14th PSCC, Sevilla, 24-28 June 2002

Session 03, Paper 2, Page 1

CAPACITOR PLACEMENT IN UNBALANCED POWER SYSTEMS

P. Varilone and G. Carpinelli

A. Abur

Dipartimento di Ingegneria Industriale

Universita degli Studi di Cassino
03043 Cassino, Italy
[email protected], [email protected]

Department of Electrical Engineering

Texas A&M University
College Station, Texas, U.S.A
[email protected]

Abstract This paper presents a capacitor placement

method for three phase unbalanced power systems. The
method aims to minimize not only the power losses and
capacitor costs, but also the distortions due to harmonics
present in the power system. The proposed method is
capable of accomplishing this objective for both balanced
and unbalanced operating conditions. One of the objectives of this paper is to demonstrate the significant differences in the results of capacitor placement studies when
unbalanced systems are approximated by their positive
sequence single phase equivalents. Furthermore, the effects of taking harmonic distortions into account during
the capacitor placement procedure are also demonstrated.
Numerical examples on a distribution test power system
are provided to illustrate the method.

Keywords: Optimization, capacitor placement, loss

minimization, harmonic distortion, unbalanced operation, distribution systems.

1 INTRODUCTION
Power distribution systems contain shunt capacitors
at various strategic locations in order to maintain a
desired voltage profile, correct power factor and reduce
power losses along feeders. When dealing with a large
scale distribution system containing several feeders and
their laterals, deciding on the best locations and sizes of
these capacitors becomes a complicated optimization
problem. In addition to the scale of the problem, there
are other issues such as the discrete nature of capacitor
sizes, operational limits on voltages and feeder loadings, that need to be addressed. Effective solution algorithms for balanced distribution feeders have been developed [1,2]. These solutions mainly utilize the positive sequence network model and the associated power
flows in formulating the problem. Hence, the results do
not directly apply for systems containing feeders with
missing phases, unevenly loaded feeders or shunt capacitors on single or double phase feeders. Three phase
unbalanced distribution systems are later studied in
[3,4] where simulated annealing and genetic algorithms
are respectively used to solve this more complicated
problem. A simplified formulation and the MINOS
optimization package are used to solve the same problem in [5]. Recently, distribution systems are populated
with nonlinear loads or control devices that generate

unwanted harmonics in the systems. Harmonics are

known to cause overvoltages under certain network
configurations involving shunt capacitors. This issue is
raised and a solution is proposed for minimizing harmonic overvoltages in [6]. A practical method that
avoids convergence problems and incorporates discrete
nature of capacitor banks along with the voltage distortions due to the installed capacitors, is developed and
presented in [7]. This method is based on the balanced
three phase operation and therefore analyzes the positive sequence network only.
In this paper, the work reported in [7] will be extended to the more general case of the three phase unbalanced operation. Several distribution feeders are
known to have line sections carrying a mixture of single, double or three phase loads. Such systems as well
as those having full three phase but unbalanced loads
can be studied using the presented method in this paper.
In addition to the cost of losses and capacitor installations, it is possible to associate a cost with the harmonic
distortions as discussed in [8-9]. Hence, the problem is
formulated in such a way that both network losses and
harmonics are minimized along with the cost of capacitors placed for this purpose.
The paper is organized such that the problem description is presented first. This is followed by the
sections describing the details of the three phase power
flow and linear harmonic analysis modules, which make
up the main computational engines of the overall algorithm. Simulation results obtained using the developed
program and a test system will be presented next. Conclusions and suggestions on future work will be presented in the final section.
2 DESCRIPTION OF THE PROBLEM
The objective of the presented method is to determine the best locations and sizes of shunt capacitors for
each phase, so that the total cost of the capacitors, of the
total power losses of the network and the harmonic
distortion of the network are minimized. This objective
should be met subject to the network three-phase power
flow equations as well as the limits on the bus voltage
and current magnitudes, harmonic indices (HI) and the
total number of capacitor units to be installed. Hence, it

14th PSCC, Sevilla, 24-28 June 2002

can be formulated as the following optimization problem:

Minimize

J(X,U) = JC + JL + JIH

Three-phase PF Eq.s at
fundamental and harmonics
V, I limits
HI limits
Limits on number of cap.s
where the terms of the objective function and constraint
list are defined below.
Subject to

2.1 The Objective function, J

The objective function assumes a given loading level
for the system. While it is possible to extend this function to a sum of similar functions, so that a number of
loading levels can be accounted for, this will not be
done here. All simulation results and the discussion of
the proposed algorithm will be based on single loading
assumption. Such an extension can however be incorporated into the presented algorithm without much
difficulty. The three terms that make up the objective
function are described below:
2.1.1

Cost of Capacitors, J C = C CT U

Cc : is the cost vector for capacitor units at each bus.

