1

A Micro-PMU Placement Scheme for Distribution

Systems Considering Practical Constraints
Reetam Sen Biswas, Student Member, IEEE, Behrouz Azimian, Student Member, IEEE, and Anamitra Pal,
Member, IEEE.

Abstract— This paper presents an innovative approach to optimal PMU placement problem for the transmission system:
micro-phasor measurement unit (micro-PMU or µPMU) (i) mixture of single-phase, two-phase, and three-phase laterals,
placement in unbalanced distribution networks. The (ii) presence of distributed loads along the length of the feeders,
methodology accounts for the presence of single-and-two- (iii) zero-injection phases (ZIPs), (iv) unknown voltage
phase laterals and acknowledges the fact that observing one regulator/transformer tap ratios, and (v) frequent changes in
phase in a distribution circuit does not translate to switch configurations. Prior literature [6]-[15] has not
observing the other phases. Other practical constraints such considered all the above-mentioned practical constraints
as presence of distributed loads, unknown regulator/ simultaneously in their problem formulation.
transformer tap ratios, zero-injection phases (ZIPs), Another limitation of prior research is the lack of distinction
modern smart meters, and multiple switch configurations between node observability and phase observability. Unlike the
are also incorporated. The proposed µPMU placement transmission network, distribution networks have three-phase,
problem is solved using integer linear programming (ILP),
two-phase, or single-phase nodes/feeders. Consequently,
guaranteeing optimality of results. The uniqueness of the
monitoring the distribution network translates to observing the
developed algorithm is that it not only minimizes the µPMU
phase voltages, and not the node voltages. Furthermore, prior
installations, but also identifies the minimum number of
phases that must be monitored by them. research on μPMU placement only provided the node locations
where the μPMUs were to be installed. Locating 𝑛 nodes does
not necessarily imply that 𝑛 μPMUs would be required, as at a
Keywords—Distribution system, Integer programming,
Micro-PMUs, Observability, Smart meter. given node, more than one μPMU might be needed. The exact
number of μPMUs to be installed at a node depends on the
I. INTRODUCTION number of outgoing phases that must be monitored from that
node location and the number of measurement channels that the
W ITH the continuous addition of distributed energy
resources (DERs) and active controllers occurring at the
distribution levels, the power flowing through the distribution
μPMU has. For the distribution system, any sensor placement
scheme is incomplete if it does not provide this vital
information. That is, an optimal μPMU placement algorithm for
feeders is becoming increasingly uncertain [1]. Instances of the distribution system must minimize the combination of the
unstable power supply, unintentional islanding, and voltage number of μPMUs, node locations, and the number of phases
stability issues are manifesting more frequently [2]. Hence, that must be observed.
there is a pressing need for real-time synchronized monitoring The primary objective of this research is to develop a μPMU
of distribution networks [3]. This has led to the creation of high placement algorithm subject to the above-mentioned practical
precision micro-PMUs (μPMUs) [4] as well as modern smart constraints of the distribution system. The proposed algorithm
meters [5] that can produce time-synchronized measurements. is also generic enough to account for the presence of pre-
Robust sensor placement methods for distribution networks installed unbundled smart meters (USMs) in the system. USMs
proposed in recent literature have focused on frequent network are modern smart meters that can report time-synchronized data
reconfigurations [6] and relay operations [7]. Ref. [8] proposed at the rate of one sample per second [5]. The relatively fast
an optimal placement scheme of μPMUs and conventional output rate of USMs compared to conventional smart meters as
smart meters to ensure observability during contingencies. well as the time-synchronized nature of their measurements
Optimal μPMU placement schemes for effective anomaly makes USMs suitable candidates for integration with μPMUs.
detection was investigated in [9]. In [10], the authors created an
optimal measurement infrastructure using different devices for II. THE NEED FOR A NEW MICRO-PMU PLACEMENT SCHEME
the distribution grid. In [11], μPMUs and intelligent electronic
devices (IEDs) were optimally allocated for the distribution Since the transmission system is usually balanced,
system using heuristic techniques. In [12]-[15], the μPMU observability of one phase translates to observing all the other
placement problem was treated in a similar way as the optimal phases. However, most distribution networks are unbalanced,
PMU placement problem for transmission systems. and often have only one-or-two phases present at a node.
The following attributes of the distribution network make Therefore, observing one phase in a distribution system does
optimal μPMU placement a more challenging problem than the not necessarily translate to observing the other phases. Fig. 1
shows the phase-connectivity between four nodes 𝑖, 𝑗, 𝑘, and 𝑙
of a distribution system. If node 𝑘 is to be indirectly observed
This research has been funded by the ARPA-E research grant DE-AR-
0001001. by a μPMU, node 𝑗 must have preference over node 𝑙, because

