Received April 8, 2022, accepted April 26, 2022, date of publication April 29, 2022, date of current version

May 5, 2022.
Digital Object Identifier 10.1109/ACCESS.2022.3171234

Service Restoration Through Microgrid Formation

in Distribution Networks: A Review
MOSAYEB AFSHARI IGDER1 , (Student Member, IEEE),
XIAODONG LIANG 1 , (Senior Member, IEEE), AND
MASSIMO MITOLO 2 , (Fellow, IEEE)
1 Department of Electrical and Computer Engineering, University of Saskatchewan, Saskatoon, SK S7N 5A9, Canada
2 School of Integrated Design, Engineering and Automation, Irvine Valley College, Irvine, CA 92618, USA
Corresponding author: Xiaodong Liang ([email protected])
This work was supported in part by the Natural Science and Engineering Research Council of Canada (NSERC) Discovery
Grant RGPIN-2016-04170.

ABSTRACT Microgrid formation is a promising solution to enhance resiliency of distribution networks. The
self-adequacy feature of a microgrid enables continuity of power supply through distributed generation (DG)
units during severe faults and natural disasters. In this paper, different methods commonly used to partition
a distribution network into multiple microgrids are presented, including the graph theory, heuristic rule-
based algorithm, cluster-based technique, and mixed integer programming. Advantages and disadvantages
of these techniques and future research directions are presented. This review provides an excellent summary
on service restoration through micrgrid formation, and offers a valuable reference for researchers working
on grid modernization of distribution networks.

INDEX TERMS Distribution networks, microgrid formation, resiliency, reliability, service restoration.

I. INTRODUCTION TABLE 1. Major blackouts across the globe.

The resiliency improvement of power systems against
extreme events is an essential aspect of the system design
and operation [1], [2]. Extreme events can be either natural
disasters or cyber-attacks, which not only affect the continuity
of electrical service for a considerable number of consumers,
but may also cause significant financial losses. For example,
more than 50,000 electricity customers were knocked out of
service due to weather disasters in the United States [1], and
over $1 billion financial losses were caused by eight weather Resiliency is defined as the power grid’s capability to with-
disasters (i.e., flooding, storms, and hurricanes) during the stand and recover quickly from severe incidents, react prop-
first half of 2016. In August 2017, the hurricane Harvey erly to changing conditions, and prevent future events [2].
caused a total of $180 billion losses, and many towns were left Fig. 1 illustrates the performance of a resilient system and
without power for several weeks [3]. Other extreme weather a conventional system under extreme events [4].
events are shown in Table 1, which have caused catastrophic The power supply of the system is P0 at time t1 when a
damages to power systems, resulting in massive power severe incidence happens. At time t2 , the power supply of the
outages [1], [4]–[6]. These blackouts within bulk power system quickly decreases to its minimum amount (Pmin−R for
networks across the globe indicate vulnerability of a resilient system and Pmin−C for a conventional system). The
power systems, and their resiliency improvement is a fun- restoration is started for resilient and conventional systems
damental task for power system operators. at time t3 and t4 , respectively. The normal power supply P0
is resumed for resilient and conventional systems at time t5
and t6 , respectively. Therefore, a resilient system equipped
The associate editor coordinating the review of this manuscript and with resiliency-boosted strategies shows better performance
approving it for publication was Gab-Su Seo . in terms of load restoration.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
46618 For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/ VOLUME 10, 2022
M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

FIGURE 1. A resilience performance curve [4].

The system’s resiliency and resilience-based models have

been recently investigated in [5]–[7]. In [8], a theoretical
tutorial system is proposed to train distribution system opera-
tors to effectively respond to emergencies. The study in [9]
proposes a cooperative agents-based system for service
restoration through artificial intelligence methods. In [10],
a framework using a fuzzy logic is developed to manage out-
ages. In [11], weather data are used to examine the probability
of blackouts. The duration and frequency of occurrences have
also been projected by selecting appropriate disaster response
approaches. FIGURE 2. (a) Resiliency improvement strategies in distribution networks,
(b) Islanding in power systems.
Based on our literature review, techniques and strategies
to improve the resiliency of distribution networks from both
planning and operation point of view are provided in Fig. 2(a).
An effective planning must be conducted prior to undesir- realized through sensors, communication networks, and
able incidents to prepare for and lessen the impact of upcom- remotely controlled switches can help distribution network
ing disasters. The disaster-specific planning may include operators to detect and separate faulty areas and recover
hardening schemes [12], resource allocation [13], [14], pre- unsupplied load by opening or closing remotely controlled
diction [15], repair crews [16], and switch design [17], [18]. switches after severe events [21].
Hardening of distribution networks refers to making the Optimal implementation of switches in distribution net-
infrastructure sturdier and consequently more durable to fail- works aim to enhance service restoration process by design-
ure, so that serious damages due to natural disasters can be ing an efficient sequence of switching operations [17]. The
minimized, and the restoration time can be reduced accord- study of optimal placement of sectionalizer in radial distribu-
ingly [12]. The system recovery capability can be improved tion networks is conducted in [18], and an algorithm based on
by availability of spare and reliable resources, and their pre- tree structure and dynamic programming is proposed to find
event allocation [14]. Prediction models are used to forecast sectionalizing switch locations while minimizing the cost of
power outages, possible damages and restoration time; util- outages and improving reliability of the system. Ref [17]
ities can use these models to plan corrective actions prior studies upgrading manual switches to remotely controlled
to incidences [15]. Repair crews play a prominent role in switches to improve service restoration in distribution net-
recovering power systems after extreme events, proper man- works, where a greedy rule-based algorithm is used to maxi-
agement and optimal number of repair crews can improve the mize load restoration and minimize the investment cost.
system resiliency [16]. However, during natural disasters, the distribution network
After a power outage occurs, the imperative mission for may lose its connection with the main grid, and is not able
system operators is to restore distribution networks as fast to supply loads. In this case, traditional restoration tech-
as possible to support critical loads and minimize financial niques may not work properly. To address this issue, micro-
losses to customers. Load restoration can be generally divided grid formation/islanding can be a promising solution because
into conventional techniques [19], [20], automation [21], loads can be fed through local distributed generators within
[22], and microgrid formation [23]–[25]. microgrids [23].
In conventional technique, the load from the off-outage Fig. 2(b) shows different types of islanding in power
area is transferred to the adjacent feeder through tie-lines systems, which can be categorized into two groups [26]:
and tie-switches [19]. Automation in distribution networks planned islanding and unplanned islanding. Planned

VOLUME 10, 2022 46619

 M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

islanding, also known as intentional islanding, is initiated of industrials and militaries [38], reduce power losses, and
by power system operators or supervisory control and data improve voltage profile, power quality, and reliability [39].
acquisition (SCADA) systems; while unplanned islanding In this paper, a comprehensive literature review is con-
usually occurs due to faults in the system [27]. Sudden ducted on service restoration through microgrid formation
unplanned islanding should be detected quickly, it may trig- techniques in distribution networks. The main contribution
ger all control operations to maintain power generation and of the paper includes:
delivery despite islanding separation [26]. 1) Various microgrid formation methods and their advan-
Both planned and unplanned islanding operations could be tages and disadvantages are discussed.
used in service restoration for distribution networks during 2) Control and economic prospects of microgrids are
severe events to supply load in their original or extended summarized.
boundary. Based on their boundary [28]: microgrids can 3) Future research directions are recommended.
be divided into pre-determined microgrids and dynamic
The paper is arranged as follows: microgrid con-
microgrids.
struction models and microgrid control are provided in
A pre-determined microgrid has a fixed boundary, which is
Sections II and III. Section IV introduces microgrid eco-
determined based on the supply adequacy, maximum distri-
nomic prospect. Section V discusses advantages and dis-
bution coverage, and reliability indices [29]–[32]. In [29], a
advantages of microgrid formation algorithms and future
systematic approach is proposed to sectionalize a distribution
research directions. The conclusion is drawn in Section VI.
network into several virtual microgrids with optimized self-
adequacy. An optimum design of microgrids in distribution
II. MICROGRID CONSTRUCTION MODELS
networks based on reliability index, and active power and
The most challenging aspect of the distribution network’s
reactive power balance for the supply-security purpose is
partitioning is to form an optimal microgrid while maintain
proposed in [30], [31]. In [32], the maximum coverage cri-
operational constraints, such as power balance and voltage
terion, and optimized communication and control infrastruc-
limits at each node [44], [45]. Existing microgrid formation
ture are used to partition a distribution network into several
strategies can be broadly categorized into four techniques
microgrids.
as shown in Fig. 3 [12]: heuristic rule-based strategy [46],
A dynamic microgrid has boundaries that can be expanded
mixed integer linear programming (MILP) [14], [47], graph
or shrunk, while still maintains a balance between power
theory [48], and cluster-based models [49]. The cluster-based
generation and load demand. To avoid imbalance between
models can be categorized into 1) spectral clustering,
local distributed generation (DG) units and loads, or to maxi-
2) hierarchical algorithms, and 3) K-Means approach.
mize the load pick-up during extreme events are main reasons
that microgrids have dynamic boundaries. In [33], microgrid
formation with flexible boundaries is proposed to improve
reliability and resiliency of distribution networks.
To solve optimization problems associated with micro-
grid formation, the genetic algorithm and mixed integer
linear programming can be used. In [34], smart switches
are used as automatic sectionalizers to determine flexible
boundaries of microgrids during natural disasters. In [35],
adaptive self-adequate microgrids using dynamic boundaries
is proposed, where clusters of nodes based on self-adequacy
measures are built first, each cluster is then assigned with an
agent with the capability of supervisory control of all power
generation sources within the cluster, communication with
customers’ smart meters within the cluster, and communi-
cation with neighbors’ agents. Afterward, desirable adaptive
microgrids can be formed by merging a group of clusters.
FIGURE 3. The summary of microgrid formation techniques.
Ref [2] proposes the formation of adaptive microgrids using
graph theory and load switching sequence. Microgrid forma-
tion based on time and location of faults using mixed integer A. GRAPH THEORY-BASED TECHNIQUE
programming is presented in [36]. Graph theory employs mathematical formulations to spec-
Planned microgrid formation/islanding can be used to sup- ify pair-wise relations between objects. Each graph is com-
ply load for expected conditions, such as maintenance in posed of vertices and edges, which are also known as nodes
upstream grids [27]. Microgrids in island mode can provide and links, respectively. In power distribution networks, the
electricity to remote communities, where the expansion of graph-based concepts, such as graph partitioning, spanning
power systems is not economical [37]. Islanded microgrids tree, and spanning forest, are used to form microgrids in
can also improve energy security for critical load demands two different topologies: 1) loop-based microgrid topology,