U : is the vector of capacitor units placed at each bus.
2.1.2

Cost of Losses, J L = C L (PG PD )T

CL: is the cost of a per unit energy loss

PG: is the total generation at (X,U)
PD: is the total load
T : is the loss duration
(X,U): is the power flow solution corresponding to
the installed capacitor vector of U.
The cost of demand lost can be also incorporated
without difficulty.
2.1.3
Cost of Distortions, JIH
The evaluation of the cost of the distortions is studied
first by the authors in [8, 9]. These costs are assumed to
be the sum of the operating costs and the aging costs.
The operating costs refer to the costs of the incremental
losses caused by the harmonics and the aging costs refer
to the incremental costs due to the premature aging of
the components caused by the harmonics. Details can be
found in [8, 9].
2.2 The Constraints
Each solution should satisfy the three-phase power
balance equations at the fundamental frequency. Thus,
a full three-phase power flow solution is to be run to
verify this constraint. This solution will also be used to
check the limit violations at fundamental on bus voltages and line currents. The effect of each capacitor unit
on the distortions of the bus voltages will have to be
also checked. This is accomplished by solving a linear
three-phase harmonic analysis problem where all
nonlinear loads are represented by their harmonic cur-

Session 03, Paper 2, Page 2

rent injections. Details of this computation are discussed in a later section.

3

SOLUTION METHOD

The above described optimization problem is solved

by using a simple yet effective procedure, which is
developed and successfully employed in solving the
single phase problem in [7]. The main idea behind the
proposed procedure is based on sequential placement of
discrete capacitor units of incremental size. It is assumed that each phase of any bus can be placed an
incremental discrete capacitor at each optimization step
(three-phase bank in case of three-phase bus, singlephase bank otherwise). If capacitor installations are
prohibited for any of the buses, then this condition can
easily be enforced by leaving out that bus from the bus
list. Furthermore, if different buses have different sizes
of unit capacitors, then placement of unit capacitors can
be accordingly modified to suit each bus. Therefore,
the physical constraints that relate to the discrete nature
of capacitor installations as well as the nonhomogeneity of the discrete units available for installations at different buses, are naturally accounted for
without any complex logic.
At each step of the optimization procedure, the three
phase power flow solution as well as the harmonics of
the bus voltages corresponding to the incremental addition of unit capacitors at each possible bus, will have to
be computed. Note that, the purpose of these computations is to obtain and compare the values of the objective function for all possible incremental installations
and choose the one that reduces the objective function
the most. It is also noted that for purposes of comparing effects of unit capacitor installations at different
buses on the objective function, approximate threephase power flow solutions can be used. Once a candidate is selected, then an accurate solution can be obtained for the selected configuration only. A fast but
approximate three phase power flow solution is implemented for the first part. A three phase linear harmonic
analysis is carried out for the calculation of distortions.
These are discussed in more detail below.
3.1 Fast Three-phase Power Flow Update
The three-phase power flow solution is affected incrementally each time a unit capacitor is added at a
system bus. This incremental change can be captured
by expressing the power flow solution by its first order
approximation. Consider the three phase power flow
equations given by:
f(X,U) = 0

(1)

The first order Taylor approximation will be:

[Fx (X0,U0) ] [ dX ] = - [ Fu(X0 , U0) ] dui

(2)

14th PSCC, Sevilla, 24-28 June 2002

Session 03, Paper 2, Page 3

Note that dui is the unit capacitor added at node i,

one phase of a bus and the functions Fx and Fu represent the gradient of f(X,U) evaluated with respect to X
and U respectively. At each optimization step, the most
recent operating point will be denoted by (X0,U0) which
will depend upon the way the optimization proceeded
up till that point. The updated power flow solution will
then be obtained as:
X = X0 + dX

start
AC Three Phase
Power Flow

(3)

