Advances in Computer Science Research, volume 50

4th International Conference on Electrical & Electronics Engineering and Computer Science (ICEEECS 2016)

Optimal PMU Placement in Multi-configuration Power Distribution

Networks
Van-khoi TRAN , He-sheng ZHANG2, b, Van-nghia NGUYEN3, c
1, a

1
School of Electrical Engineering, Beijing Jiaotong University, Beijing 100044, China
2
School of Electrical Engineering, Beijing Jiaotong University, Beijing 100044, China
3
Department of Electrical Engineering, University of Transport & Communications, Hanoi 100000,
Vietnam
a
email: [email protected], bemail: [email protected], cemail:
[email protected]

Keywords: Phasor measurement unit (PMU); Optimal placement; Integer linear programming
(ILP); Distribution system observability; Multi-configuration network

Abstract. Phasor Measurement Unit (PMU) is more and more concerned in power distribution
network due to its great benefit. In near future, it may become an essential device in the modern
distribution systems. The Optimal PMU Placement (OPP) problem on transmission network for full
system observability, was fully done. However, in distribution network the change of configuration
in normal operation and contingency modes has a very large impact to find the location of PMUs on
network. Some methods have been proposed to solve this problem however the researchers had to
spend lot of effort to calculate the system data in all configurations of network. This paper proposes
a new solution to find the optimal location of PMUs, which can ensure the full system observability
in all operation configurations of distribution network with calculating only based on one
configuration. To justify the methodology of solution, the simulations on IEEE RBTS-2 bus and 33
buses distribution systems are presented for experiment. The results of this study show that the
proposed solution is effective and feasible.

Introduction
Synchrophasor Measurements has become of interest in supervision of stability for power
distribution network systems because of more dynamic requirements for the system caused by
renewable power infeed such as wind and solar as well as energy trade [1-3]. The major limit to
PMU deployment in distribution systems is the cost, but several projects have already been in
progress worldwide [4]. In near future, PMU may become an essential device in the modern
distribution systems. The OPP problem on transmission network for full system observability, was
fully done by many different methods as Integer Linear Programming (ILP), Heuristic, or
Metaheuristic [5-9]. However, in distribution network the change of configuration in operation
modes has a very large impact to find the location of PMUs on network.
Abdelsalam et al. [10] investigated the impact of network reconfiguration on PMUs location,
then proposed one PMU placement solution to cover fully system observability in all configurations
by merging the PMU location in all configurations into one. This solution is not effective because
almost all buses need to install PMU, thus leading to high costs.
A new algorithm of PMU placement for distribution networks was proposed by Emad Jamil [11].
Although the computational time of this approach may be improved, the obtained result is not as
good as ILP method. Moreover, this method can use for one configuration, with multi-configuration
networks it is difficult to apply.
The authors in [12] proposed an overall optimization solution to specify location of PMUs that is,
each PMU is tested for all configurations and it is considered acceptable only if it allows the
accuracy limits to be met in all of them. This solution can reduce the number of PMUs, it
nevertheless has to scan through all configurations in which case that demands lot of effort in
calculating.

Copyright © 2016, the Authors. Published by Atlantis Press.

This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/). 508
 Advances in Computer Science Research, volume 50

The work was extended in [13] with considering the performance degradation of measurement
devices. With the aim of restricting the searching space, Liu et al. [14] developed a model for robust
measurement placement with the saturate number of system. Recently, the authors in reference [15]
proposed a new model for robust meter placement problem that specify the minimization of the
worst case total estimation variance. However, as similar to the solution in [12] these methods still
require very much effort to calculate for all network configurations.
This paper proposes a new solution to find the location of PMU, which bases on only one
configuration of network. From the known location of tie switches, by closing all these switches
and then applying the proposed constraint formula for bus observation in that configuration we will
find out the placement of PMU, which can ensure the full system observability in all operation
configurations of network. The proposed solution is tested on IEEE RBTS-2 bus and 33 bus
distribution systems, and results are compared with other solutions.