46620 VOLUME 10, 2022

M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

and 2) radial-based microgrid topology. In [2], both radial

and loop-based microgrid topologies are considered as part of
the load restoration strategy. The linearized DisFlow model is
employed to consider power flow and voltage characteristics
in each constructed microgrid. In [50], spanning tree and
spanning forest concepts are applied to form post-disturbance
radial-based microgrids energized by DGs. The LinDistFlow
model is also used to satisfy operational constraints. In [51],
the graph partitioning technique and linear integer program-
ming are proposed to form an optimal loop-based microgrid
to improve system reliability. In the following subsections,
radial- and loop-based models are discussed.

1) LOOP-BASED MODEL
The graph partitioning concept, which is employed to
determine potential loops based on existing DG units, sec- FIGURE 4. The coarsening first iteration in the graph partitioning
tionalizes a graph G with the vertex set V and the edge technique [52].
set E into the Q subset (V1,.,.,. Vi,... , VQ ), so that Vi ⊂
V , Vi ∩ Vj = ∅ for i 6 = j. In distribution systems,
energized buses and distribution lines are defined as the
vertex set and the edge set of the graph, respectively. Objec-
tive functions may include the maximization of load pick-
up, minimization of switching operations, generation-load
balancing, minimization of neighboring loops interactions,
and combinations of the above. The graph partitioning is
composed of three stages, 1) coarsening, 2) partitioning, and
3) uncoarsening [52].

a: STAGE1: COARSENING
The coarsening stage iteratively simplifies the distribution
network graph until it can no longer be partitioned. The
Shortest Edge Machine (SEM) is widely employed in this
process. An initial node is firstly selected randomly and
matched with the nearest adjacent node. Afterward, the nodes
are combined into a single node and the process is repeated FIGURE 5. The coarsening second iteration in the graph partitioning
technique [52].
until all possible matches in the graph have been achieved.
This process will end when the number of nodes reaches an
established percentage of the original number. The graphs
shown in Figs. 4 and 5 demonstrate the coarsening pro- approaches, such as Kernighan-Lin (KL) and Greedy Graph
cess using the IEEE 37-bus distribution network. Fig. 4(a) Growing Partitioning (GGGP) algorithms, are used in [52].
indicates the coarsening first iteration, where each red line With the GGGP algorithms, the solution is found by selecting
indicates that the two nodes should be merged according the initial vertex and expanding it to comprise a large part of
to the SEM strategy. The outcome of the first iteration is the graph. Since the main goal of service restoration is to form
shown in Fig. 4(b), in which the number of nodes is decreased microgrids with self-healing capabilities, the potential cluster
from 37 to 23. The second iteration of coarsening is illustrated is expanded around controllable DGs. Power mismatches
in Fig. 5(a), and its outcome is shown in Fig. 5(b). The and the distance between nodes are employed to ensure that
coarsening process will end in the second iteration because growing loops satisfy power balancing and the nearest nodes
the number of nodes is decreased to 16, which meets the are used [52]. The partitioning stops when all nodes are
termination criterion (obtaining less than 50% of the original covered by the loops. This process is shown in Fig. 6, which
number of nodes of the graph) [52]. includes three iterations, and the ultimate outcome is shown
in Fig. 6(c). Since four controllable DG sources are present,
four potential microgrids are formed, and four points evolve
b: STAGE2: PARTITIONING autonomously from those sources in Fig. 6(a). The process
In partitioning stage, the graph obtained in the last iteration of terminates once all nodes are covered by one of the loops
coarsening process is partitioned into Q parts. The heuristic in Fig. 6(c).

VOLUME 10, 2022 46621

 M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

FIGURE 7. Uncoarsening stage in the graph partitioning technique [52].

FIGURE 6. The partitioning stage in the graph partitioning technique [52].

c: STAGE 3: UNCOARSENING
In uncoarsening stage, the partitions obtained in previous
stage should be reversed into the original graph based on the
sequence of coarsening process. Refinement strategies, such
as KL algorithm, should be employed. In this algorithm, the
edge weight serves as a criterion to transfer a vertex between
neighboring loops. However, it is more logical to consider
FIGURE 8. The distribution network graph in a normal mode [53].
power balancing criterion in microgrid formation to trans-
fer nodes between adjacent loops. This modification should
be implemented for refinement when used in microgrid
controllable switches at the edges of this graph, and the source
formation [52].
nodes have feeders connected to the main grid, or to DG units.
Fig. 7 illustrates the uncoarsening process, which also
The fundamental loops of a graph defined by vectors, whose
contains two iterations because the graph coarsening is also
values are edges of the constructed loops, should be specified.
performed in two iterations. The first and second iterations
In Fig. 9, there are four loops, C1 to C4, defined by the
of uncoarsening are demonstrated in Figs. 7(a) and 7(c),
following vectors:
respectively; the refinement is shown in Figs. 7(b) and 7(d)
based on the power mismatch amount in per unit. In the first V1 = {10, 11, 12}
iteration of the refinement, Node 20 is transferred from the V2 = {1, 2, 3, 4, 5, 6, 8, 9, 10, 11}
lower left loop to the upper left loop to balance the power.
V3 = {1, 2, 3, 13, 14, 16, 17, 18}
V4 = {15, 16, 17, 18}.
B. RADIAL-BASED MODEL
The radial topology in distribution networks can be defined These fundamental loops can be categorized as real loops
as a graph, where all nodes are put together into one energy (C2 and C3) and virtual loops (C1 and C4). The virtual loops
source node without any loops. Two graph-based concepts, are those with DG units and the frequency control capability.
spanning tree and spanning forest, are used to model the radi- The frequency control is one fundamental requirement for
ality constraint in the microgrid formation problem. A span- power systems operation. In island mode, microgrids must
ning tree is a graph connecting all nodes with links without be able to realize voltage and frequency control through their
forming any loops. A spanning forest is a graph, whose controllers. DGs are usually connected to power systems
connected constituents are spanning trees. with interfacing power electronics converters, which enables
Distribution networks may be modelled with graphs con- advanced controllers to be designed to realize frequency con-
sisting of vertexes and edges as shown in Fig. 8 [53]. There are trol [54]. Frequency control has been widely studied for DGs

46622 VOLUME 10, 2022

M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

FIGURE 9. The distribution network graph in a self-healing mode [53].

