IET Renewable Power Generation

Research Article

Stochastic multi-objective framework for ISSN 1752-1416

Received on 26th April 2017

optimal dynamic planning of interconnected

Revised 10th July 2017
Accepted on 30th August 2017
E-First on 9th October 2017
microgrids doi: 10.1049/iet-rpg.2017.0278
www.ietdl.org

Farhad Samadi Gazijahani1, Javad Salehi1

1Department of Electrical Engineering, Azarbaijan Shahid Madani University, Tabriz, Iran
E-mail: [email protected]

Abstract: This paper proposes a new stochastic multi-objective framework for optimal dynamic planning of interconnected
microgrids (MGs) under uncertainty from economic, technical, reliability and environmental viewpoints. In the proposed
approach, optimal site, size, type, and time of distributed energy resources are determined along with optimal allocation of
section switches to partitioning conventional distribution system into a number of interconnected MGs. The uncertainties of the
problem are considered using scenario modelling and backward scenario reduction technique is implemented to deal with
computational burden. In addition, three different risk averse, risk neutral and risk seeker strategies are defined for distribution
network operator. The proposed framework is considered as two unparalleled objective functions which the first objective
minimizes the investment cost, operation and maintenance cost, power loss cost and pollutants emission cost and the second
objective is defined to minimize energy not supplied in both connected and islanded modes of MGs. Finally, multi objective
particle swarm optimization is applied to minimize the proposed bi-objective functions and subsequently fuzzy satisfying method
is accomplished to select the best solution proportional to risk based strategies. Efficiency of the proposed framework is
validated on 85-bus distribution system and obtained results are presented and discussed.

Nomenclature πsl probability of load demand at scenario s

Indices and sets πspv probability of PV generation at scenario s
πswt probability of WT generation at scenario s
i index for installed DG units
j index for installed ESS units Rl resistance of lines
k index for installed DRS units Lb length of feeders [km]
m index for installed switches Pres loads are restored during fault
n index for MGs number Prep loads are not restored during fault
NSW set of all switches, index by m Tres duration of the fault and switching time
Nb set of all network buses, index by b Trep duration of the fault repair
Nl set of all network lines, index by l IntR interest rate
Ns set of all scenarios, index by s InfR inflation rate
Nres set of all isolated nodes during fault Cint cost of interrupted load [$/kW]
Nrep set of all isolated nodes during repair ωR multiplier of heat recovery of CHP
T set of all planning years, index by t ηw whole efficiency of CHP units
ηe electrical efficiency of CHP units
Parameters ηm mechanical efficiency of CHP units
CGis, j, k generation cost of DER units [$/kW]
∂, κ and ρ shape parameter of Weibull PDF
α, β and γ CHPs coefficients MCPESS j.t
electricity price for charging/discharging of ESS
δ, ζ generator emission characteristics units
ηESS charging/discharging efficiency of ESS ICDGsi, p investment cost of DG per kW
j
k temperature coefficient ICESSsj, p investment cost of ESS per kWh
λ failure rate of distribution lines ICDRsk, p investment cost of DRS per kVAR
£ mean value of normal PDF investment cost of each switch
σ standard deviation of normal PDF ICSWsm
Ψ rate of annual load growth MCDGsi, p maintenance cost of DG per kW
φ, Ω parameters of the Beta model MCDRsk, p maintenance cost of DRS per kVAR
GING incident irradiance maintenance cost of each switch
MCSWsm
GSTC irradiance at standard test condition
PSTC rated power of PV units [kW]
Variables
Prate rated power of WT units [kW]
Vci wind turbine cut-in speed [m/s] PWT installed capacity of WT units [kW]
Vco wind turbine cut-out speed [m/s] PPV installed capacity of PV units [kW]
Vr wind turbine rated speed [m/s] PESS installed capacity of ESS units [kW]
v wind speed [m/s] PCHP installed capacity of CHP units [kW]
A, B, and C WTs coefficients in the Plant's datasheet QDRS installed capacity of DRS units [kVAR]
πs probability of each scenario

IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1749-1759 1749
© The Institution of Engineering and Technology 2017
 17521424, 2017, 14, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-rpg.2017.0278, Wiley Online Library on [22/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
PD load demand [kW]
Pls active power flow at line l
Qls reactive power flow at line l
V bs magnitude of voltage at bus b
Ils magnitude of current in line l
μ membership function
xis. j . k binary investment decision of DER units at scenario s
xms binary investment decision of switch at scenario s
θis⋅ j ⋅ k integer variable to determine the investment year
σiDER⋅ j⋅k
binary variables for installed DER units
σ ESS
j
⋅s binary variables for charging/discharging of ESS
Tr operation time of CHP units in operation conditions
Fig. 1 Main concepts of proposed approach for MGs planning problem
TC, Tref cell and reference temperatures, respectively
SLTS interrupted load in islanding mode power. The optimal objective in this study is to simultaneously
NI number of interrupted load points minimise total net present cost and pollutants emission cost in
MGs. Gu et al. [14] proposed an overall review of modelling,
Functions planning, and energy management of combined cooling, heating,
and power in MGs. Atwa et al. [15] proposed a probabilistic
F1 first objective function planning approach for optimally placement the mix of wind, solar,
F2 second objective function and biomass units to minimise annual energy losses in distribution
Z total objective function systems. Che et al. [16] discussed the optimal planning of multi-
CPV cost present value MGs. The proposed approach in this paper utilizes a probabilistic
minimal cut-set-based methodology for optimal planning of MGs
1 Introduction with variable renewable energy sources. Che et al. [17] presented a
The microgrids (MGs) comprise low voltage (LV) or medium graph partitioning and integer programming integrated
voltage (MV) distribution systems with distributed energy methodology for optimal planning of loop-based MG topology
resources (DER) and flexible loads with clearly defined electrical while considering distributed energy resources in the MG. Arefifar
boundaries [1]. Such systems can be operated in a non-autonomous et al. [18] proposed a new approach for planning a control and
way if connected to the main grid, or in an autonomous way, if communication infrastructures of MGs which is taken into account
disconnected from the grid. Main differences of MGs with both communication system and economic related aspects. The
traditional distribution networks are in reliability, economic and reconfiguration solution is proposed for loss minimisation of a
self-healing aspects. Significant benefits of MGs in comparing with distribution system using mixed-integer conic programming [19]
traditional distribution networks in the mentioned aspects make and stochastic multi-objective modelling by point estimate method
enough motivation to use MGs. According to the IEEE Std 1547.4 [20] as midterm MGs planning approach. Shen et al. [21] proposed
[2], large distribution networks can be clustered into a number of a multi-stage methodology for operational planning active
interconnected MGs. On the other hand, MG planning problem can distribution network in the presence of DGs units with considering
be categorised into two major categories namely static model and the reconfiguration of MG-based distribution system under
dynamic model [3]. In a static model, the investment decision uncertainty.
accomplishes in the first year of the planning horizon time whereas This paper presents a novel stochastic multi-objective
in the dynamic model it determines by the MG planner which may framework for optimal dynamic planning of interconnected MGs
not be necessarily the first year of the planning horizon. In fact, in from reliability, supply security as well as environmental and
the dynamic model, the investment year of DER units specifies by economic viewpoints. To do this, optimal size, type, number, and
the optimisation procedure. location of DG, ESS, and DRS units are determined simultaneously
Several kinds of literature are investigated optimal clustering of with optimal section switch allocation to specify the optimal
MGs from various aspects. In [4] investment on distributed energy electrical boundaries of multiple MGs (Fig. 1). Uncertainties of
resources and optimal partitioning of distribution system into a few load consumption and generated power by renewable energy
MGs are considered from a reliability viewpoint. Optimal planning resources are taken into account using stochastic scenario-based
of multiple MGs has been studied in the literature [5–17]. For modelling. Moreover, the backward scenario reduction technique is
instance, a model for calculating optimal size and site of energy implemented to reduce computational burden and increase the
storage systems (ESSs) in an active distribution network with optimisation efficiency. A MINLP formulation has been utilised for
considering wind turbine is presented in [5]. Moreover, two-stage modelling of the proposed problem and subsequently, multi-
robust optimisation algorithm for an optimal distributed generation objective particle swarm optimisation (MOPSO) algorithm is
(DG) allocation in MGs is investigated in [6]. Nekooei et al. [7] employed to solve the formulated framework.
presented a cost-based approach for allocation of DG units in The main contribution of the proposed framework is to
primary feeders of distribution networks via multi-objective integrated DER and section switch planning for various risk-based
harmony search algorithm. Arefifar and Mohamed [8] discussed strategies with considering diverse operating uncertainties.
the reliability-resiliency based model for optimal planning of Accordingly, three different risk-seeker, risk-neutral, and risk-
distributed reactive source (DRS) as well as DG units with the averse strategies are defined. It should be noted that the proposed
purpose to minimise the reliability costs. Khodaei and multi-objective model for MG planning problem is entirely
Shahidehpour [9] studied a transmission expansion planning that disparate in comparison with previous works because we
considers the integration of MGs. A multi objective study of the developed an integrated reliability-cost framework associated with
MG topology design which uses mixed-integer non-linear considering simultaneously DER planning and switch allocation
programming (MINLP) is proposed in [10]. The reliability-oriented under uncertain situations. However, the prior literature has not
MG planning model, which focused on the sustained and been investigated both DER planning and switch placement
temporary faults has been presented in [11]. Kalkhambkar et al. problem simultaneously and also, they have studied single aspect
[12] investigated MG generation expansion design by joint of MGs planning problem, which in this paper a comprehensive
optimisation method considering the ESS and renewable resources. model has been suggested to the optimal planning of MGs. As a
Guo et al. [13] present a two-stage optimal expansion method for result, the proposed model converts conventional distribution
MGs with a combined utilisation of cooling, heat, and electricity network (CDN) into active distribution network (ADN) and
subsequently turns into a set of interconnected MGs. The CDN is

