IEEE PES ISGT ASIA 2012 1569526945

Two-Level State Estimation Method for Power

Systems with SCADA and PMU Measurements
Hongbo Sun, Senior Member, IEEE, Zhenhua Wang and Daniel Nikovski, Member, IEEE

been proposed to utilize PMU measurements in the state

Abstract--This paper proposes a two-level hybrid state estimation procedure. Some methods used PMU
estimation method for power systems with SCADA and PMU measurements only [8], and others used both SCADA and
measurements. A power system is decoupled into PMU PMU measurements [9]-[11].
observed areas and SCADA observed areas based on its system
This paper proposes a two-level hybrid state estimation
measurement configuration and topology connectivity, and the
states of areas are solved by a two-level computational method for power systems with SCADA and PMU
procedure. The first level uses PMU measurements and pseudo measurements. Based on the system measurement
measurements derived from SCADA measurements to configuration and topology connectivity, a power system is
formulate a linear state estimation model for each PMU decoupled into PMU-observed areas and SCADA-observed
observed area, from which the states of the buses in the area areas to be solved by a two-level computational procedure.
are estimated. The second level uses the SCADA measurements
The first level uses PMU measurements and pseudo
and pseudo measurements generated by results at the first level
to formulate a nonlinear state estimation model for each measurements derived from SCADA measurements to
SCADA observed area. The weighted least square method is formulate a linear state estimation model in rectangular
used to solve the models at the two levels. Test results on an coordinates for each PMU observed area, from which the
IEEE 14 bus system and an IEEE 118 bus system are given to states of buses of the area are estimated. The second level
demonstrate the effectiveness of the proposed method. uses the SCADA measurements and pseudo measurements
generated by results at the first level to formulate a
Index Terms-- Least squares methods; Measurement;
nonlinear state estimation model in polar coordinates for
Phasor Measurement Units; Power Systems; SCADA Systems;
State Estimation each SCADA observed area. The weighted least square
method is used to solve the models at the two levels. Test
I. INTRODUCTION results on several IEEE test systems including a 14-bus test
case and a 118-bus test case are given to demonstrate the
S TATE estimation is one of the critical functions of the
power system control center of a power utility company.
The state estimator determines the best estimate of the
effectiveness of the proposed method. Compared with single
level method in which all SCADA and PMU measurements
are used simultaneously in the estimation model, the two-
current state of the system, based on the available
level method reduces the computational time of state
measurements from various measuring systems.
estimation significantly, while maintaining almost the same
Traditionally, the measurements are collected by
level of estimation accuracy.
supervisory control and data acquisition (SCADA) systems.
SCADA systems provide real-time snapshots of power
II. THE PROPOSED METHOD
flows and bus voltages, but these snapshots might not be
taken simultaneously. Various state estimation methods A. PMU and SCADA Observed Areas
based on SCADA measurements have been proposed [1],
Based on the location, type and data source of available
including weighted least square estimator [2]-[3], non-
measurements, a power system is divided into PMU-
quadratic estimator [4], least absolute value estimator [5],
observed areas and SCADA-observed areas.
leverage point estimator [6], and least median of squares
For a PMU-observed area, the state of a bus is either
estimator[7]. Weighted least square estimators are widely
directly measured by a PMU, or indirectly determined by
accepted in practice. With the increasing adoption of
the adjacent bus with PMU installation. For a SCADA-
synchronized phasor measurement units (PMU) in power
observed area, the state of a bus is determined by the
systems, PMU devices are gradually becoming a common
SCADA measurements within the area.
data source for the state estimation as well. PMUs provide
State estimation is decoupled into a two-level procedure.
synchronized measurements across dispersed locations, and
The first level is to obtain the state for the PMU-observed
furthermore, the measurement frequency and accuracy of
areas. The second level is to obtain the states for the
PMUs are much higher than those of SCADA systems.
SCADA-observed areas.
Many weighted least square based methods [8]-[11] have
Fig. 1 shows an example construction of PMU and
SCADA observed areas for a power system with PMU and
Hongbo Sun is with the Mitsubishi Electric Research Laboratories, SCADA measurements. In the figure, any single-line circle
Cambridge, MA 02139 USA (e-mail: [email protected]).
Zhenhua Wang is with the Holcombe Department of Electrical and
represents a SCADA-measured bus, and any double-line
Computer Engineering, Clemson University, Clemson, SC 29634 USA (e- circle represents a PMU-measured bus. Any single-line
mail: [email protected]). branch is SCADA measured, and double-line branches are
Daniel Nikovski is with the Mitsubishi Electric Research Laboratories,
Cambridge, MA 02139 USA (e-mail: [email protected]).
PMU measured. Buses 1-3 are PMU-measured buses, and

