1

Optimal Reactive Power Allocation for Power

System Transfer Capability Assessment
S. H. Goh, St Member, IEEE, T. K. Saha, Senior Member, IEEE, and Z. Y. Dong, Member, IEEE

different optimization techniques and their popularity based

Abstract--Power system stability issues have always been of on a compilation of over 300 publications. This paper [4] also
utmost importance and interests. Since the restructuring of provided a good list of references to over 160 papers,
power utilities, electric power systems have been strained to covering many of the methods used to solve OPF problem.
operate under much higher stress levels, and closer to their
Apart from the redistribution of system reactive power
physical and operational limits. Thus system instability issues
have become increasingly critical, and are of primary concern through ORD, series and shunt reactive power compensation
during power system operation and planning. This paper are also popularly-used to improve overall system stability or
investigates the impacts of reactive power re-dispatching and power transfer capability. These compensation usually may
reactive power compensation on static voltage stability margins include devices for eg. capacitor banks, switching capacitors,
when evaluating power transfer capability. A mix-integer Flexible AC Transmission System (FACTS) devices such as
nonlinear programming (MINLP) problem formulation is
Static Var Compensator (SVC) and Thyristor-controlled
proposed here not only to re-distribute reactive power, but also to
effectively determine optimum locations and amount of both Series Capacitor (TCSC) etc. A number of work has been
series and shunt reactive compensation. The well-known branch- published in this area which mainly focuses on optimal
and-bound based technique is being employed and tested on a placement and amount of reactive power compensation to
simple test system. Results showed significant enhancements to ease network congestions, thus leading to enhancements in
system static voltage stability margins. system stability margins or power transfer capability limits [5]
– [8].
Index Terms--Transfer capability, optimal reactive dispatch,
reactive power compensation, static voltage collapse.
Power system transfer capability as defined by the North
American Electric Reliability Council (NERC) in 1996, refers
I. INTRODUCTION to the ability of interconnected electric networks to safely
deliver or transfer bulk electrical power from one area to
V OLTAGE stability issues are critical in both power
system planning and operation procedures, whereby they
usually occur in power systems which are heavily-loaded,
another over all connecting transmission interfaces under
specified system conditions. Available Transfer Capability
(ATC) is a measure of the transfer capability remaining in the
faulted and/or have reactive power shortages [1]. These
network for further commercial activity over and above
issues have become exceptionally important in recent years
already committed uses; and determining of Total Transfer
whereby there is a need to optimally allocate available
Capability (TTC) is the key component in the context of ATC
resources in a highly constrained competitive environment to
computation. Evaluating TTC can be seen as a nonlinear
improve operational efficiencies. As such, power systems
constrained optimization problem, whereby the objective is to
have been stressed to operate near their system limits, thus
maximize inter-area active power transfers [9].
becoming very vulnerable in times of unexpected outages.
This paper formulates a mix-integer nonlinear
There have been some attempts in the past to utilize
programming (MINLP) ORD problem that re-allocates system
Optimal Reactive Dispatch (ORD) for system voltage profile
reactive power, and determines optimal locations and amount
correction and reactive power reserve maximization through
of both series and shunt reactive power compensation while
the redistribution of system reactive power and other reactive
maximizing power transfer capability. It investigates the
power related control actions. ORD is a special case of the
effects of reactive power re-dispatching and compensation on
conventional Optimal Power Flow (OPF) problem, whereby
static voltage collapse margins. It is being organized as
all active power related controls are fixed, only reactive power
follows: Section II briefly introduces the popular branch-and-
related controls are of interests [2], [3]. Since the first OPF
bound technique used to solve the proposed MINLP problem;
paper was presented in the early 1960s [4], a lot of effort has
Section III formulates the optimization model in quantifying
been devoted into how to formulate and solve the OPF
transfer capability; Section VI presents discussions from
problem. A survey from [4] presented a summary of many
simulation results; Section V concludes with possible
suggestions for future developments.
Sheng How Goh, Tapan Kumar Saha, and Zhao Yang Dong are with the
School of Information Technology and Electrical Engineering, The University of
Queensland, St. Lucia, QLD 4072 Australia (e-mail: [email protected]).

1-4244-0493-2/06/$20.00 ©2006 IEEE.

