You are on page 1of 4

2011 International Conference on Mechatronic Science, Electric Engineering and Computer

August 19-22, 2011, Jilin, China

A New Algorithm for Power System State


Estimation with PMU Measurements

Defu Ni Bing Yu
Electric Engineering Department Luoyang Power Supply Company
Electrical Engineering Northeast Dianli University Electric Power Company of Henan
Jilin, China Luoyang, China
e-mail: nidefu-84@163.com e-mail: yubingdream@sohu.com

Weimin Zhang Weiguo Gong


Power Education Center Electric Research Department
Electric Power Company of Yunnan Xuji Transformer Co. Ltd
Yunnan, China Henan, China
e-mail: ksks66@163.com gwgssh@xjgc.com

Abstract—With the development of power systems, the scale of angle measurements from PMU into vector set of the
grids has become more and more complex. Real-time traditional state estimation. Although just made a less
monitoring and control of the grid are also becoming more and changes of the program and easy to implement. However, for
more difficult. GPS-based synchronous Phasor Measurement the use of PMU measurements is not high, PMU's advantage
Unit (PMU) is a new high-precision measuring device, which cannot be fully realized. Paper [4] takes boundary conditions
can measure the bus voltage vector. The use of PMU has made as an equality constraint to improve the velocity of the state
great changes in the power system. However, how to use the estimation in distributed state estimation. In this paper, we
PMU measurement data reasonably, is the scholar most take the PMU measurements as an equality constraint in state
concerned about. A new algorithm has been put out based on
estimation of the power system for it high precision and high
the high precision of the PMU measurements as the equality
reliability. This algorithm can improve the accuracy of the
constraints to the power system. The experimental results show
that the algorithm improved the accuracy of state estimation of state estimation, and do not have to change the original state
power system, there is a strong practical. estimation module, simply on the basis of the original plus a
module, and it has a strong compatibility.
Keywords- State Estimation; Phasor Measurement Unit;
Equality Constraints; power system II. MATHEMATICAL MODEL

A. The weighted least squares state estimation model


I. INTRODUCTION
For an N node system, given the frame of network, line
The current power system state estimation which is parameters and measurement system, power system state
usually the weighted least squares (WLS) method as the estimation measurement equation can be expressed as
criterion. The main algorithms: Basic weighted least squares z = h( x ) + v (1)
state estimation, the weighted least square state estimation,
based on the rapid decomposition least squares, orthogonal Where z is the m×1 measurement vector, h(x) is a m×1
transformation method, and the state estimation of equivalent vector of nonlinear functions, x is the n×1 state vector, and v
measurement transformation. This series of algorithms are is the m×1 measurement error vector, with zero mean E(v)=0,
based on the least square method. Weighted least squares and diagonal covariance matrix R=Cov(v)E(vvT). And m is
method it has a big advantage is that its coefficient matrix is the number of the measurement, n is the number of the state
symmetric and positive definite, it make a big convenience of values, n=2N-1.
the after work [1]. The objective function of the weighted least squares
But the traditional basic weighted least squares have so (WLS) state estimation is
many deficiencies. At each iteration process, the Jacobian J ( x) = [ z − h( x)]−1 R −1[ z − h( x)] (2)
matrix needs to re-form every time, so the efficiency of the For the minimization of the objective function, from (2)
algorithm is not high. To improve these shortcomings,
optimize algorithm, many experts and scholars have made we can obtain that
some improvements in the algorithm. However, they are ­Δx ( k ) = ( H T R −1H ) −1 H T R −1ΔZ
more or less has some shortcomings. Paper [2] proposed a ® (3)
full use of PMU measurements in state estimation, although ¯ X ( k +1) = x k + Δx k
this method is simple, fast, but considering the economics of Where H is the Jacobian matrix, and
the power system, it cannot be realized in a long period of
time in China. Paper [3] proposed an algorithm that put the

