You are on page 1of 7

Electrical Power and Energy Systems 42 (2012) 7177

Contents lists available at SciVerse ScienceDirect

Electrical Power and Energy Systems


journal homepage: www.elsevier.com/locate/ijepes

An optimal PMU placement technique for power system observability


B.K. Saha Roy , A.K. Sinha, A.K. Pradhan
Indian Institute of Technology, Kharagpur 721 302, India

a r t i c l e

i n f o

Article history:
Received 25 August 2011
Received in revised form 4 January 2012
Accepted 21 March 2012
Available online 7 May 2012
Keywords:
Phasor measurement unit (PMU)
Network observability
Network connectivity
Optimal placement

a b s t r a c t
Power system state estimation with exclusive utilization of synchronous phasor measurements demands
that the system should be completely observable through PMUs only. To have minimum number of
PMUs, the PMU placement issue in any network is an optimization problem. A three stage optimal
PMU placement method is presented in this paper using network connectivity information. The method
initially considers PMU in all buses of the network. Stage I and Stage II of the algorithm iteratively determine (i) less important bus locations from where PMUs are eliminated and (ii) strategically important bus
locations where PMUs are retained. Stage III of the algorithm further minimizes the number of PMU using
pruning operation. The set of PMUs obtained after Stage III is an optimal set of PMU locations for network
observability. Proposed method is further extended for assuring complete observability under single PMU
outage cases. Simulation results for IEEE 14-bus, 24-bus, 30-bus, 57-bus, 118-bus and New England 39bus test systems are presented and compared with the existing techniques. Results show that the proposed method is simple to implement and accurate compared to other existing methods.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
Power utilities are facing increasing number of threats of security of operation due to over stressed power network in competitive power market scenario. State estimation is a tool which
provides the real time state of the system. It is an integral part of
energy management system (EMS) for security analysis and other
power system applications [1]. Pre-requisite of state estimation
is that the system must be fully observable from the available measurements. Before introduction of synchro-phasor technology,
state estimation was relying on SCADA systems. Superiority of
PMU measurements over SCADA measurements are that PMU provides the phase angle measurement directly and all the measurements are time synchronized [2]. PMU provides voltage phasor of
the bus where it is installed and current phasors of all the branches
incident to that bus. Direct measurement of all the system states is
possible by placing PMUs in all buses of a network without running
any state estimator. The PMU and its associated communication
system are costly and the voltage phasor of the incident buses to
PMU installed bus can be obtained with the help of branch parameter and branch current phasor. If the network is observable
through optimally placed PMU, a linear state estimator provides
system states in a single iteration. The main objective of optimal
PMU placement is to determine the minimal number of PMUs to
be installed at strategic locations so that the entire power system
becomes completely observable for state estimation.
Corresponding author.
E-mail address: bksr12@gmail.com (B.K. Saha Roy).
0142-0615/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijepes.2012.03.011

Some of the important contributions in PMU placement area are


bisecting search and simulated annealing based method [3], non
dominated sorting based genetic algorithm approach [4], simulated annealing based graph theoretic approach [5], integer programming (IP) based approach [6,7]. Random selection of PMU
placement sets makes bisecting search approach computationally
less efcient. The IP approach uses linear programming (LP) solver
and branch and bound algorithm [8]. Branching involves development of subproblem of the original problem and bound involves
enumeration of the subproblem through linear programming.
The process requires use of LP solver in every iteration which again
depends on the size of the system and thus the method is computationally intensive. For benchmarking of global optimal solution
the exhaustive binary search method [9] is suitable but computational burden is heavy for large size networks. In [10] a phased
installation scheme of PMUs is proposed such that the nal placement will be optimal. Tabu search metaheuristic method [11] is
proposed to obtain minimum number of PMU for complete observability of power system. A non-dominated sorting based differential evolution algorithm [12] is proposed for multi-objective
optimization of PMU placement problem. In [13] a metaheuristic
based iterative local search method is proposed to nd the optimal
solution where an initial PMU placement is considered which
makes the system fully observable. A two stage PMU placement
methodology is proposed in [14]. The stage-I takes care of topological observability and stage-II checks for numerical observability
based on exhaustive search method which increases computational burden of the method heavily. Binary particle swarm optimization (BPSO) based approach is proposed in [15]. The optimization

