energies

Article
Coordination Strategy for Optimal Scheduling of
Multiple Microgrids Based on Hierarchical System
Won-Poong Lee 1 , Jin-Young Choi 2 and Dong-Jun Won 1, *
1 Department of Electrical Engineering, Inha University, 100, Inha-ro, Nam-gu, Incheon 22212, Korea;
poong89@gmail.com
2 LG CNS, 28F, FKI Tower, 24, Yeoui-daero, Yeongdeungpo-gu, Seoul 07320, Korea; jychoi83@lgcns.com
* Correspondence: djwon@inha.ac.kr; Tel.: +82-32-860-7404; Fax: +82-32-863-5822

Received: 17 August 2017; Accepted: 31 August 2017; Published: 5 September 2017

Abstract: Research on the operation of the multiple microgrid (MMG) has been increasing as
the power system is operated through the microgrid. Some of the studies related to MMG have
introduced various operation strategies by introducing concepts such as power sharing and power
trading for power exchange between microgrids. In this paper, a strategy for obtaining optimal
scheduling of MMG systems with power sharing through coordination among microgrids that have
no cost function of generation units is proposed. There are microgrid-energy management systems
(MG-EMSs) in the lower level that determine individual schedules for each microgrid in a hierarchical
system. In the upper level, the microgrid of microgrids center (MoMC) implements the coordination
among microgrids. In order to achieve the optimal operation of the entire system, MoMC calculates
the amount of power sharing based on a predetermined limit value and allocates the command for
coordination to each MG-EMS. MG-EMS changes the individual schedule based on the command.
These processes are repeatedly performed, and when the change of the total cost becomes smaller
than a specified size, the process is terminated and the schedule is determined. The advantages
of the proposed algorithm are as follows. (1) It is a power sharing strategy of multiple microgrids
considering multiple feeder structures as well as a single feeder structure for minimizing the operation
cost of the entire system; (2) it is a power sharing strategy between microgrids that can be applied in
a microgrid where only units that do not have a cost function exist; (3) since it is the optimization of
the distributed form, the computation time decreases sharply compared with the one performed at
the central center. The verification of the proposed algorithm was performed through MATLAB.

Keywords: coordination strategy; distributed optimization; hierarchical system; multi-microgrid

1. Introduction
The microgrid has been studied in the last decade, focusing on new forms of the operation of
power systems. Because of the microgrid, the power system has shifted away from the conventional
centralized operation to a locally distributed form. Distributed generation (DG) such as the renewable
energy source and the energy storage system (ESS), which are typical components of the microgrid,
have also been continuously studied [1]. With emphasis on the importance of microgrids, the fields
where the microgrid can be applied have been diversified, such as university campuses, military service
areas, local communities, and commercial and industrial complexes. Since the above-mentioned areas
are generally medium-scale or large-scale systems, if they are composed of one microgrid they become
similar to a centralized operation method and increase the burden on management and operations.
For medium and large scale systems, therefore, it is appropriate to construct multiple microgrids rather
than a single microgrid. In accordance with this trend, in recent years there have been studies focusing
on a large number of microgrids, extending from a single microgrid operating mode. A number
of microgrid operation schemes have been proposed, including sharing or trading power between

Energies 2017, 10, 1336; doi:10.3390/en10091336 www.mdpi.com/journal/energies

Energies 2017, 10, 1336 2 of 18

microgrids when multiple microgrids are implemented. The operation of multiple microgrids is based
on extending a single microgrid operating scheme and is typically a hierarchical structure because it
is an extension of the operating structure of a single microgrid. In [2], optimal operation of a single
microgrid is proposed, but the system structure is hierarchical. Based on the model predictive control
framework, the alternating direction method of multipliers method and the dual decomposition
method are used. This means that the optimization problem of the microgrid is divided into N
sub-problems and solved in parallel. Prox-average message passing is applied in the process of solving
the problem. In this method, a specific system can be classified as net and device and handled in
a hierarchical manner. It is based on updating the output of each device by exchanging information
between net and device. Reference [3] also has a hierarchical structure using a decomposition method.
The various components in the microgrid are considered as sub-problems, and the optimum solution
for the entire microgrid is obtained while iteratively updating the multipliers. In [4], this hierarchical
distributed optimization method is applied to a large number of microgrid systems. The main purpose
here is to minimize the power loss due to the power sharing between the microgrids using Lagrange
multipliers. Reference [5] deals with the scheduling of power transactions of the multiple microgrid
with EVs. In this case, the potential price signals considering the dual variables are reflected in the time
of use (TOU) price signal, and this price signal is allocated to the respective microgrids so that the power
to deal with the utility grid is determined as the optimal value. As a result, the limit on the amount of
power trading with the utility grid at the peak is reduced, thereby reducing the cost. References [6,7]
have also proposed the method for the economic operation and power trading of multiple microgrids in
community form. Reference [6] proposes the power coordination of multiple microgrids for the optimal
operation of one community. Based on a hierarchical structure, the amount of power sharing is adjusted
through communication between the microgrid agent at the lower level and the microgrid center agent
at the upper level. Reference [7] proposes a power trading scheme in a community microgrid that
includes the AC microgrid and the DC microgrid. It is proposed that the output of the microgrid and
the amount of power trading between the utility grids by adjusting the droop curve of the converters
be adjusted. While all of the previous literature considers the grid-connected operation, references [8,9]
consider the islanded operation. In [8], a dual variable is assumed to be an electricity-selling price, and
then a power-trading scheme between the microgrids is proposed. According to the law of supply and
demand, the selling price of electricity among the microgrid is determined by considering the change
of the price by the load. Reference [9] proposes a distributed power sharing scheme using the average
consensus algorithm. Power sharing is performed only in an emergency and determines the amount
of power to be shared in proportion to the available reserve power of each microgrid. Other studies
that do not consider power sharing directly have proposed various operating strategies of multiple
microgrid systems in grid-connected or stand-alone modes [10,11], and multi-DC microgrid operation
and control strategies are also discussed recently [12,13].
Most of the studies on the operation of the multiple microgrid in the grid-connected, described
above, assume the topology with one PCC, and the mathematical models and assumptions of the
components used to solve the optimization problem are needed. Typically, it is assumed that the
objective function of the battery is arbitrarily convex quadratic. In practice, however, the topology
of multiple microgrids may have multiple feeders as well as a single feeder. For example, for two
microgrids owned by different owners, if each microgrid has a different PCC, the operating method
may be different in terms of power sharing.
Components with clearly defined mathematical models such as generators can be used as
mathematical models for optimal operation, but the elements such as the energy storage system
(ESS) are defined according to the users convenience, that is, they depend on the purpose of operation.
In recent studies, by using the relationship between the output and the efficiency of the battery,
the life-time cost or life-cycle cost of the battery can be modeled and the battery can be operated
according to these models [1416]. However, few studies have been conducted in the long-term,
considering the life cycle cost of the battery. In addition, according to a report on energy storage
 Energies 2017, 10, 1336 3 of 18

trends published in 2017, the cost of utility-scale energy storage systems is expected to gradually
decrease [17]. As a result, the operation of an ESS that considers lifetime costs is not necessarily
considered effective, and it may be better to replace the battery after obtaining the maximum benefit
because of the cost reduction of the ESS. For this reason, when an ESS runs in a microgrid, it cannot
be guaranteed to have a cost function. When a microgrid is constructed only with elements that do
not have a specified objective function, such as renewable generation or ESS, it is no longer possible
to derive an optimal operating plan considering the power sharing based on the incremental cost of
generating the components.
Therefore, a distributed operation scheme is proposed considering power sharing between
multiple microgrids, consisting of components that have a multi-feeder topology and components
that cannot have power generation costs, and cost optimization for the entire system. The proposed
operation scheme for the multi-microgrids system has a hierarchical structure. It consists of an
upper system that manages the entire microgrid called the microgrid of microgrids (MoMC) and
the microgrid energy management system (MG-EMS) that exist for each microgrid operation and
management. The proposed operating scheme achieves distributed optimal operation of the multiple
microgrid through appropriate coordination between the upper and lower systems. The purpose of
each microgrid located in the lower ones is to adjust the peak power of the microgrid, and the purpose
of the upper system is to adjust the optimum point for the entire system. In addition, in the case of
islanded operation, the scheme for performing power sharing within this hierarchical structure can be
used to increase the reliability of the multiple microgrid system. Power sharing can be accomplished
through the coordination of each microgrid EMS with the upper system, MoMC, providing power
from a microgrid with sufficient reserve power to a microgrid with deficient power.
In summary, this paper differs from previous studies in the following points.

