International Journal of Electrical Engineering.

ISSN 0974-2158 Volume 10, Number 2 (2017), pp. 157-170

© International Research Publication House
http://www.irphouse.com

Optimal Phasor Measurement Unit Placement in the

Observability of Power System by using Spanning
Tree Algorithm

P. Suresh Babu1
Assistant Professor, Dept. EEE
A.I.T.S-Rajampet, A.P, India.

Dr. M. Padmalalitha2
Professor & H.O.D, Dept. EEE
A.I.T.S-Rajampet, A.P, India.

S. Sujitha3
PG Student, Dept. EEE
A.I.T.S-Rajampet, A.P, India.

Abstract
Phasor Measurement Unit’s (PMU) are power system devices which provide
real time synchronized phasor measurements. Synchronization is achieved by
same-time sampling of voltage and current waveforms by means of timing
signals from the Global Positioning System Satellite (GPS). Synchronized
phasor measurements make higher the standards of power system monitoring,
control, and protection. Since PMU’s are expensive they need to be placed in
optimal way in power network in order to bring down the overall cost. This
paper solves the optimal placement of PMU’s problem in power network
using spanning tree algorithm. Integer linear programming algorithm is used
to determine the optimal number and location of PMUs needed to make the
network power system completely observable. Conventional techniques
assume PMU’s as multichannel in case PMU’s are single or two channel
optimal PMU locations will change. The following analyses explain optimal
158 P. Suresh Babu, Dr. M. Padmalalitha and S. Sujitha

PMU placement in single channel, two channel and multichannel cases.

Spanning tree algorithm has been applied for IEEE 14 bus system and IEEE
30 bus system for complete observability of these systems by considering and
ignoring zero injection buses separately.
Index Terms: Benchmarking, exhaustive search, measurement redundancy,
observability, optimal placement, phasor measurement units, state estimation.
Spanning tree Algorithm, PMU, Optimal PMU placement and full system.

I. INTRODUCTION
Power system plays an important role in electrifying our homes, Industries, offices
etc. So the protection and control of power system is important to engineers for safe
and reliable supply of power to consumers. Power system protection and control is
usually achieved by estimating state of power system regularly. Previously SCADA
has been used for state estimation and for monitoring power system. This
disadvantages forced engineers to discover new technology device called PMU
(phasor measurement unit). PMU eliminate this problem by measuring synchronous
values of voltage and current phasors. In the proposed method we find the optimal
placement of PMU in power system for its complete observability using Spanning tree
algorithm.
PMU placement at all substations allows direct measurement of the state of the
network. However PMU placement on each bus of a system is difficult to achieve
either due to cost factor since PMU is a highly expensive device or due to non-
existence limitation of communication facilities in some parts of substations. As a
consequence of Ohm’s law, when a PMU is placed at a bus neighboring buses also
become observable. This implies that a system can be made observable with a lesser
number of PMUs than total number of buses system to enable the power system
complete observability of power system network.

II. PHASOR MEASUREMENT UNIT

Phasor Measurement Unit (PMU) is a contemporary metering device with capability
of measuring the positive sequence of voltage and current phasors. The active power
flow in a given line in power system is proportional to the difference between the
phase angles of terminals of that line. Therefore, measuring the phase angles
difference across the power system transmission lines is important to power system
engineers. One of the greatest advantages of PMUs is the fact that in case of
unexpected incident, PMUs are capable of not only sending alerts to inform the
internal power system operator, but to have the capability of sending warnings to the
neighbouring network operator about the unexpected events as well. This can be
Optimal Phasor Measurement Unit Placement in the Observability… 159

accomplished by implementing the early warning system on operator’s desk that

would trigger simple alarms. These alarms will be initiated in case of violation of
power system operating condition. Since PMU’s are expensive they need to be placed
optimally in network to reduce overall cost. For this several different algorithms has
been proposed. Some of them are

1. Matrix manipulation algorithm

This method uses an exhaustive approach as in the binary search, but has some
differences. The algorithm is developed Matrix Manipulation convergence criterion
can therefore be simplified as to say that it compares past solution data to present
data. This cycle is perpetuated until all the Z vectors or all solutions are generated.
This method is used only on smaller power systems or large ones with little lines.