Theoretical Approach
Configuration change in distribution network
The configuration of distribution network is created by the status of sectionalizing switches and
tie switches. The location of these switches and the status of them in different operation conditions
are specified by optimal algorithms [16-19].
The change of configuration in distribution network is the process of opening or closing status of
sectionalizing switches and tie switches to transfer loads among the feeders. These switching are
performed in such a way that the radiality of the network is maintained and all the loads are
energized. This work can reduce the power losses, hold the balance of load, improve the power
quality and increase the reliability as well as the security of system. Moreover, it also help to
improve the bus voltage profile and reduce line losses, as well as avoid the overloading network
branches [20].
In modern distribution network, all sectionalizing switches and tie switches are automatic
switches and the configuration of power network may be controlled automatically at management
centre.
Power distribution network observation
In general, the problem of network observation is divided into two categories, the numerical
category and the topological category [21]. The numerical methods base on the mathematical model
of power network to solve the measurement device placement problem, whereas the topological
methods use the decoupled measurement model and graph theory.
Numerical methods require lot of network information such as line parameter, load and generator
profile, or status of switches in each network configuration; then topological method is simpler. The
objective function for full network observation is defined as follows:
n
1
min wk x k
k 1
Subject to fi(xk) =Aconfig i.xk 1
where xk is a variable which is 0 or 1 whether PMU placed at bus or not; wk is the cost of the PMU
installed at that bus; n is number of bus; fi(x) is a vector function, whose entries are non-zero if the
corresponding bus voltage is solvable using the given measurement; Aconfig_i (i=1,.., q) is adjacency
matrix of power system in each configuration, q is number of configuration. The elements of matrix
Aconfig_i are defined as,
Ak,l = 1 if k = l
= 1 if bus k is connected to bus l.
= 0 otherwise.

Proposed Solution
In distribution network, there will be up to 3m configurations if the number of tie switches is m

509
 Advances in Computer Science Research, volume 50

[22]. When m is a large number that is impossible to find out the placement of PMUs in each
configuration and then merging all of them into one, then it needs to find another solution. Example
as in Fig. 1, it supposes that the system consists of three configurations as listed as in Table 1.
According to Eq. 1, we have the constraint of bus 2 as follow:
f1 x2 x1 x2 x3 1
f2 x2 x2 x3 1 2
f3 x2 x1 x2 1
Because xk are binary variables so Eq. (2) can be written as:
f x2 f1 x2 . f2 x2 . f3 x2 1 3
Substituting f1(x2), f2(x2), f3(x2), Inequality (3) is rewritten as follows:
f x2 x1 x2 x3 x2 x3 x1 x2 1 4
After abbreviating the inequality (4) we obtain the final irreducible constraint:
f x2 x2 x1 x3 1 5
From constraint (5), it can be seen that only one PMU installed at bus 2 or two PMUs are placed
simultaneously at bus 1 and bus 3, is enough to ensure the observability for bus 2. Although xk are
binary variables, they are also counted as integer variables with their value from 0 to 1. Based on
this clue, because value of xi (i=0, 1,...,4) is only 0 or 1 so x1x3 = 1 is equivalent to x1 + x3 = 2, then
we may rewrite (5) in algebraic form as (6).
f x2 2x2 x1 x3 2 6
Comparison between (5) and (6), it is clearly that the meaning is still the same.
Similarly, we can get the constraints of other buses:
f x0 2x0 x1 x4 2
f x1 x0 2x1 x2 2
7
f x3 x2 2x3 x4 2
f x4 x0 x3 2x4 2

Table 1. 5-Bus network configurations

Sw1 Tie1 Comment
Config 1 close open Normal mode
Config 2 open close Load balancing
Config 3 close close Line 3 outage

Fig. 1. A 5-bus distribution network

Distribution network is commonly radial configuration, one bus often connects to two other
buses therefore inequality (6) seems to be a general form. However, when bus having many
connections as illustrated as in Fig. 2, this conversion also can be applied generally.
3
2 r-1

1 r

Bus k
Fig. 2. Multi-connection bus
To ensure the observability for bus in any cases, the constraint will be:
f xk xk x1 x2 x3... xr‐1 xr 1 8
then the conversion as follow:
f xk r.xk x1 x2 x3 ... xr‐1 xr r 9
where, r is total number of bus which has direct connection to bus k.
In constraint inequality (9) we do not need to concern about status of sectionalizing switches or

510
 Advances in Computer Science Research, volume 50

tie switches even the outage of lines. That means the bus observation can be remained in any
configurations of distribution network.
Through above analysis, we propose a solution by applying constraint inequality (9) for network
configuration when closing all sectionalizing switches and tie switches, in that case the PMU
placement will ensure the full system observability in all operation configurations of power network.
Applying the above proposed solution for network in Fig. 1, closing Sw1 and Tie1, then the
constraint of buses can be written as follow:
f x0 2x0 x1 x4 2
f x1 x0 2x1 x2 2
f x2 x1 2x2 x3 2 10
f x3 x2 2x3 x4 2
f x4 x0 x3 2x4 2
These constraints are the same with constraints (6) and (7). That means the PMU placement
solution for one configuration, in which case all the sectionalizing and tie switches are at closing
status that maybe ensure the full system observability in all operation configurations of network.
From above analysis, the object function and constraints of OPP problem become:
n
min wk x k 11
k 1
Subject to f x A.x b
A is the connectivity matrix of power distribution system, the elements of matrix A are specified
as follows:
.
1 . 12
0 .
Vector b expresses the expectation with regard to the observability for each bus.
13
Now we can use one of any methods as ILP, Heuristic or Metaheuristic to solve the OPP problem.
The obtained result will satisfy the requirement of robust observability in the change of network
configuration as well as in contingency modes.