 2

II. MINLP USING BRANCH-AND-BOUND TECHNIQUE follows: after solving a problem/sub-problem, if an integer
A MINLP problem is rather similar to a standard nonlinear variable x is fractional, then the problem/sub-problem will be
optimization problem with the only exception that it consists branch further into two other sub-problems. Limits will be
of a mixture of continuous, discrete or even binary imposed with the nearest upper and lower integer values of x
optimization variables. The branch-and-bound technique, for the two sub-problems; for eg., impose limit ( x ≤ x ) in one
employed in this paper, is a simple yet popular searching
method which is reasonably efficient for solving integer
( )
of the sub-problems and x ≥ x in the other. Through solving
these sub-problems, solutions that fulfill all integer
programming problems.
restrictions are then stored. An upper and lower bound on the
objective can then be determined from these stored solutions.
Nodes that give infeasible or non-optimum solutions are then
discarded. The search process continues till all nodes have
been either solved or discarded [10].
Although the branch-and-bound technique might not locate
global optimum solution at each node, it might sometimes be
more computational efficient as compared to Genetic
Algorithms (GA) [5], [10]. Thus for simplicity, the branch-
and-bound technique is chosen and implemented in this paper.

III. PROPOSED METHODOLOGY

Fig. 1. A typical tree representation of the branch-and-bound technique [10] A. Mathematical Formulation
The proposed MINLP ORD problem is a mathematically
The branch-and-bound technique branches the original complex optimization problem. The objective is to maximize
problem (node 0) into smaller sub-problems (nodes 1 to 4) active power transfers between two control areas over all
which are then being solved recursively. This resulted in a connecting transmission interfaces. It is assumed here that a
growing search tree whereby each sub-problem is linked to feasible active power dispatch solution is already available
both its parent node and its child node as shown in Fig. 1. A before solving the proposed ORD problem. As mentioned
full-grown search tree is equivalent to a complete enumeration earlier, ORD is similar to the traditional OPF problem; thus
of all possible integer points, which is exhaustive and following this concept, a multi-objective optimization
impractical for problems with numerous discrete variables. problem can be expressed as follows [2], [5], [11]:
Therefore it is very important to avoid over-growing of this
search tree which is undesirable [10]. Maximize

∑P ∑I ∑I
This technique actually solves a continuous problem Nl Nb

recursively through the relaxation of the integer restrictions f = km −A ij series C ij series − B j shunt C j shunt (1)
m∈Sink , k∉Sink i≠ j j =1
on the integer variables. It is made up of two basic operations
known as branching and bounding. Branching partitions the
continuous solution space into smaller sub-problems (nodes) where the first term represents the total active power transfers
with the purpose of eliminating parts of the continuous between two interconnected areas over all connecting
solution space that are infeasible due to the integer transmission interfaces; k denotes buses in the source area
restrictions. Each of the created sub-problems is then solved and m denotes buses in the sink area; and thus Pkm represents
as a continuous problem. Further branching is necessary if the active power transfer from bus k to bus m .
any of the integer variables of a sub-problem is not integer- For practicality, the investment cost of both series and
valued (fractional). Bounding then defines upper and lower shunt reactive compensation is being included as one of the
bounds based on these stored solutions, which are essential objectives. Since the total cost is usually related to the
for ranking optimum solutions and eliminating non-optimum amount of compensation involved. Hence, for simplicity, it is
solutions [10]. represented here by [5]:
To avoid complete enumeration of the problem, branching

∑I ∑I
and bounding have to be implemented ‘intelligently’. The Nl Nb
algorithm should only branches potential nodes that may A ij series C ij series + B j shunt C j shunt (2)
i≠ j j =1
result in an optimum solution; and discards nodes where
further growth leads to infeasible solutions or non-optimum
solutions. This is critical as poor and inefficient branching where A and B are positive constant cost coefficients;
and bounding will slow down the entire searching process C ij series is the series compensation for line connecting bus i
significantly [10]. and j ; and C j shunt is the amount of shunt compensation at
A typical branch-and-bound procedure can be described as
 3