1
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on February 24,2021 at 05:50:03 UTC from IEEE Xplore. Restrictions apply.
 buses 4-12 are SCADA measured one. estimation model for PMU-observed areas. The states for
each PMU-observed area are represented by the real and
imaginary parts of complex bus voltages for each bus in the
area.
Measurements are collected from two data sources: one is
the group of PMU devices, and the other is the set of pseudo
measurements derived from branch power flow
measurements of SCADA systems.
The PMU measurements include the real and imaginary
parts of complex bus voltages, real and imaginary parts of
branch currents for lines or transformers, and real and
imaginary parts of bus current injections. It requires all bus
voltage and branch current measurement and their standard
deviations to be provided in rectangular coordinates.
However, those values are commonly collected in polar
coordinates, and have to be converted into rectangular form
before they are used in the state estimation procedure.
If a required measurement variable is not available, but
can be determined from other known actual or pseudo
measurement variables, the following equation can be used
to determine its standard deviation:
2
⎛
m
∂f ⎞⎟
σ zi = ∑ ⎜⎜σ y j (2)
Fig. 1. PMU and SCADA Observed Areas j =1 ⎝ ∂y j ⎟⎠

The system in Fig. 1 is replaced by two PMU-observed

where, zi and σ zi are the required variable and its standard
areas and two SCADA-observed areas. deviation, y j and σ y j are the j-th associated known
PMU observed area 1 consists of buses 1-2 and buses 4-
6, in which buses 1 and 2 are equipped with PMUs. PMU variable and its standard deviation that is used to determine
observed area 2 consists of bus 3 and buses 7-8, in which the required variable, f is a scalar function to express the
bus 3 is equipped with PMUs. relationship between required variable and known variables
SCADA observed area 1 consists of bus 4 and buses 7-9, as shown in (3):
in which bus 4 is a common bus between PMU area 1 and zi = f ( y1 , y2 ,..., ym ) , (3)
SCADA area 1. Bus 4's state is determined through a linear
estimation model for PMU area 1 first, then taken as pseudo ∂f ∂y j is the derivative of the required variable with
measurements to be resolved with nonlinear estimation respect to the known variable y j , and m is the total number
model for SCADA area 1. The SCADA observed area 2
consists of buses 5-8 and buses 10-12. Buses 5-7 and bus 12 of associated variables used to determine the required one.
The pseudo measurements to be used are the branch
are boundary buses between the area and PMU area 1 and 2.
currents calculated from SCADA branch power
Similarly, the states of those buses are solved first in the
measurements, and the PMU measured bus voltage at the
first level, and then updated in the second level solution.
terminal that SCADA branch powers are measured. Based
B. Weighted least square estimation with PMU on the relationship between branch power, bus voltage and
measurements branch current, the pseudo branch current measurements can
For a PMU-observed area, the relationship between the be easily determined. After that equation (2) can be used to
measurements and state variables can be described as determine the standard deviations for the pseudo current
follows: measurements accordingly. If the PMU bus voltage at the
z = Hx + e (1) required terminal is not available, a calculated bus voltage at
where, z is the measurement vector, x is the state vector, the terminal has to be determined first, and then use together
H is a constant matrix to represent the linear relationship with SCADA measurements to create the required pseudo
of the measurements to the states, and e is the measurement branch current measurements. The PMU bus voltage and
error vector and its covariance matrix is branch current measured at the opposite terminal of the
branch are used to determine the calculated bus voltage and
R , R = E{e • eT } , where eT is the transpose of vector its standard deviation at the required terminal.
e , and E{.} stands for expectation. Assuming that the H is constructed based on the relationship between the
distribution of measurements follows a normal distribution, measurement variables and state variables. The branch
the covariance matrix is a diagonal matrix, and its i-th current equations and bus current injection equations may be
diagonal element is the variance of the i-th measurement, used to determine the values of its matrix elements.
σ z2 .
i
The following normal equation is used to solve for the
system states:
Rectangular coordinates are used to formulate the state
Gx = HT R −1z , (4)