72

B.K. Saha Roy et al. / Electrical Power and Energy Systems 42 (2012) 7177

process takes care of dual objective such as minimization of PMU


number and maximization of measurement redundancy. A modied BPSO based optimal PMU placement technique is available
in [16]. Contingency constrained optimal PMU placement techniques are reported in [17,18]. In [19] heuristics based optimal
PMU placement method is proposed, where the author addresses
issues like zero injection bus as pseudo measurements. Method
for selection of optimal PMU with available SCADA measurements
considering single and multiple measurement loss is reported in
[20].
In this paper a topological observability based three stage optimal PMU placement technique is proposed. For topological observability a set of minimum PMUs is required to make the system
completely observable. It is assumed that there are strategic buses
in every system using that the PMU placement becomes an easy
task. This is the motivation of the proposed three stage optimal
PMU placement method. The method assumes PMUs in all buses
initially. The rst two stages eliminate PMUs sequentially from less
important buses and nd the priority bus locations based on simple network connectivity where the PMUs are retained. Optimality
of the set of PMUs obtained from Stage I and Stage II is further
checked by pruning operation in Stage III. The set of PMUs obtained
after pruning is the optimal set for complete observability of the
power system. Proposed method considered that the PMU will perfectly work. Like any other metering devices PMU may malfunctions due to problem in PMU itself or with the communication
system. To increase the reliability of measurements it is of interest
to avoid such maloperating situations. It is desired that each bus
should be observable at least by two PMUs to take care of single
PMU loss in the system. Probability of loss of two PMUs simultaneously is considered to be very low or negligible here. The proposed methodology is extended to nd an overall optimal
solution to make system completely observable under normal
operating conditions and for single PMU loss situation. The proposed method is found to be simple, accurate and fast in
computation.
The paper is organized as follows. Section 2 discusses the power
system observability using PMU. Section 3 provides the problem
formulation for optimal PMU placement. Solution methodology is
discussed in Section 4. Case studies and discussions on simulation
results are presented in Section 5. Section 6 provides conclusion of
the paper. Appendix A provides the explanation of the rules used
and an illustration of the algorithm is exemplied in Appendix B.

formulation [22]. The linear state estimator model is given


below:

z hx

where h is the measurement function matrix. When the system is


completely observable by PMUs the linear state estimator provides
system state by solving linear set of equations in (2). Numerical
observability analysis checks whether the rank of measurement
matrix is full or not. For an N bus network the measurement sets
(voltage and current phasors) obtained from PMUs can make the
system numerically observable if

Rankh 2N  1

Optimal PMU placement for numerical observability of power


system can be carried out in two different ways
Method I: Introduction of PMU in the network one by one to
improve the rank of the measurement matrix. No further PMU
is introduced when (3) is satised.
Method II: Considering PMU in all buses, PMUs are sequentially
eliminated from different buses one by one. Elimination process
is overruled at a bus which makes the rank of the measurement
matrix decient.
In both approaches to reach the optimal (minimum) PMU placement solution, large number of combinations has to be checked.
Each combination involves rank testing which increases the computational burden of numerical observability analysis.
2.2. Topological observability analysis
In topological observability analysis power system is represented by a topological graph. The graph has N number of
nodes representing the network bus bar and E number of
edges representing the branches of the network connecting
the bus bars. In topological approach the optimal PMU
placement set is searched such that each bus of the network
is observable by at least one PMU. In this paper a simple topological observability analysis method is proposed which is
exclusively based on PMU measurements. Loss of single PMU
or communication link failure is also considered in the present
paper.

2. Power system observability analysis


3. PMU placement problem formulation
A power system state estimator demands complete observability of the network from the available measurements. Power system
is observable by a set of measurements when the measurements
can build a spanning tree of full rank of the system [21]. Power system observability analysis is usually carried out in two different
ways namely numerical and topological observability analysis.

For an N-bus system the optimal PMU placement problem is


formulated as follows:
N
X
wi xi

min
s:t:

2.1. Numerical observability analysis

FX P b

where X is the binary decision variable vector for PMU placement,


whose entries are dened as

The measurement model used in state estimation is

z hx e

i1

where x is the system state vector i.e. the voltage phasor of all buses
of the network, z is the measurement vector, h(x) is the non-linear
function that relates the measurement vector to the state vector of
the system and e represents the measurement error vector. Since
PMUs provide accurate measurements (voltage and current phasors) the measurement error is small and can be neglected. Exclusive use of PMU measurements results in linear state estimator

xi 1 if PMU is placed at ith bus


0

otherwise:

for i 1; . . . ; N:

b is an unit vector of length N, i.e. b = [1 1 1. . .. . .]T. wi is the cost of


PMU installed at ith bus. F(X) is the observability constraint vector
function, whose entries are nonzero if the corresponding individual
buses are observable w.r.t a given measurement set and zero otherwise. If wi is assumed unity, the optimal PMU placement problem is
represented as