bus j ; whereas I ij series and I j shunt are the installation flags (0 the above formulated problem consists of both discrete and
binary variables, for eg., the discrete steps for transformers
or 1) for the series and shunt compensation respectively; Nl
tap settings and sizes of compensation capacitors; and also the
and Nb refer to the number of lines and load bus in the
binary installation flags.
system respectively. Observe that Eqn. (2) has been added
into the objective as shown in Eqn. (1), but is made negative B. Computational Procedures
to denote it as an investment cost minimization problem. This section describes the computational procedures of the
Eqn. (2) is used to penalize the total active power transfers proposed methodology which is shown in Fig. 2.
relative to minimizing the costs of reactive compensation. In
other words, it is added to prevent the system being overly-
compensated which is impractical due to investment costs. It
should be observed that weights can be assigned to denote the
emphasis on maximizing active power transfers or minimizing
total investment costs.
The above objective shown in Eqn. (1) is then subjected to
a list of constraints which can be expressed as follows [2],
[5], [11]:

∑V V Y cos(θ
Pi −
j∈N
i j ij ij )
+ δi −δ j = 0 (3)

Q − ∑ V V Y sin (θ
i i j ij ij )
+ δi − δ j = 0 (4) Fig. 2. Flowchart of the proposed methodology
j∈N

Q g ,min ≤ Q g ≤ Q g ,max (5) First of all, a transfer case must be specified for TTC
assessment. A base case TTC can then be solved using
S ij ≤ S ij , max (6) Repeated Power Flow (RPF), Continuation Power Flow (CPF)
or OPF method. In this paper, an OPF-based technique is
Vi ,min ≤ Vi ≤ Vi ,max (7) utilized to maximize the active power transfers across all
connecting transmission interfaces between two areas, and
C ij series (min) ≤ C ij series ≤ C ij series (max) (8) also the loads in the sink area with constant power factor.
The converged transfer case is then solved again using the
C j shunt (min) ≤ C j shunt ≤ C j shunt (max) (9) proposed ORD formulation as described in the previous
section. The algorithm attempts to search for an optimal
⎧1 solution whereby only reactive power related controls are
I ij series , I j shunt ⎨ (10) being adjusted; which in this case refers to the reactive power
⎩0 re-dispatching, transformers tap settings, and optimal location
Tij min ≤ Tij ≤ Tij max (11) and amount of series and shunt compensation. If a successful
solution is found, the system static voltage collapse margin is
then studied and investigated. Following this methodology,
where Eqn. (3) and (4) are equality constraints which enforces an improved voltage collapse margin will be expected in most
power balance in the system; N is the set of all the buses in cases.
A CPF algorithm from [12] is being used here to study the
the system; Vi , V j and δ i , δ j are the voltage magnitudes and
voltage stability limits. CPF is a set of conventional power
angles of bus i and bus j respectively; Yij and θ ij are the flow equations which is reformulated to include a loading
magnitude and angle of the ij th element in the bus parameter λ that represents the generation and load
admittance matrix.; Eqn. (5) is the reactive power capability increments. It is capable of overcoming numerical difficulties
of the generators; Eqn. (6) and (7) are the thermal limits of near the nose point where conventional power flow fails due
transmission lines and bus voltages limits respectively. Eqn. to the singularity of the power flow Jacobian. This nose point
(8) is the limit of series compensation for line connecting bus is also widely-known as the voltage collapse point or the
i and j , and Eqn. (9) is the limit of shunt compensation at maximum loading point when λ is at its maximum
bus j ; Eqn. (10) is the binary installation flags for both series ( λ = λcri ). Hence, this maximum or critical lambda λcri is
and shunt compensation, whereby 0 denotes no installation often being regarded as a static voltage collapse indicator.
and 1 denotes an installation; and lastly Eqn. (11) represents Information on CPF is only briefly introduced here as it is
the tap settings limits of transformers connecting bus i and well-documented in [12] and [13].
j , which is usually range from 0.95 to 1.05 pu. Observe that
 4