2. Binary particle swarm optimization

This Algorithm is based on the natural phenomena of individuals co-operating with
each other in a swarm to find the optimal location where the swarm needs to be at.
Each trial position and velocity is continuously updated and compared to its past
condition and also if it is better than the conditions adopted presently by the entire
swarm. If it is better, then this is the new swarm position and velocity. This process is
continued until the ideal location (optimal PMU placement solution) is found.
The fitness function used to determine the fitness of all the individual particles is:
f = c1.NPMU + c2.Nunobs
Where,
NPMU and Numbs is the number of PMUs presently in the system and the number of
unobserved buses, respectively. Binary particle swarm optimization doesn’t consider
computational time .Not user friendly and not adaptable. To overcome above
disadvantages spanning tree algorithm for optimum PMU placement has been
introduced which is user friendly, adaptable and suitable for large systems irrespective
of number of connection lines.

III. SPANNING TREE APPROACH

The PMU placement technique (or alternatively the tree search) is illustrated first
before the formal algorithm is presented. Consider the spanning tree in Fig 1(a) that is
composed of 14 nodes and with 20 branches.
160 P. Suresh Babu, Dr. M. Padmalalitha and S. Sujitha

Fig.1 (a): IEEE 14 bus system

Fig.1 (b): IEEE 30 bus system

Optimal Phasor Measurement Unit Placement in the Observability… 161

A. OBSERVABILITY
System is said to be completely observable if voltage and current phasor of every bus
in network is known. Fig.1 (b) shows a completely observed system. The voltage at
buses B , E and H are directly measured buses by PMU-1, PMU-2 and PMU-3
respectively, while voltages at buses A, C, D, F, G and I can be calculated using the
measured buses. Where bus voltages and line currents are directly. We define buses B
and F as PMU buses where bus voltages and line currents are directly measured. We
define buses A, C, D, F, G and I as calculated buses because their voltages are
calculated from the PMU measurements of the buses linked to them.

Fig 2: Complete Observability using PMU’s

B. TREE SEARCH PLACEMENT TECHNIQUE

The objective is to place PMUs such that the entire system is observable. The
envisioned technique consists of a series of “walks” along branches of a spanning tree
and queries are made on each node if a PMU placement is possible. The search
procedure starts at a root node and goes down the tree until it reaches a terminal node.
At this point, it backtracks and searches for another route. Identify a walk by a bus
pair from bus-to bus. A root node is specified arbitrarily; the search for PMU
locations is terminated when the procedure returns to this root node. At this time, the
spanning tree has been fully searched. The resulting PMU placement strategy
guarantees the existence of a completely observable condition for the spanning tree. It
likewise guarantees the complete observability for the parent graph.

C. ILLUSTRATION
Consider the spanning tree in Fig 1 (a) that is composed of 14 nodes and with 20
branches. Jump-start the placement process by arbitrarily designating Node 12 as the
root node. The logical first PMU placement should obviously be one bus away from
162 P. Suresh Babu, Dr. M. Padmalalitha and S. Sujitha

the Node 12 as the root node.

Fig.3: Flow chart of Spanning Tree Algorithm

Then, we take a series of forward moves along a chosen path defined by the nodal
sequence 6-11-10-9-7-8 each time querying for possible PMU placement. The next
logical placement will be at Node 9, which makes nodes 10,7observable with depth-
of-zero unobservability. Note that the PMUs are physically separated 3 buses from
each other along the chosen path.
Optimal Phasor Measurement Unit Placement in the Observability… 163