978-1-7281-5508-1/20/$31.00 ©2020 IEEE

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SILCHAR. Downloaded on November 30,2021 at 08:32:42 UTC from IEEE Xplore. Restrictions apply.
 2

phase A of node 𝑘 cannot be observed from node 𝑙, but all the the three-phase voltage phasor at node 𝑘 (from the μPMU at
three phase voltages of node 𝑘 can be observed from node 𝑗. node 𝑗), and the active and reactive power demand at node 𝑘
(from the USM at node 𝑘), the current phasor from node 𝑘 to
node 𝑙 can be calculated. Consequently, the C-phase voltage at
node 𝑙 can be found, without placing a μPMU at node 𝑘 or node
𝑙. In essence, having one or more phase voltages monitored by
a USM has a similar effect on observability as a ZIP.

III. MATHEMATICAL FORMULATION

Let the power distribution system be represented by an
Fig. 1: Example of phase connectivity between different nodes of a distribution undirected graph 𝒢(𝑽, 𝑬), where 𝑽 is the set of nodes and 𝑬 is
network. the set of edges. The μPMU placement must be done with the
Distributed loads are electrical loads that are present at objective that every phase at every node of the system is
different lengths along a feeder and are a unique characteristic observed. To attain this objective, the original graph 𝒢 must be
of the distribution system. Fig. 1 shows the presence of modified to account for the individual phases. Hence, from the
distributed loads on phase A, along the feeder joining nodes 𝑖 graph 𝒢 we form the graph 𝒢 ′ (𝑽′, 𝑬′ ), where 𝑽′ represents the
and 𝑗. This means that if a μPMU is placed at node 𝑖, the A set of phases, and 𝑬′ represents every edge that joins a distinct
phase voltage of node 𝑗 cannot be observed from node 𝑖. As phase of one node to the same phase of another node. Each
such, the presence of distributed loads along specific phases of element of the set 𝑽′ is represented by a pair of numbers (𝑥, 𝑦)
a distribution feeder poses an additional constraint for such that 𝑥 ∈ {1,2, … , 𝑀}, where 𝑀 = |𝑽| and 𝑦 ∈ {1,2,3},
distribution system observability. The same argument holds where 1 refers to A-phase, 2 refers to B-phase, and 3 refers to
true for the presence of transformer/voltage regulators and C-phase inside node 𝑥. Next, a lexicographic ordering scheme
switches on individual phases. Since the voltage regulator/ is introduced among the phases. For the two phases 𝑣1 =
transformer tap ratios vary frequently with system conditions, (𝑥1 , 𝑦1 ) and 𝑣2 = (𝑥2 , 𝑦2 ), 𝑣1 ≺ 𝑣2 if 𝑥1 < 𝑥2 . Every edge 𝑒 ≡
especially for a distribution system, the tap ratios should be {𝑣1 , 𝑣2 } ∈ 𝑬′ joins two phases 𝑣1 = (𝑥1 , 𝑦1 ) and 𝑣2 = (𝑥2 , 𝑦2 )
treated as unknown. Therefore, placing a μPMU on one side of from two different nodes 𝑥1 and 𝑥2 . In subsequent notations, if
a voltage regulator or transformer does not translate to an edge 𝑒 ≡ {𝑣1 , 𝑣2 } is specified, it will be assumed that 𝑣1 ≺
observing the other side. For example, the presence of a voltage 𝑣2, implying that 𝑣1 is the low end and 𝑣2 is the high end of the
regulator in the B phase of the feeder joining nodes 𝑖 and 𝑗, edge 𝑒.
prevents a μPMU located at node 𝑖 to observe the B phase Every node 𝑥𝑖 ∈ 𝑽 for 1 ≤ 𝑖 ≤ 𝑀 is now associated with a
voltage at node 𝑗. Similarly, the distribution system is also binary variable 𝑧𝑖 such that
characterized by the presence of “switches”, which operate and 1 if µPMU is placed on node xi
𝑧𝑖 = { (1)
change status at a much higher frequency than the transmission 0 otherwise
system. Since μPMU placement is a planning problem, an An integer variable 𝑛𝑖 is also associated with every node 𝑥𝑖 ,
optimal μPMU placement scheme must ensure observability for such that 𝑛𝑖 denotes the number of μPMUs placed inside node
all “feasible” switch configurations (see Section III for the 𝑥𝑖 . Now, for any phase 𝑣 ∈ 𝑽′ present at a given node, let 𝑳𝑣
definition of a feasible configuration). For example, if a μPMU denote the set of edges for which 𝑣 is the “low end”. Similarly,
is placed at node 𝑖, the presence of a “switch” in phase C of the let 𝑯𝑣 be the set of edges for which 𝑣 is the “high end”; that is
feeder between nodes 𝑖 and 𝑗, hinders the observability of the 𝑬′𝑣 = 𝑳𝑣 ∪ 𝑯𝑣 . Then, every edge 𝑒 can be associated with two
C-phase voltage at node 𝑗 from node 𝑖, if the switch is open. binary valued variables 𝑔𝑒𝑙 and 𝑔𝑒ℎ such that [16]
Concept of zero-injection buses (ZIBs) has been exploited in 1 if a µPMU observes the low end of egde e
the transmission system for reducing the number of installed 𝑔𝑒𝑙 = { (2)
0 otherwise
PMUs [16]. The concept of ZIBs gives way to the concept of
1 if a µPMU observes the high end of edge e
zero-injection phases (ZIPs) for distribution networks. This is 𝑔𝑒ℎ = { (3)
0 otherwise
because at a given node among the three phases only a subset
Based on (1)-(3), the objective function for μPMU placement
of the phases might have zero-injections. Considering that a
would be expressed as shown below.
μPMU is placed at node 𝑗, and phase B of node 𝑘 does not have 𝑀 𝑀
any spot load (or injection), the phase B voltage of node 𝑙 can
𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 (∑ 𝑧𝑖 + ∑ (𝑔𝑒𝑙 + 𝑔𝑒ℎ ) + ∑ 𝑛𝑖 ) (4)
be observed from the μPMU at node 𝑗. However, if there is a
𝑒∈𝑬′
“spot load” present, as in the phase C of node 𝑘, the C-phase 𝑖=1 𝑖=1