1750 IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1749-1759
© The Institution of Engineering and Technology 2017
 17521424, 2017, 14, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-rpg.2017.0278, Wiley Online Library on [22/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
defined as a radial large-scale network without any distributed be respected between the loss of the information and decreasing the
energy resources installed on the network, which operates as computational burden. To implementation backward scenario
traditional mode by one independent operator and also, the ADN reduction technique following steps should be performed [24].
refers to a distribution network that has many DER resource on the
network and can produce power electricity and supply its own Step 1: Create the probability distance matrix with the cardinality
consumption or sell to upstream network [22]. It will become to of H (ξ, ξ′) containing the distance between pairs of generated
interconnected MGs when the section switches are allocated on the scenarios.
network. Step 2: Calculate the distances between all scenario pairs as
The main novelties and contributions of this paper are follows:
summarised as below:
Ns
• Optimally determining the size, site, type, and time of DG, ESS Ht = ∑
2
πsi − πsj , Ns = Nsl × Nspv × Nswt (3)
and DRS units to convert CDN into ADN. s=1
• Optimum allocation of section switches for clustering ADN into Step 3: Select the minimum distances as: Hi = argminHt
a number of MGs with considering maximum supply security Step 4: If the cardinality of H is sufficient then go to step 2;
and power mismatch constraints for each MGs. otherwise continue.
• Multi-objective formulation for proposed dynamic planning Step 5: Repeat steps 2–4 until the number of scenarios is decreased
problem from the economic, environmental, technical and based on the accuracy required.
reliability points of view.
• Taking into account the uncertainty in the characteristics of the 2.2 Load demand modelling
DER units and load demand using stochastic scenario based
approach and implementation backward scenario reduction One of the most important and difficult steps of MGs planning
technique to enhance a trade-off between the accuracy of the problem is to determine the load consumption level. In distribution
solution and the computational burden. systems loads are naturally stochastic as well as load prediction
• Defining three different risk-based strategies including risk- methods are accompanied by errors. Hence, it could be useful to
seeker, risk-neutral, and risk-averse structure by using fuzzy find a promising method to consider load uncertainty in
satisfying method for DNO. distribution network planning. To achieve this aim, a stochastic
representation of load is proposed in this study for DERs allocation
This paper is organised as follows: Section 2 explains procedure. Loads of system are modelled to follow the normal
mathematical model formulation for the planning problem; Section distribution function [25]. In this case, the system load is defined
3 describes the solution algorithms and Section 4 discusses the by a mean value (£) and standard deviation (σ) as below:
numerical results; and finally, Section 5 draws conclusions.
2
1 PD − £
FL(PD) = exp − (4)
2 Problem formulation 2π × σ 2 × σ2
2.1 Scenario generation and reduction
peak
Due to the high number of uncertainties encountered in distribution PtD⋅ s = Ps ⋅ d 1 + Ψ t
(5)
network planning, the proper basis of system researches should be
formed in terms of the solution of a probabilistic programming. In Also, (5) shows the load growth during the planning horizon.
this study, we supposed three uncertainties in the planning of MGs, peak
Where Ps ⋅ d is peak load of the system in the first year at scenario s
(i) network loads uncertainty, (ii) output power uncertainty of sun and PtD⋅ s is peak load of the system in year t and scenario s.
generating (PV) units and (iii) output power uncertainty of wind
generating (WT) units.
There are different methods to handle the uncertainties such as 2.3 Modelling of DER units
stochastic programming, fuzzy mathematics, interval analysis, Describing the mathematical model of DER units has been
robust optimisation and information gap decision theory [23] explained as follows.
which choosing the best technique among them depends on the
uncertainty nature and available data about the uncertain
parameters of the model. In this paper, the uncertainties of solar 3.3.1 Wind turbine: Generated power by a wind turbine is
and wind power generations and load consumption, are taken into affiliated to wind speed. The Weibull distribution function is
account using scenario modelling and a scenario reduction applied to depict the distribution of wind speed for long-term
technique is used to enhance a trade-off between the accuracy of planning targets [26]. Weibull PDF and generated power by wind
the solution and the computational burden. In this paper, a turbine are presented as
stochastic model is applied for handling the uncertainty. Each ρ−1
probability distribution function (PDF) of uncertain parameters is ∂ v
× × e(v / κ)∂ v≥0
divided into several states and the probability of each state is Fv(v) = κ ∂ (6)
specified. In each load level, the states are combined to construct 0 otherwise
the whole set of states as follows:
0 0 < ν < V ci
πst = πsl × πspv × πswt (1) 2
(Aν + Bν + C) × Prate V ci ≤ ν ≤ V r
PWT = (7)
Ns Prate V r ≤ ν ≤ V co
∑ πsl × πspv × πswt =1 (2) 0 V co ≤ ν ≤ ∞
s=1

where πsl, πspv and πswt are discrete probability distribution sets for 2.3.2 Photovoltaic: The fluctuations of the PV generation which
load consumption, solar generation and wind generation, is considered as a function of the solar radiation can be modelled
respectively. In this paper, one thousand scenarios are generated using Beta distribution function. The power generated by PV cells
using Monte Carlo simulation (MCS) and backward scenario depends on the irradiance and the environment temperature can be
reduction method is applied to reduce the computation efforts obtained by Aien et al. [27]. Beta PDF and generated power by PV
while maintaining the solution accuracy which the 1000 generated are shown as below:
scenarios are reduced to 30 scenarios. A reasonable trade off must

IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1749-1759 1751
© The Institution of Engineering and Technology 2017
 17521424, 2017, 14, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-rpg.2017.0278, Wiley Online Library on [22/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Γ Ω+φ determined based on peak load of each planning stage. These costs
× s Ω−1 ⋅ 1 − s φ−1
0 ≤ s ≤ 1
Γ Ω +Γ φ can be formulated as the following equation:
FS S =
Ω≥ 0φ≥ 0 NDG
0 otherwise ∑ θis t xisPiDGt s ICDGsi
⋅ ⋅
⋅

(8) i=1
NESS
GING + ∑ θsj ⋅ t xsjPESS ⋅s
ICESSsj
Ppv = PSTG × × 1 + k(T C − T ref ) (9) T Ns j⋅t
GSTG j=1
C1 = ∑∑ NDRS
(11)
t=1s=1
2.3.3 Combined heat and power: CHP system is an economical +∑ θks . t xksQkDRS
.t
⋅s
ICDRsk
kind of integrated power and heat generation method that provides k=1
power and heat at the same time. The power generated by CHP can NSW
be evaluated by [28] + ∑ xmsICSWsm
m=1
2
PCHP G = α + βPCHP + γPCHP (10)
Cost of present worth value of this annual cost with considering
inflation rate and interest rate [31] in planning duration is
2.3.4 Energy storage system: ESSs mainly transform electrical
calculated as below:
energy into a more convenience storable form for converting back
to the electricity energy when needed. ESS enhances flexibility, NS T
reliability, and technical specifications of the system such as power 1 + InfR t
(12)
CPV C1 = C1 ∑ ∑
loss reduction and power quality improvement [29]. In this paper, s=1t=1
1 + IntR
the battery-based ESS has been used as a storage device to improve
the reliability of MGs in both interconnected and islanded modes.
2.4.2 Maintenance cost: Another yearly cost of DERs and
Moreover, it is assumed that the ESS units charge in low load
switches is related to maintenance cost. Operation and maintenance
periods through DG units and then discharge into the network in
cost includes annual electrical and mechanical repair cost as well
peak load hour. Actually, the peak load of each year in the planning
as charged and discharged costs of ESS units that can be evaluated
period is considered in the model. In the proposed method it is
by
assumed that ESS units are fully charged in low load hours of each
day and discharged in the peak hours if necessary considering NDG
demand level of MGs. In the other word, there is a controlling
software which charges ESS units in low load hours of a day and
∑ σiDGt PiDGt sMCDGsi
⋅ ⋅
⋅

i=1
discharges them in the peak load hour. Therefore, in the peak hour, NESS
they will improve the reliability of MGs and reduce the power ⋅ s ESS ESS ⋅ s ESS ⋅ s
losses. This issue is justifiable because, in low load periods, the Ns
+ ∑ σ ESS
j ⋅ t η j P j ⋅ t MCP j ⋅ t
T
j=1
capacity of available DG units is more than the level of C2 = ∑∑ (13)
consumption thereby, ESS units can be easily charged in low load t=1s=1
NDRS
DRS ⋅ s
hours and can be used in peak load [30]. As a result, by optimal +∑ σkDRS
⋅ t Qk ⋅ t MCDRsk
arbitrage of ESS units, the reliability and technical characteristics k=1

of the system such as power losses and voltage profile will be NSW

improved. + ∑ σmSWMCSWsm
In addition, in dynamic planning determining the optimal ESS m=1
capacity in each year is related to the installed capacity of ESS
Ns
units in previous years. Therefore, the installed capacity of ESS T
1 + InfR t
units in prior stages must be calculated in each stage. In fact, in CPV C2 = C2 ∑ ∑ (14)
s=1t=1
1 + IntR
each planning stage, the network must be updated by installed
resources in prior stages. It is worth noting that in the proposed
In (14), the first part depicts the maintenance cost of renewable-
dynamic model, the planning stages are considered as annual.
based DG units; the second part denotes the cost of charged and
Therefore, in each year the planning of DER units should be
discharged powers by storage devices based on arbitrage of ESS
determined by considering the peak load of the same year.
units over operating periods. It should be stated that in this paper in
order to reduce the computational burden as well as need to be
2.4 Economic costs evaluation simple, it is assumed that the charging of ESS units is performed in
The proposed model for MG planning problem is a dynamic model low load periods and discharge in peak load periods to increase the
which the planning horizon time is divided into ten stages (T = 10). reliability of MGs. Therefore, the maintenance cost of ESS units is
Therefore, the year of DER investment is not specified whereas considered as a function of their installed capacity [32]. Also, two
that will be determined by optimisation algorithm. In this section, next parts show the maintenance costs of DRS units and installed
economic and reliability cost of DERs and switch are presented. switches during planning horizon time, respectively.
All of these duties are based on planning cost reduction,
enhancement the technical characteristics of the system and 2.4.3 Fuel cost: Since distributed energy resource shall trace
improving reliability and supply security of customer service in demand alterations, thereby, it is required to be dispatch-able units
both interconnected and islanded modes. For simplifying the on the network which their operation costs are subject to fuel cost.
operational scheduling and for reducing the curse of dimensionality In this paper, it is assumed that in the WT, PV and ESS units, fuel
issue, the ESS charging/discharging scheduling is fulfilled in two cost is zero but the fuel cost of CHP units is not zero. The present
load levels. Therefore, the planning and operation costs of DERs worth value of operation cost of CHP units calculates by
and switch allocation in the distribution system can be expressed as
below. T
1 + InfR t
Ns NDG
CPV C3 = ∑ 1 + IntR
× ∑∑
t=1 s=1i=1
2.4.1 Investment cost: The cost of DERs units including DG, (15)
ESS and DRS, site making for DG installation, construction and so ωR ηw − ηe
T rs ⋅ tCGisPis⋅ t × 1−
on are included in investment cost. In the proposed dynamic model ηm
for MGs planning problem, the investment decisions are

1752 IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1749-1759
© The Institution of Engineering and Technology 2017
 17521424, 2017, 14, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-rpg.2017.0278, Wiley Online Library on [22/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2.4.4 Pollution emission cost: The main source of air 2.6 Objective function
contamination and greenhouse gas publication are seen to be fossil
fuel consumption generating gases such as SO2, CO2, CO, and NO. As stated earlier, the proposed MG planning problem is a dynamic
model which investment decisions determine in all years of the
The quantity of emission associated with these gases is listed in
planning horizon time. The economic and reliability viewpoints
[33] for different DG technologies. The pollution emission cost
which have been explained in the previous parts are considered in
with its present worth value can be evaluated by
two unique objective functions, namely, total economic costs and
Ns T n
reliability costs that formulated as below:
t
1 + InfR
CPV C4 = ∑∑ 1 + IntR ∑ 10 −2
αi + βiPGs it + γiPGs2it
F(1) = CPV C1 + CPV C2 + CPV C3 + CPV C4 + CPV C5
s=1t=1 i=1
(16) (22)
+ ζiexp δi × PGs it
F(2) = CPV CCENS + CPV CIENS (23)