Simulation Results
To experiment the proposed solution, simulations on the IEEE RBTS-2 bus and 33 bus
distribution systems are carried out in MATLAB.
Fig. 3 illustrates the one line diagram of IEEE RBTS-2 bus distribution system. This network
consists of 22 load buses, 5 sectionalizing switches and 2 tie switches; and the placement of PMU
after applying the proposed solution is presented in Table 2. We need 11 PMUs to cover all system
in any configurations. With the exception of PMU buses, all other buses are observed by two
different PMUs, therefore the full system observability is still remained when switches change their
status or even one line outage.

511
 Advances in Computer Science Research, volume 50

Fig. 3. Diagram of IEEE RBTS-2 bus distribution system

Table 2. Placement of PMU on RBTS-2 Bus System
Number of Tie switch 2
Number of Configuration 9
Number of PMU 11
PMU location 1, 3, 5, 7, 10, 12, 14, 16, 18, 20, 22
In order to depict the performance of proposed solution, the result will be compared with the
solution in reference [10]. There are many configurations we only give three typical configurations
for experiment. The detail information of these configurations as in Table 3 then Table 4 shows the
obtained result. As we can see, it needs up to 19 PMUs for covering all system in only three
configurations. Moreover, it spends one time to find out the placement of PMU for one
configuration, and will be n times for n configurations. That is clearly not feasible and effective
when comparing to proposed solution with only one time to find out the final goal.
To illustrate more clearly the effect of proposed solution, the simulation is carried out on the 33
bus distribution system. This network contains 31 load buses, and that maybe change among 243
configurations by 11 sectionalizing switches and 5 tie switches. Fig. 4 shows the diagram of this
system. The simulated result is demonstrated in Table 5. To ensure fully system observability, 17
PMUs need to be installed on network. With this PMU placement, one bus except PMU buses is all
covered by at least two different PMUs, that can ensure reliably the full system observability in any
configurations.

Fig. 4. Diagram of 33 buses distribution network

512
 Advances in Computer Science Research, volume 50

Table 3. Three configurations of RBTS-2 Bus system

Config. 1 Config. 2 Config. 3
Sw1 close close close
Sw2 close open close
Sw3 close close close
Sw4 close close close
Sw5 close open close
Tie1 open close close
Tie2 open close close
Comment Normal Load balancing Lines 6-7 and 15-16 outage
Table 4. Placement of PMU for different configurations of RBTS-2 bus
Number of PMU PMU location
Config. 1 8 1, 4, 7, 10, 13, 16, 19, 22
Config. 2 9 2, 5, 6, 10, 12, 15, 18, 21, 22
Config. 3 9 3, 4, 7, 9, 12, 15, 18, 20, 23
Merging 19 1, 2, 3, 4, 5, 6, 7, 9, 11, 12, 13, 15,16, 18, 19, 20, 21, 22, 23
Table 5. Placement of PMU on 33 bus distribution system
Number of tie switch 5
Number of configuration 243
Number of PMU 17
PMU location 1, 3, 4, 6, 8, 10, 12, 13, 15, 17, 19, 21, 24, 27, 29, 31, 33

The simulated results on the test systems illustrate the salient effect of proposed solution. It may
only know about location of tie switches on network that can find out the final goal without the load
and generator profile or any specific configurations. This will save significantly the effort and time
to calculate the system data in each configuration as previous methods. Based on the finding of this
study, the problem of PMU placement for robust measurement in multi-configuration power
distribution networks that can be solved simply and effectively.

Conclusions
This paper proposes a new solution to solve the PMU placement problem in multi-configuration
distribution network. Based on the transforming of constraints of all configurations into one general
form, the proposed solution can find out the final goal with calculating for only one configuration.
The effect of proposed solution is demonstrated by simulations on IEEE RBTS-2 bus and 33 buses
distribution systems. The results of this study show that the proposed solution is feasible and
effective.