IV. CASE STUDY introduced to transform Eqn. (1) into a single objective
The proposed ORD problem has been tested on a simple 3- optimization problem.
machine 9-bus system from [14] as shown in Fig. 3. It is In this case study, the transformers tap settings which
divided into 2 areas with 2 tielines (line 5-6 and line 4-9). typically range between 0.95 to 1.05 pu, are assigned to have
11 discrete steps of 0.01 pu. Maximum series compensation
A. Base case TTC for each line is limited to 70%. The total series compensation
Following the flowchart shown in Fig. 2, an OPF-based for all lines is limited to 200%, which means no three lines
TTC technique is firstly used to evaluate the base case active can be fully-compensated as 70% x 3 = 210%; and the total
power transfers from Area 1 to Area 2. Table I displays the shunt compensation for all load bus is limited to a total
reactive power outputs of generators after solving TTC. resultant reactive power injection of 50 MVar. The
Table II shows the base case loads in Area 2 before solving compensation limits chosen here is only an estimation based
TTC. The base case TTC is found to be 158.29 MW; and on the total system reactive power demand. It should be clear
there is a constant power factor load increment in the sink that different limits used will yield different solutions on the
area of 1.4547 times which is shown in Table III. entire pareto-optimal set of the original multi-objective
problem.

TABLE IV
GENERATORS REACTIVE POWER OUTPUT AFTER RE-DISPACTHING
Gen Qg (MVar)
1 5.61
2 70.74
3 -37.58

TABLE V
OPTIMAL TAP SETTINGS OF TRANSFORMERS
Transformer Tij (pu)
1-4 1.03
Fig. 3. 3-machine 9-bus system
3-6 0.96
TABLE I 8-2 1.05
GENERATORS REACTIVE POWER OUTPUT
Gen Qg (MVar) TABLE VI
1 18.30 OPTIMAL LOCATION & AMOUNT OF SERIES COMPENSATION

2 18.90 Line %
3 88.89 1-4 70
5-6 60
TABLE II 9-4 70
BASE LOAD IN SINK AREA BEFORE OPF-TTC
Bus P (MW) Q (MVar) TABLE VII
7 100 35 OPTIMAL LOCATION & AMOUNT OF SHUNT COMPENSATION
9 125 50 Bus Qc (MVar)
6 20
TABLE III 8 30
LOAD IN SINK AREA AFTER OPF-TTC
Bus P (MW) Q (MVar)
The proposed branch-and-bound based methodology is
7 145.47 50.91
then used to study the active power transfers from Area 1 to
9 181.83 72.73
Area 2. TTC after reactive power re-allocation is found to be
159.63 MW. Notice that there is only a slight increment of
B. Optimal Reactive Power Allocation 1.34 MW in the TTC level, which is due to the fact that active
Solving a base case TTC as shown in the previous section power controls are not considered in the proposed ORD
will provide a reasonably good feasible starting point for formulation, as the main objective is to re-allocate system
further analysis. As the proposed methodology is a complex reactive power. However, it should be clear that TTC may
optimization problem, thus for simplicity, instead of solving increases significantly with improved system loadability.
the multi-objective optimization problem as proposed in Eqn. Table IV shows the reactive power outputs of generators
(1) to obtain the complete pareto-optimal set, limits for the after re-dispatching. Table V, VI and VII present the optimal
total shunt and series compensation in the system are being tap settings of transformers, optimal location and amount of
series and shunt compensation respectively. Although it can
 5

be seen from Table IV that generator 3 is absorbing reactive enhancements.

power of 37.58 MVar, this is still rational as limits on reactive
power absorption capability of all generators have been PV curves for bus #7

imposed to prevent under-excitation that may leads to 1.1

instability. Fig.4 presents the system bus voltage profiles
1
before and after solving the proposed ORD formulation.
Observe that the bus voltage profiles differ through the 0.9
redistribution of system reactive power. 0.8

V
System bus Profile 0.7

1.1 0.6

0.5 Base case

After VAR re-allocation
1.05 0.4
0 0.5 1 1.5 2 2.5 3
Loading parameter
V

Fig. 6. PV curves for load bus 7 before and after VAR re-allocation
1

PV curves for bus #9

1.1
0.95
Before
After 1
0.9
1 2 3 4 5 6 7 8 9
Bus No. 0.9
Fig. 4. Bus voltage profile for base case OPF-TTC
0.8
TABLE VIII V
SYSTEM CRITICAL LAMBDA BEFORE AND AFTER RE-DISPACTHING 0.7
Case λcri (pu)
0.6 Base case
Before 1.7765 After VAR re-allocation
After 2.7688 0.5
0 0.5 1 1.5 2 2.5 3
Loading parameter
PV curves for bus #5 Fig. 7. PV curves for load bus 9 before and after VAR re-allocation
1.1