2.4.5 Power losses cost: Calculation of power losses implies MinZ = F(1), F(2) s . t . (25) − (36) (24)
solving the load flow problem. In this paper, an improved
backward forward sweep load flow is utilised to compute power Therefore, DERs allocation and switch placement problem can be
losses. The power losses cost with present worth value can be solved by using MOPSO which is an appropriate optimisation
evaluated by algorithm in solving multi-objective programming especially in the
case of MINLP model. Given functions will be optimised
Ns T
1 + InfR t considering under mentioned constrains.
CPV (C5) = ∑∑ 1 + IntR
Nl
s=1t=1
(17) 2.7 Constraints
∑ Pls⋅ t2 + Qls⋅ t2 Rl /V b2 ⋅ t CPlossts ⋅ l 2.7.1 Voltage and current limits: Optimisation will be fulfilled in
l=1 order to find out optimal network buses where DER can be
installed so that the voltage profile be located in standard limits
2.5 Reliability costs evaluation (i.e. 0.95–1.05 PU) or in restoring case close to it as well as the
magnitude of lines current is less than maximum rated value. If
2.5.1 Connected mode energy not supply (CENS): Nowadays voltage and current limits are violated from acceptable values, it is
the reliability evaluation of MGs with DER units has been gained considered as a technical risk in the system. The voltage of each
an increasing attention by the researchers worldwide. The bus in each bus n and in year t and scenario s, and also current of
reliability improvement process is complicated by various output all lines in year t and scenario s should be kept within the safe
specifications of DERs in MGs. In this section, the reliability operating limits as
evaluation is done under uncertainties of load consumption and
renewable power generation which they are taken into account
using scenario modelling. The reliability evaluation in this paper is V bmin max
⋅ t ⋅ s ≤ Vb ⋅ t ⋅ s ≤ Vb ⋅ t ⋅ s ∀S ⋅ ∀T (25)
evaluated in both operating modes of MGs. In the first structure,
the MGs are interconnected as well as MGs can be sold or Il ⋅ t ⋅ s ≤ Ilmax
⋅t⋅s ∀S ⋅ ∀T (26)
purchased surplus or deficit of power generation through the
upstream network. Also, in the second structure, all MGs are in 2.7.2 Maximum capacity of installed DGs, ESSs, and
islanded mode from main grid in point of common coupling (PCC) DRS: For lack of funds and stable operation of installed capacity
as well as there is no power exchange between MGs and upstream and generated power of each DER units are restricted by lower and
network. One of the most important reliability indices is ENS upper limits as follows:
index, in which it shows the value of not supply of load (kWh). In
this paper to calculate the ENS value in both connected and Ns NDG
islanded modes, two kinds of ENS indices are used namely CENS ∑ ∑ PGsi p ≤ ⋅ PDsmax (27)
and islanded mode energy not supply (IENS). The CENS cost with s=1i=1
present worth value can be evaluated by [34]
min
PPVi
≤ PPVi, i = 1, 2, …, NPV (28)
Ns T Nl
CCENS = ∑∑ ∑ Cintsl ⋅ t × λls × Ll × min
PWT max
≤ PWTi ≤ PWT , i = 1, 2, …, NWT (29)
s=1t=1 l=1
Nres Nrep (18) i i

∑ s s
Pres ⋅ tT res ⋅t + ∑ s
Prep s
⋅ t T rep ⋅ t min
PCHP i
max
≤ PCHPi ≤ PCHP i
, i = 1, 2, …, NCHP (30)
res = 1 rep = 1

min max
Ns T t PESS ≤ PESSi ≤ PESS , j = 1, 2, …, NENS (31)
1 + InfR i i
CPV(CCENS) = CCENS × ∑∑ (19)
s=1t=1
1 + IntR min max
QDRS j
≤ QDRS j ≤ QDRS j
, k = 1, 2, …, NDRS (32)
2.5.2 Islanded mode energy not supply: Expected load energy
interrupted during the islanded mode of the main grid due to 2.7.3 Power balance constraints and load flow
available power shortage is incorporated into the objective function equations: The total electric power generation must cover the total
as security and stability indices. The IENS cost with present worth electric power demand (PD, QD) as well as the real and reactive
value can be calculated by [35] power losses in distribution lines (Ploss, Qloss), hence

Ns N Ns NDG
∑TSD = 1 SLsTS T TS
s T
CIENS = ∑ × Cints (20) ∑ ∑ ∑ PGs − PDs ⋅ t − Ploss
s
⋅t = 0 (33)
s=1
NI t=1s=1i=1
i⋅t

Ns T T Ns NDRS
t
1 + InfR (34)
CPV(CIENS) = CIENS × ∑∑ 1 + IntR
(21) ∑ ∑ ∑ QGs j⋅t
− QDs ⋅ t − Qloss
s
⋅t = 0
s=1t=1 t=1s=1 j=1

IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1749-1759 1753
© The Institution of Engineering and Technology 2017
 17521424, 2017, 14, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-rpg.2017.0278, Wiley Online Library on [22/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
N
∑i =obj1 μik
μk = M N (39)
∑ j = 1 ∑i =obj1 μij

The best agreement solution for various risk-based strategies is

related to the value of μk. With selection different values of
membership function in decision maker framework, it can be
achieved various strategies based on risk-seeking or risk aversion
of DNO. The flowchart of the proposed multi-objective algorithm
based on risk-based decision-making is demonstrated in Fig. 2.
It should be mentioned that various risk-based methods have
been developed in recent years such as conditional value-at-risk,
variance, shortfall probability, and expected shortage techniques
[38], which implementation of them is related to various
parameters and situations. However, these techniques have a few
drawbacks like they are too conservative and also impose
additional costs on investors. Therefore, in this paper, we proposed
an innovative risk-based decision-making process based on a fuzzy
satisfying mechanism for MGs operator to choose the best solution
among optimal Pareto front owing to its own strategy and asset
management programs. Although, in order to increase the
flexibility and applicability of proposed risk-based strategies, two
risk-seeking ψ(1) and risk-aversion ψ(2) parameters have been
utilised to adjust the degree of risk-based strategies. By choosing
the various values for risk-seeking and risk-aversion parameters in
different strategies can achieve diverse structures for MGs which
Fig. 2 Flowchart of the algorithms used for the proposed planning can be formulated as below:

argmin ψ(1) × F(1) ⋅ ψ(2) × F(2) (40)

Pis⋅ t + PiDER
⋅t
⋅s
− PiD⋅ t⋅ s = V is⋅ t ∑ V is⋅ t ⋅ n ψ
n
(35)
Gi j cosδi ⋅ t ⋅ j + Bi j sinδi ⋅ t ⋅ j ψ(1) ⋅ ψ() ∈ 0⋯1 (41)

s . t . (24) − (35) (42)

