1
Assessment of Reactive Power Pricing by
Controlling Generators’ Voltage under Deregulation
Seung-Wan Jung, Sung-Hwan Song, Student Member, IEEE, Yong Tae Yoon, Member, IEEE, and
Seung-Il Moon, Member, IEEE
Abstract — This paper presents the assessment schemes of
reactive power pricing through generators’ voltage control under
the assumption that a reactive power market is decomposed into a
real power market. Applying the concept of economic dispatch to
minimize the total cost of real power, the algorithm for
implementing a reactive power pricing is proposed to determine
the generators’ optimum voltage profiles. It consists of the
reactive power-voltage equation, the objective function which
minimizes the total cost of reactive power, and linear analysis of
inequality constraints in relation to the voltages at the loads.
From this algorithm, the total cost of the reactive power can be
yielded to the minimum value within network constraints. It may
provide the fair and reasonable price information for reactive
power in deregulated electricity market. The proposed algorithm
has been tested on the IEEE 14-bus system using MATLAB.
Index Terms — Reactive Power Pricing, Power Flow Equation,
Optimization Problem, Linear Analysis
I. INTRODUCTION
A s electricity power industry undergoes new environment
change in vertically integrated structure, the operation
scheme and control strategy of power systems may be altered
around the world.
The focus of power system operation mainly has been set to
adjust supply and demand balance of active power, and
reactive power has been thought as a by-product by
production of active power. Due to this recognition, the many
generator companies did not charge some monetary reward
about role of producing and consuming reactive power for
stable operation of power systems. That is, the quantity of
reactive power production through generators’ voltage control
was decided by system operators’ experience and there was no
financial compensation system about it.
However, the supply and demand of reactive power is acted
by noteworthy factor in power system operation as the load
characteristics are becoming gradually complex and dynamic.
For example, even if the problem related to active power does
not happen, voltage instability is observed by dynamic
characteristics of load, and these phenomenons can arrive in
voltage collapse.
Among the recommendations at final report about North
America blackout in 2003 proposed by NERC [1], the
Sung-Hwan Song, Seung-Wan Jung, Yong Tae Yoon, and Seung-Il Moon are
with the school of Electrical Engineering, Seoul National University,
Kwanak-Gu, Seoul, Korea
(telephone : +82-2-880-7257, e-mail:karyman@powerlab.snu.ac.kr)
strengthening of reactive power and voltage control practices
is mentioned. This is the evidence that reactive power is
significant factor in several of the outages like North America
blackout in 2003.
Consequently, reactive power plays an important role for
operating power systems stably, keeping the bus voltages
within nominal ranges, and supporting the real power transfer.
In spite of its importance, financial compensation scheme for
providing reactive power of generator is still not well
determined, compared with real power.
Of course, the problem connected with reactive power is
classified as ancillary service in some part of advanced nations
that introduce the electricity market mechanism through
restructuring and the reactive power pricing system is formed
in other structures in the each country. New York ISO,
California ISO (US), and NEMMCO (Australia) are
constructing the pricing system that compensates
economically revenue decrease by decrease in active power
output as the concept of Lost Opportunity Cost (LOC)
payment. And in case of NGC (United Kingdom), reactive
power pricing are calculated through capability component
(Price/MVAr) and utilization component (MVAr Price Curve)
[2]. Some researches present the methods of power factor
based pricing systems [3], marginal nodal pricing [4],
incremental average approach [5], reactive flow tracing
method [6], and reactive power factor adjustment based zonal
charge system [7].
Especially, it is can be interpreted that the reactive power
pricing in the most countries is defined by an equivalent
system through relation with active power output within
generators’ capacity limits.
This paper presents a scheme to compute the actual pricing
of reactive power by directly controlling generators’ voltage
under the assumption that a reactive power market is
decomposed into a real power market. This contains algorithm
that decides generators’ most suitable voltage profiles to
minimize the reactive power production cost. The proposed
algorithm is tested with IEEE 14-bus system and verified
accuracy of this algorithm.
The remaining organization of this paper is as follows. We
first formulate the objective function and analyze linearly the
non-linear constraints to get the stable convergence. And then
we propose the assessment scheme for reactive power pricing
which is based on power flow equation, minimization
procedure of objective function, and iterative linearization
 2