(1) It includes a power sharing strategy among microgrids applied in multiple microgrid systems
with a single feeder as well as multiple feeders. Due to the power sharing achieved through
the proposed coordination algorithm in the MoMC, the peaks of each feeder are adjusted to
achieve a total cost savings.
(2) For units that do not have an objective function, it is difficult to achieve power sharing by
updating dual variables such as Lagrange multipliers or by matching the incremental cost of
the objective function through the consensus algorithm. Therefore, there is a need for a way
that power sharing can be performed between microgrids in a multiple microgrids system that
is connected to the utility grid. That is, when the power sharing is performed, if the microgrid
includes only power generation units, which cannot determine the cost function, the power
sharing can be achieved by applying the proposed algorithm.
(3) The proposed power sharing strategy is based on the distributed optimization of multiple
microgrid systems in a hierarchical structure. Based on the optimization results performed by
each microgrid, a coordination algorithm in the MoMC is performed and the results are reflected
in each microgrid. The coordination algorithm performed in MoMC is a simple operation as
opposed to solving complex problems such as optimization problems, and since each microgrid
performs local optimization, the time required for determining the schedule of the entire system
is reduced significantly.

Section 2 describes the definition of power sharing and the types of power sharing. Section 3
describes the hierarchical structure of multiple microgrid systems and describes the functions and roles
of MG-EMS and MoMC, which are responsible for the operation of the system. Section 4 describes
the algorithm for distributed optimal operation for the grid connected operation of multiple microgrids,
and Section 5 shows the simulation results for the proposed algorithm. Finally, Section 6 describes
the conclusion and the possibility that this algorithm can be applied in practice.
Energies 2017, 10, 1336 4 of 18

2. Definition of Power Sharing

Before an algorithm for optimal operation in multiple microgrid conditions is proposed, power
sharing or power exchange, which are key parts of the algorithm, are defined and the need for power
sharing and the kind of power sharing is also defined. It is possible to consider various situations that
can occur when operating multiple microgrids by understanding the form of power sharing according
to situations where power sharing is required.

2.1. Definition and Necessity of Power Sharing

Power sharing or power exchange is implemented to achieve economic or reliability objectives.
This means sending power from one microgrid to another or to multiple microgrids. Conversely, it is
possible to send power from multiple microgrids to one microgrid. In an AC system that constitutes
a current power system, it is impossible to directly send power from one place to another, but it can
be considered that power sharing is achieved by indirectly changing the flow of power. The form of
power sharing can vary depending on how the topology for a single microgrid or multiple microgrid
system is configured. Also, as mentioned earlier, the form of power sharing may vary depending on
the number of feeders in the system.
There are two main cases that require power sharing. Multiple microgrid systems are operated
in connection with the utility grid, and the system is disconnected from the utility grid and islanded
operation is performed. When operated in connection with the utility grid, the main objective
is to improve the economics of the overall system. In other words, the overall cost should be
minimized considering the operating cost of each microgrid. The cost for peak power can be reduced
through power sharing between the microgrids when peak power is expected in some microgrids.
In addition, economic efficiency can be improved through coordination between a generator with
a high incremental cost and a generator with a low one. On the other hand, during islanded operation,
the main purpose is to improve the reliability of the entire system. For this purpose, the reliability
of the overall system is improved by carrying out power sharing from the microgrid, which can
share sufficient reserve power, to the microgrid with the deficiency. For this reason, power sharing is
required for multiple microgrid operations.

2.2. Types of Power Sharing

The power sharing scheme can be classified into four cases according to the conditions. The first
condition is the number of grid feeders connected with the microgrid and the second condition is
the form of power sharing. The power sharing method can be divided into a single feeder and multiple
feeders depending on the number of feeders, and can be divided into direct power sharing or indirect
power sharing depending on the form of power sharing. In the case of a single feeder, it means that
there is only one point which is connected with the utility grid in the microgrid or multiple microgrid
systems, and in the case of multiple feeders, there are several such points. On the other hand, indirect
power sharing in the form of power sharing is a common form, and this indirect power sharing implies
the power flow that changes as some power sources reduce generation and other sources increase
generation. In the case of direct power sharing, this means that power is directly shared as a form of
point-to-point (PTP) between microgrids. For example, a back-to-back (BTB) structure is constructed
between two AC systems using a power electronics facility such as a converter, and the facility can be
used to share the power directly. In a single feeder, power sharing can occur indirectly because there is
only one feeder, whereas in many feeders, if there is no physical connection through the transmission
lines between the microgrids constituting the multiple microgrid systems, it can be regarded as indirect
power sharing. If there is no connection, it can be regarded as a virtual power plant (VPP) in multiple
feeders. It is defined as VPP because it determines the optimal schedule for the whole system and it is
not the case that power sharing is implemented actually. If there is a physical connection, the system
structure must be able to change so that the connection line does not form the entire ring system.
Energies 2017, 10, 1336 5 of 18

Figure 1 shows an example of the form of power sharing in an MMG system with two feeders.
In this way, the system structures for the multiple feeders are also taken into consideration. Table 1
shows the result of classifying the power sharing method according to the economy or reliability of
the power sharing purpose mentioned above. It shows whether the form of power sharing is indirect
or direct,
Energies depending
2017, 10, 1336 on the number of feeders and the purpose of the power sharing. 5 of 19

Figure 1. An example of the form of power sharing in a multiple microgrid (MMG) system structure
Figure 1. An example of the form of power sharing in a multiple microgrid (MMG) system structure
with two feeders.
with two feeders.

Table
Table 1. Types of power sharing based on various conditions.
conditions.

Number of Feeders Single Feeder Multiple Feeder

Number of Feeders Single Feeder Multiple Feeder
Operation Mode Grid-Connected Islanded Grid-Connected Islanded
Operation Mode Grid-Connected Islanded Grid-Connected Islanded
Objectives
Objectives Economic
Economic Reliability
Reliability Optimal demand and supply(VPP)
Optimal demand and supply(VPP) Reliability
Reliability
Peak Control
Peak Control Indirect
Indirect None
None VPP-based
VPP-based NoneNone
Generation Cost
Generation CostMinimization Indirect Indirect VPP-based IndirectIndirect
or Director
(BTB)
Direct
Outage Cost Minimization Indirect
None Indirect
Indirect VPP-based
None Indirect or Direct (BTB)
Minimization (BTB)
Indirect or Direct
Outage Cost Minimization None Indirect None
3. Configuration of System Structure for Multiple Microgrid (BTB)

This section describes

3. Configuration of Systemthe system for
Structure architecture
Multipleand the main components of system required for
Microgrid
the operation of multiple microgrids. It also defines the function and role of each system component
This section
according describes
to system the system architecture and the main components of system required for
structure.
the operation of multiple microgrids. It also defines the function and role of each system component
3.1. Structure
according of Multiple
to system Microgrid Systems
structure.
Multiple microgrid is the system consisting of several single microgrids. Multiple microgrid
3.1. Structure of Multiple Microgrid Systems
systems can be distinguished in two cases, depending on whether each microgrid owns one or only
Multiple
one owner formicrogrid is the system
the entire system. If thereconsisting of several
is a microgrid owner, single
eachmicrogrids.
microgrid can Multiple
perform microgrid
its own
systems
optimal can be distinguished
operation or performin two cases,
power sharingdepending
in the formon whether
of powereach microgrid
trading owns one
with another or only
microgrid.
one owner
It is also for thetoentire
possible system. Ifoperate
cooperatively there ismicrogrids
a microgrid forowner, each microgrid
the purpose can perform
of solving problems its own
on the grid
optimal operation
side, including or perform
congestion in power sharing
the utility grid.inOn thethe
form of power
other tradingthere
hand, when withisanother
a singlemicrogrid.
owner, one It
is also possible to cooperatively operate microgrids for the purpose of solving problems
medium-scale or large-scale power system is composed of multiple microgrids. In practice, it is similar on the grid
side,
to theincluding
operatingcongestion in themicrogrid,
form of a single utility grid.butOn the other
because hand, when
it manages there system
the whole is a single
in a owner, one
distributed
medium-scale or large-scale
manner, it can compensate forpower system isofcomposed
the drawbacks centralizedofoperation
multiplesuchmicrogrids. In practice,
as the processing it is
burden
similar to the operating
of operations. form
In both cases, of a single microgrid,
a hierarchical structure can butbebecause
utilizeditfor
manages
multiplethe whole system
microgrid in a
operations.
distributed
Figure 2 shows manner, it can compensate
a hierarchical structure forfor the drawbacks
multiple microgrids. of This
centralized operation
hierarchical suchconsists
structure as the
processing
of MoMC locatedburden in of the
operations. In both
upper layer andcases,
MG-EMSa hierarchical
located in structure can be utilized
each microgrid of the for multiple
lower layer.
microgrid
The distributedoperations.
operationFigure
of the 2entire
shows a hierarchical
system structure
is possible through for multiple
a cooperative microgrids.
operation This
or specific
hierarchical
coordinationstructure
between consists
MoMC and of MoMC
MG-EMS. located
In theinexisting
the upper layer and
centralized MG-EMSMoMC
operation, locatedmanages
in each
microgrid
and operates of the lower
all the layer.elements
internal The distributed operation of
of each microgrid, butthe entire
in the system distributed
proposed is possible through
operation, a
cooperative
MoMC has aoperation or specific function
simple coordination coordination between MoMC
and monitoring function and MG-EMS.
of the In the existing
whole system. MoMC
centralized operation,
basically monitors the MoMC
microgrid manages
states inand
theoperates
MMG systemall the and
internal
giveselements
orders toofeach
eachMG-EMS
microgrid, but
under
in the proposed
certain distributed
circumstances. Whenoperation,
the schedule MoMC has a simple
is determined, thecoordination
optimizationfunction
problemand monitoring
is solved in the
function of the whole system. MoMC basically monitors the microgrid states in the MMG system and
gives orders to each MG-EMS under certain circumstances. When the schedule is determined, the
optimization problem is solved in the MG-EMS of each microgrid, not the MoMC. Therefore, all the
conditions considered when the schedule is determined are handled in the MG-EMS. As a result, the
MoMC does not solve the large-scale optimization problem, but merely performs the simple
Energies 2017, 10, 1336 6 of 18