Q js⋅ t + Q jDRS D
⋅ t ⋅ s − Qj ⋅ t ⋅ s = Vi ⋅ t ∑ Vi ⋅ t ⋅ n
s s

n
(36) ψ(1) and ψ(2), which are risk-seeking and risk-aversion
Gi j cosδi ⋅ t ⋅ j + Bi j sinδi ⋅ t ⋅ j parameters, respectively, show the degree of risk-seeking and risk-
aversion of decision makers. They can vary from 0 to 1 so that if
their values are set to be 1, it means that maximum risk-seeking
3 Solution algorithm and risk-aversion degrees are considered for DNO. By decreasing
their values to 0, the degree of risk-seeking and risk-aversion will
The multi-objective model of particle swarm optimisation called
be decreased. In fact, these parameters show the importance of
MOPSO is appropriate in case of minimising multi-objective
different objective functions for various strategies as weight
functions simultaneously. If Z(x) consists of N objective functions,
coefficients so that by increasing the value of each parameter from
then the multi-objective problem can be defined as finding the
0 to 1, the impact of its own function will be increased. Therefore,
vector X ∗ = x1∗, x2∗, …, xn∗ in order to minimise Z(x) by selection diverse values for these parameters can be obtained
different structures for MGs.
MinZ x = F1 x , F2 x , …, Fn x (37)
4 Results and discussions
Generally, multi-objective optimisation technique conclusions in a
set of optimal solutions, instead of one solution. The reason is that In this paper, a large-scale distribution system is considered for
none of the solutions can be considered to dominate than any other implementation of the proposed model. The well-known 85-bus
with regard to all objective functions. Consequently, in the distribution network [39] is chosen as the test network to show the
MOPSO approach, there is not generally one global optimum, but a applicability of the proposed model. In this paper, the planning
set of so-called Pareto-optimal solutions [36]. To extract the best problem is a dynamic model which investment decisions are
agreement solution, various methods have been implemented in the performed during planning horizon time, as well as the planning
literature, which in this paper fuzzy-based mechanism called a time horizon, is considered 10 years, thereby, ten stages are
fuzzy decision-making (FD) or fuzzy satisfying approach is considered to design DER units. Also, annual load growth rate is
performed. Due to the obscure nature of the decision maker's considered 10%. Considering prohibited operating zones of various
judgment, the ith objective function of a solution in the Pareto- DER units, power mismatch constraints, switch allocation problem
optimal set Fi is demonstrated by a membership function µi defined and the other non-linear practical constraints make the proposed
as [37] model into a complicated, non-linear, non-smooth, and non-convex
optimisation problem, which has been solved with a population-
1 Fi ≤ Fimin based meta-heuristic algorithm namely MOPSO algorithm in
MATLAB software.
Fimax − Fi The uncertainties of electric load and renewable power
μi = Fimin < Fi < Fimax (38) generation are taken into account using stochastic scenario based
Fimax − FImin
modelling. One thousand scenarios are generated using MCS and
0 Fi ≥ Fimax backward scenario reduction method is implemented to enhance a
trade-off between the accuracy of the solution and the
where Fimax and Fimin are the maximum and minimum amount of computational burden which the 1000 generated scenarios are
reduced to 30 scenarios. In this paper, it is assumed that the ESS
the ith objective function, respectively. For each non-dominated
units are fully charged in low load periods, which the value of
solution k, the normalised membership function μk is evaluated as

1754 IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1749-1759
© The Institution of Engineering and Technology 2017
 17521424, 2017, 14, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-rpg.2017.0278, Wiley Online Library on [22/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table 1 Values of parameters in DER units
PV PSTC = 200 kW GSTC = 1000 W/M2 K = 0.002 TC = 25°C DER Minimum power Maximum power
Tref = 20°C Pmin = 50 kW Pmax = 200 kW CHP 100 kW 300 kW
WT Prate = 200 kW Vci = 3 m/s Vr = 10 m/s Vco = 20 m/s ESS 50 kW 200 kW
n=2 ∂=3 κ=8 — DRS 100 kVAR 300 kVAR

Table 2 Optimum switch placement for different strategies

Strategy Optimum switch location Economic cost ($) ENS cost ($)
C1 C2 C3 C4 C5 Total CENS IENS Total ENS
risk-seeker [case 1] 1, 6, 8, 32, 66 27,674 3696 7725 1872 1076 42,176 5774 3900 9674
risk-neutral [case 2] 1, 4, 7, 10, 29, 33, 57, 67 62,338 6370 14,325 1597 914 83,720 4650 2930 7580
risk-averse [case 3] 1, 4, 7, 10, 25, 30, 33, 47, 59, 63, 67 95,120 9415 18,850 3918 725 127,540 4410 2485 6895

Table 3 Optimum DG locations and capacities for different strategies

DG type Site of DG Size of DG, kW Time of DG, year
Case 1 Case 2 Case 3 Case 1 Case 2 Case 3 Case 1 Case 2 Case 3
CHP 46, 79 14, 54, 79 6, 37, 46, 51, 75, 79 250, 225 250, 225, 200 200, 175, 175, 150, 200, 175 1, 6 1, 1, 5 1, 1, 2, 1, 3, 1
WT 17, 59 21, 31, 76 2, 12, 40, 63, 65 150, 100 100, 100, 75 50, 50, 75, 75, 100 1, 5 1, 1, 6 1, 2, 1, 1, 5
PV 21, 42, 60 6, 42, 59 15, 20, 39, 45, 57 150, 125, 100 150, 150, 125 50, 50, 50, 75, 75 1, 4, 1 1, 3, 1 1, 1, 3, 1, 1

Table 4a Continued
Resource Site of ESS Size of ESS, kWh Time of ESS, year
type Case 1 Case 2 Case 3 Case 1 Case 2 Case 3 Case 1 Case 2 Case 3
ESS 37, 53,75, 82 17, 37, 46, 74 16, 21, 42, 53, 150, 125, 125, 150, 175, 150, 150, 125, 125, 1, 3, 1, 8 1, 1, 4, 6 1, 1, 2, 2, 1, 1
74 100 150 125, 150, 125

