You are on page 1of 4

Papcr accepted for prcscntation at PPT 2001

200 1 IEEE Porto Power Tech Conference


IOth -13Ih Septcrnber, Porto, Portugal

Modeling Unknown Circuit Breakers in Generalized


State Estimators
Antonio de la Villa J a h , Antonio Gbmez Expbsito
Abstmct- Modeling unknown circuit breakers in state estimation by adding their power flows t o t h e state vector
leads t o estimated values which do not necessarily reflect
feasible statuses. In this paper, this problem is overcome by
adding two equality constraints so that the state estimator
can only converge t o either of the two mutually exclusive
statuses.
Keywordstimators.

Circuit Breaker Models, Generalized State Es-

I. INTRODUCTION

TATE estimators make use of measurement redundancy to obtain the best state vector. Subsequent
analysis of measurement residuals allows detection and
identification of bad data in the measurement set.
In conventional estimators the Topology Processor (TP)
program builds a bus-branch model using circuit breaker
(CB) information and network connectivity data [ll].However, errors in CB statuses may lead the T P to generate
an incorrect network model. Such topology errors, in turn,
give rise to large residuals in adjacent measurements [3],
[161.
More recently, the so-called generalized estimators [13],
[2] perform a detailed modeling (bus section/ switch level)
of suspected substations in which all branches are included.
The objective is to be able to detect and identify topology
errors which, otherwise, would appear as several interacting bad data. An exclusion error [12] arises when a line
actually in service is deemed as open. The opposite, i.e.,
line disconnected apparently closed, is termed an inclusion
error. Similar splitting/merging errors are posible for CBs
connecting bus sections. In order to achieve this enhanced performance the state vector is augmented with power
flows through all CBs included in the model and additional
constraints representing the status of CBs are enforced [7],
[SI, [9]. The normalized residuals and Lagrange multipliers
provided by the generalized state estimator can then be
used to detect and, if redundancy permits, identify topology errors PI, [41, [io], PI.
Any series branch (line/tranformer) whose edge CBs are
assumed t o be open is excluded from the model in conventional state estimators, but should be handled by generalized state estimators. Sometimes, the status of a branch
is unknown because the information pertaining to its associated CBs is missing or ambiguous. When this happens,
additional checks based on adjacent measurements and CB
statuses are performed a priori by the TP in conventional
A. Villa and A. G6mez are with the Department of Electrical Engineering, University of Sevilla, Sevilla, Spain. Email: avillaOesi.us.es,
expositoQcica.es.