MG-EMS of each microgrid, not the MoMC. Therefore, all the conditions considered when the schedule
is determined are handled in the MG-EMS. As a result, the MoMC does not solve the large-scale
optimization problem, but merely performs the simple coordination by receiving the schedule obtained
in the MG-EMS and sends the coordinated result to the MG-EMS. In addition, in the distributed
operation mode, the plug and play function of the microgrid is easy to expand the system, and it is
Energies 2017, 10, 1336 6 of 19
easy to construct a multi-microgrid platform.

Figure 2. Hierarchical structure diagram of multiple microgrid.

Figure 2. Hierarchical structure diagram of multiple microgrid.

3.2. Definition
3.2. Definition of
of MoMC
MoMC and
and MG-EMS
MG-EMS Functions
Functions
Since the
Since the multiple
multiplemicrogrid
microgridsystem
systemhas hasa acommon
common hierarchical
hierarchical structure
structure forfor
both cases
both which
cases are
which
the grid-connected mode and the islanded mode, the MoMC in the
are the grid-connected mode and the islanded mode, the MoMC in the upper layer and the MG-EMSupper layer and the MG-EMS in
thethe
in lower layerlayer
lower have have
basically uniqueunique
basically functions and rolesand
functions regardless of the operation
roles regardless of themode, respectively.
operation mode,
respectively. Since the MG-EMS is located in each microgrid, the optimal schedule is determined each
Since the MG-EMS is located in each microgrid, the optimal schedule is determined only for only
microgrid.
for In the case
each microgrid. of the
In the casegrid-connected mode, the
of the grid-connected optimal
mode, schedule
the optimal is set foristhe
schedule set purpose of cost
for the purpose
reduction
of and the and
cost reduction schedule is set for is
the schedule theset
purpose
for theof improving
purpose the reliability
of improving the when operating
reliability wheninoperating
islanded
in islanded mode. As a measure of reliability, the cost of power outage is used and the schedule to
mode. As a measure of reliability, the cost of power outage is used and the schedule is configured is
minimize power
configured outage costs.
to minimize powerOne outageof the main
costs. functions
One of thefunctions
of the main MoMC isoftothe collect
MoMC the data associated
is to collect the
with the schedule
data associated with after
theeach microgrid
schedule schedule
after each has been
microgrid determined.
schedule has been Based on the collected
determined. Based ondata,
the
MoMC performs a specific coordination and assigns the data
collected data, MoMC performs a specific coordination and assigns the data reflecting reflecting the coordination to each
the
MG-EMS again.
coordination The data
to each MG-EMS exchanged
again.between
The dataMoMCexchangedand MG-EMS
between depends
MoMC and on the
MG-EMSmode of operation.
depends on
In the case of the grid-connected operation, the data to be exchanged is
the mode of operation. In the case of the grid-connected operation, the data to be exchanged is the the transaction power with
the utility grid
transaction and net-demand
power with the utility of each
gridmicrogrid system, and
and net-demand MoMC
of each coordinates
microgrid net-demand
system, and MoMC data
based on thenet-demand
coordinates peak penalty cost.
data On on
based thethe
other
peakhand, in the
penalty case
cost. On ofthe
theother
islandedhand, operation,
in the case theofdata
the
exchanged is the amount of reserve power or the amount of deficiency power
islanded operation, the data exchanged is the amount of reserve power or the amount of deficiency and net-demand data of
each microgrid,
power and thendata
and net-demand MoMC directly
of each coordinates
microgrid, and then theMoMC
net-demanddirectlydata by comparing
coordinates the reserve
the net-demand
power and the deficiency. Based on the modified data, each MG-EMS
data by comparing the reserve power and the deficiency. Based on the modified data, each MG-EMS determines its own schedule
again. The re-scheduling of the MG-EMS changes the power flow
determines its own schedule again. The re-scheduling of the MG-EMS changes the power flow and, consequently, the indirect
and,
power sharing.the
consequently, Table 2 shows
indirect power thesharing.
functionsTableand2 roles
showsofthe MoMC and MG-EMS
functions and roles of andMoMCthe criteria
and MG- for
power sharing.
EMS and the criteria for power sharing.

Table 2. Definition
Table 2. Definition of
of microgrid
microgrid of
of microgrids
microgrids (MoMC)
(MoMC) and
and microgrid
microgrid energy
energy management
management system
system
(MG-EMS) functions.
(MG-EMS) functions.
Operation Mode
Operation Mode Grid-Connected
Grid-Connected Islanded
Islanded
Agent
Agent MoMCMoMC MG-EMSMG-EMS MoMC MoMC MG-EMS
MG-EMS
Objectives
Objectives Economics (Cost Minimization)
Economics (Cost Minimization) Reliability (Energy Balance)
Reliability (Energy Balance)
Coordinator (Collect Coordinator (Collect
Role/Function Coordinator (Collect andOptimal Scheduling
Optimal Coordinator (Collect and
EmergencyEmergency
Scheduling
Role/Function and Broadcast, Adjust) and Broadcast, Adjust)
Broadcast, Adjust) Scheduling Broadcast, Adjust) Scheduling
Criterion Peak Penalty Cost Reserve or Deficient Power
Criterion Peak Penalty Cost Reserve or Deficient Power
Exchanged Power to trade with Utility Grid/Net-Demand
Reserve or Deficient Power/Net-Demand Data
Exchanged
Information Power toData
trade with Utility
Reserve or Deficient Power/Net-Demand Data
Information Grid/Net-Demand Data

4. Power Sharing Algorithm for Grid-Connected Operation

This section describes the power-sharing algorithm for the grid-connected operation of multiple
microgrids. When operated as grid-connected, the amount of power sharing is indirectly calculated
by scheduling coordination rather than directly calculating the amount of power sharing. In other
words, the schedule of each microgrid is changed through adjustment at the upper level based on the
Energies 2017, 10, 1336 7 of 18

4. Power Sharing Algorithm for Grid-Connected Operation

This section describes the power-sharing algorithm for the grid-connected operation of multiple
microgrids. When operated as grid-connected, the amount of power sharing is indirectly calculated
by scheduling coordination rather than directly calculating the amount of power sharing. In other
words, the schedule of each microgrid is changed through adjustment at the upper level based on
the individual optimal schedule in each MG-EMS. Then, the difference from the individual optimal
schedule is determined as the power shares. First of all, the difference between the optimization and
power sharing algorithm of the microgrid with convex form of objective function and the algorithm
proposed in this paper is described, and the proposed algorithm will then be described in detail.