2
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on February 24,2021 at 05:50:03 UTC from IEEE Xplore. Restrictions apply.
 T
where G is the gain matrix, G = H R H . The sparse
−1 method, both the proposed two-level method and a
Cholesky's decomposition algorithm is used to solve the conventional single-level method are implemented and
state estimate expressed in (4). tested against the IEEE systems.
−1 The state variables of the single-level state estimation
The covariance of the estimate is Cov( x ˆ ) = G . For method are the voltage magnitudes and phase angles of all
simplification purposes, the standard deviation of any state buses in the systems. The measurement variables includes
estimate is approximately calculated as follows: all PMU and SCADA measurements, including PMU-
σ xˆ = xˆi /(HT R −1z)i measured bus voltage magnitudes and phase angles,
(5)
i SCADA-measured bus voltage magnitudes, SCADA-
where, x̂i is the i-th state estimate, σ x̂ i
is the standard measured bus active and reactive power injections,
SCADA-measured branch active and reactive power flows,
deviation of the estimate, and (HT R −1z )i is the i-th and the real and imaginary parts of PMU-measured branch
element of right hand side of (4). currents.
For each test system, 100 measurement samples have
C. Weighted least square estimation with SCADA
measurements been created by means of a Monte-Carlo simulation method.
The normal distribution is used to model the stochastic
For a SCADA observed area, the measurements z and distribution of measurement noise, and the used standard
state variables x can be related as follows: deviations for all measurement types are listed in Table I.
z = h(x ) + e (6)
( )
where, h x is a nonlinear relationship between the
TABLE I
STANDARD DEVIATIONS FOR MEASUREMENTS
measurement and state variables. PMU SCADA
Polar coordinates are used to formulate the state Bus Voltage Magnitude(Per Unit) 0.0002 0.003
estimation model for the SCADA observed areas. Phase Angle(Degrees) 0.01 /
Branch Magnitude(Per Unit) 0.0003 /
Measurements for the SCADA areas are mainly coming Current Phase Angle(Degrees) 0.015 /
from the SCADA systems. The measurements to be used Branch Active Power(Per Unit) / 0.003
include the bus voltage magnitude, the branch active and Power Flow Reactive Power(Per Unit) / 0.003
reactive power flows for a line or transformer, and the bus Bus Power Active Power(Per Unit) / 0.003
active and reactive power injections. Injection Reactive Power(Per Unit) / 0.003
The other measurements to be used are coming from the
first level estimation results, that is, the estimated voltage A. State Estimation of IEEE 14 Bus System
magnitude and phase angle at the boundary buses between
Fig. 2 gives the schematic diagram of IEEE 14 Bus
the area with PMU areas. Equation (5) is used to calculate
system. Its measuring points and measuring types are listed
the standard deviations of those pseudo measurements in
in Table II. For each branch, its measurements can be
rectangular coordinates, and then equation (2) is used to
provided at two different directions, the positive direction
converted them into polar coordinates.
and negative direction. The branch measurements at the
The state variables are the bus voltage magnitudes and
positive direction are measured at the first terminal bus of
the bus phase angles.
two-terminal pair of the branch, and the negative direction at
An iterative procedure is used for solving (6). The normal
the second terminal bus.
equation used for solving for the state changes at each
iteration is:
GΔx = HT R −1Δz
(7)
where, Δx is the vector of state variable
()
changes, Δz = z − h x is the vector of mismatches
between measurements and true values, H = ∂z ∂x is the
Jacobian matrix that represents the sensitivity of
measurement variables with respect to the state variables,
T −1
and G is the gain matrix, G = H R H . The branch
power flow equations, and bus power injection equations
may be used to formulate the Jacobian matrix.
The sparse Cholesky's decomposition algorithm is used
to solve for the state changes expressed in (7). The solution
of (7) is repeated until the state changes are small enough, or
a given maximum iteration number is reached.
Fig. 2. IEEE 14 Bus Test System

III. NUMERICAL EXAMPLES According to the measurement configuration and system

The proposed two-level state estimation has been tested topology, the system is divided into PMU and SCADA
on several IEEE test systems including a 14 bus system, and observed areas. Each area contains several buses and all
a 118 bus system. lines or transformers between those buses.
In order to demonstrate the effectiveness of the proposed