73

B.K. Saha Roy et al. / Electrical Power and Energy Systems 42 (2012) 7177

min

N
X
xi
i1

s:t:

ZIB

I 13

FX P b

Constraint vector function ensures full network observability. A


solution i.e. a set of minimum xi is to be found out which will satisfy (5). The constraint vector function is formed using the binary
connectivity matrix (A) of power system. The binary connectivity
matrix (A) represents the bus connectivity information of a power
system. The elements of matrix A is dened as

3
I 23

I 43
4

2
Fig. 1. Modeling of ZIB.

am;n 1 if m n
1

if bus m is connected to bus n

otherwise

Modeling of ZIB: ZIBs are the buses which have neither any generation nor any load. At ZIBs no current is injected into the system. This is used as pseudo information to make system
observable with less number of PMUs compared to the case
when information of ZIBs is not considered. Let us consider a
typical 4-bus example as in Fig. 1 for ZIB modeling. In the 4bus example, buses are numbered as 1, 2, 3 and 4. Bus 3 is a
ZIB. Application of KCL at bus 3 provides

The constraint vector function for the test system and any particular ith bus is given in (6) and (7).

FX AX P b

fi ai;1 x1 . . . ai;;i xi . . . ai;N xN

If ai,n is zero, the product ai,nxi not appear in (7). If any xi appearing in fi is nonzero, fi is observable. If all fi in F are nonzero the system is completely observable.
4. Proposed technique to PMU placement
The aim of the proposed method is to obtain minimum number
of strategic bus locations where PMUs are to be placed to satisfy
system observability (6). Strategic buses are obtained based on
network connectivity information.

I13 I23 I43 0

Among the four buses if any three bus voltages (V) are known
the 4th bus voltage can be calculated using (8). When K number
of connected ZIBs forms a super node with all known adjacent
bus voltages, the ZIBs can be solved using (9). M is the number
of bus connected to ith bus and Yij is the transfer admittance between ith bus and jth bus.
M
X
Y ij V j 0 for i 1; . . . ; K

j1

4.1. Description of the proposed method


The proposed three stage optimal PMU placement method considers PMU at each bus initially. If zero injection bus (ZIB) is used
as pseudo measurement, PMUs are not considered in ZIBs. The
PMUs are eliminated sequentially from the less important bus
locations and retained at potentially important bus locations. The
terms used in the proposed method are described below.
Bus valency: Valency of ith bus of a network is the total number
of buses connected to that bus including the ith bus itself.
MaxV: Maximum bus valency of the network.
MinV: Minimum bus valency of the network.
Radial bus (RB): Bus which is connected to only one other bus of
the network. Bus valency of radial bus is minimum and it is 2.
Equal valency (EV) bus: The set EV stores buses which have equal
bus valency at any particular valency of the test network.
Allocated PMU bus (APB): In the process of PMU elimination
from different buses there are some important bus locations
where PMUs are retained. Those PMU bus locations are called
allocated PMU buses (APB).
SB: Buses of EV those are connected to any bus in APB is stored
in set SB.
Higher valency (HV) bus: Some of the EV buses connected to a
common bus having valency higher than EV buses, forms a
group with group centre at that common bus. HV store those
higher valency buses which have maximum number of group
element. At least two EV bus is needed to form a group.
APB observable buses (APBOBs): Buses which are connected to
any APBs.
Candidate bus (CB): CB consist candidates bus locations from
where PMUs are to be eliminated. At starting CB consists all
buses except the ZIBs.

PMU elimination starts from minimum valency buses i.e. from


the radial buses. To observe a radial bus PMU is essential at radial
buses or the bus to which a radial bus is connected. If PMU is
placed at radial bus it makes two buses observable but if PMU is
placed at bus where radial bus is connected it makes more than
two buses observable. Hence PMUs are eliminated from radial
buses and retained at buses where radial buses are connected.
PMU elimination process continues for the higher valency buses
up to maximum valency. In the process of elimination PMUs are retained at important bus locations such as (1) a higher valency bus
connected to more than one running valency buses of EV, and (2) a
bus from where elimination of PMU leads to the system unobservability. At every iteration the tested buses (PMU eliminated or retained) are removed from the CB.
The three stages of the proposed algorithm are enumerated
below.
Stage-I: It nds the important bus locations where PMUs are
retained among different valency buses.
Stage-II: Some of the buses neither formed any group nor connected to any APB bus remain untested in stage-I. Among those
buses lower valency buses with higher connectivity to APBOBs
and connected to ZIBs are considered rst. PMUs are eliminated
or retained one by one from these buses subject to satisfying
system observability. On testing all lower valency buses,
untested higher valency buses are tested up to maximum
valency.
Stage-III: This is a pruning stage. Pruning checks the possible
ways for further reduction of PMUs from the set of PMU bus
locations obtained in stage-I and stage-II. It searches a non
PMU bus which is connected to more than one PMU bus.
Pruning checks whether placement of one PMU at non PMU
bus can make two or more PMU buses redundant in satisfying