4.1. Differences in the Operating Conditions of the Proposed Algorithm

Generally, the components that consist of the microgrid are generators, the battery energy storage
system (BESS), renewable energy sources, and loads. When calculating the optimal schedule of
controllable distributed sources, such as generators or BESS, the objective function of these components
is usually assumed to be a convex function, such as the quadratic function. If the objective function
is quadratic, it can be regarded as an economic dispatch problem, which means that the solution to
the problem is the coincidence of the incremental cost of the cost function. In the case of power sharing,
and considering the general multi-microgrid situation, it means converging all of the incremental
costs of each power source into the same value. Within the hierarchical operating structure described
above, all MG-EMSs build their respective schedules and send the individual incremental cost to
the higher-level MoMC. The incremental cost values collected from all MG-EMS to MoMC are adjusted
by the MoMC, and the adjusted values are again assigned to each MG-EMS. The MG-EMS coordinates
its schedule through the allocated values. As the process is repeated, the output of each microgrid is
changed, and the amount of power sharing is determined according to the changed output. The details
are presented in our previous studies [18].
Unlike the situation mentioned above, the algorithm proposed in this paper deals with the case
where there is no convex objective function. In other words, there is no distributed source such as
a generator, and only renewable sources and BESS exist. Also, in the case of BESS, it does not have any
convex cost function, such as the life-cycle cost of the batteries. A description of the variables used in
the following equations is summarized in Table 3.

Table 3. A description of the variable notation.

Variable Description Variable Description
Cgrid Price of Utility grid PLoad (t) Predicted Load at time t
Output Power of Renewable Energy
C penalty Penalty Cost Coefficient PRES (t)
Sources at time t
Power Supplied from Utility grid at Net-Demand value at time t
Pgrid (t) Pd (t)
time t (PLoad (t) PRES (t))
limit Predefined Limited Power of Power Margin to the Limited Power Value
Pgrid Pmar (t)
Supplied from Utility at time t
Excess from the Limited Power
Pbat,c (t) Charge Power of BESS at time t Pexc (t)
Value at time t
max Average Value of Margin and Excess
Pbat,c Maximum Charge Power of BESS Pavg (t)
at time t
Pbat,d (t) Discharge Power of BESS at time t Pshare (t) Power Sharing Value at time t
max
Pbat,d Maximum Discharge Power of BESS Pd,cor (t) Coordinated Net-Demand at time t
Energies 2017, 10, 1336 8 of 18

Table 3. Cont.
Variable Description Variable Description
Total Cost of MMG System at the
SOC (t) State of Charge of BESS at time t TCost(k )
iteration step k
SOCmin Minimum Constraints of SOC Scale factor of penalty function
SOCmax Maximum Constraints of SOC t Time step (1 h)
ud (t) Discharge State of BESS Convergence Criterion
The index of the microgrid that has
uc (t) Charge State of BESS n
the margin up to the limit
The index of the microgrid that has
c Efficiency of BESS when charging m
exceeded the limit
The number of the microgrid that
d Efficiency of BESS when discharging N
has the margin up to the limit
The number of the microgrid that
BESScap Capacity of BESS M
has exceeded the limit

4.2. Optimization in MG-EMS

Since there is no component that has the incremental cost, it is impossible to adjust the schedule
based on the change of the incremental cost. Therefore, the main objective of BESS is set to achieve
peak shaving and peak control by minimizing the power supplied from the system in accordance with
the peak load and utility price. An arbitrary limit value is set for the power supplied from the utility
according to the users intention, and the penalty cost is set to be paid when the supplied power
exceeds the limit value. This penalty cost may be regarded as the increasing fundamental cost based
on the peak power increased, and in this paper, the penalty cost is set equivalent to 10 times the utility
price. As will be described later, the power sharing is determined according to the set limit values.
As shown in Equation (1), the objective function is set to obtain the schedule for 24 h at intervals of 1 h.
The objective function includes terms for minimizing the cost of power supplied from the grid and
minimizing the peak increase.
24 1 (P limit
f ( Pgrid ) = [Cgrid (t) Pgrid (t) + Cpenalty e grid ( t ) Pgrid )
] (1)
t =1

limit are the utility price, the power supplied from the utility grid,
Cgrid , Pgrid , C penalty and Pgrid
the penalty cost, and the preset power limit value, respectively. The penalty cost was set at
10 times the utility price. represents a scale factor, which determines the degree of weighting
for the penalty terms.
The decision variable in the optimization problem is Pgrid , the power supplied by the grid,
but another variable determined by the constraints is the output of BESS. That is, as a result
of the scheduling, the discharge power or the charge power of BESS and Pgrid are determined.
Therefore, constraints must be constructed for this decision variable. However, since Pgrid can be
determined depending on the output of BESS, constraints include those for BESS. The BESS constraint
includes the output constraints for charging or discharging and the constraints on state-of-charge
(SOC), as well as state constraints that indicate states for charging and discharging.
max
0 Pbat,c (t) uc (t) Pbat,c (2)

max
0 Pbat,d (t) ud (t) Pbat,d (3)

SOCmin SOC (t) SOCmax (4)

c Pbat,c (t) Pbat,d (t)/d
SOC (t) = SOC (t 1) + t (5)
BESScap
(
ud = 1, when discharging
ud (t) + uc (t) 1, (6)
uc = 1, when charging
Energies 2017, 10, 1336 9 of 18

Pgrid (t) + Pbat,d (t) = Pd (t) + Pbat,c (t) (7)

where,
Pd (t) = PLoad (t) PRES (t)

Equations (2) and (3) are constraints on the battery output, and Pbat,c and Pbat,d denote the charging
power of the battery and the discharging power of the battery, respectively. Equation (4) is the upper
and lower limit for SOC, and Equation (5) is the equality constraint for SOC. In Equation (5), c and
d refer to charging and discharging efficiency, respectively, and BESScap refers to battery capacity.
Also, t means time-step, which is set to 1 h generally. Equation (6) is the inequality constraint that
indicates the state of the battery. If ud is one in Equation (6), the battery performs discharging. If uc is
one, the battery performs charging, and there is no case in which ud and uc become equal. In addition
to the constraints of BESS, Equation (7) also includes general equality constraints on the balance
between generation and demand. In Equation (7), Pd means net-demand, which is the value obtained
by subtracting the renewable power PRES from the predicted load value PLoad . With the above objective
function and constraints, each MG-EMS solves its own optimization problem and sends the data
resulted in Pgrid and Pd to the MoMC. Since the objective function is basically a nonlinear function,
the optimization problem to be solved in MG-EMS is a problem defined as nonlinear programming.
Furthermore, since the charge/discharge status of the battery must be determined, binary variables
consisting of 0 and 1 must be included. Therefore, this problem can be considered as a mixed integer
nonlinear programming problem and can be solved by NLP solver such as IPOPT or Bonmin [19,20].

4.3. Coordination in MoMC

In the MoMC, the coordination algorithm is performed using Pgrid and the Pd data of each
microgrid received from MG-EMS. It is assumed that the MoMC already knows information about
the peak limits of all microgrids. As shown in Equations (8) and (9), the excess power value Pexc
and the margin power value Pmar for each time interval can be obtained based on the limit values as
threshold values. n is the index of the microgrid that has the margin up to the limit, and m is the index
of the microgrid that has exceeded the limit.
limit
Pmar,n (t) = Pgrid,n Pgrid,n (t) (8)

limit
Pexc,m (t) = Pgrid,m (t) Pgrid,m (9)

N M
Pmar,n (t) + Pexc,m (t)
n =1 m =1
Pavg (t) = (10)
N+M
Equation (10) shows the calculation of the average power Pavg by using the obtained margin
power and excess power. Here, the average value can be used to determine the criteria for
power sharing and the power sharing value can be calculated according to this criterion as in
Equations (11) and (12).
Pshare,m (t) = Pavg (t) Pexc,m (t) (11)

Pshare,n (t) = Pavg (t) Pmar,n (t) (12)

In Equation (13), the power sharing value Pshare determined for each time slot is applied to the Pd
data received from each MG-EMS and then allocated to the MG-EMS.
Pd,cor (t) = Pd (t) + Pshare (t) (13)

The coordinated Pd,cor assigned to the MG-EMS is used to solve the optimization problem defined
above. The result is sent back to the MoMC and the same procedure described so far is repeated until
the difference between the total cost of the current step and the previous step is within the tolerance.
 Pd ,cor (t) = Pd (t) + Pshare (t) (13)