3
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on February 24,2021 at 05:50:03 UTC from IEEE Xplore. Restrictions apply.
 TABLE II
MEASUREMENT CONFIGURATION OF IEEE 14 BUS SYSTEM
PMU SCADA
Magnitude 2-5,7-8,
Bus 1,6,9 10-14
Voltage Phase /
Angle
Branch Positive 1/2,1/5,6/11,
Current Direction 6/12,6/13,
(Magnitude 9/10,9/14 /
& Phase Negative 5/6,4/9,7/9
Angle) Direction
2/3,2/4,2/5,3/4,
Positive 4/5,4/7,4/9,5/6,
Branch Direction / 10/11,12/13,
Power 13/14
(Active & 1/2,1/5, 2/3,
Reactive) 2/4,2/5,3/4,
Negative 4/5,6/11,6/12,
Direction 6/13,7/8,9/10, Fig. 3. IEEE 118 Bus Test System
9/14,10/11,
12/13,13/14 TABLE IV
MEASUREMENT CONFIGURATION OF IEEE 118 BUS SYSTEM
As shown in the table, the 14 bus system has three PMU PMU SCADA
measured buses, bus 1, bus 6 and bus 9. It is partitioned into Magnitude 1-67,69,71-76,
Bus 68,70,77, 78,79,81,82,
two areas, one PMU observed area which contains buses 1- Voltage 80,83 84-118
2,4-7, and 9-14, and one SCADA observed area which Phase /
contains buses 2-5 and 7-8. Angle
Table III gives the computation results of single-level and 70/71,70/74,
two-level methods. 70/75,77/78,
77/80,77/80,
Positive 68/81,77/82,
TABLE III Direction 83/84,83/85,
COMPARISON OF COMPUTATION Branch 80/96,80/97,
PERFORMANCES ON IEEE 14 BUS SYSTEM Current 80/98,80/99, /
Maximum Average CPU (Magnitude 68/116,68/69
Mismatches Mismatches Time & Phase Negative 65/68,69/70,
Method Voltage Phase Voltage Phase (Sec.) Angle) Direction 24/70,76/77,
(Per Angle (Per Angle 69/77,75/77,
Unit) (Deg.) Unit) (Deg.) 77/80,77/80,
Single- 82/83
level 0.00142 0.08773 0.00015 0.00842 0.019 Positive 8/5,26/25,30/17,
method Direction 38/37,63/59,
Two- Branch / 64/61,65/66,
level 0.00555 0.11386 0.00237 0.03480 0.003 Power Negative 81/80, and all line
method (Active & Direction branches except
Reactive) branches with
The accuracy of two-level method is slightly lower than PMU branch
current
the single-level method. If we look at the maximum measurements
estimation error, the maximum errors of the single-level
method are 0.00142 per unit in magnitude, and 0.08773 Table V shows computation results on the single-level
degrees in phase angle. In comparison, the maximum errors and two-level methods of the 118-bus system.
of two-level method are 0.00555 per unit in magnitude, and
0.11386 degrees in phase angle. TABLE V
COMPARISON OF COMPUTATION
But the slight sacrifice of estimation accuracy is well PERFORMANCES ON IEEE 118 BUS SYSTEM
rewarded by the gains in computation efficiency. As shown Method Maximum Average CPU
in the table, the computation time for two-level method was Mismatches Mismatches Time
0.003 seconds compared to 0.019 seconds for single-level Voltage Phase Voltage Phase (Sec.)
(Per Angle (Per Angle
method. Unit) (Deg.) Unit) (Deg.)
B. State Estimation of IEEE 118 Bus System Single-
level 0.00134 0.10779 0.00016 0.01850 30.434
Fig. 3 is a schematic diagram of the IEEE 118 Bus method
system. Its measuring points and measuring types are listed Two-
in Table IV. level 0.00243 0.12362 0.00035 0.02167 4.819
method
Based on the measurement configuration, the system is
divided into three areas: one PMU-observed area which
contains buses 24, 65, 68-71, 74-85, 96-99 and 116, and two The maximum errors of the single-level method are
SCADA-observed areas, such that the first SCADA area 0.00134 per unit in magnitude, and 0.10779 degrees in
contains buses 1-10,11-67, 69,71-76, 113-115,117, and 118, phase angle. The maximum errors of the two-level method
and the second SCADA area contains buses 82, and 84-112. are 0.00243 per unit in magnitude, and 0.12362 degrees in