Table 4b Optimum ESS and DRS locations and capacities risk-based strategies. With respect to these tables, case 1 is the
for different strategies most economical structure because of low investment costs and
Resource Site of DRS Size of DRS, Time of DRS, avoid to paying additional costs in order to increase reliability of
type kVAR year system, case 2 is a structure with trade-off between economic and
Case Case Case Case Case Case Case Case Case reliability aspects and case 3 shows the most reliable as well as the
1 2 3 1 2 3 1 2 3 most expensive construction due to policy of risk aversion of this
strategy that is willing to pay additional cost to reduce the
DRS 3, 11, 2, 9, 3, 9, 250, 150, 150, 1, 1, 1, 1, 1, 1,
reliability cost. Table 5 illustrates the value of membership
28, 12, 18, 225, 150, 100, 2, 1, 1, 2, 1, 1,
function (μ) for proposed different risk-based strategy. According
34, 18, 28, 225, 125, 125, 2 11, 1, 1, 2,
69 32, 33, 175, 125, 100, 2 1, 1,
to this table, the first solution has minimum membership function
34, 44, 150 175, 125, 2, 1,
for economic related objective and also it has a maximum value of
63, 49, 150, 100, 1
μ for reliability related objective. Therefore, it is appropriate to
71 61, 125, 125, risk-averse DNO.
72, 150 100, In contrast, the 40th solution has maximum μ value for
76, 150, economic related objective and it has a minimum value of μ for
84 125, reliability related objective. Therefore, it is suitable to risk-seeker
100 DNO. On the other hand, the seventeenth solution has the medium
and prudent amount for both its membership function, so it is
applied for risk-neutral DNO.
Using MOPSO to minimise the multi-objective function as an
electricity price is low, thereby, in the peak load period which the
optimisation algorithm, the trade-off Pareto curve can be plotted as
reliability of the system is low and electricity price is high, they
shown in Fig. 3. Moreover, an approach based on fuzzy set theory
can be discharged into the network.
is extended to evoke one of the Pareto optimal solutions for each
In Table 1, the parameters of DER units are given that are used
strategy. By using FD approach, Pareto curve and data provided by
in a simulation study. Moreover, to consider construction
the DNO, demonstrating the importance of cost versus reliability,
constraints in practical mode, the steps of generation power for
for example taking values between 1 and 0 (1 shows maximum
DER units is set to be 25 kW. Table 2 shows the obtained results
importance while 0 shows minimum importance), the different
for optimal switch placement problem for various strategies.
planning strategy can be selected. A risk-neutral DNO considers
According to this table selection, different risk-based strategies
the trade-off between economic and reliability costs. It evaluates
based on various values of membership function (μ) affect to
different scenarios and makes decisions based on the expected cost
location and number of switches. So that the risk-seeker strategy
but risk-averse DNO pays additional cost in order to reduce the
has the lowest number of switches and risk-averse strategy has the
reliability cost of system.
most switches number to installation on the network. This shows
Fig. 4 shows the obtained structure in last year of planning
the fact that by increasing the number of switches, the reliability of
horizon time (T = 10 years) for the risk-seeker DNO, which shows
system increases but the economic cost of system increases.
the optimum locations of DER units and electrical boundaries of
Therefore, with regard to policies and strategies of MGs owner and
MGs. As regards to this figure, the number of MGs and DER units
operators by changing the values of membership function different
are set to be least due to the economic viability of the proposed
structures can be gained for different strategies.
strategy. Instead, the reliability of this system is low due to the low
Tables 3 and 4 show the results of the optimal site, size, and
number of installed switches and DER on the network. The
time of DG, ESS, and DRS units planning for proposed different

IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1749-1759 1755
© The Institution of Engineering and Technology 2017
 17521424, 2017, 14, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-rpg.2017.0278, Wiley Online Library on [22/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table 5 Value of membership function for different risk-based strategy
Solution μ(1) μ(2) F(1) F(2) Solution μ(1) μ(2) F(1) F(2)
1 0.205 0.795 127,540 6895 21 0.398 0.602 80,466 7780
2 0.214 0.786 126,330 6950 22 0.412 0.588 79,860 7845
3 0.228 0.772 122,840 7052 23 0.425 0.575 79,774 7956
4 0.232 0.768 118,850 7120 24 0.441 0.559 78,330 7990
5 0.246 0.754 118,026 7275 25 0.462 0.538 75,640 8095
6 0.254 0.746 115,792 7290 26 0.473 0.527 72,456 8187
7 0.269 0.731 114,249 7310 27 0.482 0.518 70,634 8294
8 0.275 0.725 105,820 7388 28 0.499 0.501 67,945 8420
9 0.281 0.719 102,196 7449 29 0.531 0.469 65,325 8496
10 0.299 0.701 99,428 7498 30 0.572 0.428 63,845 8500
11 0.312 0.688 98,374 7504 31 0.596 0.404 60,127 8514
12 0.320 0.680 97,668 7523 32 0.625 0.375 57,635 8525
13 0.335 0.665 95,094 7540 33 0.638 0.362 53,630 8542
14 0.339 0.661 92,633 7545 34 0.672 0.328 51,126 8555
15 0.347 0.653 89,250 7563 35 0.689 0.311 49,136 8574
16 0.355 0.645 86,160 7570 36 0.699 0.301 48,429 9582
17 0.368 0.632 83,720 7580 37 0.732 0.268 46,367 9595
18 0.372 0.628 82,536 7600 38 0.745 0.255 44,205 9633
19 0.379 0.621 81,875 7675 39 0.766 0.234 43,360 9660
20 0.390 0.610 80,652 7737 40 0.784 0.216 42,176 9674

planning problem. It should be mentioned that these figures are

obtained when the values of risk-aversion and risk-seeking
parameters are set to be 1. Note that by changing these parameters
different structures could be obtained for MGs. As a result, it can
be concluded that by increasing the MGs number as well as DER
units in the system, the reliability of MGs will be significantly
increased. However, this reliability enhancement will be
accompanied by increasing the investment costs. Therefore, there
is a direct relationship between MGs number and reliability
improvement.
The proposed framework for dynamic planning of MGs is done
under uncertainty by stochastic scenario-based technique. Owing to
the fact that input parameters were as PDF mode, hence the output
parameters such as economic and reliability cost will be as PDF
form. The probability distribution of MGs planning cost including
reliability and economic costs is presented in Fig. 7 for all
proposed risk-based strategies.
Fig. 3 Trade-off Pareto curve of different cases According to the first part of Fig. 7a, it can be seen that the
risk-seeker strategy has the minimum reliability level, so it gives
the most economical structure of MGs and it is obvious that the
risk-averse strategy has the maximum reliability level between
proposed strategies that gives the most reliable structure. On the
other hand, the risk-neutral strategy has intermediate cost for
reliability improvement and it gives the realistic structure which
gives a trade-off between economic and reliability costs. The
second part of this Fig. 7b illustrates the economic cost of MG
planning problem in different risk-based strategies. With regard to
obtained results of proposed strategies, it can be seen that risk-
seeker strategy is the most economical strategy among the
proposed strategies. Hence, it is appropriate for parsimonious
owners that have a low budget to investment. On the other hand,
the risk-averse strategy is the most expensive strategy among the
proposed strategies. Therefore, it is suitable for wealthy owners
that have a high budget to the investment of MGs and risk-neutral
Fig. 4 Optimum DER and switch locations for risk-seeker strategy (case
strategy is the mediocre strategy to MG planning. Also, in this
1)
figure, the obtained results for planning costs by MOPSO
algorithm are compared with MCS approach to verify the
optimum infrastructure, in this case, divides the network into five performance of proposed method.
MGs, in which each MG has the maximum supply security in both In MG planning problem, improvement of technical
interconnected and islanded modes. Fig. 5 shows the achieved characteristics of the system such as voltage profile, power losses,
topology for a risk-neutral strategy that divides the network into and ENS value is very important for DNO as responsible for MGs
eight MGs and Fig. 6 presents a single-line diagram of risk-averse operation. Therefore, to demonstrate the advantages and benefits of
strategy which it shows the most reliable construction as well as the proposed model, Fig. 8 depicts the technical specifications of a
the most expensive model. Owing to this shape, the distribution case study in different cases. Fig. 8a compares the voltage profiles
network is clustered into 11 MGs; this case has the largest number for three proposed planning strategies and traditional method for
of MGs and DERs and it is the most reliable topology for MG MG planning problem in [22]. According to this figure, case 3 has