voltage of node 𝑙 cannot be observed from node 𝑗. Equation (4) simultaneously minimizes the number of affected
Modern smart meters such as USMs also influence network nodes, the number of monitored phases, and the total number of
observability. Consider now that node 𝑘 of Fig. 1 is indirectly μPMUs. The optimization problem is formulated as an integer
linear programming (ILP) problem that guarantees globally
monitored by a μPMU placed at node 𝑗, and a USM is placed at
optimum solutions. The different constraints applied to the
node 𝑘 that provides active and reactive power injection
optimization problem are described below.
information of the spot loads at node 𝑘. With the knowledge of

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SILCHAR. Downloaded on November 30,2021 at 08:32:42 UTC from IEEE Xplore. Restrictions apply.
 3

equation. If for each of the 𝑓 network topologies,

1. Phase observability constraint (binary variables 𝑔𝑒𝑙 , 𝑔𝑒ℎ ): 𝒢1′ , 𝒢2′ , … , 𝒢𝑓′ , the respective A and B matrices are
For every phase 𝑣 ∈ 𝑽′, the constraint for phase 𝑨1 , 𝑨2 , … , 𝑨𝑓 , and 𝑩1 , 𝑩2 , … , 𝑩𝑓 , the μPMU placement
observability is shown below [16]. solution will satisfy all 𝑓 switch configurations if the
∑ (𝑔𝑒𝑙 + 𝑔𝑒ℎ ) ≥ 1 (5) following holds true.
𝑒∈𝑬′𝑣 𝑨1 𝑩𝟏
𝑨2 𝑩𝟐
2. Constraint for number of μPMUs (integer variable 𝑛𝑖 ): [ ⋮ ]𝑿 ≥ [ ⋮ ] (10)
Let a μPMU have 𝐾 measurement channels, where one
channel can measure one voltage and one current phasor. Also, 𝑨𝑓 𝑩𝑓
let 𝑽′𝑖 ⊆ 𝑽′ contain the phases that are present at node 𝑥𝑖 . Then, 3. Handling of ZIPs and USMs: The concept of zero-
number of μPMUs, 𝑛𝑖 , to be placed at node 𝑥𝑖 is given by injections is handled in a manner similar to what was done
∑𝑣∈𝑽′ ∑𝑒∈𝑳𝑣 𝑔𝑒𝑙 + ∑𝑣∈𝑽′ ∑𝑒∈𝑯𝑣 𝑔𝑒ℎ in [16]. Let 𝑽𝑍𝐼 denote the set of all phases whose
𝑛𝑖 ≥ 𝑖 𝑖
(6) injections are known (either by them being a ZIP or
𝐾 through a USM). Let the neighborhood of a phase 𝑣 ∈ 𝑽′ ,
3. Constraint for the affected nodes (binary variable 𝑧𝑖 ): denoted by 𝑵𝑣 , contain the phase 𝑣 itself and all phases that
For every phase 𝑣 ∈ 𝑽′, and every edge 𝑒 ∈ 𝑳𝑣 or 𝑒 ∈ 𝑯𝑣 , are adjacent to 𝑣. Now, we define 𝑽𝑍𝐼𝑆 ⊆ 𝑽𝑍𝐼 , such that,
following constraints must be added for 𝑧𝑖 (which corresponds
𝑽𝑍𝐼𝑆 contains only those known injection phases that are at
to the node 𝑥𝑖 ) [16].
the same voltage level as other nodes in their
𝑧𝑖 ≥ 𝑔𝑒𝑙 , ∀𝑒 ∈ 𝑳𝑣 neighborhood, and do not contain distributed loads
} (7)
𝑧𝑖 ≥ 𝑔𝑒ℎ , ∀𝑒 ∈ 𝑯𝑣 between themselves and any other phase in their
It is important to highlight here that the binary variable 𝑧𝑖 gives neighborhood. Let 𝑚 = |𝑽𝑍𝐼𝑆 |. An object set 𝑹 defined for
the node locations where the μPMUs must be installed, the modeling the observability constraints, is described below:
variables 𝑔𝑒𝑙 and 𝑔𝑒ℎ give the outgoing phase laterals that must For 𝑚 = 1: If 𝑖 is a single element of the set 𝑽𝑍𝐼𝑆 and 𝑵𝑖
be monitored from a given node, and the integer variable 𝑛𝑖 represents the neighborhood of phase 𝑖, for every pair of
gives the number of μPMUs contained at the node 𝑥𝑖 . elements 𝑝, 𝑞 ∈ 𝑵𝑖 , the object set 𝑹 contains the two-
The additional constraints (introduced in Section II) are element set {𝑝, 𝑞}.
modeled as follows: For 𝑚 ≥ 2: Let 𝑗 and 𝑘 be any two elements of set 𝑽𝑍𝐼𝑆 ,
1. Handling of voltage regulators/transformers and such that sets 𝑵𝑗 and 𝑵𝑘 denote the neighborhoods of 𝑗 and
distributed loads: It has been shown in Section II that the 𝑘, respectively. It is important to note that the sets 𝑵𝑗 and
presence of unknown voltage regulator/transformer tap 𝑵𝑘 may have elements in common. Let 𝑵𝑗,𝑘 represent the
ratios and distributed loads can influence network
common elements of the sets 𝑵𝑗 and 𝑵𝑘 ; i.e., 𝑵𝑗,𝑘 = 𝑵𝑗 ∩
observability. Therefore, such additional constraints must
𝑵𝑘 . Let 𝑵𝑗′ contain the elements that are present only in set
be accounted for during the μPMU placement formulation
itself. If the edge set 𝑬𝑉𝑅 ⊆ 𝑬′ includes the voltage 𝑵𝑗 , but not in set 𝑵𝑘 ; i.e., 𝑵𝑗′ = 𝑵𝑗 − 𝑵𝑗,𝑘. Similarly, let
regulators or transformers, then, all the edges 𝑒 ∈ 𝑬𝑉𝑅 must 𝑵′𝑘 contain the elements that are present only in set 𝑵𝑘 , but
be removed from the edge set 𝑬′ of the modified graph 𝒢 ′ not in the set 𝑵𝑗 ; i.e., 𝑵′𝑘 = 𝑵𝑘 − 𝑵𝑗,𝑘 . For each pair of
before (1)-(7) are implemented. Similarly, if distributed elements 𝑝 and 𝑞 in 𝑵𝑗′ or 𝑵′𝑘 or 𝑵𝑗,𝑘 , the two-element set
loads are present on the set of edges 𝑬𝐷𝐿 , where 𝑬𝐷𝐿 ⊆ 𝑬′ , {𝑝, 𝑞} is added to 𝑹𝑗,𝑘 . Next, the cross-product set 𝑸𝑗,𝑘 =
it implies that all edges 𝑒 ∈ 𝑬𝐷𝐿 must be removed from the 𝑵𝑗′ × 𝑵′𝑘 × 𝑵𝑗,𝑘 is calculated, which consists of all the
edge set 𝑬′ of the modified graph 𝒢 ′ before the triplets (𝑝, 𝑞, 𝑟) such that 𝑝 ∈ 𝑵𝑗′ , 𝑞 ∈ 𝑵′𝑘 and 𝑟 ∈ 𝑵𝑗,𝑘 .
optimization problem is solved.
Each triplet (𝑝, 𝑞, 𝑟) is added to the object set 𝑹𝑗,𝑘 . For each
2. Handling of different switch configurations: Here, the
pair of elements 𝑗 and 𝑘, the object set 𝑹𝑗,𝑘 is created for
objective is to ensure complete phase observability for all
feasible switch configurations. If there are 𝑠 switches, 2𝑠 every pair. The total collection of objects stored in the
switch configurations are possible. However, in the context object set 𝑹 is given below [16].
of this research, only those switch configurations are
𝑹= ⋃ 𝑹𝑗,𝑘 (11)
deemed feasible, which do not result in an islanded mode
of operation. For a given system, let there be 𝑓 such 1≤𝑗≤𝑘≤𝑚