1
C. Contingency Studies
This section demonstrates the impacts of unexpected line
0.9 outages on the system static voltage collapse margins using
static power flow solutions from the base case OPF-TTC and
V

0.8 the proposed ORD formulation. Due to the limitation of

space, only two line outages are presented here for analyses.
0.7
Similar to previous section, only load bus 5, 7 and 9 are being
Base case investigated in each of the line outages. Observe that the
After VAR re-allocation voltage collapse margins for both line outages decrease as
0 0.5 1 1.5 2 2.5 3 compared to those from normal case shown in the previous
Loading parameter
Fig. 5. PV curves for load bus 5 before and after VAR re-allocation
section which fulfills expectations. It is important to
understand that the static power flow solutions from the base
A CPF algorithm from [12] is used in this paper to study case OPF-TTC and the proposed ORD algorithm may not be
the effects on the system static voltage collapse margins n-1 compliance, and thus additional operating constraints may
before and after solving the proposed ORD formulation. be essential.
Table VIII presents the value of the critical lambda λcri
1) Line 4-5 outage
before and after solving the proposed ORD formulation. Only
Fig. 8, 9 and 10 present the system static voltage collapse
load bus 5, 7 and 9 are being investigated, which are
margins for load bus 5, 7 and 9 respectively with the outage of
illustrated by Fig. 5, 6 and 7 respectively. Analyses show that
line 4-5. Table IX presents the value of the critical lambda
there is a significant improvement to the system static voltage
λcri before and after solving the proposed ORD formulation.
collapse margin, which is most likely due to the series and
shunt reactive compensation. It should be noted that further
dynamic simulations may be required to verify these
 6

lambda λcri before and after solving the proposed ORD

TABLE IX
SYSTEM CRITICAL LAMBDA BEFORE AND AFTER RE-DISPACTHING
formulation.
Case λcri (pu)
TABLE X
Before 1.3439 SYSTEM CRITICAL LAMBDA BEFORE AND AFTER RE-DISPACTHING
After 2.2418 Case λcri (pu)
Before 1.3175
PV curves for bus #5 2.4713
After

1.1 PV curves for bus #5

1.1
1
1.05
0.9
1
V

0.8
0.95
0.7

V
0.9
0.6 Base case
After VAR re-allocation 0.85
0.5
0 0.5 1 1.5 2 2.5
0.8 Base case
Loading parameter
After VAR re-allocation
Fig. 8. PV curves for load bus 5 (with line 4-5 out) 0.75
0 0.5 1 1.5 2 2.5
Loading parameter
PV curves for bus #7
Fig. 11. PV curves for load bus 5 (with line 8-9 out)
1.1
PV curves for bus #7

1 1.1

0.9 1
V

0.8 0.9
V

0.7 0.8
Base case
After VAR re-allocation

0 0.5 1 1.5 2 2.5 0.7

Loading parameter Base case
Fig. 9. PV curves for load bus 7 (with line 4-5 out) After VAR re-allocation
0 0.5 1 1.5 2 2.5
Loading parameter
PV curves for bus #9
1.1 Fig. 12. PV curves for load bus 7 (with line 8-9 out)
Base case
PV curves for bus #9
1 After VAR re-allocation
1.1
Base case
1 After VAR re-allocation
0.9
V

0.9
0.8
0.8
V

0.7 0.7

0.6
0 0.5 1 1.5 2 2.5
Loading parameter 0.5
Fig. 10. PV curves for load bus 9 (with line 4-5 out)
0.4
0 0.5 1 1.5 2 2.5
Loading parameter
2) Line 8-9 outage Fig. 13. PV curves for load bus 9 (with line 8-9 out)
Fig. 11, 12 and 13 present the system static voltage
collapse margins for load bus 5, 7 and 9 respectively with the V. CONCLUSIONS
outage of line 8-9. Table X presents the value of the critical
A branch-and-bound technique based MINLP problem is
 7