4
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on February 24,2021 at 05:50:03 UTC from IEEE Xplore. Restrictions apply.
 phase angle. It took 30.434 seconds for the single-level V. BIOGRAPHIES
method to find the solution, and 4.819 seconds for the two-
level method. Hongbo Sun(SM' 2000) was born in Liaoning,
Compared with single-level method, the two-level China in 1966. He received the B.S. degree in
Electrical Engineering from Southwest Jiaotong
method has reduced the computational time of state
University, China in 1986, and the M.S. and Ph.D.
estimation significantly, while maintaining almost the same degrees in Electrical Engineering from Chongqing
level of estimation accuracy. University, China in 1988 and 1991 respectively.
He is a senior member of IEEE, and a registered
professional engineer. He is currently working at
IV. CONCLUSIONS Mitsubishi Electric Research Laboratories in
A two-level hybrid state estimation method is proposed Cambridge, Massachusetts, USA. His research
interests include power system planning and analysis, power operation and
for power systems with SCADA and PMU measurements. control, and smart grid applications.
Based on the system measurement configuration and
topology connectivity, a power system is decoupled into
PMU observed areas and SCADA observed areas to be Zhenhua Wang was born in Shanxi, China in
1981. He received BS degree in Electrical
solved. The first level uses PMU measurements and pseudo Engineering from Zhengzhou University, China
measurements derived from SCADA power flow in 2003 and MS degree in Electrical Engineering
measurements to formulate a liner state estimation model for from Tianjin University, China in 2007
respectively. Now he is pursuing the Ph.D.
each PMU observed area, and the states of buses of the area degree at the Holcombe department of Electrical
are determined. The second level uses the SCADA and Computer Engineering, Clemson University,
measurements and pseudo measurements generated by USA. His research interest includes power
system transient stability, power system state
results at first level to formulate an nonlinear state estimation and wind energypenetration.
estimation model for each SCADA observed area. The
weighted least square method is used to solve two level Daniel Nikovski(M' 2003) was born in Plovdiv,
models. Bulgaria, in 1969. He received a B.Sc. degree in
Computer Systems and Control from the
The test results of several IEEE test systems have
Technical University Sofia, Bulgaria, in 1992, a
demonstrated the effectiveness of the proposed method. M.Sc. degree in Computer Science from the
Compared with single-level method in which all SCADA Southern Illinois University in Cabondale, USA,
and PMU measurements are used simultaneously in the in 1995, and a Ph.D. degree in Robotics from
estimation model, the two-level method reduces the Carnegie Mellon University in Pittsburgh, USA,
computational time of state estimation significantly, while in 2002. He is currently Senior Member of
maintaining almost the same level of estimation accuracy. Research Staff and Group Manager of the Data Analytics group at
Mitsubishi Electric Research Laboratories in Cambridge, Massachusetts,
USA. His research interests include machine learning, optimization and
IV. REFERENCES
control, and numerical methods for analysis of complex industrial systems.
[1] A. Monticelli," Electric Power System State Estimation", Proceedings
of the IEEE, vol. 88, no.2, pp.262-282, February 2000.
[2] F. C. Schweppe and E. J. Handschin, “Static state estimation in
electric power systems”, Proceedings of the IEEE, vol. 62, no.7, pp.
972–982, July 1974.
[3] F. C. Schweppe, J.Wildes, and D. Rom, “Power system static state
estimation: Parts I, II, and III,” IEEE Transactions on Power
Apparatus and Systems, vol. PAS-89, no.1, pp. 120–135, January
1970.
[4] H. M. Merril and F. C. Schweppe, “Bad data suppression in power
system state estimation”, IEEE Transactions on Power Apparatus and
Systems, vol. PAS-90, no.6, pp. 2718–2725, November/December
1971.
[5] M. R. Irving, R. C. Owen, and M. Sterling, “Power system state
estimation using linear programming,” Proc. Inst Elect.Eng., vol. 125,
no.9, pp. 978–885, September 1978.
[6] L. Mili, M. G. Cheniae, N. S. Vichare, and P. J. Rousseeuw, “Robust
state estimation based on projection statistics”, IEEE Transactions on
Power Systems, vol. 11, no.2, pp. 1118–1127, May 1996.
[7] L. Mili, V. Phaniraj, and P. J. Rousseeuw, “Least median of squares
estimation in power-systems”, IEEE Transactions on Power Systems,
vol. 6, no.2, pp. 511–523, May 1991.
[8] A.G. Phadke, J.S. Thorp, and K.J. Karimi, “State Estimation with
Phasor Measurements”, IEEE Transactions on Power Systems, vol. 1,
no.1, pp. 233-241, February 1986.
[9] A.G. Phadke, J.S. Thorp and K.J. Karimi, “ Real Time Voltage Phasor
Measurements for Static State Estimation”, IEEE Transactions on
Power Apparatus and Systems, vol. PAS-104, no.11, pp.3098-3107,
November 1985.
[10] R. Avila-Rosales, M.J. Rice, J. Giri, L. Beard, and F. Galvan, “Recent
experience with a hybrid SCADA/PMU on-line state estimator”, in
Proc. 2009 IEEE Power & Energy Society General Meeting, pp.1-8.
[11] S. Chakrabarti, E. Kyriakides, G. Ledwich, and A. Ghosh, “Inclusion
of PMU current phasor measurements in a power system state
estimator”, IET Generation, Transmission, and Distribution, vol. 4,
no.10, pp. 1104-1115, September. 2010.

5
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on February 24,2021 at 05:50:03 UTC from IEEE Xplore. Restrictions apply.