This procedure will be repeated as many times as the

number of nodes excluding those where no capacitors
are allowed to be placed. Upon computing the objective function J(X,U) for all cases, the solution which
yields the smallest objective function will be selected.
A full three phase AC power flow solution will be obtained and all operating constraints will be checked for
this case. In case the solution violates any one of the
limits, the same procedure will be repeated for the second best solution. This procedure will continue until a
feasible solution is obtained. If no solution can be
found satisfying all the constraints, then the optimization procedure will be terminated. Otherwise, it will
proceed to the next optimization iteration.
3.2 Calculation of distortions
In parallel with the fundamental frequency power
flow solution, a solution for each dominant harmonic
present in the network will have to be obtained so that
the corresponding effects of the power system harmonics can be evaluated. This is accomplished by solving
the networks linear harmonic equations given by:
[Yn] [Vn ] = [In ]

section will present several representative cases which

are simulated using the developed program.

set k=1

Approximate operating point

when a cap. unit is added at bus k
k=k+1
Compute the objective
function J

K > Nbus

Y
Select the capacitor that yields the
minimal J

stop

Capacitor
exists ?
Y
AC Three Phase
Power Flow

stop

Select the next

qualifying bus

Is J reduced ?
Y
Check the constraints

(4)
Satisfied ?

where [Yn], [Vn ] and [In ] are the three phase network
admittance matrix, bus voltage vector and independent
current source vector evaluated at the n-th harmonic
frequency. The admittance matrix is modified each time
a unit capacitor is added at a bus. Also note that, the
harmonic injections of nonlinear devices are modeled
according to the recommendations of the IEEE PES
Working Group on Harmonics Modeling and Simulation [10-11] where it is suggested that the phase angles
of the injected harmonic current sources be adjusted
according to the phase angle of the fundamental with
respect to reference.
A Matlab-based program is developed in order to
evaluate the above described procedure for general
three phase unbalanced system operation. At this time,
the program is implemented for a single loading level
and therefore the capacitor placement can be tested only
for fixed capacitor types. However, it can be easily
modified to account for seasonal or daily load variations
of the bus loads. A flow chart of the procedure implemented by this program is shown in Figure 1. Next

Y
N

Maximal # of
capacitor reached ?
Y
stop

Figure 1: Flowchart of the proposed procedure

4 SIMULATION RESULTS
The simulations are run using the three phase unbalanced IEEE 13-bus test system shown in Figure 2 [11].
This system contains a mixture of single, double and
three phase lines and loads, making it quite suitable for
evaluating the proposed method. Note that the system
lines 684-653 and 684-611 are single phase; lines 671684, 632-645, 645-646 are two phase and the remaining
lines are three phase. The complete network data and
parameters can be found in [11].
The system contains loads that are linear, non linear,
and/or unbalanced. The harmonic producing loads include fluorescent lighting, adjustable speed drives for

14th PSCC, Sevilla, 24-28 June 2002

Session 03, Paper 2, Page 4

heat pumps and non-specific sources such as PCs, TVs,

etc. The load data reported in [11] are reduced by 50%,
since the voltage distortions are found to exceed the
IEEE-519 recommended limits [12] before the placement of any capacitors.
Unit capacitors available at any bus are assumed to
come in discrete sizes of 50 kVAR. The unit cost of
capacitors are reported in [13] while the unit cost of
energy are reported in [14].
The constraint limits are assumed as 105% for the
voltages at fundamental, 3% for the single voltage harmonics and 5% for the total harmonic distortion.
Several cases are simulated. The following results
will be presented as representative cases.
4.1 Case 1
In this case the sensitivity of capacitor placement to
the objective function structure is investigated. The
objective function is first considered as the sum of the
losses and the capacitor costs only. The harmonic costs
are then introduced into the objective function, to observe the significance of this consideration in the outcome of the optimization process. In order to simplify
the analysis, the harmonic index is assumed to be simply the total harmonic distortion and its associated cost
is evaluated so as to have the same cost of losses and
650

