I. INTRODUCTION
Manuscript received May 02, 2011; revised August 29, 2011 and October 04,
2011; accepted November 02, 2011. Date of publication December 26, 2011;
date of current version April 18, 2012. Paper no. TPWRS-00404-2011.
S. Wang is with the Center for Energy Systems Research, Tennessee Technological University, Cookeville, TN 38501 USA.
W. Gao is with the Department of Electrical and Computer Engineering, University of Denver, Denver, CO 80208 USA (e-mail: Wenzhong.Gao@du.edu).
A. P. S. Meliopoulos is with the School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA 30332 USA (e-mail:
sakis.m@gatech.edu).
Color versions of one or more of the gures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identier 10.1109/TPWRS.2011.2175255
WANG et al.: ALTERNATIVE METHOD FOR POWER SYSTEM DYNAMIC STATE ESTIMATION
943
using
(6)
As can be seen from the above steps, the mean and covariance are calculated using standard vector and matrix operations;
it is not necessary to evaluate the Jacobians which are usually
needed in existing state estimation of nonlinear systems. This
means that the algorithm can be implemented rapidly and suitable for online application. Moreover, the algorithm can guarantee a higher order of accuracy than traditional linearization
algorithm [20][22]. The more detailed derivation and properties of the unscented transformation algorithm can also be found
in [20][22].
III. FILTERING ALGORITHM BASED
UNSCENTED TRANSFORMATION
ON THE
In the previous section, we introduced a method for determining the mean and covariance of a random vector that undergoes a nonlinear transformation. The method is superior to
linearization methods in many important respects [20][22]. In
this section, we describe how UT is embedded into the recursive
prediction and update structure of Kalman lter.
A system with nonlinear functions of state transition and observation models is shown in (7):
(3)
(7)
is the th row or column of the
where
is the weight which is
matrix square root of
associated with
. Here
is a scaling factor; is another parameter of the method,
if is Gaussian [20],
and can be chosen as
[22]; the matrix square root of positive denite matrix
means that a matrix
exists such that
.
2) Instantiate each point through the nonlinear function to
yield the set of transformed sigma points as shown in
(4)
3) Calculate the mean of
(3):
by (5), where
is dened by
(5)
where
is the state variable at the time step
is the measurement at the time step
and are vectors consisting of nonlinear functions;
is the Gaussian
is the
process noise at the time step , and
and
are
Gaussian measurement noise at the time step
covariance of and at the time step . Using UT described
in the previous section, the prediction and update steps of the
new lter algorithm can be formulated in standard matrix operations [23] as follows:
as the state
1) Prediction: according to (2)(6), use
mean to yield sigma points and calculate the predicted state
and the predicted covariance
shown in (8)
mean
is
at the bottom of the page, where
dened in (3),
is an
matrix,
is an
-dimensional vector.
as the state mean
2) Update: according to (2)(6), use
to yield sigma points and calculate the predicted mean
and covariance
of the measurement, and the cross-coshown in (9),
variance of the state and measurement
(8)
944
where
is dened in (10),
is an
mais similar to
trix, the operation of
in (8). Here is a parameter of the method, and
if
is Gaussian [19]. Then calculate the lter gain
and
and covariance
as shown
the updated state mean
in (11):
(10)
(11)
(16)
(13)
2) Calculate
(17)
(18)
by solving
(14)
WANG et al.: ALTERNATIVE METHOD FOR POWER SYSTEM DYNAMIC STATE ESTIMATION
945
946
result from the method in this paper. Fig. 8 gives the comparison of the estimation errors of two methods. Both methods use
the same measured variables with noises are shown in Fig. 9 in
which a Gaussian noise with a zero mean and variance of 0.001
p.u. added to the measured electric power, and another Gaussian
noise with a zero mean and variance of 0.01 rad/s added to the
measured rotor velocity.
It can be seen from Fig. 6 that a phase shift (lag) emerges
between the actual values and the estimated values from the
method in [6], while Fig. 7 shows that the new method can
still work well. This lag, when the estimated values are used
as feedback signals, can deteriorate controller performance, or
even threat stability of closed-loop systems [24][26].
B. Case 2
In this section, WSCC three-machine nine-bus system [27]
shown in Fig. 10 is studied as a case to test the proposed method.
In this example system, the dynamics of generators are represented with a 4th order model. The state variables are rotor angle
, velocity , and internal voltages
and . Each generator
has its own excitation system which is modeled with state variables:
. The entire system has 21 state variables, and
the state transition equations are given in (19) where
corresponding to the three generators, respectively. The related
parameters can be found in [27]. Now we discuss measurements
for this case study.
WANG et al.: ALTERNATIVE METHOD FOR POWER SYSTEM DYNAMIC STATE ESTIMATION
G.
947
G.
(19)
Given
G.
0.01 p.u. variance for electric power; 0.00001 p.u. variance for
; 0.001 rad for ;
rotor velocities; 0.0001 p.u. variance for
948
Fig. 15. Estimation results for rotor angle difference of the system.
G.
G.
G.
(20)
WANG et al.: ALTERNATIVE METHOD FOR POWER SYSTEM DYNAMIC STATE ESTIMATION
949
VI. CONCLUSION
Fig. 21. Performance under the data update rate of 80 ms.
Fig. 22. Performance under the data update rate of 100 ms.
Figs. 2022 in which the solid lines represent actual values and
the dashed lines represent the estimated values. From these
gures, we can see that the method is still able to maintain a
good performance for update rate up to 80 ms for this case
system; the method starts to show deteriorated performance for
update rates beyond 100 ms.