procedure of inequality constraints. Following that, we

demonstrate the proposed algorithm through the numerical V dj = h ( X ; Q d 1 , Q d 2 , Q d 3 , L , Q dM )
(4)
examples. Finally, we summarize it briefly.
= h' ( X )
Similarly, Assuming that generator voltages are defined as a
II. MATHEMATICAL FORMULATION vector like g i = g1 , g 2 , g 3 , L g N , the reactive power of
generator buses ( Qgi ) can be expressed like equation (5).
A. Reactive Power and Total Cost Formulation
In power flow equation, reactive power injected to each bus
is as following. Q gi = u ( X ; Vd 1 , Vd 2 , Vd 3 , L , VdM )
= u ( X ; h ' ( X )) (5)
n
= u'( X )
Qi = ∑ V i V k ( G ik sin θ ik − B ik cos θ ik )
(1)
k =1 Finally, reactive power at the generator bus becomes
n
formulated by generators’ voltages.
≈ − ∑ V i V k B ik cos θ ik (Q G ik << 1 )
k =1 Using above equations, total cost for producing reactive
where, power at the generators can be formulated as following,
Qi : reactive power injected to the i th bus
TC gi = f ( Q gi )
n : the number of buses (6)
= f ( u ' ( X ))
Vi : voltage of the i th bus
= f '( X )
Vk : volage of the k th bus The total cost for producing reactive power at the generators
Bik : line susceptance between the i th bus and the k th bus can be formulated by vector X , that is generator voltages.
θ ik : angle difference between the i th bus and the k th bus, And under the assumption that generators are the only reactive
power supplier in the power system, the supply and demand of
that is ( θ i − θ k )
reactive power in the power system can be performed by
controlling generator voltages.
Assuming that generator voltages are defined as a vector It is noted that (∗)′ indicates the function expressed by
like X = [V g 1 , V g 2 , V g 3 , L V gN ]' , equation (1) can be
control variables X .
expressed as a function of both generator voltages and load
voltages as following,
B. Objective Function
Qi = g (V g1 , Vg 2 , Vg 3 , LVgN −1 , VgN , Vd 1 , Vd 2 , Vd 3 , L , VdM ) The total cost for producing reactive power at the generators
can be expressed as below.
= g ( X , Vd 1 , Vd 2 , Vd 3 , L , VdM )
(2)
where, min
V gi
∑ TC
gi
gi ( Q gi ( X )) (7)

M : the number of load buses where,

N : the number of generator buses TC gi :total cost for reactive power at the generator i
and the n th generator bus is assigned as a slack bus.

Load voltages should be kept within rated range necessarily

Assuming that load voltages are defined as a vector like
d j = d1 , d 2 , d 3 , L d M , the reactive power of load buses
( Qdj ) can be expressed like equation (3). So the reactive
power injected to the load buses is a function of the
generators’ voltage ( X ) and the load voltages
( V d 1 , V d 2 , L , V dM ) .
in power system. If load voltage escapes rated range,
consumer's electrical appliances may cause damage.
Therefore, inequality constraints condition such as equation
(8) exists. Load voltages are function of generator voltage, so
load voltages can be kept within desirable voltage range by
controlling reactive power at the generators.

≤ Vdj ( X ) ≤ Vdj
min max
Vdj (8)
Q dj = g ( X , V d 1 , V d 2 , V d 3 , L , V dM ) (3)
where,
In power flow equation, load bus is assumed to be PQ bus, V dj : load voltage
and the reactive power is a given value at PQ bus. Therefore, Vdj
max
: maximum load voltage
load voltage can be calculated by equation (4) as following.
 3

min
Vdj : minimum load voltage
−1
∆Vdj = −( J dd J dg )∆Vgi (14)

C. Linear Analysis of Inequality Constraints ∆Qgi = J gg ∆Vgi + J gd ∆Vdj

Objective function in (7) has inequality constraints such as (15)
−1
(8). In this inequality constraints, uncertainty of convergence = ( J gg − J gd J dd J dg )∆Vgi
can be existed in the process that finds optimized total cost of
objective function. Therefore, linearization analysis Equation (14) and (15) mean that the deviation of loads’
procedure is needed for the stable convergence as following. voltage and generators’ reactive power can be linearized by
the deviation of generators’ voltages.
Load voltage is a function of generator voltage as following,