feasible switch configurations, implying that there exist 𝑓 In (11), the ∪ operator eliminates duplicate entries. For
connected graph topologies, 𝒢1′ , 𝒢2′ , … , 𝒢𝑓′ . The constraint every set (𝑝, 𝑞) ∈ 𝑹 and (𝑝, 𝑞, 𝑟) ∈ 𝑹 the modified
equations (5)-(7) for a given topology can be grouped observability constraints are given below [16].
together and written in the form shown below.
∑ {𝑔𝑙𝑒 + 𝑔ℎ𝑒 } + ∑ {𝑔𝑙𝑒 + 𝑔ℎ𝑒 } ≥ 1
𝑨𝑿 ≥ 𝑩 (9) 𝑒∈ 𝑬′𝑝 𝑒∈𝑬′𝑞
where 𝑨 contains the coefficients of the different integer (12)
variables (𝑧𝑖 , 𝑔𝑒𝑙 , 𝑔𝑒ℎ , and 𝑛𝑖 ), 𝑿 contains all the integer ∑ {𝑔𝑙𝑒 + 𝑔ℎ𝑒 } + ∑ {𝑔𝑙𝑒 + 𝑔ℎ𝑒 } + ∑ {𝑔𝑙𝑒 + 𝑔ℎ𝑒 } ≥ 1
variables, and 𝑩 contains the constants of each constraint 𝑒∈𝑬′𝑝 𝑒∈𝑬′𝑞 𝑒∈𝑬′𝑟 }

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SILCHAR. Downloaded on November 30,2021 at 08:32:42 UTC from IEEE Xplore. Restrictions apply.
 4