978-1-61284-722-1/11/$26.00 ©2011 IEEE 114


∂h( x) ªGA BT º ªΔx º ª H T R −1Δz º
CT
H ( x) = (4)
∂x « » « »
«C 0 0 » ««Δc »» = « Δc » (14)
B. Take the measurement of PMU as equality constrain «B
¬ 0 E »¼ ¬« 0 ¼» «¬ 0 »
¼
The values of voltage magnitudes and Phasor angles from
the PMU, for its high precision and reliability, can be used as Where μ = − BΔx = − Δx p , is the correction vector, and
an constraint of the estimation to correct the state variables. E is the identity matrix.
So the state estimation problem is formulated as an equality
constrained WLS minimization problem. Constraints are: G A = G + BT B (15)
c( x) = Ax = zc (5) Absolutely, the matrix GA is positive definite and
nonsingular and its Cholesky decomposition is:
Where A is the l×n matrix, l is the number of the PUM. So
the objective function is come to GA = L11LT11 (16)
min{J ( x) = [ z − h( x)]−1 R −1[ z − h( x)]} (6) ª L(′n − p )×( n − p ) 0 ( n − p )× p º
L11 = «
S .t. c( x) = zc (7) ¬ L′p′× ( n − p ) E p × p »¼
The extreme value function of (6) and (7) is
ª0 ( n − p ) × ( n − p ) 0( n − p ) × p º
L( x) = J ( x) + λT c( x) (8) =L+« (17)
We can obtain the minimization with equality constraint ¬ 0 p×( n − p ) E p× p »¼
by Lagrange multiplier method. We need to make a Cholesky factorization to the
Where L(x) is the objective function with equality coefficient matrix
constraint, Ȝ is the l×1 Lagrange multiplier. According to the
steps for solving extremum, we can obtain that
ªGA CT BT º ª L11 0 0 º
« » «
ª H T R −1H C T º ªΔx º ª H T R −1Δz º «C 0 0 » = « L21 − L22 0 »»
« »« » = « » (9) «B 0 E »¼ «¬ L31 − L32 L33»¼
¬ C 0 ¼ ¬ λ ¼ ¬ − Δc ¼ ¬
Or ª LT11 LT21 LT31 º
­ H T R −1H ⋅ Δx + C T = H T R −1Δz « »
(10)
⋅ « 0 LT22 LT32 » (18)
®
¯ C ⋅ Δx = −Δc «0 0 LT33 »¼
¬
∂c( x) From (18) we can get
Where C is the l×n matrix, and C = , − Δc is
∂x ªG C T º ª L11 0 º ª LT11 LT21 º
residual vector of the PMU measurements. U =« A =
» « »« » (19)
¬C 0 ¼ ¬ L21 − L22 ¼ ¬ 0 LT22 ¼
∂h( x)
Δz = z − h(x) , Δc = zc − c(x) , H= Combining (12) and (16) we have
∂x −1 T
L11 B = BT (20)
∂c( x) Since the rows of B are linearly independent from the
C= is the Jacobian matrix, we set
∂x rows of H, we have the following properties
−1 T
ª H T R −1H C T º LT31 = L11 B (21)
T −1
U =« », G = H R H . T −1 T
L =L L L (22)
¬ C 0¼ 32 22 21 31
T T
For an observable system, the coefficient matrix is L L =L L
33 33 32 32 (23)
nonsingular. To get the solution of (8), we need to make a Combining (19) (20) (21) we have
Cholesky of HTR-1H .
T LT31 = BT (24)
G = LL (11)
T −1 T
The L has following form: L = L L21B
32 22 (25)
ª L(′n − p )× ( n − p ) 0( n − p ) × p º The linear system (13) can be solved by a forward and a
L=« (12) back substitution as follows:
¬ L′p′×( n − p ) 0 p× p »¼
ª L11 0 º ª y1 º ª H T R −1Δz º
We define the matrix as:
»« y » = « » (26)
[
B = 0 p ×( n − p ) E p× p ] (13)
«L
¬ 21 − L22 ¼ ¬ 2 ¼ ¬ Δc ¼
From [5] we know that (8) is equal to: L33 y3 = L32 y2 − By1 (27)
L μ = y3
T
33 (28)

115
In order to verify the performance of this algorithm,
ª LT11 LT21 º ªΔx º ª y1 − BT μ º comparing with the result of properties of the traditional
« »« » = « » (29)
¬0 LT22 ¼ ¬ λ ¼ ¬ y2 − l32
t
μ¼ weighted least squares (WLS) method.
It is to be noted that G has full rank, then B = 0, ȝ=0, and
(13) equal to (8), the solution of (8) would be obtained. TABLE III. THE COMPARATIVE RESULT OF PROPERTIES OBTAINED BY
THE TWO DIFFERENT ALGORITHM (IEEE-14 NODE NETWORK)

III. NUMERICAL SIMULATION P. deviation


Run time Iterations
The IEEE 14-bus system is used as the test bed for our P Q
algorithm. IEEE 14-bus system and associated measurement
Est. 113 5 0.0041 0.0225
configuration as follows: WLS 99 4 0.0002 0.0009