0-7803-7 139-9/01/$10.00 0200 I IEEE


~~

~~~

estimators in order t o reach a conclusion on the most plausible branch status. This extra logic is unnecessary if the
power flows through the unknown branch are added to the
state vector, but then the local redundancy deteriorates as
no new information compensates for these additional variables.
The customary two-stage procedure, in which a standard
state estimator is first run and then suspected substations
are modeled in detail, is not always successful, particularly
in the presence of exclusion errors. This has led some researches to develop alternative methods [14], [15], [SI intended somewhat to reinforce the performance of conventional
TP. The goal under this approach is t o determine, as accurately as possible, the actual network topology before
running the state estimator, for which the past history and
approximate linear models are useful.
11. CIRCUITBREAKER

MODELING

The status of a circuit breaker (CB) connected between


bus sections i and j can be modeled by adding the power
flows Pij, Q i j to the state vector [8], [2]. If the CB is
assumed open, then the constraints

Pij = 0
Qij

=0

should be added to the model, whereas

must be enforced when it is assumed closed. The advantage of doing so is that bad data detection and identification techniques can be applied to the augmented model by
means of the normalized multipliers [5], [4]. This way, the
risk for a hidden topology error to appear as several wrong
measurements is prevented.
No constraints are imposed, however, when there is no
information about the CB status. In such cases, owing to
measurement errors, the estimated power flow through an
open CB and the estimated voltage drop across a closed
CB are not exactly zero. This extra degree of freedom
(i.e., decreased redundancy) leads t o less accurate estimated measurements.

TABLE I

CASE1: MEASUREMENTS

Magnitude
c

I
1

-6

Open
Unknown

Fig. 1. Substation example

111.

P Flow 18-5
Q Flow 18-5
P Flow 19-2
Q Flow 19-2
P Inject. 15
Q Inject. 15
P Inject. 16
Q Inject. 16

Measurements
Exact
Noisy
1,13261 1,15339
0,05030 0,04530
0.78490 0.75881
-0,09739 -0,09602
1.13261 1.14257
0,05029 0,04864
0.78490 0.81700
-0,09739 -0,10673

PROPOSED TECHNIQUE

The key idea is to enforce the following two constraints


for every unknown CB.

TABLE I1
CASE

1:

Magnitude

This way, the state estimator is forced to converge to either of the two mutually exclusive CB statuses, avoiding the
"almost" closedlopen status that would otherwise result.
When building the Jacobian, the following new terms
must be accomodated:

P Flow 18-5
Q Flow 18-5
P Flow 19-2
Q Flow 19-2
P Inject. 15
Q Inject. 15
P Inject. 16
Q Inject. 16

ESTIMATED VALUES

Estimated Values
Conv. I ProDosed
1,14798
0,04697
0,04358
0.76808
0.78790
-0,09799 -0,10138
1.14798
1.12816
0,04697
0,05036
0,80772
0.78790
-0,10138

A . Case 1: Unknown CB actually open.


Gaussian noise is added to exact measurements obtained from a load flow. Tables I and I1 show the exact and
noisy power measurements related with substation 1 as well
as the estimated values when the conventional approach is
These terms are either null or very small at flat start,
which means that they are useful only after the first iteration.
Note that the status of a CB forming a loop with closed
CB's is irrelevant. The same happens with a CB belonging
t o a cut-set where the remaining CB's are open. In such
cases, there is no need to add extra variables and constraints.

Iv. TESTRESULTS
The proposed method has been tested by modeling in
detail several substations of the IEEE l 4 b u s network. R e
sults corresponding to substation 1, modeled as shown in
figure 1, will be presented.
Note that, in addition to the original bus 1, buses 15
to 19 must be modeled in order t o be able t o check for
topology errors. The six resulting buses contrast with the
two electrical nodes generated by conventional Topology
Processors if the unknown CB is finally labeled as open.
The set of measurements comprises all bus voltages and
injections plus the power flow at one edge of all lines.
Two cases are considered.

adopted and when the proposed constraints are enforced.

Table I11 shows the estimated values corresponding to


the state variables related with CB 15-16. Note that, when
the two constraints are not employed, the power flow across
CB 15-16 is not null.
Obviously the results depend on the accuracy of available measurements. In order to show this influence, the
experiment is repeated 200 times for different accuracy levels and the results corresponding to P15-16 are shown, in
the form of histograms, in figures 2 and 3.
In all cases, for the same measurement values, the estimated value of P15-16 is null when the proposed constraints
are enforced after the second iteration.
These results confirm that using the proposed constraints
allows accurate identification of the CB status, which may
not be clearly defined otherwise.
Figure 4 represents the average error of the estimated
measurements versus the input measurement error, for the
set contained in table I and different accuracy classes. In
turn, every point in the diagram is the average of the 200
experiments, and the relative errors refer t o full scales of
1.2 P.u., 2 p.u. and 0.2 p.u. for voltage, active and readive power measurements respectively. Clearly, the filtering

TABLE 111

60

BREAKER
15-16

CASE 1: STATE VARIABLES RELATED TO CIRCUIT

Method

z.2

-0.15

-0.1

-0.05

0.05

0.1

0.15

Conv.

1 Proposed

0.2

Active power flow

Fig. 2. Flow

p1&16

histogram with 0.8 % mean measurement error


0

0.5

1,5

2.5

3,5

46 Measurement error
50 -

+Conventional --cWith constraints


40 -

Fig. 4. Case 1: Estimated measurement error

.g

g30-

V. CONCLUSIONS

t
-

Active I)owerfIow

Fig. 3. Flow

P15-16

histogram with 3 % mean measurement error

capability of the state estimator is enhanced when the constraints are added.