in microgrids [55], [56]. In [55], a dual-stage fractional order

proportional–integral–derivative (PID) controller is used to
improve the frequency control of microgrids when operating
in island mode, and the imperialist competitive algorithm
is implemented to optimize the PID gains. The Fuzzy tilt
integral derivative using a filter and the double integral con-
trol are employed for frequency control of DGs in [56], and
coefficients of the controller are optimized through the Whale
optimization algorithm.
Following the designation of fundamental loops, spanning
tree and spanning forest algorithms are employed to deter-
mine microgrid formation, taking into account the radiality FIGURE 10. The flowchart of microgrid construction using the heuristic
constraint and the load supply. The spanning tree in a graph method [46].
is not used with all nodes connected, thus some edges should
be eliminated. A spanning forest, which is a graph with in the first step. In the second step, the non-isolated nodes
several trees, is used to model a distribution network [50]. are clustered into microgrids using k-means method and DG
To create a spanning forest and ensure the radiality of the nodes are considered as centroids. The network configura-
network, it is sufficient to open only one switch in each loop tion for the constructed microgrids must meet the total load
that is not shared with any adjacent loop. In addition, if the demand. In the third step, the dispatch assessment is imple-
switch is selected from virtual loops, a microgrid energized mented, where the capacity of the constructed microgrids
by DGs is formed. For instance, if switches 10 and 17 are is evaluated based on power system operation constraints.
opened, two microgrids (MG1 and MG2) are formed (Fig. 9). Fig. 10 illustrates the heuristic method process in microgrid
Accordingly, by considering the switch status as a decision formation.
variable in the optimization problem, optimal microgrids are In [57], [58], a decentralized multiagent system (MAS)
formed in load restoration process. strategy and the heuristic rule-based algorithm are employed
to form microgrids using load priority and switching oper-
C. HEURISTIC RULE-BASED ALGORITHM ations as objective functions. The following ten steps are
Heuristic rule-based algorithms employ heuristics or rules to proposed, and the controlled DG unit is used as the power
find solutions. The goal of this method is to solve the problem source for critical load restoration [58]:
within an acceptable time frame. The solution may not be • Step 1: entire nodes = a set of all nodes, which is
the best, but it is near the optimal one. As this algorithm is demanded by the DG agent for restoration.
computationally efficient, it can be used in the optimization • Step 2: node to restore = a set of the nodes to be
problem with many decision variables. The microgrid for- restored is selected by calculating the objective function
mation problem based on the heuristic rule-based algorithm (i.e., load priority order) and following the branch
utilizes rules to achieve solutions. current limits, voltage limits, and consumed power
In [46], a post-disturbance microgrid construction solu- constraints.
tion is proposed for medium to large distribution networks • Step 3: min priority load = the least priority node in the
using the heuristic algorithm in three steps. In the first step, node to restore set.
DG units are placed optimally without considering micro- • Step 4: lower priority nodes = a set of nodes in the entire
grid formation constraints. Load dispatch, nodal balance, nodes with a priority less than min priority node.
line flow, generation placement, and voltage constraints are • Step 5: new loads to restore are initialized as a set of
considered in the optimization problem with the objective of the nodes from nodes to restore by removing the min
maximizing the load pick-up. Locations of DGs are provided priority node.

VOLUME 10, 2022 46623

 M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

• Step 6: the new load to restore is chosen from lower systems [59]. Since this method clusters buses by using the
priority nodes, based on the load priority objective and affinity matrix, buses with a greater affinity become a cluster.
operational constraints. It is required that the affinity matrix must be obtained with
• Step 7: max priority node = the highest priority node high accuracy. The affinity between any two nodes is deter-
chosen from the new nodes to restore set. mined by the Hessian matrix related to AC optimal power
• Step 8: priority factor = the priority order of min priority flow. The Hessian matrix is the second derivative of the
node divided by the priority order of max priority node. Lagrange function, and a larger amount of entry in this matrix
• Step 9: Compute the number of switching operations for indicates a stronger coupling.
nodes to restore and new nodes to restore as X and Y , To calculate the Hessian matrix, AC optimal power flow
respectively. must be performed. After computing the Hessian matrix, the
• Step 10: If Y smaller than X multiplied by priority spectral clustering method is used to group buses with greater
factor, then nodes to restore is new nodes to restore and affinity together. The process of partitioning a distribution
go to Step 4. Else, lower priority nodes = lower priority network with B buses into N clusters using spectral clustering
nodes – max priority nodes. If lower priority nodes is an is given by [59] as follows:
empty set, the algorithm ends, else go to Step 6. • Determine the components of the affinity matrix based
m P n
Fig. 11 shows the islanding of the 119-bus test system using on Ai,j = (1 − w)
P
|Hk.l | + w ∗ Yi,j if i 6= j and
the above-mentioned heuristic rule-based algorithm. In out of k=1 l=1
service areas, each dispatchable DG unit with the assistance set Aii = 0.
B
of other types of renewable-based DGs build an individual • Form the diagonal matrix D based on Di,i =
P
Ai,n and
microgrid to restore critical loads with an optimum number n=1
of switching operations. build the matrix P = D−1/2 AD−1/2 .
• Specify the N largest eigenvalues associated with the
matrix P and construct the matrix V by stacking the
eigenvectors in columns. To have unit length, normalize
the V ’s rows.
• Consider each row of V as a data point and group these
data points into N partitions using an algorithm, such as
k-means or hierarchical.
• Give bus i to cluster A if row i of V was given to
cluster A.
Yi,j is the component of the admittance matrix, and w is the
affinity weight.
In [62], an adaptive spectral splitting technique is pro-
posed. The primary splitting of the distribution network
is achieved through the spectral clustering strategy first.
When the primary partitioning is obtained, the boundary
nodes/buses transfer from the present location to the neigh-
boring partition in every iteration to acquire the correspond-
ing power balance ratio of each partition, which is defined
as a ratio of the total power demand to the total generation
FIGURE 11. Microgrid formation of the 119-bust test system [57]. capacity, while maintaining power balance constraints.

2) HIERARCHICAL ALGORITHM
D. CLUSTERING ALGORITHM In hierarchical clustering, nodes are grouped into hierar-
The clustering analysis splits a set of objects into uni- chical structures based on power system specifications of
form groups based on similarity measures, so the simi- the line including average power flow or line impedance
larity of objects in one constructed group is greater than (admittance), and the obtained results are known as tree or
that in another group. Three clustering algorithms includ- dendrogram. A dendrogram’s root point corresponds to the
ing spectral clustering [59], hierarchical algorithm [60], and entire set of nodes, whereas each leaf represents a separate
k-means method [61] are widely used in distribution network node. To what degree the nodes are similar to each other is
partitioning. shown by intermediate points. The distance between clusters
or objects can be determined by the dendrogram height.
1) SPECTRAL CLUSTERING Dendrograms can be cut at various levels to gain the final
Spectral clustering is a type of graph partitioning that uses the clustering results [63]. Fig. 12 depicts the dendrogram related
affinity between two components within the dataset, which to IEEE 39-bus test system in which at the height level one,
is the computational coupling between two nodes in power the blue, green, yellow and red clusters are formed; cutting

46624 VOLUME 10, 2022

M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

FIGURE 12. Hierarchical algorithm, the dendrogram of IEEE 39-bus test

system [49].

dendrogram at the height level three will result in combining

the green and red clusters and forming three islands.

3) K-MEANS METHOD
The K-means algorithm falls into centroid or distance-based
algorithms, and the distances are computed to assign an
object to a cluster with its own centroid point [61]. The
purpose of K-means method is to partition the network with
n nodes into k clusters and to ensure that the distances are
minimal within each cluster. This strategy begins by select-
FIGURE 13. The K-means algorithm [64].
ing k nodes randomly as initial centroid points within the
networks. The remaining nodes are assigned to the closest
among them. After that, the centroids are repositioned from • Splitting constraints.
each cluster to ensure that there is a minimum distance • Power system physical constraints.
between the centroid and any other node within the cluster. • Subgraph connectivity constraints.
Afterward, the distance between each node and k points is cal- Mathematically, MILP optimization can be handled
culated, and the node is assigned to a cluster with the nearest by branch-and-bound, branch-and-cut, or cutting plane
centroid. approaches. Currently, several commercial optimizers, such
In this method, k centroid points are moved in succession as CPLEX, GUROBI, and MOSEK, are available to provide
until they reach a minimal threshold, and a stable cluster flexible, parallel-processing and high-performance solvers
is obtained. In the distribution network’s partitioning, nodes for MILP.
with controllable DGs are regarded as centroid points, and the According to our literature review, a summary on micro-
electrical distance is utilized for distance metric. The process grid formation for service restoration in distribution networks
of the K-means algorithm is illustrated in Fig. 13 [64]. can be found in Table 2. Different studies are compared from
application, information discovery, construction approach,
E. MIXED INTEGER LINEAR PROGRAMMING objective function, and optimization problem aspects.
Mixed integer linear programming (MILP) is a mathematical
optimization problem with integer decision variables, linear III. MICROGRID CONTROL
objective functions and constraints. This approach is broadly Renewable energy-based DG units in microgrids use inter-
used in load restoration and microgrid formation [23], [33], facing power electronics converters to connect to the sys-
[65]. Ref. [21] develops a MILP to form a microgrid ener- tem, and controllers are designed and implemented on these
gized by DGs through controlling the status of remotely power electronics converters to achieve power, voltage and
controlled switches after natural disasters, where critical load frequency control of DGs and the microgid. In grid-connected
pick-up is maximized with self-adequacy and operational mode, the voltage and frequency of microgrids are governed
constraints satisfied. by the utility grid, controllers are used for real and reactive
In [65], a MILP-based method is proposed to form multiple power or power factor control. In island mode, the microgid
microgrids to restore prioritized load in distribution networks must be able to control its voltage and frequency through
after extreme events. Flexible microgrid formation is investi- advanced DG controllers. Hierarchical control techniques are
gated in [33], where MILP is used to solve the optimization extensively utilized for microgrid control and power manage-
problem based on utility profits and customer satisfactions. ment [70], [71].
Implementation of MILP-based microgrids requires a large
number of control variables. To address this problem, a MILP A. TRADITIONAL HIERARCHICAL CONTROL FOR
with radiality constraints is proposed in [66]. In the MILP- MICROGRIDS
based method, the following constraints must be satisfied The traditional hierarchical control framework is illustrated
when forming a microgrid: in Fig. 14, including primary, secondary, and tertiary control.