1756 IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1749-1759
© The Institution of Engineering and Technology 2017
 17521424, 2017, 14, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-rpg.2017.0278, Wiley Online Library on [22/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
the best characteristic. Cases 2 and 1 are in the next ranks and it is reduced. However, the trade-off between obtained profits from the
clear that the proposed approach has better performance compared reduction of power losses and cost of installed DRS units
to previous traditional methods. determines the capacity of DRS to installation on the network.
Fig. 8b presents power losses of the distribution network for Since the DRSs do not generate active power, their effects on the
traditional MG planning approaches and proposed method of this total transferred power between the MGs is not as much as DERs.
paper based on different strategies. According to this figure, case 3 DRSs are added to the distribution networks to reduce power losses
has the best specification from power losses viewpoint. Cases 2 and improve voltage profile. Also, Fig. 8c shows the ENS values of
and 1 are in the next ranks. It is seen that by inserting the DRSs the system in different risk-based strategy and traditional system. It
into the distribution system, the power losses of the network are is clear that by applying the proposed method on the distribution
system the ENS value is decreased considerably as well as its
values are different in various risk strategies which case 3 has
minimum ENS and case 1 has maximum ENS among proposed
strategies.
The obtained results show that the MGs planning problem
highly depends on the planning strategies, which can vary from one
case to another case by changing risk-aversion and risk-seeking
parameters. The proposed strategy for planning reliable and
economic interconnected MGs in a distribution network is a step
towards for having a more reliable and cost efficient smart
distribution network. The proposed model was analysed further
through numerical simulations, where it was demonstrated that MG
investments in the distribution network can provide significant
economic and reliability benefits to owners and consumers. In
addition, the obtained results show that by clustering distribution
system into the number of interconnected MGs, the economic and
Fig. 5 Optimum DER and switch locations for risk-neutral strategy (case technical characteristics of the system will be considerably
2) improved. Also, it is obvious that by increasing the MGs number as
well as DER units, the reliability of system will be increased. The
numerical results also verified that the proposed methodology is
adaptive and can be implemented easily for practical applications.
Also, the obtained results show that by clustering distribution
system into the number of interconnected MGs, the economic and
technical characteristics of distribution system will be improved.
Furthermore, to investigate the impact of risk-seeking and risk-
aversion parameters on the proposed problem, a sensitivity analysis
based on fluctuations of MGs number proportion to variations of
risk-seeking and risk-aversion parameters has been fulfilled, as
shown in Fig. 9. According to this figure, it can be seen that by
increasing the value of risk-aversion parameters (which adjust the
degree of risk-aversion), the MGs number will be increased. In the
other words, it shows the fact that with increasing the number of
MGs the robustness and flexibility of system against uncertainty
will be improved. On the other hand, it is obvious that the small
Fig. 6 Optimum DER and switch locations risk-averse strategy (case 3) MGs are more reliable than large MGs while large MGs are more

Fig. 7 Probability distribution of planning cost in all strategies

(a) Reliability cost (F(2)), (b) economic cost (F(1))

IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1749-1759 1757
© The Institution of Engineering and Technology 2017
 17521424, 2017, 14, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-rpg.2017.0278, Wiley Online Library on [22/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
system as a case study. With respect to simulation results, it can be
concluded that the topology of MGs is highly depended on risk-
based strategies so that by increasing risk-aversion in decision-
making the number of MGs will be increased and vice versa. In the
other words, the robustness and flexibility of system against
operating uncertainties will be augmented proportional to
increasing the MGs number. Increasing reliability and robustness
of system will be accompanied to impose additional expenses in
the planning costs. Therefore, the MGs operator can be selected
various MGs topologies proportional to its own risk-based
strategies as well as available funds. As a result, the risk-seeker
strategy provides the most economical structure for MGs in such a
way that the economic costs be minimised, the risk-neutral strategy
is a compromise structure which creates a trade-off between
economic and reliability perspectives, as well as the risk-averse
strategy, is the most reliable and also the most expensive
construction that gives the maximum number for MGs.
Furthermore, the obtained numerical results shown that clustering
conventional distribution system into a number of interconnected
MGs will be improved the technical characteristics of the system.
Therefore, the proposed framework is a comprehensive model
which considers various aspects of MG planning problem, thereby,
it is appropriate for practical applications.
Fig. 8 Voltage profile, power losses and ENS values based on expected
value for traditional method and proposed strategies 6 References
[1] Chen, C., Duan, S., Cai, T., et al.: ‘Smart energy management system for
optimal microgrid economic operation’, IET Renew. Power Gener., 2011, 5,
(3), pp. 258–267
[2] Masaud, T.M., Nannapaneni, G., Challoo, R.: ‘Optimal placement and sizing
of distributed generation-based wind energy considering optimal self VAR
control’, IET Renew. Power Gener., 2017, 11, (3), pp. 281–288
[3] Soroudi, A., Afrasiab, M.: ‘Binary PSO-based dynamic multi-objective model
for distributed generation planning under uncertainty’, IET Renew. Power
Gener., 2012, 6, (2), pp. 67–78
[4] Arefifar, S.A., Mohamed, Y.A., El-Fouly, T.M.: ‘Supply adequacy-based
optimal construction of microgrids in smart distribution systems’, IEEE
Trans. Smart Grid, 2012, 3, (3), pp. 1491–1502
[5] Zhang, Y., Dong, Z.Y., Luo, F., et al.: ‘Optimal allocation of battery energy
storage systems in distribution networks with high wind power penetration’,
IET Renew. Power Gener., 2016, 10, (8), pp. 1105–1113
[6] Wang, Z., Chen, B., Wang, J., et al.: ‘Robust optimization based optimal DG
placement in microgrids’, IEEE Trans. Smart Grid, 2014, 5, (5), pp. 2173–
2182
[7] Nekooei, K., Farsangi, M., Pour, N.H., et al.: ‘An improved multi-objective
harmony search for optimal placement of DGs in distribution systems’, IEEE
Fig. 9 Sensitivity analysis of variations of MGs number in various risk- Trans. Smart Grid, 2013, 4, (1), pp. 557–567
[8] Arefifar, S.A., Mohamed, Y.I.: ‘DG mix, reactive sources and energy storage
seeking and risk-aversion parameters units for optimizing microgrid reliability and supply security’, IEEE Trans.
Smart Grid, 2014, 5, (4), pp. 1835–1844
economical proportion to small MGs. Therefore, depending on [9] Khodaei, A., Shahidehpour, M.: ‘Microgrid-based co-optimization of
network operator policies and available budgets, various structures generation and transmission planning in power systems’, IEEE Trans. Power
Syst., 2013, 28, (2), pp. 1582–1590
can be obtained with diverse technical and economic specifications [10] Guo, L., Liu, W., Jiao, B., et al.: ‘Multi-objective stochastic optimal planning
for MGs by setting the value of risk-aversion as well as risk- method for stand-alone microgrid system’, IET Gener. Transm. Distrib., 2014,
seeking parameters based on its own strategy. 8, (7), pp. 1263–1273
[11] Arefifar, S.A., Mohamed, Y.A.I., El-Fouly, T.H.M.: ‘Optimum microgrid
design for enhancing reliability and supply-security’, IEEE Trans. Smart
5 Conclusion Grid, 2013, 4, (3), pp. 1567–1575
[12] Kalkhambkar, V., Kumar, R., Bhakar, R.: ‘Joint optimal allocation
This paper presented a new stochastic multi-objective framework methodology for renewable distributed generation and energy storage for
for optimal dynamic planning of interconnected MGs with economic benefits’, IET Renew. Power Gener., 2016, 10, (9), pp. 1422–1429
considering economic, reliability, technical, and environmental [13] Guo, L., Liu, W., Cai, J., et al.: ‘A two-stage optimal planning and design
method combined cooling, heat and power microgrid system’, Energy
aspects. Various concerns including economics, reliability, Convers. Manage., 2013, 74, (3), pp. 433–445
generation and load variability are addressed in the optimal [14] Gu, W., Wu, Z., Bo, R., et al.: ‘Modeling planning and optimal energy
interconnection planning. The proposed approach is applied for management of combined cooling, heating and power microgrid: a review’,
optimally determining the size, site, type, and time of DG, ESS and Int. J. Electr. Power Energy Syst., 2014, 54, (2), pp. 26–37
[15] Atwa, Y., El-Saadany, E., Salama, M., et al.: ‘Optimal renewable resources
DRS units planning simultaneously with optimal allocation of mix for distribution system energy loss minimization’, IEEE Trans. Power
section switch to clustering conventional distribution systems into Syst., 2010, 25, (1), pp. 360–370
a number of interconnected MGs. The partitioning distribution [16] Che, L., Zhang, X., Shahidehpour, M., et al.: ‘Optimal interconnection
systems into a number of MGs is applied with maximum supply planning of community microgrids with renewable energy sources’, IEEE
Trans. Smart Grid, 2015, 4, (3), pp. 1–10
security and power mismatch constraints. To consider the [17] Che, L., Zhang, X., Shahidehpour, M., et al.: ‘Optimal planning of loop-based
uncertainties of wind speed, solar radiation, and load demand, a microgrid topology’, IEEE Trans. Smart Grid, 2015, 4, (3), pp. 1–10
stochastic scenario-based model is employed and backward [18] Arefifar, S.A., Mohamed, Y.A.R., El-Fouly, T.H.M.: ‘Optimized multiple
scenario reduction technique is fulfilled to enhance the trade-off microgrid-based clustering of active distribution systems considering
communication and control requirements’, IEEE Trans. Ind. Electron., 2015,
between the accuracy of the solution and the computational burden. 62, (4), pp. 1–11
Moreover, three different strategies based on risk-seeker, risk- [19] Jabr, R-A., Singh, R., Pal, B-C.: ‘Minimum loss network reconfiguration
neutral, and risk-averse are defined for DNO by the fuzzy using mixed-integer convex programming’, IEEE Trans. Power Syst., 2012,
satisfying method. The effectiveness and efficiency of the proposed 27, (2), pp. 1106–1115
[20] Malekpour, A-R., Niknam, T., Pahwa, A., et al.: ‘Multi-objective stochastic
framework are validated by applying it to the 85-bus distribution distribution feeder reconfiguration in systems with wind power generators and