When doubtful circuit breakers are included in the state


estimation, the measurement noise leads to estimated statuses which are not fully compatible with the true status.
In other words, the power flow (voltage drop) through open
(closed) devices is not null.
Two quadratic constrains are proposed in this paper t o
prevent this problem. Experiments show that more accurate results are obtained when the proposed technique is
employed, particularly when the local redundancy is low.
In the presence of severe bad data, a few cases have been
detected in which the wrong status is reached or the active/reactive submodels converge t o different statuses. This

B. Case 2: Unknown CB actually closed

A new set of measurements, compatible with the closed


status, is obtained. Table IV and V are the counterpart of
tables I and I1 for this case.
Table VI shows that the voltage drop across CB 15-16 is
not null when the two constraints are not employed, leading
t o different power flow values.
The same set of 200 experiments discussed in case 1 is
performed for this case. Figure 5 is the counterpart of
figure 4. Again, the noise filtering capability is improved
when the proposed constraints are added, and the other
conclusions of case 1 remain valid.

Magnitude

P Flow 18-5
Q Flow 18-5
P Flow 19-2
Q Flow 19-2
P Inject. 15
Q Inject. 15
P Inject. 16
Q Inject. 16

Measurements
Exact
Noisy
0,75507 0,75467
0,03353 0,03212
1,56979 1,65770
-0,19478 -0,18813
0,46497 0,46906
-0,03225 -0,03331
1,85989 1,93596
-0,12900 -0,14137

TABLE V
CASE2: ESTIMATED
VALUES

Magnitude

P Flow 18-5
Q Flow 18-5
P Flow 19-2
Q Flow 19-2
P Inject. 15
Q Inject. 15
P Inject. 16
Q Inject. 16

Estimated Values
Proposed
Conv.
0,73599
0,74336
0,02725
0,03214
1,54906
1,55619
-0,19301 -0,19637
0,45090
0,45086
-0,02946 -0,02880
1,84153
1,84132
-0,13630 -0,13543

TABLE VI
CASE2: STATEVARIABLES
RELATED TO CIRCUITBREAKER
15-16

I Method I

Conv.

I Proposed I

has nothing t o do, however, with the constraints enforced,


as they are only intended to more precisely define the state t o which the state estimator is anyway converging after
two iterations.

REFERENCES
[1] A. Abur, H. Kim, and M. Celik, Identifying the Unknown Circuit Breaker Statuses in Power Networks, IEEE Transactions on
Power Systems, Vol. 10, No. 4, pp. 2029-2037, August 1998.
[2] 0. Alsac, N. Vempati, B. Stott, and A. Monticelli, Generalized
State Estimation, IEEE Transactions on Power Systems, Vol.
13, No. 3, pp. 1069-1075, August 1998.
[3] K.A. Clements, and P. W. Davis, Detection and Identification of
Topology Errors in Electric Power Systems, IEEE Transactions
on Power Systems, Vol. 3, No. 4, pp. 1748-1753, November 1988.
[4] K.A. Clements, and A. Simoes Costa, Topologv Error Identifica-

tion Using Normalized Lagrange Multipliers, IEEE Transactions


on Power Systems, Vol. 13, No. 2, pp. 347-353, May 1998.
[5] A. Gjelsvik, The Significance of the Lagrange Multipliers in WLS
State Estimation with Equality Constraints, Proceedings PSCC,
Vol. 11, pp. 619-625, Avignon, August-September 1993.
[6] Mili L., Steeno G . , Dobraca F., French D. A Robust Estimation Method for Topology Error Identification. IEEE 7tnns. o n
Power Systems, Vol. 14, No. 4, pp. 1469-1476, Nov 1999.
[7] A. Monticelli, Modeling Circuit Breakers in Weighted Least
Squares State Estimation, IEEE Transactions on Power Systems,
Vol. 8, No. 3, pp. 1143-1149, August 1993.
[8] A. Monticelli, The Impact of Modeling Short Circuit Branches
in State Estimation, IEEE Transactions on Power Systems, Vol.
8, No. 1, pp. 364-370, February 1993.
[9] A. Monticelli, and A. Garcia, Modeling Zero Impedance Branches in Power System State Estimation, IEEE Transactions on
Power Systems, Vol. 6, No. 4, pp. 1561-1570, November 1991.
[lo] A. Monticelli, Testing Equality Constraint Hypothesis in Weighted Least Squares State Estimation, Santa Clara, PICA 1999.
[11] M. Prais, and M. Bose, A Topology Processor that Tracks Network Modifications Over Time, IEEE Transactions on Power
Systems, Vol. PWRS-3, No. 3, pp. 992-998, August 1988.
[12] A. Simoes Costa, and J:A. Leao Identification of Topology
Errors in Power System State Estimation, IEEE Transactions
on Power Systems, Vol. 13, No. 2, pp. 347-353, May 1998.
[13] I. W. Slutsker, and S. Mokhtari, Comprehensive Estimation
in Power Systems: State, Parameter and Topology Estimation,
Proc. of the American Power Conference, Vol. 57-1, pp. 149-155,
April 1955.
[14] Singh N., Glavitsch H., Detection and Identification of Topological Errors in Online Power System Analysis, IEEE f i n s a c tions o n Power Systems, Vol. 6(1), pp. 324-331, Febrero 1991.
[15] Souza J.C.S., Leite da Silva A.M., Alves da Silva A.P., On line
topology determination and bad data suppression in power system
operation using artificial neural networks. IEEE Zhnsactions on
Power Systems, Vol. 13, No. 3, pp. 796-803, Aug. 1998.
[16] F.F. Wu, and W. H. E. Liu, Detection of Topology Errors by
State Estimation, IEEE Transactions on Power Systems, Vol. 4,
No. 1, pp. 176-183, February 1989.

Antonio de la Villa JaGn was born in Riotinto, Spain, in 1960.


He received his Electrical Engineering degree from the University of
Seville where he is pursuing his Ph.D. degree a t the Dep. of Electrical
Engineering. He has been a High-School teacher and he is presently
an Assistant Professor at University of Seville. His primary areas
of interest are computer methods for power system state estimation
problems.

Antonio GBmez ExpBsito was born in Andujar, Spain, in 1957.


He received his electrical engineering degrees from the University of
Seville. Since 1982 he h a s been with the Dep. of Electrical Engineering, University of Seville, where he is currently a h l l Professor. His
primary areas of interest are sparse matrices, parallel computation,
reactive power optimization, state estimation and computer relaying.

% Measurement error

+- Conventional

--c With constraints

Fig. 5. Case 2: Estimated measurement error

You might also like