D. Bad Data Detection
Detecting and handling the effects of gross measurement
errors is another important task in power system state estimation. The unscented lter identies the gross errors through
a normalized innovation vector. Each element of the vector
corresponds to a unique measurement. The th measurement
corresponds to the th element shown in (21) where subscript
represents the th time step; subscript represents the th
measurement;
is the th diagonal element of
. The
other symbols are the same as dened in previous sections.
The results from [29] show that the measurement should be
is greater than 1.5:
discarded if
(21)
By letting
in (21), now we test if this index can work
well by adding a gross measurement error 0.05 p.u. to the measured rotor velocity of generator No. 1 at
s. We get
for
this measurement shown in Fig. 23 from which we can see the
gross error can be detected.
950
[9] A. P. Meliopoulos, G. J. Cokkinides, and G. K. Stefopoulos, Numerical experiments for three-phase state estimation performance and evaluation, presented at the 2005 IEEE PowerTech Conf., St. Petersburg,
Russia, Jun. 2730, 2005.
[10] S. Mohagheghi, R. H. Alaileh, G. J. Cokkinides, and A. P. S. Meliopoulos, Distributed state estimation based on the SuperCalibrator
conceptLaboratory Implementation, in Proc. IREP 2007 Symp. Bulk
Power System Dynamics and Control, Charleston, SC, Aug. 2007.
[11] A. P. Meliopoulos, G. J. Cokkinides, F. Galvan, B. Fardanesh, and
P. Myrda, Advances in the supercalibrator conceptPractical implementations, in Proc. 40th Annu. Hawaii Int. Conf. System Sciences
(HICSS), Waikoloa, Big Island, HI, Jan. 36, 2007.
[12] G. J. Cokkinides, A. P. Meliopoulos, R. Alaileh, and A. Mohan, Visualization and characterization of stability swings via GPS-synchronized
data, in Proc. 40th Annu. Hawaii Int. Conf. System Sciences (HICSS),
Waikoloa, Big Island, HI, Jan. 36, 2007.
[13] L. Zhao and A. Abur, Multiarea state estimation using synchronized
phasor measurement, IEEE Trans. Power Syst., vol. 20, no. 2, pp.
611617, May 2005.
[14] A. Ghassemian and B. Fardanesh, Phasor assisted state estimation
for NYS transmission system-implementation & testing, in Proc.
IEEE/PES Power Systems Conf. Expo., Seattle, WA, Mar. 1518, 2009.
[15] A. G. Phadke, J. S. Thorp, R. F. Nuqui, and M. Zhou, Recent developments in state estimation with phasor measurements, in Proc.
IEEE/PES Power Systems Conf. Expo., Seattle, WA, Mar. 1518, 2009.
[16] J. Amit and N. R. Shivakumar, Impact of PMU in dynamic state estimation of power systems, in Proc. 40th North American Power Symp.,
Calgary, AB, Canada, Sept. 2830, 2008.
[17] A. Z. Gamm, Y. A. Grishin, I. N. Kolosok, A. M. Glazunova, and E. S.
Korkina, New EPS state estimation algorithms based on the technique
of test equations and PMU measurements, in Proc. IEEE Power Tech,
Lausanne, Switzerland, Jul. 15, 2007, pp. 16711675.
[18] J. Zhu and A. Abur, Effect of phasor measurements on the choice of
reference bus for state estimation, presented at the IEEE/PES General
Meeting, Tampa, FL, Jun. 2428, 2007.
[19] S. Julier and J. Uhlmann, Unscented ltering and nonlinear estimation, Proc. IEEE, vol. 92, no. 3, pp. 401422, Mar. 2004.
[20] S. Julier, J. Uhlmann, and H. F. Durrant-Whyte, A new method for
the nonlinear transformation of means and covariances in lters and
estimators, IEEE Trans. Autom. Control, vol. 45, no. 3, pp. 477482,
Mar. 2000.
[21] S. Julier, The scaled unscented transformation, in Proc. American
Control Conf., Anchorage, AK, May 810, 2002.
[22] S. Julier, J. Uhlmann, and H. F. Durrant-Whyte, A new approach for
lter nonlinear systems, in Proc. American Control Conf., Seattle,
WA, Jun. 2123, 1995.
[23] S. Srkk, Recursive Bayesian inference on stochastic differential
equations, Ph.D. dissertation, Laboratory of Computational Engineering, Helsinki Univ. Technol., Helsinki, Finland, 2006.
[24] H. Wu, K. Tsakalis, and G. Heydt, Evaluation of time delay effects to
wide-area power system stabilizer Design, IEEE Trans. Power Syst.,
vol. 19, no. 4, pp. 19351941, Nov. 2004.
[25] B. Chaudhuri, R. Majumder, and B. Pal, Wide-area measurement
based stabilizing control of power system considering signal transmission delay, in Proc. IEEE/PES General Meeting, San Francisco, CA,
2005, pp. 17.
[26] B. Naduvathuparambil, M. Valenti, and A. Feliachi, Communication
delays in wide area measurement systems, in Proc. 34th Southeastern
Symp. System Theory, Huntsville, AL, 2002, pp. 118122.
[27] P. Sauer and M. A. Pai, Power System Dynamics and Stability. Englewood Cliffs, NJ: Prentice-Hall, 1998.
[28] Q. Yang, T. Bi, and J. Wu, WAMS implementation in china and the
challenges for bulk power system protection, in Proc. IEEE Power
Eng. Soc. General Meeting, Tampa, FL, Jun. 2428, 2007.
[29] G. Valverde and V. Terzija, Unscented Kalman lter for power
system dynamic state estimation, IET Gen., Transm., Distrib., vol. 5,
pp. 2937, 2011.