variations of bus voltages and feeder power flow (real and
B. Uncertainties in Feeder Parameters Only
reactive) do not change appreciably.
The intervals of variations of the phase voltages obtained by In this case, the load parameters have been kept fixed at their
both the methods are shown graphically in Figs. 2–4. In these respective values corresponding to the “base loading condi-
figures, the legends “uin” and “lin” denote the upper and lower tion,” and uncertainties are assumed to be present only in the
limits of the voltages, respectively, obtained by ILFM. Similarly, feeder parameters. The uncertainties in the feeder parameters
the legends “urp” and “lrp” denote the upper and lower limits of have been taken into account by assuming that the feeder
the voltages, respectively, obtained by RPFS. parameters are varying over a certain range or interval. For
From these figures, an interesting observation can be made. this purpose, it has been assumed that for each phase of each
The intervals obtained by RPFS are always contained within the feeder, the uncertainties in its parameters values are limited
intervals depicted by ILFM. In other words, the solutions ob- to % variation with respect to the corresponding values
tained by ILFM contain all the solutions given by RPFS. Thus, w.r.t. its “base feeder parameter.” For example, for phase a of
1092 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 21, NO. 3, AUGUST 2006

Fig. 5. Phase-a voltage profile for feeder parameter uncertainties only. (a) Mag- Fig. 7. Phase-c voltage profile for feeder parameter uncertainties only. (a) Mag-
nitude. (b) Angle (degree). nitude. (b) Angle (degree).

Fig. 8. Phase-a voltage profile for both feeder parameter and load parameter
uncertainties. (a) Magnitude. (b) Angle (degree).
Fig. 6. Phase-b voltage profile for feeder parameter uncertainties only.
(a) Magnitude. (b) Angle (degree).

intervals obtained by ILFM, and the intervals obtained by these


feeder , the resistance (reactance) values are assumed to vary two methods are quite close to each other.
over intervals of ohm ohm),
C. Uncertainties in Both Feeder Parameter and
respectively, where are the resistance (reactance) of
Load Parameter
phase a of feeder corresponding to the “base feeder param-
eter.” Similarly, for the remaining two phases of feeder l and In this case, it is assumed that the uncertainties are present
for all the three phases of all the other feeders, the intervals both in the feeder parameters and load parameters simultane-
of variation of feeder parameters have been decided. Subse- ously. For this purpose, similar to the cases considered in parts
quently, with these interval feeder parameters, ILFM has been A and B, the uncertainties both in the load parameters and the
carried out. For this case also, the voltage error tolerance limit feeder parameters have been assumed to be limited to %
has been chosen as 0.000001 p.u., and the algorithm took nine variation with respect to the corresponding values at “base
iterations to converge. Moreover, following a procedure similar loading condition” and “base feeder parameter,” respectively.
to that described in part A of this section, a large number of For a voltage limit tolerance limit of 0.000001 p.u., the ILFM
operating points (in the range of lacs) has been created by method took ten iterations to converge in this case. Again, sim-
randomly varying the feeder parameters within the intervals ilar to the studies carried out in parts A and B, RPFS has been
used in ILFM, and RPFS has been carried out for these 1 carried out for a large number of (in the range of lacs) operating
million operating points. points, which had been created by randomly varying the load
The results obtained by these two methods are shown in parameters and the feeder parameters within the intervals used
Figs. 5–7. in ILFM.
These results also affirm the observations of Figs. 2–4, that The results obtained by these two methods are shown in
is, the intervals obtained by RPFS are always contained in the Figs. 8–10.
DAS: CONSIDERATION OF INPUT PARAMETER UNCERTAINTIES IN LOAD FLOW SOLUTION 1093