function of capacitor bank additions. The objective

functions are normalized to make the comparison easy.
It should be noted that in the case without THD term the
placement procedure is terminated at the fifth step due
to a constraint violation. The violated constraint is the
value of the fundamental voltage at bus #675, which
exceeds the specified limit of 1.05 p.u., when an additional unit capacitor is attempted to be placed after the
fourth step.
Note that, in Fig.3, the solid lines represent the variation of the objective function J until the optimal solution
is reached. The dashed lines indicate how J would have
varied, had the optimization procedure not been terminated. In the top curve, the procedure terminates because the minimum is reached. In the bottom curve, it
is terminated by the constraint violation after step 4 and
no other solution yielding a lower J within the constraints can be found at the fifth step.
4.2 Case 2
In this case the sensitivity of the capacitor placement
procedure to the specified constraints is considered.
Initially, constraints identical to those used in Case 1,
are considered. These limits are subsequently relaxed
(110% for the voltages at fundamental, 5% for the single voltage harmonics and 8% for the total harmonic
distortion), in order to observe their significance on the
Table 1:Capacitor Placement Results of Case 1

646

645

632

633

634

Bus code

675

Phase code

611

684

671

692

675

611
3

Without THD cost 2x50 2x50

(kVAr)

2x50

With THD cost 2x50 2x50

(kVAr)

2x50

3
2x50

J [p.u.]
652

680

Figure 2: Network Diagram for the 13-bus test system

1.05
1
0.95

harmonics in the system without capacitors.

The locations and sizes of the capacitors that yield
the minimum objective function in both cases are listed
in Table 1. As evident from the table, two unit capacitor banks are located at the most distant three phase load
bus (note that bus #680 is unloaded) and two other unit
banks are located at a single phase bus. On the other
hand, the capacitor banks at the single phase bus are no
longer allocated when the objective function takes into
account the THD cost.
Fig. 3 shows the variation of the two objective functions one with and one without the THD cost term, as

0.9
+ J includes THD term

0.85

o J excludes THD term

0

Number of capacitors bank

Figure 3: Variation of objective functions versus number of

capacitors bank

14th PSCC, Sevilla, 24-28 June 2002

Session 03, Paper 2, Page 5

Table 2: Capacitor Placement Results of Case 2

Bus code

675

Phase code

611
3

Without THD cost

(kVAr)

2x50 2x50 2x50

With THD cost

(kVAr)

2x50 2x50 2x50

645

646

2x50

1x50

3
1x50

Table 3: Capacitor Placement Results of Case 3

Bus code

675

611

692

671

652

645

646

Without THD cost

(kVAr)

2x150

1x150

2x150

1x150

1x150

1x150

1x150

With THD cost

(kVAr)

2x150

1x150

2x150

1x150

1x150

1x150

1x150

outcome of the optimization process. The locations

and sizes of the capacitors that yield the minimum objective function in case of relaxed limits are listed in
Table 2.
It follows from the analysis of the results in Tables 1
and 2 without THD cost that the same capacitor banks
are located in the three phase bus #675; two more banks
are located at the single phase busses #645 and #646.
When the objective function takes into account the
THD cost, none of the capacitor banks in single phase
busses are allocated.
4.3 Case 3
In the absence of a three phase capacitor placement
program such as the one described here, an approximate
solution may be attempted by using a single phase program that works for balanced systems. This means that
all unbalanced loads and structurally non-symmetric
lines will be replaced by balanced components and
hence the solution will be different. This case investigates the errors involved in such an approximation.
The case compares the capacitor placement results of
Case 1 with that of a fully balanced version of the same
system. Thus, the procedure of Section 3 is applied to a
balanced system which is created by approximating the
unbalanced system of Case 1 by a balanced version.
This is accomplished by balancing all the lines (all lines
are assumed three phase and structurally symmetric)
and loads (all loads are assumed three phase balanced
loads) of the unbalanced system. Moreover, harmonics
of order 5, 7, 11 and 13 only have been taken into account with consequent reduction of harmonic distortion
levels.
The locations and sizes of the capacitors (all 150
kVAR three phase banks) that yield the minimum objective function for this fully balanced case are listed in
Table 3. Results shown in the table indicate that in both
cases capacitor banks are located in the same three
phase buses #675, #611, #692, #671, #652, #645, #646.

A comparison of the results shown in Tables 1 and 3

implies that in the case of the balanced system a much
larger number of capacitors are required to achieve the
minimum J. This is due to both the reduced level of
voltage harmonics and the larger amount of total reactive load in the balanced system.