74

B.K. Saha Roy et al. / Electrical Power and Energy Systems 42 (2012) 7177

observability constraints. It also checks redundancy of the individual PMUs to satisfy system observability constraints.

b
Set: V=2;

Elimination and retention of PMU obey the following rule.

Find EV with valency V=V+1 from updated CB


excluding ZIBs. Find APBOB buses.

Rule-1: Eliminate PMU from SB buses one by one subject to satisfaction of (6).
Rule-2: Retain PMU at any highest valency bus among HV buses.

Order the EV buses connected to maximum APBOBs


st
and connected to ZIBs 1 followed by the rest.

4.2. Algorithm for optimal PMU placement under normal operating


condition

Eliminate one PMU from ordered EV buses. Retain PMU


when system observability violated. Update EV, CB and APB.

Flow charts of three stages of the proposed algorithm are detailed in Figs. 2a2c.

No

Yes
EV nil ?

Yes

V<maxV ?
No

4.3. Algorithm for optimal PMU placement considering single PMU loss
Fig. 2b. Flow chart of Stage II.

The optimal set of PMUs obtained for normal operating condition is not sufcient to observe all buses under single PMU loss
or any communication failure. Extra PMUs are needed along with
the optimal set obtained for normal operating condition to take
care single PMU loss. Additional PMUs are selected in such a way
that nally an overall optimal set of PMUs obtained that make
the system completely observable under normal operating condition as well as under single PMU loss condition. To obtain the overall optimal set the optimization problem is restated as

c
Do pruning and check
observability
Observable ?

No

Dont prune

Yes
No

Pruning complete ?
Yes
Optimal solution obtained

Start

Stop
Form system A matrix, compute valency of all
CB; find MaxV , MinV. Set V=MinV=2 ;

Fig. 2c. Flow chart of Stage III.

Consider PMU in all buses except ZIBs when


ZIBS are considered as pseudo measurements

min

Eliminate PMU from RB and retain PMU at


bus connected to RB. Retained PMU buses are
stored in APB. Remove tested buses from CB.

i1

s:t:

Find buses with valency V=V+1 from CB and


store in EV.

Eliminate PMUs as per rule 1. Retain PMU at


bus violating observability. Update EV, APB
and CB. n = nos of bus in EV.

No

nh=0 ?
No

Apply rule 2. Update APB, CB. Reset EV, HV


No

PMU retained ?
Yes
V=2

FX P b

n =0 ?

Find HV from CB ; nh = nos of bus in HV


Yes

10

where b = [2 2 2. . .. . .]T. Length of b is N. Doubly observable buses are


either observable from two PMUs or from a single PMUs and also
observable using KCL at ZIBs. Singly observable buses are either observable from a PMU or using KCL at ZIB. When all but one bus involved with any ZIB are doubly observable, KCL at ZIB makes the
unknown bus doubly observable. The following steps successfully
provide the overall optimal set of PMUs required for making the
system observable under single PMU loss contingency.

Find SB buses from EV buses.

Yes

N
X
xi

V< MaxV?
Yes

Fig. 2a. Flow chart of Stage I.

Step 1: Place PMU at all the radial buses those are not connected
to any ZIBs along with the optimal set of PMU buses obtained
for complete observability under normal operating condition.
Step 2: Find the singly observable buses with the already placed
PMUs, store these buses in set {SOB}.
Step 3: If the number of buses in SOB is zero, go to step 6 else go
to step 4.
Step 4: Find the maximum valency bus from SOB excluding the
PMU buses.
Step 5: Select any bus from the list of bus obtained in step 4 as
PMU bus. Remove the buses which become doubly observable
due to the additional PMU bus from SOB and go to step 2.
Step 6: Perform pruning to discard any redundant PMU for making the system buses double observable.