VOLUME 10, 2022 46625

 M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

TABLE 2. Summary of literature review on microgrid formation for service

restoration in distribution networks.

FIGURE 14. Traditional hierarchical control for microgrids [71].

The deviation of output voltage and frequency caused by

primary control can be eliminated by secondary control [72].
The frequency and voltage restoration controllers are demon-
strated in Fig. 14 [71].
Z
δω = kpw ωMG − ωMG + kiw
∗
ωMG
∗
− ωMG dt + 1ωs

Z (3)
δE = kpE EMG
∗ ∗



− EMG + kiE EMG − EMG dt (4)

The frequency and voltage amplitude values, ωMG and EMG ,

in a microgrid are identified and evaluated by references,
ωMG
∗ and EMG ∗ , the obtained errors through compensators,

δω and δE, are send to each DG unit to adjust frequency and

voltage. The secondary control parameters are represented
by kpw , kiw , kpE , and kiE . 1ωs indicates the synchronization
term.
Centralized control [73], [74] and decentralized con-
trol [65], [76] are two major techniques used in secondary
control. For centralized control, the main drawback is its
high reliance on the microgrid control center and the bidi-
rectional communication structure. When the microgrid con-
trol center is faulted or the communication system fails,
The widely used primary control is droop control to adjust
the centralized control no longer works well. Accordingly,
the local voltage and power, prevent system instability, and
the stability of the microgrid is decreased and its cost is
handle proper power sharing among DGs [70], [71]. The
increased [77].
following well-recognized droop control technique is used to
Decentralized control, on the other hand, can overcome
reach primary control objectives [71]:
above issues, errors caused by one DG will not cause a
ωMG = ω∗ − m.(P − P∗ ) (1) whole system’s failure, and it does not rely on communication
EMG = E ∗ − n.(Q − Q∗ ) (2) networks, and can be simply expanded to several DGs, which
improves the system’s scalability [78].
where, the frequency and output voltage amplitude are det- Tertiary control is employed to identify power flow to
onated by ωMG and EMG , respectively. ω∗ and E ∗ represent achieve optimal operation for economic and service restora-
the reference frequency and voltage amplitude, respectively. tion purposes [79]. In Fig. 14, by calculating P/Q via the
Droop coefficients are represented by m and n. static bypass switch, PG and QG can be evaluated by the

46626 VOLUME 10, 2022

M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

preferred references, P∗G and Q∗G .

Z
ωMG
∗
kpP P∗G P∗G − PG dt



= − PG + KiP (5)
Z
∗
= kpQ Q∗G − QG + KiQ Q∗G − QG dt



EMG (6)

The tertiary control parameters are represented by kpP , KiP ,

kpQ , and KiQ .
Power quality adjustment and economic operation can be
achieved by the hierarchical control strategy, which enhances
the flexibility of microgrids. The microgrid control center
is utilized in this strategy to manage DGs and loads, which
contributes to reliable operation of multiple microgrids [80].

B. MULTIAGENT SYSTEM-BASED DISTRIBUTED CONTROL

Multiagent control strategy splits a large system into a num-
ber of autonomous subsystems, which can communicate with
each other [71]. Each agent has intelligent features [81], [82].
Using these intelligent agents, the multiagent system-based
FIGURE 15. The structure of multiagent-based microgrid [71].
distributed control can achieve coordinated operation of the
entire system. Fig. 15 demonstrates the structure of a multi-
agent control-based microgrid [71]. Various electrical com-
ponents, such as wind turbine generation units, loads, gas
turbines, and energy storage systems, are assigned to each
agent. These agents observe control operations and the sta-
tus of each electrical component, and the microgrid control
center coordinates activities among all agents. Once there is a
command from an agent, the microgrid control center notifies
and coordinates all agents [83]. Communication and coordi-
nation are crucial during the entire decision-making process.
Ref [84] uses the contact net protocol in such process.
Each agent generally contains two-level control blocks: FIGURE 16. Peak shaving and load leveling [95].
the upper-level control block identifies the power supply
reference and demand, and measures optimal increment cost; reduction [94], power quality improvement [95], and
the lower-level control block applies the power reference resiliency enhancement [96].
tracking of related electrical components [85]. Accordingly, DGs and energy storage along with advanced control tech-
each agent controls local load and power generation, and nologies enables flexible power management within a micro-
exchanges information with other agents. grid. It can be especially economically influential when the
There are many reported multiagent system-based dis- utility’s Time-of-Use tariff comprises power and volumet-
tributed control schemes in the literature [86]–[91]. A multi- ric charges [95]. This type of tariff is usually a powerful
agent distributed control with a frequency control framework motivation to facilitate peak shaving and load leveling as
is proposed in [86] by employing consensus method. Power shown in Fig. 16. In this example, an optimized load profile
sharing among distributed energy resources in microgrids is achieved by implementing load shifting and peak shaving,
through a multiagent-based technique is suggested in [87]. which minimizes power demand and volumetric charges.
Ref [90] can overcome the weakness of droop control by Exporting electricity to power systems is one main source
using multiagent distributed control, and realize voltage of income brought by microgrid formation [93]. Net metering
and frequency control and proportional reactive power shar- and feed-in tariffs are two general methods used to specify
ing among DGs. Ref [91] can regulate frequency, where surplus power generated by microgrids [95]. Net metering
each local controller shares information with neighboring utilizes a bidirectional meter to calculate a customer’s net
controllers. power consumption [95]. If the generation is more than the
consumption, the meter turns backward. In the feed-in tariff
IV. MICROGRID ECONOMICS method, all power producers receive a payment when they
Economic benefits from microgrid formation is an essen- inject power into the system [97].
tial feature to address. Microgrids have several economic As microgrids are located locally, power generation
benefits, such as load leveling and peak shaving [92], can be consumed locally, which avoids long distance
power export and net metering [93], loss and emission transmitting electricity, and thus, power losses along the

VOLUME 10, 2022 46627

 M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

TABLE 3. Cost details of various power generation technologies [100].

feeders are reduced [98]. Power quality and the system’s management of the electricity market, and aims to satisfy
reliability can be also improved because the decentralized local demand during the system operation through optimal
power supply can better match power supply and demand allocation of local energy sources [106]. Microgrids with
locally, and the influence due to transmission and generation renewable energy sources and energy storage can also partic-
outages can be reduced [99]. Due to increasing penetration of ipate in the emission trading market, where the energy price
renewable energy-based DGs, microgrids can reduce green- and emission data are sent to the microgrid control center, and
house gas (GHG) emission compared to conventional power the microgrid will be paid based on these data.
generation.
Table 3 provides a summary of cost details for various V. ADVANTAGES AND DISADVANTAGES OF MICROGRID
power generation technologies, including their technology FORMATION ALGORITHMS, AND FUTURE RESEARCH
cost, operation and maintenance (O & M) cost, capacity DIRECTIONS
factor, fuel cost, efficiency, GHG emission, renewable poten- A. ADVANTAGES AND DISADVANTAGES OF MICROGRID
tial, and the life time [100]. DGs in microgrids are mainly FORMATION ALGORITHMS
renewable energy sources, which makes microgrid formation A fast and effective service restoration strategy is vital to
economically viable. improve resiliency of distribution networks. Due to high
In [101], the economic analysis is conducted for penetration of DGs, microgrid formation can be an effective
a microgrid with PV and battery storage in Northampton, strategy to enhance the system’s resiliency via critical load
Massachusetts, USA by considering the outage mitigation, restoration during contingency. To form optimal microgrids,
emission reduction and resiliency improvement. The system different algorithms have been reported in the literature. Each
is modeled using the battery storage evaluation tool for a approach has its advantage and disadvantage, which will be
one-year period, and its efficiency is demonstrated through discussed below:
historical data and randomly generated large outages. • The mixed integer linear programming-based tech-
A lifecycle analysis is conducted for a microgrid with niques can provide a complete picture of microgrid for-
wind turbines, PV, diesel generators, and energy storage to mation by modeling all components of a distribution
evaluate its commercial aspect in [102], and it shows sig- network in details, but an optimal solution can be com-
nificant reduction in costs of GHG emission and loss, and putationally expensive or practically infeasible when the
improvement in reliability indices. An industrial microgrid size of the system is large.
with PV in China is analyzed from the economic aspect • The heuristic rule-based algorithm can quickly find
in [103] regarding emission reduction costs, levelized energy feasible microgrid formation after faults, but it needs
costs, and the payback period. This study demonstrates the problem-dependent information and may not guarantee
economic benefit gained by a PV-based microgrid through an optimal solution.
real microgrid output data. In [104], economic benefits of • The graph theory can be successful to find optimal solu-
microgrids are assessed according to reliability improvement, tion rapidly in a small system, however, its efficiency
emission reduction, power quality of services, and the less- degrades for medium to large systems since the number
ened peak loading. of trees is increasing, and implementing the graph par-
Participation in electricity market is considered one ben- titioning concept takes time to form a microgrid, which
efit behind microgrid formation [105]. In the restructured makes this method unattractive in the service restoration
electricity market, microgrids can participate in both energy problem.
and ancillary service markets as autonomous entities. Power • In the spectral algorithm, to determine affinity between
generated by microgrids can be traded in electricity markets. two components, the Hessian matrix needs to be calcu-
The microgrid control center conducts the optimal power lated for AC power follow at a specific operating point.