IV. RESULTS AND DISCUSSION node 9, implying that the three-phase voltage of node 11
The ILP problem was solved in MATLAB using Gurobi as would be the same as that of node 9, because of zero
the optimizer. For the simulations done in this paper, we have voltage drop in the feeder 9-11. Consequently, there is no
assumed that (i) a subset of the phases that have relatively large need to monitor the phase laterals (9,1)-(11,1), (9,2)-(11,2)
spot loads are monitored by USMs, and (ii) a μPMU has three and (9,3)-(11,3), which were monitored when ZIPs were
measurement channels and can therefore measure three voltage not considered (see Table I).
and three current phasors [17]. The proposed μPMU placement • The presence of a USM at node 5 allows the phase voltages
algorithm not only finds the node locations where the μPMUs of node 6 to be observed by a μPMU placed at node 2.
must be installed, but also the phases that must be monitored. Similarly, the presence of a USM at node 10 allows the
This is explained in Table I using the IEEE 13-node distribution μPMU placed at node 9 to observe node 10.
feeder shown in Fig. 2. In this system, without considering • The A and C phases of node 12 do not have any net
ZIPs, seven μPMUs are placed at five node locations, namely, injection. One μPMU placed at node 9 indirectly monitors
2, 4, 7, 9, and 12 with the consideration that USMs are located the A and C phases of node 12. Therefore, we do not need
at nodes 5, 6, 8, and 10. a separate μPMU for observing the edge (12,3)-(1,3) as the
The μPMU placement results for four IEEE distribution current that flows from (12,3) towards (1,3), must be the
feeders are summarized in Table II. The second column same as the current that flows from (9,3) to (12,3) (refer
indicates whether ZIPs were considered in the analysis or not. Fig. 2). Since, edge (9,3)-(12,3) is monitored by a μPMU,
The third column indicates the total number of ZIPs in C phase voltage of node 1 can be effectively observed. The
comparison to the total number of phases present in the system. same reasoning applies for edges (9,1)-(12,1) and (12,1)-
In IEEE 13-node distribution feeder there exists 18 ZIPs and 32 (8,1), because the A phase of node 12; i.e., (12,1) is a ZIP.
phases. The fourth and the fifth columns give the number of Table I: μPMU installation for IEEE 13-node feeder (without ZIP)
node locations required for μPMU installation and the number Node Phases # USM
of μPMU devices, respectively. The last column of Table II Location Monitored** μPMU location
2 (2,1)-(3,1); (2,2)-(3,2); (2,3)-(3,3); 2
gives the locations of USMs present in the systems. It is (2,2)-(5,2); (2,3)-(5,3)
observed that for the IEEE 13, 34, 37, and 123 node systems, 4 (4,1)-(3,1); (4,2)-(3,2); (4,3)-(3,3) 1
on considering ZIPs, the number of μPMUs reduced from 7 to 7 (7,1)-(2,1);(7,2)-(2,2);(7,3)-(2,3) 1 5,6,8,10
6, 25 to 22, 19 to 15, and 54 to 43, respectively. 9 (9,1)-(11,1);(9,2)-(11,2);(9,3)-(11,3); 2
(9,1)-(13,1);(9,2)-(13,2);(9,3)-(13,3)
12 (12,1)-(8,1);(12,3)-(1,3) 1
** The first and second term inside the bracket denote the node number and
phase numbers (1 for Phase-A, 2 for Phase-B, 3 for Phase-C), respectively.
Table II: μPMU placement results for IEEE distribution feeders
Test ZIP+ #ZIP # # USM Locations
System (#Phases) Nodes μPMU
13- No 18 (32) 5 7 5,6,8,10
node Yes 4 6
34- No 68 (86) 22 25 22
node Yes 21 22
37- No 79 (111) 14 19 1,15,17,24,26,30,
node Yes 13 15 31,34
123- No 176 (273) 51 54 22,43,47,48,64,7
node Yes 41 43 7,80,90,106,107

Table III: Optimal node locations (not considering ZIP)