The coordinated Pd,cor assigned to the MG-EMS is used to solve the optimization problem
Energies 2017, 10, 1336 10 of 18
defined above. The result is sent back to the MoMC and the same procedure described so far is
repeated until the difference between the total cost of the current step and the previous step is within
If thethe tolerance. If the iteration is terminated, the schedule of the power supplied from the grid and the
iteration is terminated, the schedule of the power supplied from the grid and the ESS in each
ESS in each microgrid are finally determined.
microgrid are finally determined.
The processes described so far are summarized in the flow chart of Figure 3. In Figure 3,
The processes described so far are summarized in the flow chart of Figure 3. In Figure 3, TCost(k )
TCost (the
represents k ) total
represents
cost ofthethe
total cost of the
multiple multiplesystem
microgrid microgrid system
at the k th at theand
step
kth step
means and asmall
means a
number.
In thesmall
MoMC, number. In thetransaction
the utility MoMC, thepowerutilityexceeding
transaction thepower exceeding
limit value set inthe limit
each value setisin
microgrid each
distributed
microgrid is distributed to another microgrid, and the coordination is carried out
to another microgrid, and the coordination is carried out in the net-demand. That is, the coordinated in the net-demand.
That is, the coordinated schedule means the result of power sharing between microgrids. In this
schedule means the result of power sharing between microgrids. In this paper, the results of power
paper, the results of power sharing when the schedule is planned through penalty terms are shown.
sharing when the schedule is planned through penalty terms are shown. However, the proposed
However, the proposed algorithm is effective without penalty terms. Even if a real-time schedule is
algorithm
appliedis effective
instead without penalty
of a full-day terms. Even
schedule, the if a real-time
algorithm is schedule
considerably is applied
effective instead of a the
because full-day
schedule,
computation burden is small and the time required for obtaining the result is small. The verification time
the algorithm is considerably effective because the computation burden is small and the
required foralgorithm
of the obtaining forthe result
these is small.
parts, includingThethe
verification of the of
basic verification algorithm for these
the proposed parts,isincluding
algorithm, also
the basic verification
presented of the
in the next proposed algorithm, is also presented in the next section.
section.

Figure 3. The flowchart of the proposed algorithm.

Figure 3. The flowchart of the proposed algorithm.
5. Numerical Results
5. Numerical Results
In this section, algorithm verification is performed on multiple microgrids. The multiple
microgrid
In system
this section, can be assumed
algorithm to have
verification a system structure
is performed with amicrogrids.
on multiple PCC as shown Theinmultiple
Figure 4a,microgrid
or a
system structure with multiple feeders as shown in Figure 4b. In the case of multiple
system can be assumed to have a system structure with a PCC as shown in Figure 4a, or a system feeder systems
like Figure
structure with 4b, there must
multiple be a separate
feeders as shownline in
forFigure
power sharing.
4b. In Inthethiscase
case,of
themultiple
connections between
feeder systems
the microgrids must be electrically isolated through the power electronics in order to prevent the
like Figure 4b, there must be a separate line for power sharing. In this case, the connections
between the microgrids must be electrically isolated through the power electronics in order to
prevent the problem of system stability, such as protection cooperation according to the ring topology.
The number of the microgrid set is three, to verify the basic performance of the algorithm. As mentioned
earlier, each microgrid has only renewable energy sources and battery energy storage systems because
the microgrid that does have the units without the cost function is focused. For simplicity of analysis,
each microgrid was assumed to have one photovoltaic (PV) generation and one battery energy storage
system (BESS). The PV curve of each microgrid is shown in Figure 5. The system price parameter
is based on PJM market data [21], and Figure 6 shows the price curve. The load curves are shown
in Figure 7 and the load capacities are summarized in Table 4. The parameters such as the rated output
 problem of system stability, such as protection cooperation according to the ring topology. The
problem
number of system
of the microgridstability, such as
set is three, to protection cooperation
verify the basic according
performance to algorithm.
of the the ring topology. The
As mentioned
number of the microgrid set is three, to verify the basic performance of
earlier, each microgrid has only renewable energy sources and battery energy storage systems the algorithm. As mentioned
earlier,the
because each microgrid
microgrid thathasdoesonly renewable
have the unitsenergy
withoutsources
the costand batteryisenergy
function storage
focused. systems of
For simplicity
because the microgrid that does have the units without the cost function
analysis, each microgrid was assumed to have one photovoltaic (PV) generation and one battery is focused. For simplicity of
Energies analysis,
10, each microgrid was assumed to have one photovoltaic (PV) generation
energy storage system (BESS). The PV curve of each microgrid is shown in Figure 5. The system price11 of 18
2017, 1336 and one battery
energy storage
parameter is basedsystem
on PJM (BESS).
marketThedata
PV curve
[21], of
andeach microgrid
Figure is shown
6 shows in Figure
the price curve.5.The
Theload
system priceare
curves
parameter is based on PJM market data [21], and Figure 6 shows the price curve. The load curves are
shown in Figure 7 and the load capacities are summarized in Table 4. The parameters such as the
power shown
and theinrated
Figurecapacity
7 and the load thecapacities are summarized in Table 4. The parameters suchand
as the
rated output power and theof rated power sources
capacity are set
of the power appropriately
sources for the loads
are set appropriately are listed
for the loads
in Table rated
4 output
together power
with and
the the
other rated capacity
parameter of the
values.power
All sources are
simulations set appropriately
are done with for
1-hthe loadsstep size
time
and are listed in Table 4 together with the other parameter values. All simulations are done with 1-h
and are listed in Table 4 together with the other parameter values. All simulations are done with 1-h
and all
timeparameters
step size and setall
above can beset
parameters changed
above can according
be changedto the systemto
according configuration. The simulation
the system configuration.
time step size and all parameters set above can be changed according to the system configuration.
TheThe
was also simulation
performed was also
using a performed
computer using
with a computer
an Intel with
Core an Intel
i7-6800K Core i7-6800K
3.4GHz
simulation was also performed using a computer with an Intel Core i7-6800K 3.4GHz CPU CPU 3.4GHz
and CPU
32GB andand
memory
32GB
and MATLAB32GB memory and MATLAB (2016a, Mathworks, Natick, MA, USA). Bonmin was used as an an and
memory
(2016a, and MATLAB
Mathworks, (2016a,
Natick, Mathworks,
MA, USA). Natick,
Bonmin MA,
was USA).
used Bonmin
as an was used
optimization as
solver
optimization
the settings for the
optimization solver and
solver
solver the
are
and settings
settingsfor
summarized
the for theinsolver
the are
Tableare
solver summarizedininTable
5. summarized Table5.5.

Figure 4. Examples of multiple microgrid system architecture for application of the proposed
Figure
Figure 4. 4. Examples
Examples
algorithm. (a) of of multiple
multiple
Multiple microgrid
microgrid
microgrids withsystemsystem architecture
architecture
a single feeder; for application
forMultiple
and (b) application of theofproposed
microgrids themultiple
with proposed
algorithm.
algorithm. (a) Multiple microgrids with a single feeder; and (b) Multiple microgrids
(a) Multiple microgrids with a single feeder; and (b) Multiple microgrids with multiple feeders.
feeders. with multiple
feeders.

70

70
60

60
50
Power [kW]

50 40
PV1
Power [kW]

40 30
PV2PV1

30 20 PV3PV2

20 10 PV3

10 0
0 5 10 15 20 25

Time of the day [hour]

0
0 5 10 15 20 25

Time of the day [hour]

Energies 2017, 10, 1336 12 of 19
Figure 5. Renewable
Figure 5. Renewablegeneration photovoltaic
generation photovoltaic (PV)
(PV) profile.
profile.

Figure
0 .6 5. Renewable generation photovoltaic (PV) profile.
i e Electricity price [$/kWh]

0 .5

0 .4

0 .3

0 .2
Real-t m

0 .1

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1 7 18 19 20 21 22 23 24
Time of the day (hour)

Figure6.6. Real-time
Figure priceprofile.
Real-time price profile.

45 0
40 0
35 0
30 0
Power [kW]

25 0
20 0
15 0
 Re
0
1 2 3 4 5 6 7 8 9 10 11 12 13 1 4 15 16 17 18 19 20 21 22 23 24
Time of the day (hour)

Energies 2017, 10, 1336 12 of 18

Figure 6. Real-time price profile.

45 0
40 0
35 0
30 0

Power [kW]
25 0
20 0
15 0
10 0
50
0
1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 13 14 15 16 1 7 1 8 19 20 21 22 23 24
Time of the day [hour]

MG1 MG2 MG3

Figure
Figure 7.
7. Load
Loadprofiles
profiles of
of each
each microgrid.
microgrid.