46628 VOLUME 10, 2022

M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

Although this algorithm provides a reliable solution for REFERENCES

a particular operating point, it may not be a promis- [1] M. N. Ambia, K. Meng, W. Xiao, and Z. Y. Dong, ‘‘Nested formation
ing solution for service restoration problem because the approach for networked microgrid self-healing in islanded mode,’’ IEEE
Trans. Power Del., vol. 36, no. 1, pp. 452–464, Feb. 2021.
system operating condition varies and the partitioning [2] L. Che and M. Shahidehpour, ‘‘Adaptive formation of microgrids with
should be run for many operating points. mobile emergency resources for critical service restoration in extreme
• The hierarchical clustering technique employs structural conditions,’’ IEEE Trans. Power Syst., vol. 34, no. 1, pp. 742–753,
Jan. 2019.
characteristics of distribution networks rather than the [3] United States Environmental Protection Agency. (Mar. 15, 2009). U.S.
operating point, which makes partitioning more rea- Billion-Dollar Weather and Climate Disasters, 1980—Present (NCEI
sonable than the spectral algorithm. K-means method Accession 0209268). Accessed: Mar. 16, 2022. [Online]. Available:
https://hero.epa.gov/hero/index.cfm/reference/details/reference
clusters objects by minimizing the distance within each _id/7310466
cluster, but it may not be able to guarantee radiality [4] D. K. Mishra, M. J. Ghadi, A. Azizivahed, L. Li, and J. Zhang, ‘‘A review
constraints of distribution networks. on resilience studies in active distribution systems,’’ Renew. Sustain.
Energy Rev., vol. 135, Jan. 2021, Art. no. 110201.
Based on advantages and disadvantages of various algo-
[5] J. Lopez, J. E. Rubio, and C. Alcaraz, ‘‘A resilient architecture for the
rithms for microgrid formation, the combination of differ- smart grid,’’ IEEE Trans. Ind. Informat., vol. 14, no. 8, pp. 3745–3753,
ent strategies may contribute to a more reliable solution for Aug. 2018.
service restoration problem. For example, the combination [6] A. Shaker, A. Safari, and M. Shahidehpour, ‘‘Reactive power manage-
ment for networked microgrid resilience in extreme conditions,’’ IEEE
of k-means method with mixed integer programming tech- Trans. Smart Grid, vol. 12, no. 5, pp. 3940–3953, Sep. 2021.
nique can result in less decision variables, which improves [7] Q. Zhou, M. Shahidehpour, A. Alabdulwahab, and A. Abusorrah, ‘‘Flexi-
convergence speed of the optimization problem and satisfies ble division and unification control strategies for resilience enhancement
in networked microgrids,’’ IEEE Trans. Power Syst., vol. 35, no. 1,
operational constraints. pp. 474–486, Jan. 2020.
[8] L. Faria, A. Silva, Z. Vale, and A. Marques, ‘‘Training control cen-
B. FUTURE RESEARCH DIRECTIONS ters’ operators in incident diagnosis and power restoration using intel-
ligent tutoring systems,’’ IEEE Trans. Learn. Technol., vol. 2, no. 2,
In the area of service restoration using microgrid forma- pp. 135–147, Apr. 2009.
tion, the distributed optimization technique can be a suitable [9] H. Liu, R. A. Davidson, and T. V. Apanasovich, ‘‘Statistical forecasting
method to tackle microgrid formation problem, as it decom- of electric power restoration times in hurricanes and ice storms,’’ IEEE
Trans. Power Syst., vol. 22, no. 4, pp. 2270–2279, Nov. 2007.
poses a large optimization problem into several subproblems [10] P.-C. Chen and M. Kezunovic, ‘‘Fuzzy logic approach to predictive risk
and handles them in a parallel fashion, and thus, the con- analysis in distribution outage management,’’ IEEE Trans. Smart Grid,
vergence speed increases and the global optimal solution is vol. 7, no. 6, pp. 2827–2836, Nov. 2016.
[11] R. B. Duffey and T. Ha, ‘‘The probability and timing of power system
obtained. restoration,’’ IEEE Trans. Power Syst., vol. 28, no. 1, pp. 3–9, Feb. 2013.
To develop efficient service restoration strategies through [12] X. Liang, M. A. Saaklayen, M. A. Igder, S. M. R. H. Shawon, S. O. Faried,
microgrid formation, more realistic models of distribution and M. Janbakhsh, ‘‘Planning and service restoration through microgrid
formation and soft open points for distribution network modernization: A
networks are needed. For example, most distribution net- review,’’ IEEE Trans. Ind. Appl., vol. 58, no. 2, pp. 1843–1857, Mar. 2022.
works have unbalanced configurations, which has not been [13] T. Ding, Z. Wang, W. Jia, B. Chen, C. Chen, and M. Shahidehpour,
sufficiently investigated in microgrid formation. In addi- ‘‘Multiperiod distribution system restoration with routing repair crews,
tion, supplying power to different types of loads (static and mobile electric vehicles, and soft-open-point networked microgrids,’’
IEEE Trans. Smart Grid, vol. 11, no. 6, pp. 4795–4808, Nov. 2020.
dynamic loads) needs to be further investigated following out- [14] K. S. A. Sedzro, A. J. Lamadrid, and L. F. Zuluaga, ‘‘Allocation
ages as each constructed microgrid must have the capability of resources using a microgrid formation approach for resilient elec-
to manage motor starting transients as an efficient service tric grids,’’ IEEE Trans. Power Syst., vol. 33, no. 3, pp. 2633–2643,
May 2018.
restoration strategy. [15] Y. Wang, C. Chen, J. Wang, and R. Baldick, ‘‘Research on resilience of
Artificial intelligent and machine learning-based methods power systems under natural disasters—A review,’’ IEEE Trans. Power
can be developed in service restoration to realize intelligent Syst., vol. 31, no. 2, pp. 1604–1613, Mar. 2016.
[16] S. Lei, C. Chen, Y. Li, and Y. Hou, ‘‘Resilient disaster recovery logistics
control actions. For example, deep reinforcement learning has of distribution systems: Co-optimize service restoration with repair crew
a big potential to realize microgrid formation and intelligent and mobile power source dispatch,’’ IEEE Trans. Smart Grid, vol. 10,
service restoration in distribution networks. no. 6, pp. 6187–6202, Nov. 2019.
[17] Y. Xu, C.-C. Liu, K. P. Schneider, and D. T. Ton, ‘‘Placement of remote-
controlled switches to enhance distribution system restoration capabil-
VI. CONCLUSION ity,’’ IEEE Trans. Power Syst., vol. 31, no. 2, pp. 1139–1150, Mar. 2016.
In this paper, service restoration through microgrid formation [18] Z. Galias, ‘‘Tree-structure based deterministic algorithms for optimal
switch placement in radial distribution networks,’’ IEEE Trans. Power
techniques in the literature is extensively reviewed. Various Syst., vol. 34, no. 6, pp. 4269–4278, Nov. 2019.
approaches to construct mircorgrids are introduced, such as [19] S. P. Singh, G. S. Raju, G. K. Rao, and M. Afsari, ‘‘A heuristic method for
graph theory, heuristic rule-based algorithm, clustering algo- feeder reconfiguration and service restoration in distribution networks,’’
Int. J. Electr. Power Energy Syst., vol. 31, nos. 7–8, pp. 309–314, 2009.
rithm, and mixed integer linear programming. Control and
[20] M. B. Jorge, V. O. Héctor, L. G. Miguel, and P. D. Héctor, ‘‘Multi-
economic aspects of microgrids are summarized. The future fault service restoration in distribution networks considering the operat-
research directions are recommended in the paper. The paper ing mode of distributed generation,’’ Electr. Power Syst. Res., vol. 116,
offers valuable information to engineers and researchers pp. 67–76, Nov. 2014.
[21] F. Shen, Q. Wu, and Y. Xue, ‘‘Review of service restoration for distribu-
working on renewable energy sources and distribution system tion networks,’’ J. Mod. Power Syst. Clean Energy, vol. 8, no. 1, pp. 1–14,
modernization. 2020.