D. THEORETICAL FORMULATION
We can now pose the following rule that: given a desired depth of unobservability, the
next candidate PMU placement node must be of distance.
Dp =U+3
Where, -Dp is the number of buses away from the current PMU placement the next
one will be;
-U is the desired level of unobservability (For complete observability U=0).
1. The next move is to the terminal node 8 but PMU at this place observes only
one bus so the next PMU location will be at location 7.
2. By placing PMU’s 6,9,7 the entire span 5,6,11,10,9,7,8 observable and also
the nodes connected to the PMU buses(6,9,7) i.e.,12,13,14,4.is also
observable. We backtrack until we reach a node where an unobservable path
can be taken.
3. Then move to nodes 1, 2, 3 where here 1 act as sub root node and we again
apply spanning tree algorithm by placing PMU at node directly connected to
root node i.e., 2 which makes unobservable path 1, 2, 3 observable.
4. So by placing PMU’S at 2,6,7,9 entire networks is observable.
5. To ensure minimum number of PMU placements, it is necessary to perform
another search from a different root node from 1 to 14.
6. Compare all solutions for one with least number of PMU’s. The same
approach is applied to IEEE 30 bus system also.
7.
III. MINIMUM CHANNEL PMU
PMU’s are different types such as
1. Single channel PMU
2. Two channel PMU
In case there is only single channel the depth of observability condition Dp=U+3
changes to Dp=U+2 as shown in .Moreover, for single channel placing PMU starts
from root node itself as it provide less number of PMUs than placing at one bus away
from root node. If there is only single channel PMU, observable buses in this
condition is just the entire span of traverse, but not the other buses that also connected
to PMU buses as that in case of multichannel.
164 P. Suresh Babu, Dr. M. Padmalalitha and S. Sujitha

Fig.4. PMU placement for single channel PMU

In case there is only two channel the depth of observability condition Dp=U+3 is
same. If there is only two channels PMU, observable buses in this condition is just the
entire span of traverse, but not the other buses that also connected to PMU buses as
that in case of multichannel.

Fig.5: PMU placement for two channel PMU

IV. ILLUSTRATION OF SPANNING TREE APPROACH TO SINGLE

CHANNEL PMU
Consider the spanning tree in Fig 1 that is composed of 14 nodes and with 20
branches.
We can now pose the following rule that known a desired intensity of unobservability,
the after that candidate PMU placement node must be of distance.
Dp =U+2
Where,
- Dp is the number of buses away from the current PMU placement the next
one will be;
-U is the desired level of unobservability (For complete observability U=0).
Optimal Phasor Measurement Unit Placement in the Observability… 165

The next move is to place PMU at bus 14 according to Dp=U+2 rule and next will be
7 by applying same rule. Note that PMU’s are separated by two buses.

By placing PMU’s 12,14,7 the entire span 12, 13, 14,9,7,8 observable. Backtrack
until we reach a node where an unobservable path can be taken.

Then move to nodes 1,2,3,4,5,6,11,10 where here 1 act as sub root node and we again
apply spanning tree algorithm by placing PMU at sub root node and applying above
rule PMU placement will be at 3,5,11.

So by placing PMU’s at 1,3,5,7,11,12,14 entire networks is observable.

To ensure minimum number of PMU placements, it is necessary to perform another

search from a different root node from 1 to 14.

Compare all solutions for one with least number of PMU’s.

For two channel PMU placement will be at locations as given by Dp=u+3. Only buses
present in span which has been traversed is taken as observable buses but not buses
linked to PMU buses as that in case of multichannel.

V. SIMULATION RESILTS
It is shown that spanning tree algorithm has less average compared to other
algorithms. Algorithm is said to be better if its average is less.
For example: For spanning tree optimum locations is 1,7,10,12,15,19,30 means it
require 2,2,6,5,4,2,2 channel PMU’s and for ILP optimum locations is 1,2,10,
12,15,18,25 Which require 2,4,6,5,4,2,3 channel PMUs. As cost of PMUs is in order
6>5>4>3>2>1 lesser the channel of PMU required lesser will be overall cost. So
spanning tree algorithm reduces total overall cost more efficiently than other
algorithms. Spanning tree algorithm has been applied to IEEE 14 and IEEE 30 bus
system and optimal PMU placement for 14 bus system is 2,6,7,9 and for 30 bus
system it is 1,7,9,10,12,15,19,25,28,30 and if zero injection buses is considered it is
2,6,9 and 1,7,10,12,15,19,30 for 14 and 30 bus systems respectively. In case there is
only single channel or two channels PMU’S the possible locations for optimal PMU
placement is found as shown in below tables.
166 P. Suresh Babu, Dr. M. Padmalalitha and S. Sujitha