75

B.K. Saha Roy et al. / Electrical Power and Energy Systems 42 (2012) 7177
Table 1
Optimal PMU placement results for normal operating condition without considering ZIB.
System

Optimal PMU locations

Optimal number

IEEE 14-bus
IEEE 24-bus
IEEE 30-bus
NE 39-bus
IEEE 57-bus
IEEE 118-bus

2,67,9
2,3,8,10,16,21,23
2,3,6,9,10,12,15,19,25,27
2,6,9,10,12,14,17,19,20,22,23,25,29
1,4,9,20,24,27,29,30,32,36,38,39,41,45,46,51,54
1, 5,9, 12,13,17,21,23,26,28,34,37,41,45,49,53,56,62,63,68,71,75,77,80,85,86,90,94,101,105,110,114

4
7
10
13
17
32

Table 2
Comparison of optimal PMU placement results with available techniques for normal operating condition without considering ZIB.
Method

IEEE 14-bus (Nos.)

IEEE 24-bus (Nos.)

IEEE 30-bus (Nos.)

NE-39 (Nos.)

IEEE 57-bus (Nos.)

IEEE 118-bus (Nos.)

Proposed
Xu and Abur [7]
Chakrabarti and Kyriakides [9]
Hurtgen and Maun [13]

4
4
4
4

7
N/A
7
N/A

10
10
10
N/A

13
N/A
13
N/A

17
17
N/A
17

32
32
N/A
32

N/A: Not available.

Table 3
Optimal PMU placement results for normal operating condition considering ZIB.
System

ZIBs locations

Optimal PMU locations

Optimal number

IEEE 14-bus
IEEE 24-bus
IEEE 30-bus
NE 39-bus
IEEE 57-bus
IEEE 118-bus

7
11,12,17,24
6,9,22,25,27,28
1,2,5,6,9,10,11,13,14,17,19,22
4,7,11,21,22,24,26,34,36,37,39,40,45,46,48
5,9,30,37,38,63,64,68,71,81

2,6,9
1,2,8,16,21,23
2,3,10,12,18,24,30
3,8,12,16,20,23,25,29
1,6,13,19,25,29,32,38,51,54,56
1,6,8,12,15,17,21,25,29,34,40,45,49,53,56,62, 72,75,77,80,85,86,
90,94,101,105,110,114

3
6
7
8
11
28

Table 4
Comparison of optimal PMU placement results with available techniques for normal operating condition considering ZIB.
Method

IEEE 14-bus (Nos.)

IEEE 24-bus (Nos.)

IEEE 30-bus (Nos.)

NE 39-bus (Nos.)

IEEE 57-bus (Nos.)

IEEE 118-bus (Nos.)

Proposed
Xu and Abur [7]
Chakrabarti and Kyriakides [9]
Hajian [16]

3
3
3
3

6
N/A
6
N/A

7
7
7
7

8
N/A
8
8

11
12
N/A
11

28
28
N/A
28

Table 5
Optimal PMU placement results for single PMU loss condition considering ZIB.
System

Optimal PMU locations

Optimal number

IEEE 14-bus
IEEE 24-bus
IEEE 30-bus
NE 39-bus
IEEE 57-bus
IEEE 118-bus

1,2,4,6, 9,10,13
1,2,7,8,9,10,11,15,16,17,20,21,23
1,2,3,5,6,10,12,13,15,16,18,19,24,27,30
2,3,5,6,8,13,16,17,20,22,23,25,26,29,34,36,37,38
1,2,4,6,9,12,15,18,19,22,24,25,27, 29,30, 32,33, 36,38,41,47,50, 51,53,54,56
1,2,5,6,8,9,11,12,15,17,19,20,21,23,25,27,28,29,32,34,35,37, 40,41,43,45,46,49,50,51,52,53,56,59,62,66,68,70,71,72,75,76,
77,78,80,83,85,86,87,89,90,92,94,96,100,101,105,106,108, 110, 111,112,114,117

7
13
15
18
26
64

Table 6
Comparison of optimal PMU placement results with topological observability based approach for single PMU loss condition considering ZIBs.
Method

IEEE 14-bus (Nos.)

IEEE 24-bus (Nos.)

IEEE 30-bus (Nos.)

NE-39 (Nos.)

IEEE 57-bus (Nos.)

IEEE 118-bus (Nos.)

Proposed
Xu and Abur [7]

7
7

13
N/A

15
17

18
N/A

26
26

64
65

Table 7
Comparison of computation time for obtaining optimal solution by the proposed method with two other simulated methods.
Method

IEEE 14-bus

IEEE 24-bus

IEEE 30-bus

NE-39

IEEE 57-bus

IEEE 118-bus

Proposed
BIP
Exhaustive search

0.66s
1.16s
1.3s

0.76s
1.34s
7.4 m

0.83s
1.24s
2.6hr

0.84s
1.27s

0.87s
1.94s

1.34s
1.55s

76

B.K. Saha Roy et al. / Electrical Power and Energy Systems 42 (2012) 7177

18

21

22

17
20

19

16
15

14

24

11

13
12

3
4
1

23

10
8

Fig. B.1. Single line diagram of IEEE 24-bus test system.