Vdj = f (Vgi ) (9) III. ALGORITHM FOR REACTIVE POWER PRICING

This paper presents an algorithm for the purpose of
The deviation of load voltage can be as following. evaluating the total cost for reactive power at the generator.
The proposed solution method has the following algorithm in
the procedure of cost minimization.
⎡ ∂f ⎤ (10)
∆Vdj = ⎢ ⎥ ∆Vgi
⎢⎣ ∂Vgi
Vgi ,Vdj
⎥⎦ Stage 1

Initial Value
And reactive power at the load bus is a function of generator
Power Flow
voltage and load voltage as following,
θ gi , θ dj , Vgi , Vdj
Qdj = g (Vgi ,Vdj ) (11)
Stage 2
Using above equations, the deviation of reactive power at
Minimization Procedure
generator bus and load bus can be linearized by the deviation
of generator voltage and load voltage. Objective Function

min ∑ TC gi (Qgi ( X ))
⎡Qgi + ∆Q gi ⎤ J gd ⎤ ⎡ ∆Vgi ⎤
Vgi
⎡ J gg
[ ]
gi

⎢Q + ∆Q ⎥ = g (Vgi , Vdj ) + ⎢ J J dd ⎥⎦ ⎢⎣ ∆Vdj ⎥⎦

(12)
⎣ dj dj ⎦ ⎣ dg Constraints

- Power Flow Balance

- Load Voltage
where, Vdj
min
≤ Vdj (Vgi ) ≤ Vdj
max

⎡ ∂Q ⎤ ⎡ ∂Q ⎤
J gg = ⎢ gi ⎥ , J gd = ⎢ gi ⎥ ,
⎣⎢ ∂Vgi ⎦⎥ ⎣⎢ ∂Vdj ⎦⎥ Linearization of Inequality
Constraints
⎡ ∂Q ⎤ ⎡ ∂Q ⎤
J dg = ⎢ dj ⎥ , J dd = ⎢ dj ⎥ - Jacobian

⎣⎢ ∂Vgi ⎦⎥ ⎣⎢ ∂Vdj ⎦⎥
Jgg, Jgd, Jdg, Jdd

Equation (12) can be simplified by following procedure. Stage 3 ∆Vgiopt

Relinearization Procedure
⎡Q gi + ∆Q gi ⎤ ⎡ J gg J gd ⎤ ⎡ ∆V gi ⎤
⎢ Q dj ⎥ [
= g (V gi , Vdj ) + ⎢ ] J dd ⎥⎦ ⎢⎣ ∆ Vdj ⎥⎦
V g Update
⎣ ⎦ ⎣ J dg
V d Update
(13)
(Q ∆Q dj = 0) Jacobian Update

⎡ ∆ Q gi ⎤ ⎡ J gg J gd ⎤ ⎡ ∆V gi ⎤
⎢ 0 ⎥ = ⎢J J dd ⎥⎦ ⎢⎣ ∆Vdj ⎥⎦
⎣ ⎦ ⎣ dg Fig. 1. Algorithm for reactive power pricing

From (13), we finally get the ∆Vdj and ∆Qgi like (14) and (15). Proposed algorithm is divided into 3 procedures. Initial
values for optimization are obtained in Stage I, and total cost
 4

for reactive power is minimized in Stage II by iterative method.
Iteration steps are performed to correct the error generated by
linear analysis of nonlinear constraints in Stage III.
The detailed algorithm can be explained as following:
like capability price structure and utilization price structure
in NGC (United Kingdom) [8]. So, the total cost can be
calculated with cost curve coefficient product reactive
power at the generator. The total cost equation is as
following,
TC = ∑ a gi Q gi
2

▪ Stage I : gi

Where,
1. Input system data
2. Construct Y matrix a gi : cost curve coefficient
3. Solve power flow, and set the initial condition
( θ gi , θ dj , Vgi , Vdj ) ⋅ It is assumed that the reactive power at the generator is
defined in reactive capability curves including field current
limit, armature current limit, and under-excitation limit [9].
▪ Stage II : min max
⋅ The voltage limits at load buses, Vdj and Vdj , are
4. Set the deviation of generation voltages ( ∆Vg ) as
0.98 [pu] and 1.02 [pu] respectively.
control
variables G G
5. Linearize inequality constraints using Jacobian 2 3