Test System Node Locations where μPMUs are placed
Fig. 2: Phase-connectivity between different nodes of the IEEE 13 node
13-node 2,4,7,9,12
distribution feeder (nodes renumbered in ascending order).
34-node 1,4,5,6,8,10,11,13,14,16,18,19,20,21,23,24,27,28,29,30,
Tables III and Table IV provide the μPMU locations for the 32,33
37-node 2,3,5,7,8,9,10,11,14,19,30,35,36,37
test systems, without considering ZIPs and with considering
1,3,6,8,13,14,15,18,19,23,27,28,29,32,36,39,40,44,46,49,
ZIPs, respectively. The way in which the consideration of ZIPs 123-node 51,52,54,55,58,60,63,65,67,70,74,76,78,82,84,87,91,95,
reduces the number of μPMUs is explained below, with regards 97,99,101,103,105,110,113,115,117,123,124,127,128
to the IEEE 13-node distribution feeder.
• The μPMUs that are positioned at nodes 2, 4, and 7 monitor Table IV: Optimal node locations (considering ZIP)
the same phases that were monitored when ZIPs were not Test System Node Locations where μPMUs are placed
13-node 2, 4,7, 9
considered (compare Table I with Table V). The presence
34-node 1,4,5,8,10,11,13,14,16,17,19,20,23,24,27,28,29,30,31,32,
of a transformer and a voltage regulator between nodes 2,7, 33
and 3,4, respectively, necessitates the installation of 37-node 2,3,5,6,7,8,10,11,14,27,35,36,37
μPMUs at each of the nodes 4 and 7. 5,8,14,15,18,20,21,25,27,31,35,38,42,46,49,52,54,58,60,
• The A, B, and C phases of node 11 do not have any load. 123-node 63,65,67,70,75,76,78,82,85,87,91,95,97,100,103,105,109,
113,116,118,122,124
Therefore, there is zero incoming current at node 11 from

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SILCHAR. Downloaded on November 30,2021 at 08:32:42 UTC from IEEE Xplore. Restrictions apply.
 5