VOLUME 10, 2022 46629

 M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

[22] H. Sekhavatmanesh and R. Cherkaoui, ‘‘Distribution network restoration [43] H. Hou, X. Yin, Q. Chen, D. You, G. Tong, and D. Shao, ‘‘Review on the
in a multiagent framework using a convex OPF model,’’ IEEE Trans. wide area blackout of 500 kV main power grid in some areas of south
Smart Grid, vol. 10, no. 3, pp. 2618–2628, May 2019. China in 2008 snow disaster,’’ Autom. Electr. Power Syst., vol. 32, no. 11,
[23] C. Chen, J. Wang, F. Qiu, and D. Zhao, ‘‘Resilient distribution system by pp. 12–15, 2008.
microgrids formation after natural disasters,’’ IEEE Trans. Smart Grid, [44] S. Cai, Y. Xie, Q. Wu, M. Zhang, X. Jin, and Z. Xiang, ‘‘Distributionally
vol. 7, no. 2, pp. 958–966, Mar. 2016. robust microgrid formation approach for service restoration under random
[24] M. Khederzadeh and S. Zandi, ‘‘Enhancement of distribution system contingency,’’ IEEE Trans. Smart Grid, vol. 12, no. 6, pp. 4926–4937,
restoration capability in single/multiple faults by using microgrids as Nov. 2021.
a resiliency resource,’’ IEEE Syst. J., vol. 13, no. 2, pp. 1796–1803, [45] W. T. El-Sayed, H. E. Z. Farag, H. H. Zeineldin, and E. F. El-Saadany,
Jun. 2019. ‘‘Formation of islanded droop-based microgrids with optimum load-
[25] G. Patsakis, D. Rajan, I. Aravena, and S. Oren, ‘‘Strong mixed-integer ability,’’ IEEE Trans. Power Syst., vol. 37, no. 2, pp. 1564–1576,
formulations for power system islanding and restoration,’’ IEEE Trans. Mar. 2022.
Power Syst., vol. 34, no. 6, pp. 4880–4888, Nov. 2019. [46] K. S. A. Sedzro, X. Shi, A. J. Lamadrid, and L. F. Zuluaga, ‘‘A heuristic
[26] A. Llaria, O. Curea, J. Jiménez, and H. Camblong, ‘‘Survey on micro- approach to the post-disturbance and stochastic pre-disturbance micro-
grids: Unplanned islanding and related inverter control techniques,’’ grid formation problem,’’ IEEE Trans. Smart Grid, vol. 10, no. 5,
Renew. Energy, vol. 36, no. 8, pp. 2052–2061, Aug. 2011. pp. 5574–5586, Sep. 2019.
[27] P. Fuangfoo, T. Meenual, W.-J. Lee, and C. Chompoo-inwai, ‘‘PEA
[47] L. H. Macedo, G. Muñoz-Delgado, J. Contreras, and R. Romero, ‘‘Opti-
guidelines for impact study and operation of DG for islanding operation,’’
mal service restoration in active distribution networks considering micro-
IEEE Trans. Ind. Appl., vol. 44, no. 5, pp. 1348–1353, Sep. 2008.
grid formation and voltage control devices,’’ IEEE Trans. Ind. Appl.,
[28] L. Zhu, C. Zhang, H. Yin, D. Li, Y. Su, I. Ray, J. Dong, F. Wang, vol. 57, no. 6, pp. 5758–5771, Nov. 2021.
L. M. Tolbert, Y. Liu, Y. Ma, B. Rogers, J. Glass, L. Bruce, S. Delay,
[48] J. Li, X.-Y. Ma, C.-C. Liu, and K. P. Schneider, ‘‘Distribution system
P. Gregory, M. Garcia-Sanz, and M. Marden, ‘‘A smart and flexible
restoration with microgrids using spanning tree search,’’ IEEE Trans.
microgrid with a low-cost scalable open-source controller,’’ IEEE Access,
Power Syst., vol. 29, no. 6, pp. 3021–3029, Nov. 2014.
vol. 9, pp. 162214–162230, 2021.
[29] S. A. Arefifar, Y. A.-R. I. Mohamed, and T. H. El-Fouly, ‘‘Supply- [49] R. J. Sánchez-García, M. Fennelly, S. Norris, N. Wright, G. Niblo,
adequacy-based optimal construction of microgrids in smart distribu- J. Brodzki, and J. W. Bialek, ‘‘Hierarchical spectral clustering of
tion systems,’’ IEEE Trans. Smart Grid, vol. 3, no. 3, pp. 1491–1502, power grids,’’ IEEE Trans. Power Syst., vol. 29, no. 5, pp. 2229–2237,
Sep. 2012. Sep. 2014.
[30] M. Barani, J. Aghaei, M. A. Akbari, T. Niknam, H. Farahmand, and [50] S. Lei, C. Chen, Y. Song, and Y. Hou, ‘‘Radiality constraints for
M. Korpas, ‘‘Optimal partitioning of smart distribution systems into resilient reconfiguration of distribution systems: Formulation and appli-
supply-sufficient microgrids,’’ IEEE Trans. Smart Grid, vol. 10, no. 3, cation to microgrid formation,’’ IEEE Trans. Smart Grid, vol. 11, no. 5,
pp. 2523–2533, May 2019. pp. 3944–3956, Sep. 2020.
[31] S. A. Arefifar, Y. A.-R. I. Mohamed, and T. H. M. El-Fouly, ‘‘Optimum [51] C. A. Cortes, S. F. Contreras, and M. Shahidehpour, ‘‘Microgrid
microgrid design for enhancing reliability and supply-security,’’ IEEE topology planning for enhancing the reliability of active distribution
Trans. Smart Grid, vol. 4, no. 3, pp. 1567–1575, Sep. 2013. networks,’’ IEEE Trans. Smart Grid, vol. 9, no. 6, pp. 6369–6377,
[32] S. A. Arefifar, Y. A.-R.-I. Mohamed, and T. El-Fouly, ‘‘Optimized multi- Nov. 2018.
ple microgrid-based clustering of active distribution systems considering [52] L. Che, X. Zhang, M. Shahidehpour, A. Alabdulwahab, and Y. Al-Turki,
communication and control requirements,’’ IEEE Trans. Ind. Electron., ‘‘Optimal planning of loop-based microgrid topology,’’ IEEE Trans.
vol. 62, no. 2, pp. 711–723, Feb. 2015. Smart Grid, vol. 8, no. 4, pp. 1771–1781, Jul. 2017.
[33] A. Mohsenzadeh, C. Pang, and M.-R. Haghifam, ‘‘Determining optimal [53] M. Zadsar, M. R. Haghifam, and S. M. M. Larimi, ‘‘Approach for self-
forming of flexible microgrids in the presence of demand response in healing resilient operation of active distribution network with micro-
smart distribution systems,’’ IEEE Syst. J., vol. 12, no. 4, pp. 3315–3323, grid,’’ IET Gener., Transmiss. Distrib., vol. 11, no. 18, pp. 4633–4643,
Dec. 2018. Dec. 2017.
[34] Y. Du, H. Tu, X. Lu, J. Wang, and S. Lukic, ‘‘Black-start and service [54] J. Morren, S. W. H. de Haan, and J. A. Ferreira, ‘‘Contribution of DG
restoration in resilient distribution systems with dynamic microgrids,’’ units to primary frequency control,’’ Eur. Trans. Electr. Power, vol. 16,
IEEE J. Emerg. Sel. Topics Power Electron., early access, Apr. 8, 2021, no. 5, pp. 507–521, 2006.
doi: 10.1109/JESTPE.2021.3071765. [55] K. Singh, M. Amir, F. Ahmad, and S. S. Refaat, ‘‘Enhancement of
[35] M. E. Nassar and M. M. A. Salama, ‘‘Adaptive self-adequate microgrids frequency control for stand-alone multi-microgrids,’’ IEEE Access, vol. 9,
using dynamic boundaries,’’ IEEE Trans. Smart Grid, vol. 7, no. 1, pp. 79128–79142, 2021.
pp. 105–113, Jan. 2016.
[56] Zaheeruddin, K. Singh, and M. Amir, ‘‘Intelligent fuzzy TIDF-II con-
[36] R. Hemmati, H. Mehrjerdi, and S. M. Nosratabadi, ‘‘Resilience-oriented
troller for load frequency control in hybrid energy system,’’ IETE Tech.
adaptable microgrid formation in integrated electricity-gas system with
Rev., pp. 1–17, Nov. 2021.
deployment of multiple energy hubs,’’ Sustain. Cities Soc., vol. 71,
[57] A. Sharma, D. Srinivasan, and A. Trivedi, ‘‘A decentralized multi-agent
Aug. 2021, Art. no. 102946.
approach for service restoration in uncertain environment,’’ IEEE Trans.
[37] M. E. Khodayar, ‘‘Rural electrification and expansion planning of off-grid
Smart Grid, vol. 9, no. 4, pp. 3394–3405, Jul. 2018.
microgrids,’’ Electr. J., vol. 30, no. 4, pp. 68–74, May 2017.
[38] P. Siritoglou, G. Oriti, and D. L. Van Bossuyt, ‘‘Distributed energy- [58] A. Sharma, D. Srinivasan, and A. Trivedi, ‘‘A decentralized multiagent
resource design method to improve energy security in critical facilities,’’ system approach for service restoration using DG islanding,’’ IEEE Trans.
Energies, vol. 14, no. 10, May 2021, Art. no. 10. Smart Grid, vol. 6, no. 6, pp. 2784–2793, Nov. 2015.
[39] A. Naderipour, Z. Abdul-Malek, M. Hajivand, Z. M. Seifabad, [59] J. Guo, G. Hug, and O. K. Tonguz, ‘‘Intelligent partitioning in distributed
M. A. Farsi, S. A. Nowdeh, and I. F. Davoudkhani, ‘‘Spotted hyena optimization of electric power systems,’’ IEEE Trans. Smart Grid, vol. 7,
optimizer algorithm for capacitor allocation in radial distribution system no. 3, pp. 1249–1258, May 2016.
with distributed generation and microgrid operation considering different [60] X. Xu, F. Xue, S. Lu, H. Zhu, L. Jiang, and B. Han, ‘‘Structural and hier-
load types,’’ Sci. Rep., vol. 11, no. 1, Feb. 2021, Art. no. 1. archical partitioning of virtual microgrids in power distribution network,’’
[40] R. Yan, N.-A. Masood, T. K. Saha, F. Bai, and H. Gu, ‘‘The anatomy of IEEE Syst. J., vol. 13, no. 1, pp. 823–832, Mar. 2019.
the 2016 South Australia blackout: A catastrophic event in a high renew- [61] E. Cotilla-Sanchez, P. D. Hines, C. Barrows, S. Blumsack, and M. Patel,
able network,’’ IEEE Trans. Power Syst., vol. 33, no. 5, pp. 5374–5388, ‘‘Multi-attribute partitioning of power networks based on electrical
Sep. 2018. distance,’’ IEEE Trans. Power Syst., vol. 28, no. 4, pp. 4979–4987,
[41] G. Liang, S. R. Weller, J. Zhao, F. Luo, and Z. Y. Dong, ‘‘The 2015 Nov. 2013.
Ukraine blackout: Implications for false data injection attacks,’’ IEEE [62] C. Shah and R. Wies, ‘‘Adaptive day-ahead prediction of resilient power
Trans. Power Syst., vol. 32, no. 4, pp. 3317–3318, Jul. 2017. distribution network partitions,’’ in Proc. IEEE Green Technol. Conf.
[42] S. Sreekumar, D. S. Kumar, and J. S. Savier, ‘‘A case study on self (GreenTech), Apr. 2021, pp. 477–483.
healing of smart grid with islanding and inverter volt–VAR function,’’ [63] R. Xu and D. Wunsch, ‘‘Survey of clustering algorithms,’’ IEEE Trans.
IEEE Trans. Ind. Appl., vol. 56, no. 5, pp. 5408–5416, Oct. 2020. Neural Netw., vol. 16, no. 3, pp. 645–678, May 2005.