6. Calculate the total cost G

7. Check the following convergence condition 1

5 4
k +1
TC − TC k
<ε
8. If converged then go to Stage III with ∆V g opt which is the G
8 7
voltage variation after optimization process, otherwise G
repeat from step 4 6 11 10 9

12

▪ Stage III :
9. Check the following convergence condition 13 14
k +1
−Vg < ε
k Fig. 2. IEEE 14 bus system
Vg
10. If converged then terminate, otherwise update V g with Two cases are considered in this paper, the one is that all
generators have the identical cost curve coefficient, and the
∆V g
opt
other is that one generator has a different cost curve
11. Solve Q-V flow, and update Vd . coefficient.

12. Update Jacobian ( J gg , J gd , J dg , J dd ). A. Case 1 : identical cost curve coefficient

13. Repeat from Stage II.
The minimum total cost for reactive power through
generators’ voltage control can be yielded from the proposed
algorithm including consideration of power flow balance and
load voltage constraints.
IV. CASE STUDY
The proposed algorithm has been tested on IEEE-14 bus
system which has 5 generator buses (bus #1-slack bus, #2, #3,
#6 and #8) and 9 load buses. In this paper, following
conditions and assumptions are utilized to demonstrate the
numerical solutions.
⋅ The cost function is given in the form of quadratic curve in
order to compensate for the region of supply and demand
Table I shows the identical cost curve coefficient at each
generator.
TABLE I. GENERATOR’S COST CURVE COEFFICIENT
Bus #1 #2 #3 #6 #8
Cost curve
1 1 1 1 1
Coefficients
Table II shows the initial and final values of voltage,
reactive power and cost at each bus when all generator buses
have 1.02[pu] as the initial voltage. All the voltages at the load
buses should be maintained within the network constraints
( 0.98 [ pu ] ≤ Vd ≤ 1.02 [ pu ] ). The left table reveals the initial
value given as a power flow result, and the right table reveals
the final values after optimization with 16 iterations. All
generator voltages except bus 8 have been decreased below
1.0[pu] to reduce the total reactive power cost. But bus 8 has a
unique characteristic in terms of network topology that is
located near many load buses and it has an important role to
 5