Table
Table 4.
4. Parameter
Parameter setting
setting of
of each
each microgrid.
microgrid.
Parameter (Unit) MG1 MG2 MG3
ParameterPeak
(Unit)
Load (kW) MG1
300 500MG2 350 MG3
Peak Rated
Load Power
(kW) of PV (kW) 30025 60 500 30 350
Rated Output
Rated of battery
Power ofenergy
PV (kW)storage system (BESS) (kW) 25180 500 60 270 30
Capacity of BESS
Rated Output of battery energy storage system (kWh) (BESS) (kW) 180180 2000500 540 270
Capacity of BESS (kWh) (%)
Efficiency of BESS 18095 97 2000 96 540
Upper of
Efficiency limit of BESS
BESS (%) output (kW) 95180 500 97 270 96
Lower limit of BESS
Upper limit of BESS output (kW) output (kW) 1800 0 500 0 270
Upper limit of
Lower limit of BESS output (kW)BESS f (p.u.) 00.9 0.9 0 0.9 0
Lower limit of BESS state-of-charge
Upper limit of BESS f (p.u.) (SOC) (p.u.) 0.90.2 0.2 0.9 0.2 0.9
Initial and end SOC of BESS (p.u.)
Lower limit of BESS state-of-charge (SOC) (p.u.) 0.20.5 0.4
0.2 0.4 0.2
Predefined Power Limit Value (kW)
Initial and end SOC of BESS (p.u.) 0.5245 550
0.4 200 0.4
Penalty Cost Coefficient 10 10 10
Predefined Power Limit Value (kW) 245 550 200
Scale factor 81.67 183.33 100
Penalty Cost Coefficient 10 10 10
Time step (h) 1 1 1
Scale factor 81.67 183.33 100
The Number of Microgrids (N + M) 3
Time step (h) 1 1 1
Convergence Criterion 0.01
The Number of Microgrids (N + M) 3
Convergence Criterion 0.01

Table 5. Parameter settings for the solver.

Parameter Description Values

Maxiter The number of maximum iteration 1500
Maxfeval The number of maximum function evaluation 10,000
Maxnodes The number of maximum nodes for the mixed integer solver 10,000
Maxtime The number of maximum execution time of the solver 1000
Tolrfun The desired relative convergence tolerance of the solver 1.0 107
Tolafun The desired absolute convergence tolerance of the solver 1.0 107
Tolint The absolute tolerance used to define whether a solution is an integer value 1.0 105

Figure 8 shows the daily schedule results of the power supplied from the utility grid for each
microgrid. The solid bars are the results before the coordination are applied, the shaded bars show
the results of the coordination applied, and the dotted line is the predetermined limit value for each
microgrid. In the figure of the schedule without the coordination, in the case of MG1, the schedule is
configured to be smaller than the limit value at all times, and only the amount of the power transaction
with utility grid at 1:00, 2:00, and 24:00 are exceeded the limits in MG2, and the limit was exceeded in
most of the times except for 12:00 to 15:00 in MG3. This result shows that MG1 has some margin in all
times until the limit is exceeded and MG2 has a margin in most time periods, including peak time, and
microgrid. In the figure of the schedule without the coordination, in the case of MG1, the schedule is
configured to be smaller than the limit value at all times, and only the amount of the power
transaction with utility grid at 1:00, 2:00, and 24:00 are exceeded the limits in MG2, and the limit was
exceeded in most of the times except for 12:00 to 15:00 in MG3. This result shows that MG1 has some
Energies 2017,
margin 10,times
in all 1336 until the limit is exceeded and MG2 has a margin in most time periods, including 13 of 18

peak time, and MG3 needs the power sharing from other MGs in the majority of the times. In other
words,
MG3 needs sincetheMG3powerpays a penalty
sharing fromfor exceeding
other MGs in the the limit,
majority it has a cost
of the loss.
times. InThe
otherexpected lossesMG3
words, since are
resolved somewhat through power sharing between the microgrids, as shown
pays a penalty for exceeding the limit, it has a cost loss. The expected losses are resolved somewhat in Figure 8 (shaded
bars).
through Since eachsharing
power microgrid determines
between the schedule
the microgrids, asinshown
consideration
in Figure of8the price curve
(shaded bars).ofSince
the utility
each
system, the result of the coordination in MoMC can be interpreted as follows.
microgrid determines the schedule in consideration of the price curve of the utility system, the result(1) Net-demand is smallof
in
themost of the time
coordination including
in MoMC peak
can be time, so as
interpreted that MG2, (1)
follows. which has enough
Net-demand margin
is small fromofthe
in most thelimit,
time
changes
including the
peakschedule
time, soandthat
shares
MG2,some
whichor all
hasofenough
the power supplied
margin from from 3:00 to
the limit, 23:00 to
changes other
the MGs
schedule
as power sharing; (2) in the case of MG1, it is supplied from the grid at
and shares some or all of the power supplied from 3:00 to 23:00 to other MGs as power sharing; the lowest price of 24:00, 1:00,
and
(2) in2:00,
the and
case is ofshared
MG1, itwith other MGs;
is supplied from(3) ingrid
the caseatofthe
MG3, sinceprice
lowest it exceeds
of 24:00,the1:00,
limitand
in most of the
2:00, and is
time, it does not pay the penalty cost at all times except 24:00, 1:00, and 2:00
shared with other MGs; (3) in case of MG3, since it exceeds the limit in most of the time, it does not through power shared
from
pay theMG1 and MG2.
penalty cost atFigure 9 shows
all times exceptthe state1:00,
24:00, of power
and 2:00 sharing
throughby time
power forshared
each microgrid.
from MG1Each and
microgrid constructs a schedule according to its own limit line and calculates
MG2. Figure 9 shows the state of power sharing by time for each microgrid. Each microgrid constructs a value for power
sharing
a schedule by according
varying itstoschedule
its own according
limit line and to coordination
calculates a valuefrom the for MoMC, and performs
power sharing by varyingpower its
sharing according to these values. When the schedule is changed through
schedule according to coordination from the MoMC, and performs power sharing according to these coordination, the schedule
of the ESSs
values. When belonging to eachismicrogrid
the schedule also changes.
changed through Figures 10
coordination, theand 11 show
schedule ofthe
thechange of SOC and
ESSs belonging to
the output of each ESS.
each microgrid also changes. Figures 10 and 11 show the change of SOC and the output of each ESS.

350 800
Power transaction with Utility Grid[kW]

Power transaction with Utility Grid[kW]

300 700

250 600
500
200
400
150
300
100 200
50 100
0 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time of the day [hour] Time of the day [hour]

Without Coordination With Coordination Limit Without Coordination With Coordination Limit

Energies 2017, 10, 1336 14 of 19

(a) (b)

400
Power transaction with Utility Grid[kW]

350
300
250
200
150
100
50
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time of the day [hour]

Without Coordination With Coordination Limit

(c)
Figure 8.
Figure Power transaction
8. Power transaction with
with utility
utility grid
grid (a)
(a) MG1,
MG1, (b)
(b) MG2,
MG2, (c)
(c) MG3.
MG3.

300 500
The amount of the Power Sharing[kW]

The amount of the Power Sharing[kW]

400
200
300
100
200
0 100
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 0
-100
-100 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
-200
-200
-300 -300
Time of the day [hour] Time of the day [hour]

Sharing (Send) Sharing (Receive) Sharing (Send) Sharing (Receive)

(a) (b)

0
W]

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

trans
PowerPower
0
1 2 3 4 5 6 7 8 Time
9 10 of
11the day
12 13 14[hour]
15 16 17 18 19 20 21 22 23 24
Time of the day [hour]
Without Coordination With Coordination Limit

Without Coordination With Coordination Limit

(c)
Energies 2017, 10, 1336 14 of 18
(c)
Figure 8. Power transaction with utility grid (a) MG1, (b) MG2, (c) MG3.
Figure 8. Power transaction with utility grid (a) MG1, (b) MG2, (c) MG3.