CONCLUSIONS

This paper presents a capacitor placement procedure

for three phase unbalanced power systems. The proposed procedure not only accounts for system losses
and capacitor costs, but also for harmonic distortion of
bus voltages and the effects of unbalanced operation.
The recent proliferation of nonlinear loads in distribution systems makes the issue of harmonic pollution an
important one to be considered in any of the distribution
automation functions such as capacitor sizing and location. This paper presents a simple yet effective solution
to the problem. Numerical examples demonstrating the
effectiveness of the procedure under both balanced and
unbalanced operating conditions are provided.
Future work will involve more realistic representation of the harmonic cost in the objective function.
REFERENCES
[1] M. Baran, F.F. Wu, Optimal Capacitor Placement
on Radial Distribution systems, IEEE Trans. on
Power Delivery, Vol.4, No.1, Jan. 1989, pp.725734.
[2] J. J. Grainger, and S. H. Lee, Optimum Size and
Location of Shunt Capacitors for Reduction of
Losses on Distribution Feeders, IEEE Trans. on
PAS, Vol.100, Mar.1981, pp.1105-1118.
[3] H-D. Chiang, J-C. Wang, J. Tong and G. Darling,
Optimal Capacitor Placement, Replacement and
Control in Large-Scale Unbalanced Distribution

14th PSCC, Sevilla, 24-28 June 2002

[4]

[5]

[6]

[7]

[8]

[9]

Systems: System Modeling and A New Formulation, IEEE Trans. on Power Systems, Vol.10,
No.1, February 1995, pp.356-362.
C.S. Chen, C.T. Hsu and Y.H. Yan, Optimal Distribution Feeder Capacitor Placement Considering
Mutual Coupling Effect of Conductors, IEEE
Trans. on Power Delivery, Vol.10, No.2, April
1995, pp.987-994.
K.N. Miu, H-D. Chiang and G. Darling, Capacitor
Placement, Replacement and Control in LargeScale Distribution Systems by a GA-Based TwoStage Algorithm, IEEE Trans. on Power Systems,
Vol.12, No.3, August 1997, pp.1160-1166.
Y. Baghzouz, S. Ertem, Shunt Capacitor Sizing
for Radial Distribution Feeders with Distorted Substation Voltages, IEEE Trans. on Power Delivery,
Vol.5, April 1990, pp.650-657.
Bei Gou and A. Abur, Optimal Capacitor Placement for Improving Power Quality, Paper SM 011.
Proceedings of IEEE/PES Summer Meeting, July
18-22, 1999, Edmonton, Canada.
P. Caramia, G. Carpinelli, E. Di Vito, A. Losi and
P. Verde, Probabilistic Evaluation of the Economical Damage due to the Harmonic Losses in Industrial Energy Systems, IEEE Trans. on Power
Delivery, Vol.11, No.2, 1996, pp.1021-1031.
P. Caramia, G. Carpinelli, A. Russo, P. Varilone, P.
Verde: An Integrated Probabilistic Harmonic Index, IEEE PES 2002 Winter Meeting, New York
(USA), 27-31 January 2002

Session 03, Paper 2, Page 6

[10] IEEE Task Force on Harmonics Modeling and

Simulation, Modeling and Simulation of the
Propagation of Harmonics in Electric Power Networks Part I: Concepts, Models and Simulation
Techniques, IEEE Trans. on Power Delivery, Vol.
11, No. 1, 1996, pp. 452-465
[11] IEEE Task Force on Harmonics Modeling and
Simulation, Modeling and Simulation of the
Propagation of Harmonics in Electric Power Networks Part II: Sample Systems and Examples,
IEEE Trans. on Power Delivery, Vol. 11, No. 1,
1996, pp. 466-474
[12] IEEE Recommended Practices and Requirements
for Harmonic Control in Electrical Power Systems,
IEEE Std 519-1992, IEEE, New York, NY, April
1993.
[13] A.E. Emanuel, C. Kawann, Passive Shunt Harmonic Filters for Low and Medium Voltage: A Cost
Comparison Study, IEEE Trans. on Power Systems, Vol. 11, No. 4, 1996, pp. 1825-1831
[14] T.T. Chang, H. Chang, Application of Differential
Evolution to Passive Shunt Harmonic Filter Planning, 8th IEEE ICHQP, Athens (Greece), 1998, pp.
149-153