maintain the load voltages within the load voltage constraints times as large as the other generators, and it can be expected
( 0.98 [ pu ] ≤ Vd ≤ 1.02 [ pu ] ). As shown in this table, the for the voltage at bus 8 to be changed compared to the case
voltage values from bus 10 to bus 13 are close to 0.98[pu] and that all generators have the identical cost curve coefficients.
bus 14 has a lower bound constraints voltage, therefore, Table IV shows the initial and final value of voltage,
generator at bus 8 has to increase its voltage to support load reactive power and cost at each bus and all the generator
voltages within constraints. voltages are also assigned to be 1.02[pu] as the initial voltage.
The reactive power at each generation bus has been also All the voltages at the load bus are maintained within the
decreased but bus 8 as shown in the table II. So the total cost inequality constraints ( 0.98 [ pu ] ≤ Vd ≤ 1.02 [ pu ] ) as shown.
has been decreased from 158.33 to 6.43 through the proposed The left table indicates the initial value as a result of power
algorithm. flow, and the right table indicates the final value after 15
iterations. All generators’ voltages have been decreased from
TABLE II. INITIAL AND FINAL VOLTAGE, REACTIVE POWER AND COST 1.02[pu] to around 1.0[pu]. But, compared to the case that all
Initial values Final values
Bus V(pu) Q(pu) Cost Bus V(pu) Q(pu) Cost generators have the identical cost curve coefficients, the
1
2
1.0000
1.0200
-0.9247
0.7000
85.51
49.00
1
2
1.0000
0.9969
0.1692
0.0300
2.86
0.09
voltage at bus 8 has been decreased from 1.0289[pu] to
3 1.0200 0.4037 16.30 3 1.0007 0.0516 0.27 1.0036[pu] as shown in the Table IV. This is why the
4 0.9996 0.0390 4 0.9981 0.0390
5 0.9997 -0.0160 5 0.9953 -0.0160 coefficient of bus 8 changed from 1 to 5 and it has 5 times
6 1.0200 0.2603 6.78 6 0.9926 0.1124 1.26 expensive than other generators, so generator at bus 8 has to
7 1.0050 0 7 1.0050 0
8 1.0200 0.0868 0.75 8 1.0289 0.1397 1.95 drop its voltage as low as possible to reduce the cost to
9 0.999 0.0240 9 0.9951 0.0240
10 0.9956 -0.0580 10 0.9892 -0.0580
produce reactive power within the load voltage constraints
11 1.004 -0.0180 11 0.9887 -0.0180 ( 0.98 [ pu ] ≤ Vd ≤ 1.02 [ pu ] ).
12 1.004 -0.0160 12 0.9854 -0.0160
13
14
0.9986
0.9803
-0.0580
-0.0500
13
14
0.9827
0.9800
-0.0580
-0.0500
And the reactive power at each generator bus has been also
Total Cost 158.33 Total Cost 6.43 decreased. And the total cost has been decreased from 158.33
to 7.30 by the proposed algorithm. The total cost has been
Fig. 3 shows generator voltages at each iteration step. increased because of the change of cost curve coefficient at
During iteration step 1 through 3, the voltage at bus 8 has a bus 8.
rapid increase and decrease, after that, the value is converged
stably to 1.028pu. That is why the minimum total cost is TABLE IV. INITIAL AND FINAL VOLTAGE, REACTIVE POWER AND COST
Initial values Final values
computed by linearization process, so the voltage resulted Bus V(pu) Q(pu) Cost Bus V(pu) Q(pu) Cost
from linearization process has an error compared to actual 1 1.0000 -0.9247 85.51 1 1.0000 0.1474 2.17
2 1.0200 0.7000 49.00 2 0.9981 0.0475 0.23
value. And the same explanation can be applies to the case of 3 1.0200 0.4037 16.30 3 1.0041 0.0804 0.65
4 0.9996 0.0390 4 0.9978 0.0390
bus 3. 5 0.9997 -0.0160 5 0.9962 -0.0160
6 1.0200 0.2603 6.78 6 1.0006 0.1781 3.17
7 1.0050 0 7 0.9954 0
Generator Voltages at each step 8 1.0200 0.0868 0.75 8 1.0036 0.0466 0.22
9 0.999 0.0240 9 0.9904 0.0240
1.05
10 0.9956 -0.0580 10 0.9867 -0.0580
1.04 11 1.004 -0.0180 11 0.9912 -0.0180
1.03 12 1.004 -0.0160 12 0.9927 -0.0160
13 0.9986 -0.0580 13 0.9889 -0.0580
1.02
14 0.9803 -0.0500 14 0.9800 -0.0500
Voltage(pu)

1.01 Total Cost 158.33 Total Cost 7.30

1

0.99
bus 2 bus 3 bus 6 bus 8
0.98
Generator Voltages at each step
1.03
0.97

0.96 1.02
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Iteration
1.01
Voltage(pu)

bus 2 bus 3 bus 6 bus 8

Fig. 3. Generator Voltages at each step 1

0.99

B. Case 2 : different cost curve coefficient 0.98

Table III shows the different cost curve coefficient at the 0.97
generator. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Iteration
bus 2 bus 3 bus 6 bus 8
TABLE III. GENERATOR’S COST CURVE COEFFICIENT Fig. 4. Generator Voltages at each step
Bus #1 #2 #3 #6 #8
Cost curve
Coefficients
1 1 1 1 5 Fig. 4 shows generators’ voltages at each iteration step. At
iteration step 3, the voltage at bus 3 had the minimum voltage
The reactive power coefficient at bus 8 is assigned to be 5 and after that, it has been increased to 1.0041[pu] stably. That