formulated in this paper for optimal reactive power allocation T. K. Saha (SM’97) was born in Bangladesh and immigrated to Australia in
1989. Currently he is a professor in the School of Information Technology and
when evaluating power transfer capability between two Electrical Engineering, The University of Queensland, Australia. Before joining
control areas. The proposed ORD algorithm maximizes the University of Queensland, he taught at the Bangladesh University of
active power transfers while considering the total investment Engineering and Technology, Dhaka for three and a half years; and then at
costs of both series and shunt reactive power compensation. James Cook University, Australia for two and a half years. His research
interests include power systems, power quality, and condition monitoring of
This proposed MINLP problem can handle both discrete and electrical plants. Dr. Saha is a Fellow of the Institute of Engineers Australia.
binary variables which also enable the modeling for optimal
placement and amount of series and shunt compensation. Z. Y. Dong (M’99) received his Ph.D degree in Electrical and Information
Results from case study show significant improvements to Engineering from The University of Sydney, Australia in 1999. He is now a
senior lecturer at the School of Information Technology and Electrical
system static voltage collapse margins during both the normal Engineering, The University of Queensland, Australia. His research interests
and contingency cases. However, it should be noted that include power system security, electricity market, artificial intelligence and its
power flow solutions obtained from the proposed ORD application in power engineering, power system planning, and power system
stability and control.
algorithm may not be n-1 contingency compliance, and further
dynamic simulations are essential. Further work can be aimed
at developing n-1 contingency constraints into the proposed
ORD model.

VI. REFERENCES
[1] “Voltage Stability Assessment: Concepts, Practices and Tools”, IEEE/PES
Power System Stability Subcommittee Special Publication, 2002.
[2] S. Granville, “Optimal reactive dispatch through interior point methods,”
IEEE Transactions on Power Systems, vol. 9, no. 1, pp. 136-146, 1994.
[3] G. R. M. da Costa, “Optimal reactive dispatch through primal-dual
method,” IEEE Transactions on Power Systems, vol. 12, no. 2, pp. 669-
674, 1997.
[4] M. Huneault and F. D. Galiana, “A survey of optimal power flow
literature,” IEEE Transactions on Power Systems, vol. 6, no. 2, pp. 762-
770, 1991.
[5] M. Delfanti, G. P. Granelli, P. Marannino, and M. Montagna, “Optimal
capacitor placement using deterministic and genetic algorithms,” IEEE
Transactions on Power Systems, vol. 15, no. 3, pp. 1041-1046, 2000.
[6] A. C. Z. de Souza, L. M. Honorio, G. L. Torres, and G. Lambert-Torres,
“Increasing the loadability of power systems through optimal-local-control
actions,” IEEE Transactions on Power Systems, vol. 9, no. 1, pp. 188-
194, 2004.
[7] T. Orfanogianni and R. Bacher, “Steady-state optimization in power
systems with series FACTS devices,” IEEE Transactions on Power
Systems, vol. 18, no. 1, pp. 19-26, 2003.
[8] W. Ongsakul and P. Jirapong, "Optimal allocation of FACTS devices to
enhance total transfer capability using evolutionary programming," in
Proc. of IEEE International Symposium on Circuits and Systems, vol. 5,
pp. 4175-4178, 2005.
[9] Transmission Transfer Capability Task Force (NERC), "Available transfer
capability definitions and determination," 1996.
[10] H. A. Taha, Integer Programming: Theory, Applications and
Computation. New York, Academic Press, 1975.
[11] S. H. Goh, Z. Xu, Z. Y. Dong, and T. K. Saha, "Economic constrained
transfer capability assessment," in Proc. of IEEE Power Engineering
Society General Meeting, pp. 318-325, 2005.
[12] F. Milano, “An open source power system analysis toolbox,” IEEE
Transactions on Power Systems, vol. 20, no. 3, pp. 1199-1206, 2005.
[13] V. Ajjarapu and C. Christy, "The continuation power flow: a tool for
steady state voltage stability analysis," IEEE Transactions on Power
Systems, vol. 7, no.1, pp. 416-423, 1992.
[14] P. M. Anderson and A. A. Fouad, Power system control and stability.
Ames: Iowa State University Press, 1977.

VII. BIOGRAPHIES

S. H. Goh (Student M’02) received his B.Eng honours degree in electrical

engineering from the University of Queensland, Australia in 2001. Currently, he
is pursuing a Ph.D degree in electrical engineering at the University of
Queensland, Australia. His research interests include available transfer
capability and power system stability and security.