Table B.1
Bus valencies of different buses of IEEE 24bus system.
Valency (V)

Bus locations

2
3
4
5
6

7
4,5,6,14,18,19,20,22,24
1,2,3,8,13,15,17,21,23
11,12,16
9,10

Simulation results for optimal number and corresponding bus


locations for system observability considering ZIBs information
are given in Table 3. For IEEE 14-bus system 3 PMUs are sufcient
and the bus numbers for placement are 2, 6, and 9. Comparative results of the proposed method considering ZIB with available PMU
placement methods are given in Table 4. Comparative results show
that the proposed method successfully nds the global optimal
solution for different systems.
The methodology is extended for an overall optimal solution to
take care of single PMU loss contingencies. Table 5 shows the optimal number of PMUs required for different systems and corresponding bus locations for complete observability under single
PMU loss case considering ZIBs. For IEEE 14-bus system 7 PMUs
are essential and the bus numbers for placement are 1, 2, 4, 6, 9,
10 and 13. The result of the proposed method for single PMU loss
case is compared with the results reported by Xu and Abur [7] and
provided in Table 6. Results show that proposed method is capable
of producing optimal set of PMUs for single PMU loss condition.
The CPU time taken by proposed method under normal operating condition is shown in Table 7 for different test systems. Computation time of the proposed method is compared with the time
taken in matlab binary integer programming (BIP) and exhaustive
search approach without considering ZIBs. Performance comparison shows the computational efciency of the proposed method
compared to the other methods. The simulation platform used
for implementation of the methods is Intel Pentium 4, 3.0-GHz
CPU with 512 MB of RAM.

6. Conclusion
Step 7: Optimal set of PMUs obtained for complete observability
of power system under normal operating condition as well as
for single PMU loss contingency.

5. Case studies and discussion on simulation results


The proposed three stage algorithm determines minimum number of strategic bus locations where PMU must be placed for complete observability of the power system. First and second stage of
the algorithm determines the important bus locations for allocating PMUs. The third or Pruning stage checks the possible ways to
further reduce any PMUs from the set obtained after Stage I and
Stage II. The set of PMUs obtained after pruning stage provides
an optimal (minimum) solution for PMU placement for complete
observability of the system. The proposed method considered only
PMU measurements for complete observability of the system.
The algorithm has been tested on IEEE 14-bus, 24-bus, 30-bus,
New England 39-bus, IEEE 57-bus and 118-bus test system to validate the effectiveness of the proposed method. Detail system
information and single line diagram for each of the above networks
is available in [23,24]. Table 1 shows the optimal number of PMUs
required for different systems and corresponding bus locations. For
IEEE 14-bus system 4 PMUs are essential and the bus numbers for
placement are 2, 6, 7 and 9 without considering ZIBs. For IEEE 118bus system 32 PMUs are required for complete observability. The
results of the proposed method have been compared with the results of three other topological observability based PMU placement
methods reported by Xu and Abur [7], Chakrabarti and Kyriakides
[9] and Hurtgen and Maun [13]. Table 2 shows the comparative results of minimum number of PMUs required for complete observability of different systems. All the three methods show that
minimum 4 PMUs are required for complete observability of IEEE
14-bus system and so does the proposed method.

A new methodology for optimal placement of PMU is presented


in this paper. The new iterative method makes the test systems
topologically observable by placing a set of minimum PMUs. The
three stage algorithm is simple, fast and easy to implement. The
present method obtains optimal solution using simple network
connectivity information. The overall optimal solution obtained is
sufcient to take care of system observability under normal operating condition as well as for single PMU loss cases. Simulation results for different networks show the effectiveness of the proposed
method in obtaining the minimum number of PMUs required for
complete observability of power systems and also its advantage
of computational efciency.