is why linearization process caused an error compared to
actual value resulted from the nonlinear constraint.
V. CONCLUSIONS
This paper suggests algorithm that compute the actual
pricing of reactive power by directly controlling generators’
voltage and optimum generators’ voltage profiles to minimize
the reactive power pricing are decided from the algorithm. The
summary is as follows.
▪ The proposed algorithm is divided by 3 stages. Stage I is
the step to yield the necessary initial value for minimization
of total reactive power cost, and the optimized values of
generators’ reactive power are obtained in stage II by
iterative method. Stage III is the step to update voltage
values and Jacobian in each bus to reduce the error by
linear analysis.
▪ In the case that the coefficients of all generators’ reactive
power cost are same, generators’ optimal voltage profiles
can be decided to minimize reactive power cost within the
voltage range at load buses that is the network constraint.
▪ In the case that the coefficients of generators’ reactive
power cost are different, generators’ voltage are adjusted
so that the fewest reactive power is produced in the most
expensive generator. This can be similar to economic
dispatch concept into reactive power allocation.
In some countries, like Korea, ISO (Independent System
Operator) receives bids about only active power. And it offers
reactive power as ancillary service according to system
situations and its pricing system is not still defined reasonably.
Therefore, it is necessary that the reactive power market
systems may have to be constructed not through ancillary
service but through bids similar to active power. If these
systems are established, the proposed algorithm can be
applied as the fundamental principles for deciding reactive
power and assessing its pricing under deregulation
environment. Furthermore, it may provide the fair and
reasonable price information for reactive power, since it sets
up a scheme to compute the actual pricing of reactive power
by directly controlling generators’ voltages.
VI. REFERENCES
[1] U. S. – Canada Power System Outage Task Force, “Final report on the
August 14, 2003 Blackout in the United States and Canada : Causes and
Recommendations”, April 2004.
[2] Kankar Bhattacharya, Jin Zhong, “Reactive Power as an Ancillary
Service”, IEEE Transactions on Power Systems, Vol. 16, No. 2, May
2001.
6
[3] John W. Lamont, Jian Fu, “Cost Analysis of Reactive Power Support”
IEEE Transactions on Power Systems, Vol. 14, No.3, August 1999
[4] A.A. El-Keib, X. Ma et al, “Calculating Short-Run Marginal Costs of
Active and Reactive Power Production”, IEEE Transactions on Power
Systems, Vol. 12, No. 2 May 1997.
[5] Huang, G.M.; Zhang, H., "Pricing of generators reactive power delivery
and voltage control in the unbundled environment," Power Engineering
Society Summer Meeting, 2000. IEEE , Vol. 4 , 16-20 July 2000
[6] D. Kirschen, R. Allan et al, “Contributions of Individual Generators to
Loads and Flows”, IEEE Transactions on Power Systems, Vol. 12, No. 1
February 1997.
[7] S.Hao, A. Papalexopoulos, “Reactive Power Pricing and Management”,
IEEE Transactions on Power Systems, Vol. 12, No. 1 February 1997.
[8] Jin Zhong, Kankar Bhattacharya, “Reactive Power Management in
Deregulated Electricity Markets – A Review”, Power Engineering Society
Winter Meeting, 2002. IEEE , Vol. 2 , 27-31 Jan. 2002
pp. 1287 - 1292
[9] P.Kunder, “Power System Stability and Control”, McGraw-Hill, Inc.,
pp.191-197, 1994.
VII. BIOGRAPHIES
Seung-Wan Jung was born in Korea, on May 31,
1979. He received the B.S. degree from Korea
University, Korea in 2003. Now he is a M.S. student
with Seoul National University, Korea. His research
field of interest includes HVDC and power system
operation.
Sung-Hwan Song was born in Korea, on May 30, 1978.
He received the B.S. degree from Pusan National
University, Korea in 2001, and M.S. degree from Seoul
National University, Korea in 2003. Now he is a Ph.D.
student with Seoul National University, Korea. His
research field of interest includes the power system
operation in a deregulated environment and FACTS.
Yong Tae Yoon was born in Korea, on April 20, 1971.
He received the B.S. degree, M.Eng. and Ph.D. degrees
from M.I.T., USA in 1995, 1997 and 2001, respectively.
Currently, he is an assistant professor in the School of
Electrical Engineering at Seoul National University,
Korea. His special field of interest includes electric
power network economics, power system reliability, and
the incentive regulation of independent transmission
company.
Seung-Il Moon was born in Korea, on February 1, 1961.
He received the B.S. degree from Seoul National
University, Korea in 1985 and M.S. and Ph.D. degrees
form The Ohio State University in 1989 and1993,
respectively. Currently, he is an associate professor of
School of Electrical Engineering at Seoul National
University, Korea. His special field of interest includes
analysis, control and modeling of the power system,
FACTS and power quality.