Acknowledgment
The authors of this paper would like to acknowledge the support of the Science and Technology
Development Plan of China Railway (Grant 2015Z005-D) and the National Natural Science
Foundation of China (Grant 61174179).

References
[1] Wache, M. and D.C. Murray. Application of Synchrophasor Measurements for distribution
networks [C]. 2011 IEEE Power and Energy Society General Meeting, 2011.
[2] Ochoa, L.F. and D.H. Wilson. Using synchrophasor measurements in smart distribution
networks [C]. 21st International Conference on Electricity Distribution CIRED, 2011.
[3] Muscas, C., et al., Smart electric energy measurements in power distribution grids [J]. IEEE
Instrumentation & Measurement Magazine, 2015 18(1) 17-21.

513
 Advances in Computer Science Research, volume 50

[4] Meier, A.v., et al. Micro-Synchrophasors for Distribution Systems [C]. IEEE Power & Energy
Society Innovative Smart Grid Technologies Conference, 2014.
[5] Xu, B. and A. Abur, Observability Analysis and Measurement Placement for Systems with
PMUs [C]. Power Systems Conference & Exposition, 2004. 943-946.
[6] Gou, B., Optimal Placement of PMUs by Integer Linear Programming [J]. IEEE Transaction
on Power Systems, 2008 23( 3) 1525-1526.
[7] Shewale, R.H., et al., Optimal placement of phasor measurement unit for power system
observability by heuristic search method [J]. International journal of Advanced Technology &
Engineering Research, 2012 2( 2) 128-133.
[8] Aghaei, J., et al., Multi-objective Phasor Measurement Unit Placement in Electric Power
Networks: Integer Linear Programming Formulation [J]. Electric Power Components & Systems,
2015 43(17) 1902-1911.
[9] Babu, R. and B. Bhattacharyya, Optimal allocation of phasor measurement unit for full
observability of the connected power network [J]. Electrical Power and Energy Systems, 2016 79
89–97.
[10] Abdelsalam, H.A., A.Y. Abdelaziz, and V. Mukherjee. Optimal PMU placement in a
distribution network considering network reconfiguration [C]. International Conference on Circuit,
Power and Computing Technologies, 2014.
[11] Jamil, E., M. Rihan, and M.A. Anees. Towards optimal placement of phasor measurement
units for smart distribution systems [C]. Power India International Conference (PIICON), 2014 6th
IEEE, 2014. Delhi.
[12] Liu, J., et al., Trade-Offs in PMU Deployment for State Estimation in Active Distribution
Grids [J]. IEEE Transactions on Smart Grid, 2012 3(2) 915-924.
[13] Liu, J., et al. Optimal placement for robust distributed measurement systems in active
distribution grids [C]. IEEE International Instrumentation & Measurement Technology Conference,
2013.
[14] Wang, H., et al. A robust measurement placement method in distribution system state
estimation [C]. 2014 IEEE PES Asia-Pacific Power and Energy Engineering Conference (APPEEC),
2014.
[15] Damavandi, M.G. and V. Krishnamurthy, Robust Meter Placement for State Estimation in
Active Distribution Systems [J]. IEEE Transactions on Smart Grid, 2015 6(4) 1972-1982.
[16] Celli, G. and F. Pilo, Optimal sectionalizing switches allocation in distribution networks [J].
IEEE Transactions on Power Delivery, 1999 14(3) 1167-1172.
[17] Lim, I.-H., et al., An Algorithm on Optimal Placement Decision of Automatic Switches for 6
Sections/3 Links Configuration in DAS [J]. Journal of Electrical Engineering & Technology, 2011
6(3) 328-333.
[18] Ray, S., A. Bhattacharya, and S. Bhattacharjee, Optimal placement of switches in a radial
distribution network for reliability improvement [J]. Electrical Power and Energy Systems, 2016 76
53-68.
[19] Baran, M.E. and F.F. Wu, Network reconfiguration in distribution systems for loss reduction
and load balancing [J]. IEEE Transactions on Power Delivery, 1999 4(2) 1401-1407.
[20] Mishra, S., D. Das, and S. Paul, A comprehensive review on power distribution network
reconfiguration [J]. Energy Systems, 2016 1-58.
[21] MohammadiIvatloo, B., Optimal Placement of PMUs for Power System Observability Using
Topology Based Formulated Algorithms [J]. Journal of Applied Sciences, 2009 9(13) 2463-2468.
[22] Abul’Wafa, A.R., A new heuristic approach for optimal reconfiguration in distribution
systems [J]. Electric Power Systems Research, 2011 81 282-289.

514