Académique Documents
Professionnel Documents
Culture Documents
Abstract — This paper presents the assessment schemes of strengthening of reactive power and voltage control practices
reactive power pricing through generators’ voltage control under is mentioned. This is the evidence that reactive power is
the assumption that a reactive power market is decomposed into a significant factor in several of the outages like North America
real power market. Applying the concept of economic dispatch to blackout in 2003.
minimize the total cost of real power, the algorithm for
implementing a reactive power pricing is proposed to determine
Consequently, reactive power plays an important role for
the generators’ optimum voltage profiles. It consists of the operating power systems stably, keeping the bus voltages
reactive power-voltage equation, the objective function which within nominal ranges, and supporting the real power transfer.
minimizes the total cost of reactive power, and linear analysis of In spite of its importance, financial compensation scheme for
inequality constraints in relation to the voltages at the loads. providing reactive power of generator is still not well
From this algorithm, the total cost of the reactive power can be determined, compared with real power.
yielded to the minimum value within network constraints. It may
provide the fair and reasonable price information for reactive
Of course, the problem connected with reactive power is
power in deregulated electricity market. The proposed algorithm classified as ancillary service in some part of advanced nations
has been tested on the IEEE 14-bus system using MATLAB. that introduce the electricity market mechanism through
restructuring and the reactive power pricing system is formed
Index Terms — Reactive Power Pricing, Power Flow Equation, in other structures in the each country. New York ISO,
Optimization Problem, Linear Analysis California ISO (US), and NEMMCO (Australia) are
constructing the pricing system that compensates
I. INTRODUCTION
economically revenue decrease by decrease in active power
≤ 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)
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 ⎤
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
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. like capability price structure and utilization price structure
Iteration steps are performed to correct the error generated by in NGC (United Kingdom) [8]. So, the total cost can be
linear analysis of nonlinear constraints in Stage III. calculated with cost curve coefficient product reactive
power at the generator. The total cost equation is as
The detailed algorithm can be explained as following: 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
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.
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)
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)
0.99
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
6
is why linearization process caused an error compared to [3] John W. Lamont, Jian Fu, “Cost Analysis of Reactive Power Support”
IEEE Transactions on Power Systems, Vol. 14, No.3, August 1999
actual value resulted from the nonlinear constraint. [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.
V. CONCLUSIONS [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
This paper suggests algorithm that compute the actual [6] D. Kirschen, R. Allan et al, “Contributions of Individual Generators to
Loads and Flows”, IEEE Transactions on Power Systems, Vol. 12, No. 1
pricing of reactive power by directly controlling generators’ February 1997.
voltage and optimum generators’ voltage profiles to minimize [7] S.Hao, A. Papalexopoulos, “Reactive Power Pricing and Management”,
the reactive power pricing are decided from the algorithm. The IEEE Transactions on Power Systems, Vol. 12, No. 1 February 1997.
[8] Jin Zhong, Kankar Bhattacharya, “Reactive Power Management in
summary is as follows. Deregulated Electricity Markets – A Review”, Power Engineering Society
Winter Meeting, 2002. IEEE , Vol. 2 , 27-31 Jan. 2002
pp. 1287 - 1292
▪ The proposed algorithm is divided by 3 stages. Stage I is [9] P.Kunder, “Power System Stability and Control”, McGraw-Hill, Inc.,
the step to yield the necessary initial value for minimization pp.191-197, 1994.
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 VII. BIOGRAPHIES
values and Jacobian in each bus to reduce the error by
linear analysis. 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
▪ In the case that the coefficients of all generators’ reactive with Seoul National University, Korea. His research
power cost are same, generators’ optimal voltage profiles field of interest includes HVDC and power system
operation.
can be decided to minimize reactive power cost within the
voltage range at load buses that is the network constraint.