used in the initial stages effectively to save on the time, effort,


and resources required. However, before practical implementa-
tion, several related issues need to be considered.
1) In this paper, as mentioned earlier, the transformers have
been omitted. However, their inclusion in this method is
quite straightforward. As any three-phase transformer can
be represented as an equivalent 3 3 impedance matrix,
it can be included in the interval calculation in the same
manner as shown in (14).
2) The developed method can also easily accommodate the
model of distributed generators. In most countries, ac-
cording to the grid codes, any distributed generator (DG)
is generally required to supply a contracted amount of real
and reactive power to the local distribution grid. Therefore,
for modeling purposes, any such DG can be represented as
Fig. 9. Phase-b voltage profile for both feeder parameter and load parameter
uncertainties. (a) Magnitude. (b) Angle (degree).
a constant PQ source (or as a constant negative PQ load),
and its effect can be included in the interval calculation in
the same manner as in (12) and (13).
3) The proposed interval arithmetic-based method can take
care of any load characteristics. Essentially, in a given time
period, the minimum and maximum power consumed by
any load is governed by its characteristics. As the proposed
method needs only these two extreme (minimum and max-
imum) values for calculation, it can be said that the pro-
posed method incorporates the load characteristics in the
load flow computation implicitly.
4) In this paper, only 10% variation in the parameters has
been taken for illustration purposes. However, the devel-
oped method is generalized enough to be able to handle any
amount of uncertainties that might occur in the distribution
system. Now, the proposed method does not take into ac-
count the diversity of the loads. As a result, it can be ar-
Fig. 10. Phase-c voltage profile for both feeder parameter and load parameter
gued that use of intervals for modeling different loads may
uncertainties. (a) Magnitude. (b) Angle (degree). lead to an over-conservative design. Indeed, if the max-
imum and minimum power consumptions (during the 24-h
period) of every load are utilized in the interval calcula-
Again, these results affirm the observations of Figs. 2–7, that tion, then the results would surely be over-conservative.
is, although the intervals obtained by these two methods are However, if the time window (for determining the max-
quite close to each other, the intervals obtained by ILFM always imum and minimum power consumptions) is shortened de-
contain the intervals obtained by RPFS. pending upon the diversity of loads, then the over-conser-
The above observation has two important implications for op- vative nature of the results would reduce to a large extent.
erational and planning studies of a distribution system. First, A simple, hypothetical example would probably help to il-
while considering uncertainties, ILFM produces the most pes- lustrate this point.
simistic results, although these most pessimistic results are quite Let, in a given power distribution system, the loads be
close to the results obtained by RPFS. grouped into three distinct classes, and also let these three
Therefore, any strategy adopted for improving the system per- different classes follow distinct load curves (the peak loads
formance, taken on the basis of the results obtained from ILFM, of these three load curves occur at different periods during
would be able to handle all possible scenarios in the distribution the 24-h period). Also assume that the peak loads for these
network arising out of the uncertainties in the system parame- three classes occur during the periods of 09–12 hours,
ters. Second, in a planning and design studies, where a large 13–16 hours, and 17–20 hours, respectively. During the
number of operating conditions need to be considered, ILFM remaining periods (other than these three peak loading
can be used effectively. As ILFM encompasses all the solutions periods), the maximum and minimum loading of all the
of RPFS, in the initial stages (of planning or design studies), use three classes follow more or less the same pattern. Table I
of ILFM instead of RPFS (for obtaining the outer bounds of all shows the maximum and minimum real power loads (in
possible situations) can save a lot of time, effort, and resources. KW) of the three classes, depending upon the time periods.
Although in the final stage, after converging to a particular plan- Now, with reference to the above table, if the interval
ning or design strategy, repeated load flow simulations may still values of loading are taken covering the entire 24-h pe-
be necessary before reaching the ultimate decision, ILFM can be riod, then, for these three classes, the interval loading
1094 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 21, NO. 3, AUGUST 2006