1758 IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1749-1759
© The Institution of Engineering and Technology 2017
 17521424, 2017, 14, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-rpg.2017.0278, Wiley Online Library on [22/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
fuel cells using the point estimate method’, IEEE Trans. Power Syst., 2013, [31] Abapour, S., Zare, K., Mohammadi-Ivatloo, B.: ‘Dynamic planning of
28, (2), pp. 1483–1492 distributed generation units in active distribution network’, IET Gener.
[21] Shen, X., Shahidehpour, M., Zhu, S., et al.: ‘Multi-stage planning of active Transm. Distrib., 2015, 9, (12), pp. 1455–1463
distribution networks considering the Co-optimization of operation [32] Atwa, Y-M., El-Saadany, E.F.: ‘Optimal allocation of ESS in distribution
strategies’, IEEE Trans. Smart Grid., 2016, 1, (99), pp. 1–10 systems with a high penetration of wind energy’, IEEE Trans. Power Syst.,
[22] Chowdhury, S., Chowdhury, S.P., Crossley, P.: ‘Microgrids and active 2010, 25, (4), pp. 1815–1822
distribution networks’ (IET, London, UK, 2009) [33] Abido, M.A.: ‘Multi objective evolutionary algorithm for electric power
[23] Aien, M., Hajebrahimi, A., Fotuhi-Firuzabad, M.: ‘A comprehensive review dispatch problem’, IEEE Trans. Evol. Comput., 2006, 10, (3), pp. 315–329
on uncertainty modeling techniques in power system studies’, Renew. Sustain. [34] Khalesi, N., Rezaei, N., Haghifam, M.R.: ‘DG allocation with application of
Energy Rev., 2016, 57, (3), pp. 1077–1089 dynamic programming for loss reduction and reliability improvement’, Int. J.
[24] Cai, Y., Lin, J., Wan, C., et al.: ‘Stochastic bi-level trading model for an active Elect. Power Energy Syst., 2011, 33, (8), pp. 288–295
distribution company with DGs and interruptible loads’, IET Renew. Power [35] Wang, S., Li, Z., Wu, L., et al.: ‘New metrics for assessing the reliability and
Gener., 2017, 11, (2), pp. 278–288 economics of microgrids in distribution systems’, IEEE Trans. Power Syst.,
[25] González, M.G., Rodriguez, F.J.R., Jurado, F.: ‘Metaheuristic and 2013, 28, (3), pp. 2852–2861
probabilistic techniques for optimal allocation and size of biomass distributed [36] Ma, G., Xu, G., Chen, Y., et al.: ‘Multi-objective optimal configuration
generation in unbalanced radial systems’, IET Renew. Power Gener., 2015, 9, method for a standalone wind-solar-battery hybrid power system’, IET Renew.
(6), pp. 653–659 Power Gener., 2017, 11, (1), pp. 194–202
[26] Vos, K.D., Driesen, J.: ‘Active participation of wind power in operating [37] Ameli, A., Bahrami, S., Khazaeli, F., et al.: ‘A multi objective particle swarm
reserves’, IET Renew. Power Gener., 2015, 9, (6), pp. 566–575 optimization for sizing and placement of DGs from DG owner's and
[27] Aien, M., Fotuhi-Firuzabad, M., Aminifar, F.: ‘Probabilistic optimal power distribution company's viewpoints’, IEEE Trans. Power Deliv., 2014, 29, (4),
flow in correlated hybrid wind-photovoltaic power systems’, IEEE Trans. pp. 1831–1840
Smart Grid, 2013, 4, (4), pp. 934–943 [38] Shahidehpour, M., Yamin, H., Li, Z.: ‘Market operations in electric power
[28] Moradi, M.H., Eskandari, M., Hosseinian, S.M.: ‘Operational strategy systems: forecasting, scheduling, and risk Management’ (Wiley, New York,
optimization in an optimal sized smart microgrid’, IEEE Trans. Smart Grid, USA, 2002)
2015, 6, (3), pp. 1087–1095 [39] Das, D., Kothari, D.P., Kalam, A.: ‘Simple and efficient method for load flow
[29] Korpaasa, M., Holena, A-T., Hildrum, R.: ‘Operation and sizing of energy solution of radial distribution network’, Electr. Power Energy Syst., 1995, 17,
storage for wind power plants in a market system’, Electr. Power Energy (5), pp. 335–346
Syst., 2003, 25, pp. 599–606
[30] Bahramirad, S., Reder, W., Khodaei, A.: ‘Reliability-constrained optimal
sizing of energy storage system in a microgrid’, IEEE Trans. Smart Grid,
2012, 3, (4), pp. 2056–2062

IET Renew. Power Gener., 2017, Vol. 11 Iss. 14, pp. 1749-1759 1759
© The Institution of Engineering and Technology 2017