300 500
Sharing[kW]

Sharing[kW]
300 400
500
200
Sharing[kW]

Sharing[kW]
300
400
200
100
200
300
the Power

the Power
1000 100
200
of theofPower

of theofPower
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
0
-100 1000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 -100 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
0
The amount

The amount
-100
-200
-200
-100 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
The amount

The amount
-200
-300 -300
-200
Time of the day [hour] Time of the day [hour]
-300 -300
Time of the day [hour] Time of the day [hour]
Sharing (Send) Sharing (Receive) Sharing (Send) Sharing (Receive)

Sharing (Send) Sharing (Receive) Sharing (Send) Sharing (Receive)

(a) (b)
(a) (b)
0
Sharing[kW]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
0
-50
Sharing[kW]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
-50
-100
the Power

-100
of theofPower

-150

-150
The amount

-200
The amount

-200
-250
Time of the day [hour]
-250
Time of the day [hour]
Sharing (Send) Sharing (Receive)

Sharing (Send) Sharing (Receive)

(c)
(c)
Figure 9. The amount of the power sharing in each microgrid (a) MG1, (b) MG2, (c) MG3.
Figure 9. The amount of the power sharing in each microgrid (a) MG1, (b) MG2, (c) MG3.
Figure 9. The amount of the power sharing in each microgrid (a) MG1, (b) MG2, (c) MG3.
1 1

0.8 0.8
SOC [P.U]

SOC [P.U]

1 1

0.6
0.8 0.6
0.8
SOC [P.U]

SOC [P.U]
Battery

Battery

0.4
0.6 0.4
0.6

0.2 0.2
Battery

Battery

0.4 0.4

0.20 0.20
1 2 3 4 5 6 7 8 9 10111213141516171819202122232425 1 2 3 4 5 6 7 8 9 10111213141516171819202122232425
0 0
1 2 3 4 5 6 7 8 9Time of the day [hour]
10111213141516171819202122232425 1 2 3 4 5 6 7 8 9Time of the day [hour]
10111213141516171819202122232425
Time of the day [hour] Time of the day [hour]
Without Coordination With Coordination Without Coordination With Coordination

Without Coordination With Coordination Without Coordination With Coordination

Energies 2017, 10, 1336 (a) (b) 15 of 19

(a) (b)

0.8
Battery SOC [P.U]

0.6

0.4

0.2

0
1 2 3 4 5 6 7 8 9 10111213141516171819202122232425
Time of the day [hour]

Without Coordination With Coordination

(c)

Figure 10. Battery SOC in each microgrid (a) MG1, (b) MG2, (c) MG3.
Figure 10. Battery SOC in each microgrid (a) MG1, (b) MG2, (c) MG3.

80 600
60
40 400
Output Power [kW]

Output Power [kW]

20
0 200
-20 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 0
-40
-60 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
-200
-80
-100 -400
-120
-140 -600
Time of the day [hour] Time of the day [hour]
 0
1 2 3 4 5 6 7 8 9 10111213141516171819202122232425
Time of the day [hour]

Without Coordination With Coordination

Energies 2017, 10, 1336 (c) 15 of 18

Figure 10. Battery SOC in each microgrid (a) MG1, (b) MG2, (c) MG3.

80 600
60
Output Power [kW] 40 400

Output Power [kW]

20
0 200
-20 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 0
-40
-60 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
-200
-80
-100 -400
-120
-140 -600
Time of the day [hour] Time of the day [hour]

Without Coordination With Coordination Without Coordination With Coordination

(a) (b)

150
100
Output Power [kW]

50
0
-50 1 2 3 4 5 6 7 8 9 10 11 12 13 1415 16 17 1819 20 21 22 2324
-100
-150
-200
-250
Time of the day [hour]

Without Coordination With Coordination

(c)

Figure
Figure 11.11.Battery
Batteryoutput
output in each
eachmicrogrid
microgrid(a)(a)
MG1, (b)(b)
MG1, MG2, (c) MG3.
MG2, (c) MG3.

TheThe change
change in in theschedule
the schedule of
of each
eachmicrogrid
microgrid through
throughthisthis
coordination is dueistodue
coordination the change
to the ofchange
of the net-demand shown in Figure 12. The MoMC calculates the amount of the powerand
the net-demand shown in Figure 12. The MoMC calculates the amount of the power sharing sharing
and reflects
reflectsthese
thesevalues
valuesto to
thethe
net-demand
net-demandof each
of MG;
eachitMG;
delivers the coordinated
it delivers net-demand
the coordinated to each
net-demand to
MG-EMS, and then the MG-EMS uses the coordinated net-demand to change its own schedule.
each MG-EMS, and then the MG-EMS uses the coordinated net-demand to change its own schedule.
Therefore, the amount of the power sharing shown in Figure 9 reflects the change of the power
Therefore, the amount of the power sharing shown in Figure 9 reflects the change of the power schedule
schedule supplied from the utility grid and the change of the ESS schedule.
Energies 2017, 10, 1336 16 of 19
supplied from the utility grid and the change of the ESS schedule.

350 800
300 700
Net Demand [kW]

Net Demand [kW]

250 600
500
200
400
150
300
100 200
50 100
0 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time of the day [hour] Time of the day [hour]

Original Net Demand Coordinated Net Demand Original Net Demand Coordinated Net Demand

(a) (b)

300
250
Net Demand [kW]

200
150
100
50
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time of the day [hour]

Original Net Demand Coordinated Net Demand

(c)

Figure 12. Changes in net-demand for each microgrid (a) MG1, (b) MG2, (c) MG3.
Figure 12. Changes in net-demand for each microgrid (a) MG1, (b) MG2, (c) MG3.
Table 6 shows the cost for each microgrid and the total MMG according to the conditions. The
cost results before and after the coordination are applied, and the cost results of the centralized
method considering all the microgrids are compared. This table shows two things. First, the results
before and after the coordination is applied are different for each microgrid. In the case of MG1 and
MG2, the cost was reduced due to coordination and the cost for MG2 increased. However, the cost of
Energies 2017, 10, 1336 16 of 18

Table 6 shows the cost for each microgrid and the total MMG according to the conditions. The cost
results before and after the coordination are applied, and the cost results of the centralized method
considering all the microgrids are compared. This table shows two things. First, the results before and
after the coordination is applied are different for each microgrid. In the case of MG1 and MG2, the cost
was reduced due to coordination and the cost for MG2 increased. However, the cost of the entire
MMG is reduced by about 1.34% when the coordination is applied. In other words, the cost of some
microgrids will increase, but other microgrids will benefit, which will benefit the entire MMG system.
Secondly, the difference between the cost of the coordinated MMG and the cost of the centralized
method is only about 0.3%. That is, even if an optimal schedule is constructed in a distributed form, it
is close to the result of the centralized method and achieves a near optimum.
Table 6. A summary of the results comparison with penalty term.

without with Coordination Centralized Cost

Subject Cost Gap (%)
Coordination ($) ($) Method ($) Reduction (%)
MG1 $1587.52 $1265.76 $1204.68 20.27% 5.07%
MG2 $2390.90 $3760.05 $3830.86 57.27% 1.85%
MG3 $2033.91 $905.85 $878.47 55.46% 3.12%
MMG $6012.30 $5931.70 $5914.00 1.34% 0.30%

Table 7 shows the results of applying the same algorithm, except for the penalty term, in
the objective function defined in the previous section. Since the penalty was not applied here, the MMG
cost after coordination was about 0.16% higher. The proposed algorithm stops the iteration when
the difference of total cost in the iteration is within a convergence criterion. Since it is set to 0.01,
the error can be canceled if it is set to a smaller value. As in Table 5, the difference from the centralized
method is about 0.3% in Table 6, which can be regarded as having an approximate optimal value.
Table 7. A summary of the results comparison without penalty term.

Without With Centralized Cost

Subject Cost Gap (%)
Coordination ($) Coordination ($) Method ($) Reduction (%)
MG1 $1537.90 $1224.80 $1751.50 20.36% 30.07%
MG2 $2351.00 $3704.80 $2086.90 57.58% 77.53%
MG3 $1893.80 $862.08 $1936.60 54.48% 55.48%
MMG $5782.70 $5791.68 $5775.00 0.16% 0.29%

Figure 13 shows the computation time as the number of microgrids increases. As a result,
the conventional central optimization method increases the computation time dramatically as
the number of microgrids increases. On the other hand, in the case of the proposed algorithm,
the computation time does not increase significantly even if the number of microgrids increases.
When the proposed algorithm is applied, it can be achieved within a few minutes because it is
computed in parallel and in a distributed manner. As a result of the analysis under the same solver,
the computation time was about 66 s for the seven microgrids, whereas the conventional centralized
method took about 21 h. Even if considering the actual data communication time, the required time is
within a few minutes. Therefore, the proposed algorithm is more effective because it does not increase
the burden of computation even if the number of microgrids increases in multiple microgrid systems.
 the proposed algorithm is applied, it can be achieved within a few minutes because it is computed in
parallel and in a distributed manner. As a result of the analysis under the same solver, the
computation time was about 66 s for the seven microgrids, whereas the conventional centralized
method took about 21 h. Even if considering the actual data communication time, the required time
is within a few minutes. Therefore, the proposed algorithm is more effective because it does not
increase
Energies the
2017, 10, burden of computation even if the number of microgrids increases in multiple microgrid17 of 18
1336
systems.

1400

1200

Computational time [min]

1000

800

600

400

200

0
1 3 5 7

The number of MG

Centralized method Proposed algorithm

Figure 13. Computation

Figure time
13. Computation timeaccording
according to
to the numberofofmicrogrids
the number microgrids using
using centralized
centralized method.
method.