TABLE I 6) In modern power distribution systems, different voltage


LOADING IN THE HYPOTHETICAL SYSTEM control devices, such as voltage regulators, etc., are
placed at strategic locations to improve/control the overall
voltage profile of the system. Therefore, the models of
these voltage control devices also need to be considered
in the interval arithmetic-based method. However, in the
course of our work, it was found that the convergence
property of the basic interval arithmetic-based method
described in this paper deteriorates substantially when the
models of the voltage control devices are incorporated in
values would be [25 500] KW, [30 400] KW, and [40 600] the calculation. Therefore, it is felt that advanced interval
KW, respectively. Let these three intervals be termed as calculation techniques need to be used for accommodating
“global” intervals. If these three “global” interval quanti- the control device models. We are working on it presently,
ties are used for load flow calculation, the results indeed and we will report it as soon as we get acceptable results
would be over-conservative. However, if four different in a separate paper.
time windows are chosen (corresponding to the first
column of Table I), then each row of this table would V. CONCLUSION
give the corresponding values of the interval loading for In this paper, a method for considering the uncertainties of the
the three different classes. As the intervals represented in input parameters in the power flow solution of three-phase un-
each row are quite narrow in comparison with the “global” balanced radial distribution systems has been presented. Based
intervals, the corresponding interval power flow solutions on interval arithmetic, the proposed methodology can consider
would also be quite narrow in comparison with the power the uncertainties in both the load demand and the feeder parame-
flow solution obtained with the “global” intervals. Of ters successfully. The solutions obtained from the interval arith-
course, in this case, four different interval load flow solu- metic-based power flow method encompass all the solutions ob-
tions need to be obtained, but this is a small price to pay tained from repeated load flow simulations. Consequently, any
for reducing the width of the solution intervals. strategy for system improvement taken on the basis of interval
The above example has been set up to be extremely simple power flow study would hopefully be effective over all the pos-
just to illustrate the main idea. The actual load variations sible situations. Moreover, in the initial stages of planning and
as well as the load curves in any practical power distribu- design studies, the proposed technique can be a useful tool to
tion system would be much more complex (as compared save on the time, effort, and resources required.
to the hypothetical case presented in Table I). However,
following the same principle as presented above, the time REFERENCES
windows can be chosen properly to narrow down the in- [1] S. K. Goswami and S. K. Basu, “Direct solution of distribution sys-
tervals of the load variations. By this process, one may tems,” Proc. Inst. Elect. Eng., Gen., Transm., Distrib., vol. 138, no. 1,
have to compute several interval load flow computations, pp. 78–88, Jan. 1991.
[2] T. H. Chen, “Distribution system power flow analysis—a rigid ap-
but this number (of load flow computations) would surely proach,” IEEE Trans. Power Del., vol. 6, no. 3, pp. 1146–1152, Jul.
be much less than the number of repeated load flow calcu- 1991.
lations needed to obtain the intervals of variations of dif- [3] C. S. Cheng and D. Shirmohammadi, “A three phase power flow
method for real time distribution system analysis,” IEEE Trans. Power
ferent quantities. Syst., vol. 10, no. 2, pp. 671–679, May 1995.
5) In the literature, fuzzy logic has been proposed to deal with [4] R. D. Zimmerman and H. D. Chiang, “Fast decoupled power flow for
the uncertainties, although to date, to the best of the knowl- unbalanced radial distribution system,” IEEE Trans. Power Syst., vol.
10, no. 4, pp. 2045–2052, Nov. 1995.
edge of the author, no paper has used fuzzy logic to com- [5] A. V. Garcia and M. G. Zago, “Three phase fast decoupled power flow
pute the three-phase power flow analysis of the radial dis- for distribution networks,” Proc. Inst. Elect. Eng., Gen., Transm., Dis-
tribution system. Now, in fuzzy set, each uncertain vari- trib., vol. 143, no. 2, pp. 188–192, Mar. 1996.
[6] W. M. Lin, “Three phase unbalanced distribution power flow solution
able needs to be defined in terms of a suitable member- with minimum data preparation,” IEEE Trans. Power Syst., vol. 14, no.
ship function within its corresponding maximum and min- 3, pp. 1178–1183, Aug. 1999.
imum values. Generally, the choice of the fuzzy member- [7] P. A. N. Garcia, “Three phase power flow calculations using the cur-
rent injection method,” IEEE Trans. Power Syst., vol. 15, no. 2, pp.
ship function is often subjective as there is no concrete al- 508–514, May 2000.
gorithm (or guideline) for choosing the membership func- [8] J. C. M. Vieira, Jr., W. Freitas, and A. Morelato, “Phase decoupled
tion. As a result, depending upon the choice of fuzzy mem- method for three phase power flow analysis of unbalanced distribution
system,” Proc. Inst. Elect. Eng., Gen., Transm., Distrib., vol. 151, no.
bership functions (even within the same maximum and 5, pp. 568–574, Sep. 2004.
minimum values), different solutions can be obtained for [9] J. H. Teng, “A direct approach for distribution system load flow solu-
the power flow analysis problem. On the other hand, the tion,” IEEE Trans. Power Del., vol. 18, no. 3, pp. 882–887, Jul. 2003.
[10] T. H. Chen, “Three phase co-generator and transformer models for dis-
proposed interval arithmetic-based technique does not as- tribution system analysis,” IEEE Trans. Power Del., vol. 6, no. 4, pp.
sume any specific variation of the uncertain quantities (it 1671–1681, Oct. 1991.
only works with the maximum and minimum values) and [11] T. H. Chen and J. D. Chang, “Open Wye open delta and open delta
open delta transformer models for rigorous distribution system anal-
therefore can be considered as more general method as ysis,” Proc. Inst. Elect. Eng., Gen., Transm., Distrib., vol. 139, no. 3,
compared to fuzzy set theory. pp. 227–234, May 1992.
DAS: CONSIDERATION OF INPUT PARAMETER UNCERTAINTIES IN LOAD FLOW SOLUTION 1095