Appendix A
Explanation of rule 1: SB buses are connected to an already allocated PMU bus. Elimination of PMU from any of SB bus does not
violate its observability. Hence this is the choice of elimination.
However this elimination may violate observability of other
bus. In that case PMU is retained to that particular SB bus.
Explanation of rule 2: When maximum number of EV bus is connected to a higher valency (HV) bus, PMU is allocated to that
bus. In case large number of higher valency buses has connection to similar number of EV bus, highest valency bus among
the HV buses is the choice for PMU retention. If more than
one bus has highest valency, any of them is the choice. Allocation of such PMU can make more than one PMUs of EV buses
redundant.
In stage-II, a lower valency bus with higher connectivity to APBOBs and connected to ZIB is the rst choice for PMU elimination.
This is supposed to be a redundant PMU bus in satisfying system
observability.

B.K. Saha Roy et al. / Electrical Power and Energy Systems 42 (2012) 7177

Appendix B
The proposed algorithm is illustrated on IEEE 24-bus test system. Single line bus connection diagram of IEEE 24-bus system is
given in Fig. B.1. ZIBs are bus number 11, 12, 17 and 24.
Stage I
Bus valeny for IEEE-24 bus system is computed and given in Table B.1.
Set V = MinV = 2 and MaxV = 6;
Bus 7 is the radial or minimum valency bus and connected to
bus 8. PMU is eliminated from 7 and retained at bus 8 to make
bus 7 observable. Bus 7 and 8 is removed from CB. APB = {8}.
Iteration 1
V = V + 1 = 3. EV = {4, 5, 6, 14, 18, 19, 20, 22, 24}. None of EV is
connected to APB i.e. SB = {nil}.
Higher valency buses i.e. HV = {2, 10, 16, 17, 21} are connected to
more than one bus of EV. Bus 17 is a ZIB and there is no PMU hence
not considered. PMU is retained at highest valency bus 10.
APB = {8, 10}. Bus 10 is removed from CB.
PMU is retained in bus 10. Set V = 2.
Iteration 2
V = V + 1 = 3. EV = {4, 5, 6, 14, 18, 19, 20, 22, 24}.
SB = {5, 6} of EV is connected to APB bus 10. PMU is eliminated
from SB one by one and system found observable. Bus 5 and 6 is
removed from CB. EV = {4, 14, 18, 19, 20, 22, 24}
HV = {16, 21}. PMU is retained at highest valency bus 16. Bus 16
is removed from CB; APB = {8, 10, 16}. PMU is retained in bus 16.
Set V = 2.
Iteration 3
V = V + 1 = 3. EV = {4, 14, 18, 19, 20, 22, 24}.
SB = {14, 19} connected to APB bus 16. CB is updated;
EV = {4, 18, 20, 22, 24}.
HV = {21}. PMU is retained at bus 21. Bus 21 is removed from
CB; APB = {8, 10, 16, 21}. Set V = 2.
Iteration 4
V = V + 1 = 3. EV = {4, 18, 20, 22, 24}.
SB = {18, 22} connected to APB bus 21. EV = {4, 20, 24}.
HV = {nil}. V < maxV.
Iteration 5
V = V + 1 = 4. EV = {1, 2, 3, 13, 15, 17, 23}.
SB = {15, 17} is connected to APB bus 16. EV = {1, 2, 3, 13, 23}.
HV = {nil}. V < maxV.
Iteration 6
V = V + 1 = 5. EV = {11, 12}.
SB = {11, 12} connected to APB bus 10. EV = {nil} HV = {nil}.
V < maxV.
Iteration 7
V = V + 1 = 6. EV = {9}.
SB = {9} is connected to APB bus 8. EV = {nil}. HV = {nil}.
V = maxV.
Stage II
V = 2;
Iteration 1:
V = V + 1 = 3; buses EV = {4, 20, 24}. Bus 24 is a ZIB. Hence
EV = {4, 20}. Both the buses have same connectivity to APBOB buses
and none of them is connected to ZIBs. PMUs are eliminated from
bus 4 and bus 20, system found observable. CB is updated. EV = {nil} and V < maxV.
Iteration 2:
V = V + 1 = 4; buses EV = {1, 2, 3, 13, 23}. Bus 13 has maximum
connection to OBs. Bus 1, 2, 3 and 23 have similar connection to
OBs. Bus 3 and 23 are connected to ZIBs. Ordered
EV = {13, 3, 23, 1, 2}. PMUs are eliminated one by one from EV and
found system unobservable at bus 23, 1 and at 2. Hence PMUs

77

are retained at bus 23, 1 and at 2. CB is updated.