6. Conclusions
6. Conclusions
In this
In this paper,
paper, a distributedcoordination
a distributed coordination strategy
strategy isisproposed
proposed between
between microgrids
microgridsfor the
foroptimal
the optimal
operation of the whole system under a hierarchical multiple microgrid system. The optimal operation
operation of the whole system under a hierarchical multiple microgrid system. The optimal operation
of the entire microgrid system is achieved through the determination of individual schedules in each
of the entire microgrid system is achieved through the determination of individual schedules in each
microgrid and the iterative execution of coordination is performed in terms of the overall system. The
microgrid
proposedandalgorithm
the iterative execution
is verified of coordination
by simulation is performed
using MATLAB. in terms
As a result ofsimulation,
of the the overallit system.
is
The proposed
shown that the total cost has been reduced by paying less penalty cost after coordinationsimulation,
algorithm is verified by simulation using MATLAB. As a result of the at MoMC. it is
shown that thewith
Compared totalthe
cost has of
results been reduced
applying the by paying method,
centralized less penalty cost afterbetween
the difference coordination
the totalatcost
MoMC.
Compared with thealgorithm
of the proposed results ofandapplying thecentralized
that of the centralized method,
method thethan
is less difference between
1%, regardless thepenalty
of the total cost of
term of the
the proposed objectiveand
algorithm function
that of of the
eachcentralized
microgrid. method
This means thatthan
is less the 1%,
proposed powerofsharing
regardless the penalty
algorithm through the coordination is valid without a special penalty cost
term of the objective function of each microgrid. This means that the proposed power sharing function, which is practical
algorithm
in multiple
through microgrids is
the coordination consisting of onlyaESSs
valid without or renewable
special energy
penalty cost resources
function, without
which an energy
is practical incost
multiple
function. Finally, comparing the computation time with the increase of the number of microgrids
microgrids consisting of only ESSs or renewable energy resources without an energy cost function.
shows that the proposed algorithm has a considerable advantage in terms of computational burden
Finally, comparing the computation time with the increase of the number of microgrids shows that
and time required. Therefore, this algorithm can be applied not only to determine the day-ahead
the proposed
schedule but algorithm has a considerable
also to determine the real-timeadvantage
schedule. Inin terms ofthe
summary, computational
advantages of the burden and time
proposed
required. Therefore,
algorithm this algorithm
are as follows: can besharing
(1) it is a power applied not only
strategy to determine
of multiple microgridsthe considering
day-ahead multiple
schedule but
also to determine the real-time schedule. In summary, the advantages of the proposed algorithm are as
follows: (1) it is a power sharing strategy of multiple microgrids considering multiple feeder structures
as well as a single feeder structure for minimizing the operation cost of the entire system; (2) it is
a power sharing strategy between microgrids that can be applied in a microgrid where only units that
do not have a cost function exist; (3) since it is the optimization of the distributed form, the computation
time decreases sharply compared with the one performed at the central center. In future works, we
will study the operational strategy of improving the economics through power sharing for peak power
control when constructing real-time schedules.

Acknowledgments: This work was supported by the Korea Institute of Energy Technology Evaluation and
Planning (KETEP) and the Ministry of Trade, Industry & Energy (MOTIE) of the Korea (No. 20151210200080).
This research was supported by Basic Science Research Program through the National Research Foundation of
Korea (NRF), funded by the Ministry of Education, Science and Technology (Grant no. NRF-2013R1A1A1012667).
Author Contributions: Won-Poong Lee carried out the main research tasks and wrote the full manuscript, and
Jin-Young Choi provided technical support to verify the proposed algorithm in simulation software. Dong-Jun Won
validated and double-checked the proposed algorithm, the results, and the whole manuscript.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Olivares, D.E.; Mehrizi-Sani, A.; Etemadi, A.H.; Canizares, C.A.; Iravani, R.; Kazerani, M.; Hajimiragha, A.H.;
Gomis-Bellmunt, O.; Saeedifard, M.; Palma-Behnke, R.; et al. Trends in microgrid control. IEEE Trans.
Smart Grid 2014, 5, 19051919. [CrossRef]
Energies 2017, 10, 1336 18 of 18

2. Wang, T.; Neill, D.O.; Kamath, H. Dynamic Control and Optimization of Distributed Energy Resources in
a Microgrid. IEEE Trans. Smart Grid 2015, 6, 28842894. [CrossRef]
3. Yang, Z.; Wul, R.; Yang, J.; Long, K.; You, P. Economical Operation of Microgrid with Various Devices
Via Distributed Optimization. IEEE Trans. Smart Grid 2016, 7, 857867. [CrossRef]
4. Wang, Y.; Wul, R.; Mao, S.; Nelms, R.M. On Hierarchical Power Scheduling for the Macrogrid and
Cooperative Microgrids. IEEE Trans. Ind. Inform. 2015, 11, 15741584. [CrossRef]
5. Wang, D.; Guan, X.; Wu, J.; Li, P.; Zan, P.; Xu, H. Integrated Energy Exchange Scheduling for Multimicrogrid
System With Electric Vehicles. IEEE Trans. Smart Grid 2016, 7, 17621774. [CrossRef]
6. Tian, P.; Xiao, X.; Wang, K.; Ding, R. A Hierarchical Energy Management System Based on Hierarchical
Optimization for Microgrid Community Economic Operation. IEEE Trans. Smart Grid 2016, 7, 22302241.
[CrossRef]
7. Che, L.; Shahidehpour, M.; Alabdulwahab, A.; Al-Turki, Y. Hierarchical Coordination of a Community
Microgrid with AC and DC Microgrids. IEEE Trans. Smart Grid 2015, 6, 30423051. [CrossRef]
8. Gregoratti, D.; Matamoros, J. Distributed Energy Trading: The Multiple-Microgrid Case. IEEE Trans.
Ind. Electron. 2015, 62, 25512559. [CrossRef]
9. Wang, Z.; Chen, B.; Wang, J.; Chen, C. Networked Microgrids for Self-Healing Power Systems. IEEE Trans.
Smart Grid 2016, 7, 310319. [CrossRef]
10. Wang, Z.; Wang, J. Self-Healing Resilient Distribution Systems Based on Sectionalization into Microgrids.
IEEE Trans. Power Syst. 2015, 30, 31393149. [CrossRef]
11. Liu, W.; Gu, W.; Sheng, W.; Meng, X.; Xue, S.; Chen, M. Pinning-Based Distributed Cooperative Control for
Autonomous Microgrids under Uncertain Communication Topologies. IEEE Trans. Power Syst. 2016, 31,
13201329. [CrossRef]
12. Moayedi, S.; Davoudi, A. Distributed Tertiary Control of DC Microgrid Clusters. IEEE Trans. Power Electron.
2016, 31, 17171733. [CrossRef]
13. Meng, L.; Dragicevic, T.; Roldan-Perez, J.; Vasquez, J.C.; Guerrero, J.M. Modeling and Sensitivity Study of
Consensus Algorithm-Based Distributed Hierarchical Control for DC Microgrids. IEEE Trans. Smart Grid
2016, 7, 15041515. [CrossRef]
14. Xu, Y.; Zhang, W.; Hug, G.; Kar, S.; Li, Z. Cooperative control of distributed energy storage systems in
a microgrid. IEEE Trans. Smart Grid 2015, 6, 238248. [CrossRef]
15. He, G.; Chen, Q.; Kang, C.; Pinson, P.; Xia, Q. Optimal Bidding Strategy of Battery Storage in Power Markets
Considering Performance-Based Regulation and Battery Cycle Life. IEEE Trans. Smart Grid 2016, 7, 23592367.
[CrossRef]
16. Shen, J.; Dusmez, S.; Khaligh, A. Optimization of Sizing and Battery Cycle Life in Battery/Ultracapacitor
Hybrid Energy Storage Systems for Electric Vehicle Applications. IEEE Trans. Ind. Inform. 2014, 10, 21122121.
[CrossRef]
17. Eller, A.; Gauntlett, D. Energy Storage Trends and Opportunities in Emerging Markets; Energy Sector Management
Assistance Program, International Finance Corporation World Bank Group: Washington, DC, USA, 2017;
pp. 1011.
18. Choi, B.R.; Lee, W.P.; Choi, J.Y.; Shin, Y.H.; Won, D.J. Optimal Power Exchange of Multi-Microgrids with
Hierarchical Coordination. In Proceedings of the 6th International Youth Conference on Energy 2017,
Budapest, Hungary, 2124 June 2017.
19. Ipopt Home Page. Available online: https://projects.coin-or.org/Ipopt (accessed on 2 June 2017).
20. Bonmin Home Page. Available online: https://projects.coin-or.org/Bonmin (accessed on 2 June 2017).
21. Website of the Regional Transmission Organization PJM. Available online: http://www.pjm.com
(accessed on 2 June 2017).

2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).