Table V: μPMU installation for IEEE 13-node feeder (considering ZIP) unknown tap ratios of voltage regulators/transformers, different
Node Phases # USM
switch configurations, and ZIPs, simultaneously. Prior
Location Monitored μPMU Locations
2 (2,1)-(3,1); (2,2)-(3,2); (2,3)-(3,3) 2 placement schemes only identified the node locations for
(2,2)-(5,2); (2,3)-(5,3) μPMU installation. The proposed research not only finds the
4 (4,1)-(3,1);(4,2)-(3,2);(4,3)-(3,3) 1 5, 6, 8,10 optimal node locations, but also the minimum number of μPMU
7 (7,1)-(2,1); (7,2)-(2,2); (7,3)-(2,3) 1
9 (9,1)-(12,1); (9,3)-(12,3); (9,1)- 2
devices required at a node, and the least number of phases that
(13,1);(9,2)-(13,2);(9,3)-(13,3) must be monitored by them. The importance of considering “the
phases to be monitored” for distribution system state estimation
The unique advantage of simultaneously minimizing the will be described in a future publication. Multistage placement
number of affected nodes and the total number of μPMUs is
of μPMUs will also be investigated as a future topic of research.
discussed next. Consider the 5-node system depicted in Fig. 3.
The squares marked with letter “p” denote the position of the
REFERENCES
μPMU channels as obtained using the proposed methodology.
It is observed that two nodes (nodes 2 and 3) were disrupted for [1] A. Primadianto, and C. Lu, “A review on distribution system state
estimation,” IEEE Trans. Power Syst., vol. 32, no. 5, pp. 3875-3883, Sep.
μPMU installations, the number of monitored phase laterals 2017.
were eight, and the number of μPMUs required is three (two at [2] H. Wu, P. Dong, and M. Liu, “Distribution network reconfiguration for
node 3 and one at node 2). The results that one will obtain if the loss reduction and voltage stability with random fuzzy uncertainties of
two above-mentioned objectives are not minimized renewable energy generation and load,” accepted in IEEE Trans.
Industrial Informatics.
simultaneously are described below. [3] A. von Meier, E. Stewart, A. McEachern, M. Andersen, and L.
Case A-Minimizing only the number of node locations: If the Mehrmanesh, “Precision micro-synchrophasors for distribution systems:
number of μPMUs are not included inside the objective a summary of applications,” IEEE Trans. Smart Grid, vol. 8, no. 6, pp.
function, then the placement solution that will be obtained is 2926-2936, Nov. 2017.
[4] E. Stewart, “Micro-synchrophasors for distribution systems”, Lawrence
denoted by the squares marked with letter “a” in Fig. 3. This Berkeley National Laboratory. Available Online:
placement scheme also affects two nodes (nodes 2 and 3) and https://www.naspi.org/sites/default/files/2016-
monitors eight phase laterals. However, four μPMUs are 09/lbnl_stewart_micro_synchrophasors_for_distribution_20150324.pdf
required in this scenario: three at node 3 and one at node 2 [5] D. Stanescu, M. Sanduleac, and C. Stanescu, “Unbundled meters can
boost smart city project,” in CIRED - Open Access Proceedings Journal,
(since a μPMU has three measurement channels). vol. 2017, no. 1, pp. 2931-2934, Oct. 2017.
Case B-Minimizing only the number of μPMUs: If the number [6] H. Wang, W. Zhang, and Y. Liu, “A robust measurement placement
of affected node locations are not included inside the objective method for active distribution system state estimation considering
function, then the placement solution that will be obtained is network reconfiguration,” IEEE Trans. Smart Grid, vol. 9, no. 3, pp.
2108-2117, May 2018.
given by the squares marked with letter “b” in Fig. 3. This [7] Z. Wu et al., “Optimal PMU placement considering load loss and relaying
placement scheme uses three μPMUs to monitor eight phases; in distribution networks,” IEEE Access, vol. 6, pp. 33645-33653, 2018.
but it affects three nodes, as one μPMU is installed at each of [8] S. Teimourzadeh, F. Aminifar, and M. Shahidehpour, “Contingency-
the nodes 2, 3, and 4. Installing three μPMUs at two different constrained optimal placement of micro-PMUs and smart meters in
microgrids,” IEEE Trans. Smart Grid, vol. 10, no. 2, pp. 1889-1897, Mar.
node locations is better than installing three μPMUs at three 2019.
separate node locations. This is because the associated [9] M. Jamei et al., “Anomaly detection using optimally placed μPMU
infrastructure costs are proportional to the number of sites sensors in distribution grids,” IEEE Trans. Power Syst., vol. 33, no. 4, pp.
(nodes) that are disrupted for device placement [16]. 3611-3623, Jul. 2018.
[10] J. Liu, J. Tang, F. Ponci, A. Monti, C. Muscas, and P. A. Pegoraro,
“Trade-offs in PMU deployment for state estimation in active distribution
grids,” IEEE Trans. Smart Grid, vol. 3, no. 2, pp. 915-924, June 2012.
[11] S. Prasad, and D. M. Vinod Kumar, “Robust meter placement for active
distribution state estimation using a new multi-objective optimisation
model,” IET Science Meas. Technol., vol. 12, no. 8, pp. 1047-1057, Nov.
2018.
[12] E. Jamil, M. Rihan, and M. A. Anees, “Towards optimal placement of
phasor measurement units for smart distribution systems,” in Proc. 6th
IEEE Power India Int. Conf. (PIICON), Delhi, India, pp. 1-6, Dec. 2014.
[13] A. Tahabilder, P. K. Ghosh, S. Chatterjee, and N. Rahman, “Distribution
system monitoring by using micro-PMU in graph-theoretic way,” in Proc.
4th Int. Conf. Advances Elec. Eng. (ICAEE), Dhaka, Bangladesh, pp. 159-
163, Jan. 2017.
[14] X. Chen, T. Chen, K. J. Tseng, Y. Sun, and G. Amaratunga, “Customized
optimal μPMU Placement method for distribution networks,” in Proc.
Fig. 3: μPMU placement solutions subject to different objectives. IEEE Asia-Pacific Power Energy Eng. Conf. (APPEEC), Xi'an, China,
pp. 135-140, Oct. 2016.
[15] H. Nazaripouya, and S. Mehraeen, “Optimal PMU placement for fault
V. CONCLUSIONS AND FUTURE WORK observability in distributed power system by using simultaneous voltage
This paper proposed an optimal μPMU placement scheme and current measurements,” in Proc. IEEE Power Energy Soc. Gen.
Meeting, Vancouver, BC, pp. 1-6, Jul. 2013.
that is subject to the practical constraints of a distribution [16] A. Pal, A. K. S. Vullikanti, and S. S. Ravi, “A PMU placement scheme
system. The optimization problem was formulated as an ILP, considering realistic costs and modern trends in relaying,” IEEE Trans.
which guaranteed optimality of results. The proposed Power Syst., vol. 32, no. 1, pp. 552-561, Jan. 2017.
[17] “Micro PMU installation and User’s manual.” [Online]. Available:
optimization methodology can handle the presence of single- https://www.powerstandards.com
and-two-phase laterals, distributed loads, modern smart meters,

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SILCHAR. Downloaded on November 30,2021 at 08:32:42 UTC from IEEE Xplore. Restrictions apply.