Results for Minimum Channel PMU for IEEE 14 bus System (1 channel)

Results for Minimum Channel PMU for IEEE 14 bus System (2 channel)
Optimal Phasor Measurement Unit Placement in the Observability… 167

Results for Minimum Channel PMU for IEEE 30 bus System (1 channel)
168 P. Suresh Babu, Dr. M. Padmalalitha and S. Sujitha

Results for Minimum Channel PMU for IEEE 30 bus System (2 channel)
Optimal Phasor Measurement Unit Placement in the Observability… 169

Results for Optimum PMU placement using Spanning Tree Algorithm

Comparison of Results obtained with other Algorithm

VI. CONCLUSION
From the results it can be concluded that spanning tree can be used for big and small
systems. This makes spanning tree algorithm a feasible option for use in locating the
PMUs. The average of spanning tree algorithm is less compared to other algorithms
which means spanning tree algorithm efficiently reduce overall cost of system. Not
only 14, 30 bus system spanning tree can also be used for any of large system within
short elapsed time. The results obtained in this project can be used practically for
placing PMU in large power systems so that the power system is secured and it can be
reliably operated.

REFERENCES
[1] Abdul-Aziz Fish, S.Chowdhury, and S.P.Chowdhury, “optimal PMU
placement in a power network for full system observability” proc IEEE, vol
978.no.1.pp. 4577-1022.nov. 2011.
[2] Reynaldo F. Nuqui, Member, IEEE, and Arun G. Phadke, Life Fellow, IEEE
‘’Phasor Measurement Unit Placement Techniques for Complete and
Incomplete Observability’’ IEEE TRANSACTIONS ON POWER
DELIVERY, VOL. 20, NO. 4, OCTOBER 2005.
170 P. Suresh Babu, Dr. M. Padmalalitha and S. Sujitha

[3] Sanjay Dambhare Devesh Dua Rajeev Kumar Gajbhiye S. A. Soman

‘’Optimal Zero Injection Considerations in PMU Placement: An ILP
Approach’’.
[4] D.Dua, S.Dambhare, R.K.Gajbhiye and S.A.Soman, “Optimal Multistage
Scheduling of PMU Placement: An ILP Approach”, IEEETransactions on
Power Delivery, Vol.23, Issue 4, pp.1812-1820, 2008’’.
[5] D. Novolsel, K.Vu, V. Centeno, S. Skok, M. Begovic, “Benefits of
synchronized measurement technology for power grid application”,
Proceeding of the 40th international conference on system sciences, 2007.
[6] M. Hajian, A. M. Ranjbar, T. Amraee and A. R. Shirani, “Optimal Placement
of Phasor Measurement Units: Particle Swarm Optimization Approach”, Proc.
of Int. Conf. on Intelligent Systems Applications to Power Systems 2007, Toki
Messe, Niigata, pp.1-6, November 2007.
[7] R. F. Nuqui, “State estimation and voltage security monitoring using
synchronized phasor measurementSSSs,” Ph.D. dissertation, Dept. Elect.
Computer. Eng., Virginia Polytechnic Inst. State Univ., Blacksburg, VA,
2001.
[8] T. L. Baldwin, L. Mili, M. B. Boisen, and R. Adapa, “Power system
observability with minimal phasor measurement placement,” IEEE Trans.
Power Syst., vol. 8, no. 2, pp. 701–715, May 1993.
[9] F.Aminifar, C.Lucas, A.Khodaei and M.Fotuhi-Firuzabad, “Optimal
Placement of Phasor Measurement Units Using Immunity Genetic
Algorithm”, IEEE Transactions on Power Delivery, Vol.24, Issue 3, pp.1014-
1020, July 2009.
[10] M.Hurtgen and Jean-Claude Maun, “Advantages of power system state
estimation using Phasor Measurement Units”, Proc. of 16th PSCC, Glasgow,
Scotland, pp.1-7, July 2008.

Author’s Profile:
S. SUJITHA: She was born in 1993. She obtained her Bachelor degree in Electrical
and Electronics Engineering in 2014 from NIST, Rajampet. Currently pursuing her
Post Graduation in Electrical Power Engineering in AITS, Rajampet, Kadapa (dist.).