46630 VOLUME 10, 2022

M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

[64] D. M. L. K. Cheong, T. Fernando, H. C. Iu, M. Reynolds, and J. Fletcher, [85] Y. Xu and Z. Li, ‘‘Distributed optimal resource management based on the
‘‘Review of clustering algorithms for microgrid formation,’’ in Proc. consensus algorithm in a microgrid,’’ IEEE Trans. Ind. Electron., vol. 62,
IEEE Innov. Smart Grid Technol. Asia (ISGT-Asia), Dec. 2017, pp. 1–6. no. 4, pp. 2584–2592, Apr. 2015.
[65] M. A. Gilani, A. Kazemi, and M. Ghasemi, ‘‘Distribution system [86] W. Liu, W. Gu, W. Sheng, X. Meng, Z. Wu, and W. Chen, ‘‘Decen-
resilience enhancement by microgrid formation considering distributed tralized multi-agent system-based cooperative frequency control for
energy resources,’’ Energy, vol. 191, Jan. 2020, Art. no. 116442. autonomous microgrids with communication constraints,’’ IEEE Trans.
[66] T. Ding, Y. Lin, G. Li, and Z. Bie, ‘‘A new model for resilient distribution Sustain. Energy, vol. 5, no. 2, pp. 446–456, Apr. 2014.
systems by microgrids formation,’’ IEEE Trans. Power Syst., vol. 32, [87] T. Morstyn, B. Hredzak, and V. G. Agelidis, ‘‘Cooperative multi-
no. 5, pp. 4145–4147, Sep. 2017. agent control of heterogeneous storage devices distributed in a DC
[67] T. Zhao, B. Chen, S. Zhao, J. Wang, and X. Lu, ‘‘A flexible operation microgrid,’’ IEEE Trans. Power Syst., vol. 31, no. 4, pp. 2974–2986,
of distributed generation in distribution networks with dynamic bound- Jul. 2016.
aries,’’ IEEE Trans. Power Syst., vol. 35, no. 5, pp. 4127–4130, Sep. 2020. [88] C. Dou, M. Lv, T. Zhao, Y. Ji, and H. Li, ‘‘Decentralised coordinated con-
[68] R. A. Osama, A. F. Zobaa, and A. Y. Abdelaziz, ‘‘A planning framework trol of microgrid based on multi-agent system,’’ IET Gener., Transmiss.
for optimal partitioning of distribution networks into microgrids,’’ IEEE Distrib., vol. 9, no. 16, pp. 2474–2484, Dec. 2015.
Syst. J., vol. 14, no. 1, pp. 916–926, Mar. 2020. [89] C. M. Colson and M. H. Nehrir, ‘‘Comprehensive real-time microgrid
[69] Z. N. Popovic, S. D. Knezevic, and B. S. Brbaklic, ‘‘A risk manage- power management and control with distributed agents,’’ IEEE Trans.
ment procedure for island partitioning of automated radial distribution Smart Grid, vol. 4, no. 1, pp. 617–627, Mar. 2013.
networks with distributed generators,’’ IEEE Trans. Power Syst., vol. 35, [90] Z. Li, C. Zang, P. Zeng, H. Yu, and H. Li, ‘‘MAS based distributed
no. 5, pp. 3895–3905, Sep. 2020. automatic generation control for cyber-physical microgrid system,’’
[70] L. Meng, E. R. Sanseverino, A. Luna, T. Dragicevic, J. C. Vasquez, and IEEE/CAA J. Autom. Sinica, vol. 3, no. 1, pp. 78–89, Jan. 2016.
J. M. Guerrero, ‘‘Microgrid supervisory controllers and energy manage- [91] J. Yang, J. Dai, H. B. Gooi, H. Nguyen, and A. Paudel, ‘‘A proof-
ment systems: A literature review,’’ Renew. Sustain. Energy Rev., vol. 60, of-authority blockchain based distributed control system for islanded
pp. 1263–1273, Jul. 2016. microgrids,’’ IEEE Trans. Ind. Informat., early access, Jan. 13, 2022, doi:
[71] Y. Han, K. Zhang, H. Li, E. A. A. Coelho, and J. M. Guerrero, ‘‘MAS- 10.1109/TII.2022.3142755.
based distributed coordinated control and optimization in microgrid and [92] M. M. Rana, A. Rahman, M. Uddin, M. R. Sarkar, S. A. Shezan,
microgrid clusters: A comprehensive overview,’’ IEEE Trans. Power M. F. Ishraque, S. M. S. H. Rafin, and M. Atef, ‘‘A comparative analysis
Electron., vol. 33, no. 8, pp. 6488–6508, Aug. 2018. of peak load shaving strategies for isolated microgrid using actual data,’’
[72] J. M. Guerrero, J. C. Vasquez, J. Matas, L. G. de Vicuna, and M. Castilla, Energies, vol. 15, no. 1, p. 330, Jan. 2022.
‘‘Hierarchical control of droop-controlled AC and DC microgrids—A [93] A. Rauf, A. T. Awami, M. Kassas, and M. Khalid, ‘‘Optimizing a residen-
general approach toward standardization,’’ IEEE Trans. Ind. Electron., tial solar PV system based on net-metering approaches,’’ in Proc. CIRED
vol. 58, no. 1, pp. 158–172, Jan. 2011. 26th Int. Conf. Exhib. Electr. Distrib. IET, Sep. 2021, pp. 1–5.
[73] M. M. A. Abdelaziz, M. F. Shaaban, H. E. Farag, and E. F. El-Saadany, [94] L. Zhang, Y. Yang, Q. Li, W. Gao, F. Qian, and L. Song, ‘‘Economic
‘‘A multistage centralized control scheme for islanded microgrids with optimization of microgrids based on peak shaving and CO2 reduction
PEVs,’’ IEEE Trans. Sustain. Energy, vol. 5, no. 3, pp. 927–937, effect: A case study in Japan,’’ J. Cleaner Prod., vol. 321, Oct. 2021,
Jul. 2014. Art. no. 128973.
[74] N. L. Díaz, A. C. Luna, J. C. Vasquez, and J. M. Guerrero, ‘‘Centralized [95] M. Stadler, G. Cardoso, S. Mashayekh, T. Forget, N. DeForest,
control architecture for coordination of distributed renewable generation A. Agarwal, and A. Schönbein, ‘‘Value streams in microgrids: A literature
and energy storage in islanded AC microgrids,’’ IEEE Trans. Power review,’’ Appl. Energy, vol. 162, pp. 980–989, Jan. 2016.
Electron., vol. 32, no. 7, pp. 5202–5213, Jul. 2017. [96] Y. Wang, A. O. Rousis, and G. Strbac, ‘‘On microgrids and resilience: A
[75] Q. Shafiee, V. Nasirian, J. C. Vasquez, J. M. Guerrero, and A. Davoudi, comprehensive review on modeling and operational strategies,’’ Renew.
‘‘A multi-functional fully distributed control framework for AC micro- Sustain. Energy Rev., vol. 134, Dec. 2020, Art. no. 110313.
grids,’’ IEEE Trans. Smart Grid, vol. 9, no. 4, pp. 3247–3258, [97] R. Ma, H. Cai, Q. Ji, and P. Zhai, ‘‘The impact of feed-in tariff degression
Jul. 2018. on R&D investment in renewable energy: The case of the solar PV
[76] X. Lu, X. Yu, J. Lai, Y. Wang, and J. M. Guerrero, ‘‘A novel distributed industry,’’ Energy Policy, vol. 151, Apr. 2021, Art. no. 112209.
secondary coordination control approach for islanded microgrids,’’ IEEE [98] Y. Parag and M. Ainspan, ‘‘Sustainable microgrids: Economic, environ-
Trans. Smart Grid, vol. 9, no. 4, pp. 2726–2740, Jul. 2018. mental and social costs and benefits of microgrid deployment,’’ Energy
[77] A. Bidram, A. Davoudi, F. L. Lewis, and J. M. Guerrero, ‘‘Dis- Sustain. Develop., vol. 52, pp. 72–81, Oct. 2019.
tributed cooperative secondary control of microgrids using feedback [99] S. Chowdhury, S. P. Chowdhury, and P. Crossley, Microgrids and Active
linearization,’’ IEEE Trans. Power Syst., vol. 28, no. 3, pp. 3462–3470, Distribution Networks (The Institution of Engineering and Technology
Aug. 2013. (IET)), London, U.K.: 2009.
[78] Y. Han, H. Li, P. Shen, E. A. A. Coelho, and J. M. Guerrero, ‘‘Review [100] S. R. Thangavelu, A. M. Khambadkone, and I. A. Karimi, ‘‘Long-term
of active and reactive power sharing strategies in hierarchical controlled optimal energy mix planning towards high energy security and low GHG
microgrids,’’ IEEE Trans. Power Electron., vol. 32, no. 3, pp. 2427–2451, emission,’’ Appl. Energy, vol. 154, pp. 959–969, Sep. 2015.
Mar. 2017. [101] P. Balducci, K. Mongird, D. Wu, D. Wang, V. Fotedar, and R. Dahowski,
[79] S. M. Mohiuddin and J. Qi, ‘‘Optimal distributed control of AC ‘‘An evaluation of the economic and resilience benefits of a microgrid
microgrids with coordinated voltage regulation and reactive power in Northampton, Massachusetts,’’ Energies, vol. 13, no. 18, p. 4802,
sharing,’’ IEEE Trans. Smart Grid, vol. 13, no. 3, pp. 1789–1800, Sep. 2020.
May 2022. [102] T. Adefarati and R. C. Bansal, ‘‘Reliability, economic and environ-
[80] X. Yu, A. M. Khambadkone, H. Wang, and S. T. S. Terence, ‘‘Control of mental analysis of a microgrid system in the presence of renew-
parallel-connected power converters for low-voltage microgrid—Part I: A able energy resources,’’ Appl. Energy, vol. 236, pp. 1089–1114,
hybrid control architecture,’’ IEEE Trans. Power Electron., vol. 25, no. 12, Feb. 2019.
pp. 2962–2970, Dec. 2010. [103] M. Mao, P. Jin, Y. Zhao, F. Chen, and L. Chang, ‘‘Optimal allocation and
[81] A. Mehrizi-Sani, ‘‘Chapter 2 - Distributed Control Techniques in Micro- economic evaluation for industrial PV microgrid,’’ in Proc. IEEE Energy
grids,’’ in The Book of Microgrid: Advanced Control Methods and Renew- Convers. Congr. Expo., Sep. 2013, pp. 4595–4602.
able Energy System Integration, Amsterdam, The Netherlands: Elsevier, [104] G. Y. Morris, C. Abbey, S. Wong, and G. Joós, ‘‘Evaluation of the costs
Imprint Butterworth-Heinemann, 2017, pp. 43–62. and benefits of microgrids with consideration of services beyond energy
[82] M. Yazdanian and A. Mehrizi-Sani, ‘‘Distributed control techniques in supply,’’ in Proc. IEEE Power Energy Soc. Gen. Meeting, Jul. 2012,
microgrids,’’ IEEE Trans. Smart Grid, vol. 5, no. 6, pp. 2901–2909, pp. 1–9.
Nov. 2014. [105] M. J. Ghadi, A. Rajabi, S. Ghavidel, A. Azizivahed, L. Li, and J. Zhang,
[83] V. N. Coelho, M. W. Cohen, I. M. Coelho, N. Liu, and F. G. Guimarães, ‘‘From active distribution systems to decentralized microgrids: A review
‘‘Multi-agent systems applied for energy systems integration: State-of- on regulations and planning approaches based on operational factors,’’
the-art applications and trends in microgrids,’’ Appl. Energy, vol. 187, Appl. Energy, vol. 253, Nov. 2019, Art. no. 113543.
pp. 820–832, Feb. 2017. [106] M. H. Imani, P. Niknejad, and M. R. Barzegaran, ‘‘The impact of cus-
[84] C.-H. Yoo, I.-Y. Chung, H.-J. Lee, and S.-S. Hong, ‘‘Intelligent control of tomers’ participation level and various incentive values on implementing
battery energy storage for multi-agent based microgrid energy manage- emergency demand response program in microgrid operation,’’ Electr.
ment,’’ Energies, vol. 6, no. 10, pp. 4956–4979, Sep. 2013. Power Energy Syst., vol. 96, pp. 114–125, Mar. 2018.