APB = {1, 2, 8, 10, 16, 21, 23}. EV = {nil} and V < maxV.
Iteration 3:
V = V + 1 = 5; buses EV = {nil}. and V < maxV.
Iteration 4:
V = V + 1 = 6; buses EV = {nil};V = maxV.
Stage III. Pruning stage.
Set of PMU bus obtained are {1, 2, 8, 10, 16, 21 and 23}. There
are different possible PMU bus groups such as {2, 10} and
{16, 21}. which are connected to a non-PMU bus 6, 15, etc. All possible groups are checked by pruning operation. None of the nonPMU bus makes any group of PMU bus redundant. Pruning for individual PMU nd bus 10 as redundant in satisfying observability.
Hence optimal set is {1, 2, 8, 16, 21 and 23}.
References
[1] Phadke AG. Synchronized phasor measurements in power systems. IEEE Trans
Comput Appl Power 1993;6(2):105.
[2] Phadke AG. Synchronized sampling and phasor measurements for relaying and
control. IEEE Trans Power Deliver 1994;9(1):44252.
[3] Baldwin TL, Mill L, Boisen MB, et al. Power system observability with minimal
phasor measurement placement. IEEE Trans Power Syst 1993;8(2):70715.
[4] Milosevic B, Begovic M. Nondominated sorting genetic algorithm for optimal
phasor measurement placement. IEEE Trans Power Syst 2003;18(1):6975.
[5] Nuqui RF, Phadke AG. Phasor measurement unit placement techniques for
complete and incomplete observability. IEEE Trans Power Deliver 2005;20(4):
23818.
[6] Xu B, Abur A. Observability analysis and measurement placement for systems
with PMUs. In: Proceedings of the IEEE PES power systems conference and
exposition 2004;2: p. 9436.
[7] Xu B, Abur A. Optimal placement of phasor measurement units for state
estimation. Final project report, PSERC 2005.
[8] Momoh JA. Electric power system applications of optimization. Marcel Dekkar;
2001.
[9] Chakrabarti S, Kyriakides E. Optimal placement of phasor measurement units
for power system observability. IEEE Trans Power Syst 2008;23(3):143340.
[10] Dua D, Dambhare S, Gajbhiye RK, Soman SA. Optimal multistage scheduling of
PMU placement: an ILP approach. IEEE Trans Power Deliver 2008;23(4):
181220.
[11] Peng J, Sun Y, Wang HF. Optimal PMU placement for full network observability
using Tabu search algorithm. Int J Electr Power Energy Syst 2006.
[12] Peng C, Sun H, Guo J. Multi-objective optimal PMU placement using a nondominated sorting differential evolution algorithm. Int J Electr Power Energy
Syst 2010;32:88692.
[13] Hurtgen M, Maun JC. Optimal PMU placement using iterated local search. Int J
Electr Power Energy Syst 2010;32(8):85760.
[14] Sodhi R, Srivastava SC, Singh SN. Optimal PMU placement method for
complete topological and numerical observability of power system. Int J
Electr Power Syst Res 2010;80(9):11549.
[15] Ahmadi A, Beromi YA, Moradi M. Optimal PMU placement for power system
observability using binary particle swarm optimization considering
measurement redundancy. Expert Syst Appl Sci Direct 2011;38(6):72639.
[16] Hajian M, Ranjbar AM, Amraee T, Mozafari B. Optimal placement of PMUs to
maintain network observability using a modied BPSO algorithm. Int J Electr
Power Energy Syst 2011;33(1):2834.
[17] Rakpenthai C, Premrudeepreechachacharan S, Uatrongjit S, Watson N. An
optimal PMU placement method against measurement loss and branch
outage. IEEE Trans Power Deliver 2007;22(1):1017.
[18] Aminifar F, Khodaei A, Firuzabad MF, Shahidehpour M. Contingency
constrained PMU placement in power networks. IEEE Trans Power Syst
2010;25(1):51623.
[19] SahaRoy BK, Sinha AK, Pradhan AK. Optimal phasor measurement unit
placement for power system observability a heuristic approach. In:
Proceedings IEEE Symposium Series on, Computational Intelligence; 2011.
[20] Jamuna K, Swarup KS. Optimal placement of PMU and SCADA measurements
for security constrained state estimation. Int J Electr Power Energy Syst
2011;33(10):165865.
[21] Krumpholz GR, Clements KA, Davis PW. Power system observability: a
practical algorithm using network topology. IEEE Trans Power Ap Syst
1980;99(4):153442.
[22] Yoon YJ. Study of utilization and benets of phasor measurement units for
large scale power system state estimation, MS dissertation, Texas A&M
University; December 2005.
[23] Christie R. Power system test archive; August 1999. <http://www.ee.
washington.edu/research/pstca>.
[24] Pai MA. Energy function analysis for power system stability. Norwell: Kluwer
Academic Publishers; 1989.

You might also like