Table 2 shows that the algorithm has great improved the


precision for the use of measurements from the PMU as the
equality constraint, because of the high-precision of the PMU
measurements. From the table 3 we know, for the increasing
number of measurements, result to the increase of the system
redundancy. It makes the run time longer for the increase of
the dimension of the Jacobian matrix. Although, we got the
power deviation decreased clearly. So it is worth to sacrifice
a little time to get more high-precision in the system.
IV. Conclusion
In this paper, we presented a method by use of the high-
precision of the PMU measurements to solution of the WLS
state estimation problem with equality constraints. In the
proposed method the coefficient matrix is sparse, symmetric
. and indefinite, but has a unique singed-Cholesky
factorization. The proposed algorithm is numerical stable,
Figure 1. IEEE 14-bus system configuration
simple to implement. The algorithm is illustrated with the
IEEE 14-bus system.
The bus 5, 7 and 11 has equipped with PMU device. We The established algorithm for use of PMU measurement
compared this algorithm with the weighted least squares data as an equality constraint in power system state
(WLS). Table 1 shows the values of the bus measurements estimation, can improve the system precision at the expense
in IEEE 14-bus system. Table 2 shows that it takes PMU of less efficient. And it does not change the original state
measurements as equality constraint has more high-precision estimation module, simply on the basis of the original plus a
compared with traditional weighted least squares in Phasor simple module. The algorithm also has a high flexibility.
angle of the bus line. Table 3 shows comparative result of
properties obtained by the two different algorithms. REFERENCES

[1] Yu Erkeng. “State Estimation for Electric Power System,” Bejing:


TABLE I. MEASUREMENT VALUES Hydraulic Power Publishing Company, 1985
true meas. true meas. [2] PHADKE A G, THORP J S, KARIMI K J. “State Estimation with
p1-2 0.0903 0.1129 p2 0.2496 0.2205 Phasor Measurements.”IEEE Trans on Power System, 1986, 1(1):
p1-5 0.1289 0.1106 p3 -0.2105 -0.1998 233-241.
p4-7 0.2879 0.2507 p4 -0.3619 -0.3178 [3] THORP J S. PHADKE A G. KARIMI K J. “Real Time Voltage
p4-9 0.7260 0.6811 p9 -0.0628 -0.0649 Phasor Measurements for Static State Estimation.” IEEE Trans on
p6-11 0.0108 0.0207 p10 -0.0082 -0.0136 Power Apparatus and Systems, 1985, 104 (11): 3098-3106.
p7-8 -0.0039 0.0106 p12 -0.1326 -0.1602 [4] Li Yanglin, Wei Zhinong, Wan Junbiao, “A new algorithm for the
p7-9 0.3106 0.2855 p13 -0.0129 -0.0419 distributed state estimation of power systems,” Relay, 2007, 35(20):
p1 0.2659 0.2601 p14 -0.0372 -0.0300 13-16.
[5] F.C. Aschmoneit, N.M. Peterson, and E.C. Adrian, “State estimation
TABLE II. THE COMPARATIVE RESULT OF THE TWO DIFFERENT with equality constraints,” In Proc. 10th PICA Conf., May 1977, pp.
ALGORITHM’S PHASE ANGLE (IEEE-14 NODE NETWORK) 427-430.
true Est. WLS true Est. WLS [6] F.F. Wu, W. -H, E, Liu, and S. –M. Lun, “Observability analysis and
į2 -5.11 -6.23 -6.48 į9 -14.20 -12.89 -12.44 bad data processing for state estimation with equality constraints,”
į3 -12.23 -12.10 -12.45 į10 -14.78 -12.78 -15.62 IEEE Trans. Power Syst,. Vol. 3, no. 2, pp. 541-548, May 1988.
į4 -11.75 -12.57 -13.48 į11 -15.12 -15.96 -16.32 [7] LI Qiang, ZHOU Jingyang, YU Erkeng, et al. “Power system linear
į5 -9.56 -9.01 -8.18 į12 -15.06 -14.34 -14.21 state estimation based on Phasor measurements.” Automation of
į6 -13.87 -12.28 -11.20 į13 -14.98 -13.49 -12.10 Electric Power Systems, 2005, 29(18): 18-24.
į7 -12.96 -10.45 -10.10 į14 -14.87 -14.01 -15.43 [8] LI Qiang, ZHOU Jingyang, YU Erkeng, et al. “A hybrid algorithm for
į8 -13.00 -14.38 -12.27 power system state estimation based on PMU measurements and

116
SCADA measurement.” Automation of Electric Power Systems,2005, [10] G.N.Korres, “A new method for treatment of equality constraints in
26(2): 58-63 power system state estimation,” in Proc. 16th PSCC, Glasgow, U.K.,
[9] F.Alvarado and W. Tinney, ”State estimation using augmented Jul. 14-18,20.
blocked matrices,” IEEE Trans. Power Syst., vol. 5, no. 3, pp. 911-921,
Aug,1990.

117

You might also like