VOLUME 10, 2022 46631

 M. A. Igder et al.: Service Restoration Through Microgrid Formation in Distribution Networks: A Review

MOSAYEB AFSHARI IGDER (Student Member, MASSIMO MITOLO (Fellow, IEEE) received the
IEEE) received the M.S. degree in electrical engi- Ph.D. degree in electrical engineering from the
neering from the Shiraz University of Technology, University of Napoli ‘‘Federico II,’’ Italy.
Shiraz, Iran, in 2016. He is currently pursuing the He is currently a Full Professor of electrical
Master of Science degree with the University of engineering at the Irvine Valley College, Irvine,
Saskatchewan, Saskatoon, SK, Canada. He had CA, USA, and a Senior Consultant in the matter
a research collaboration with the Department of of failure analysis and electrical safety with Engi-
Electrical and Electronics Engineering, Shiraz neering Systems Inc., ESi. He has authored over
University of Technology, from 2016 to 2020. His 140 journal articles and the books Electrical Safety
research interests include power system operation, of Low-Voltage Systems (McGraw-Hill), Labora-
renewable energy, and marine power systems. tory Manual for Introduction to Electronics: A Basic Approach (Pearson),
and Analysis of Grounding and Bonding Systems (CRC Press). His research
XIAODONG LIANG (Senior Member, IEEE) was interests include the analysis and grounding of power systems, and electrical
born in Lingyuan, Liaoning, China. She received safety engineering.
the B.Eng. and M.Eng. degrees in electrical engi- Dr. Mitolo is a fellow of IEEE ‘‘for contributions to the electrical safety
neering from Shenyang Polytechnic University, of low-voltage systems,’’ and a fellow of the Institution of Engineering and
Shenyang, China in 1992 and 1995, respectively, Technology (IET) of London, U.K. He was a recipient of the IEEE Region 6
the M.Sc. degree in electrical engineering from ‘‘2015 Outstanding Engineer Award,’’ he has earned nine best paper awards,
the University of Saskatchewan, Saskatoon, SK, numerous achievements and recognitions, among which are the IEEE ‘‘Ralph
Canada, in 2004, and the Ph.D. degree in elec- H. Lee I&CPS Department Prize Paper Award,’’ the IEEE ‘‘I&CPS 2015
trical engineering from the University of Alberta, Department Achievement Award,’’ and the James E. Ballinger Engineer of
Edmonton, AB, Canada, in 2013. the Year 2013 Award from the Orange County Engineering Council. He is a
From 1995 to 1999, she worked as a Lecturer at Northeastern University, Registered Professional Engineer in the state of California and in Italy. He is
Shenyang. In October 2001, she joined Schlumberger, Edmonton, and was active within the Industrial and Commercial Power Systems Department of
promoted to be a Principal Power Systems Engineer with this world’s leading the IEEE Industry Applications Society (IAS) in numerous committees and
oil field service company, in 2009. After serving Schlumberger for almost working groups.
12 years, from 2013 to 2019, she was with Washington State University,
Vancouver, WA, USA, and the Memorial University of Newfoundland,
St. John’s, NL, Canada, as an Assistant Professor and later an Associate
Professor. In July 2019, she joined the University of Saskatchewan, where
she is currently an Associate Professor and a Canada Research Chair in
Technology Solutions for Energy Security in Remote, Northern, and Indige-
nous Communities. Her research interests include power systems, renewable
energy, and electric machines.
Dr. Liang is a Registered Professional Engineer in the province of
Saskatchewan, Canada.

46632 VOLUME 10, 2022