[12] T. H. Chen and Y. L. Liang, “Integrated models of distribution trans- [18] G. Alefeld and J. Herzberger, Introductions to Interval Computa-
formers and their loads for three phase power flow analysis,” IEEE tions. New York: Academic, 1983.
Trans. Power Del., vol. 11, no. 1, pp. 507–513, Jan. 1996. [19] W. H. Kersting, “Radial distribution test feeders,” IEEE Trans. Power
[13] P. A. N. Garcia, J. L. R. Pereira, and S. Carneiro, “Voltage control de- Syst., vol. 6, no. 3, pp. 975–985, Aug. 1991.
vices models for distribution power flow analysis,” IEEE Trans. Power [20] H. Wang and N. N. Schulz, “A revised branch current based distribution
Syst., vol. 16, no. 4, pp. 586–594, Nov. 2001. system state estimation algorithm and meter placement impact,” IEEE
[14] K. N. Miu and H. D. Chiang, “Existence, uniqueness and monotonic Trans. Power Syst., vol. 19, no. 1, pp. 207–213, Feb. 2004.
properties of the feasible power flow solution for radial three phase
distribution network,” IEEE Trans. Circuits Syst. I, Fundam. Theory
Appl., vol. 47, no. 10, pp. 1502–1514, Oct. 2000.
[15] Y. Zhu and K. Tomsovic, “Adaptive power flow method for distribution
systems with dispersed generation,” IEEE Trans. Power Del., vol. 17, Biswarup Das (M’02) received the B.E.E. (Hons.) and M.E. degrees from
no. 3, pp. 822–827, Jul. 2002. Jadavpur University, Calcutta, India, in 1989 and 1991, respectively, and the
[16] Z. Wang, F. Chen, and J. Li, “Implementing transformer nodal admit- Ph.D. degree in electrical engineering from IIT Kanpur, Kanpur, India, in 1998,
tance matrices into backward/forward sweep based power flow analysis with specialization in electric power systems.
for unbalanced radial distribution system,” IEEE Trans. Power Syst., Since 1998, he has been with Department of Electrical Engineering, IIT
vol. 19, no. 4, pp. 1831–1836, Nov. 2004. Roorkee, Roorkee, India, where he is presently an Associate Professor.
[17] Z. Wang and F. L. Alvarado, “Interval arithmetic in power flow anal- His current research interests are in the area of FACTS, distribution automa-
ysis,” IEEE Trans. Power Syst., vol. 7, no. 3, pp. 1341–1349, Aug. 1992. tion, distributed generation, and renewable energy sources.

You might also like