Manuscript Details

Manuscript number SWEVO_2019_196

Title Nature-inspired Hybrid Optimization Algorithms for Load Flow Analysis of Droop
Controlled Islanded Microgrids

Short title Hybrid Nature-inspired Algorithms for Load Flow of Islanded Microgrids

Article type Full Length Article

Abstract
Load flow analysis is a significant tool for dynamic analysis of a conventional power system which provides the steady
state values of voltage magnitudes and angles at fundamental frequency. However, due to the absence of slack bus in
an islanded microgrid, modified load flow algorithms should be adopted considering the system frequency as solution
variables. This work proposes the application of nature inspired hybrid optimization algorithms for solving the load flow
problem of islanded microgrids. Several nature-inspired algorithms such as, Genetic Algorithm (GA), Differential
Evolution (DE) algorithm, Flower Pollination Algorithm (FPA) and Grasshopper Optimization Algorithm (GOA) are
separately merged with Imperialistic Competitive Algorithm (ICA) to form four hybrid algorithms named as ICGA, ICDE,
ICFPA and ICGOA and their performances are tested on a modified IEEE 37- Bus microgrid system. To identify the
algorithm with better performance among the proposed ones, a number of statistical tests were conducted using SPSS
statistical analysis software. From the statistical analysis, it was identified that on an average, ICDE takes less number
of iterations and consequently needs lesser execution time compared to others for solving the load flow problem of
islanded microgrid system.

Keywords nature-inspired hybrid optimization; islanded microgrid; load flow analysis

Corresponding Author Saad Mohammad Abdullah

Corresponding Author's Islamic University of Technology (IUT)

Institution

Order of Authors Saad Mohammad Abdullah, Ashik Ahmed, Quazi Nafees Ul Islam

Submission Files Included in this PDF

File Name [File Type]
Cover_Letter.pdf [Cover Letter]

Manuscript_Saad_Mohammad_Abdullah.pdf [Manuscript File]

To view all the submission files, including those not included in the PDF, click on the manuscript title on your EVISE
Homepage, then click 'Download zip file'.
 AUTHOR DECLARATION

We the undersigned declare that this manuscript is original, has not been published before and is not
currently being considered for publication elsewhere.
We wish to draw the attention of the Editor to confirm that there are no known conflicts of interest
associated with this publication and there has been no significant financial support for this work that could
have influenced its outcome.
We confirm that the manuscript has been read and approved by all named authors and that there are no
other persons who satisfied the criteria for authorship but are not listed. We further confirm that the order
of authors listed in the manuscript has been approved by all of us.
We confirm that we have given due consideration to the protection of intellectual property associated with
this work and that there are no impediments to publication, including the timing of publication, with respect
to intellectual property. In so doing we confirm that we have followed the regulations of our institutions
concerning intellectual property.
We understand that the Corresponding Author is the sole contact for the Editorial process (including
Editorial Manager and direct communications with the office). He is responsible for communicating with
the other authors about progress, submissions of revisions and final approval of proofs. We confirm that
we have provided a current, correct email address which is accessible by the Corresponding Author.

Signed by all authors as follows:

Author 1 (Correspoding Author)

Signature :
Date : 24.03.2019
Name : Saad Mohammad Abdullah
(E-mail Address: saadabdullah@iut-dhaka.edu)

Author 2
Signature :
Date : 24.03.2019
Name : Ashik Ahmed

Author 3

Signature :

Date : 24.03.2019
Name : Quazi Nafees Ul Islam
 Nature-inspired Hybrid Optimization Algorithms for Load Flow Analysis of Droop
Controlled Islanded Microgrids

Saad Mohammad Abdullah∗, Ashik Ahmed, Quazi Nafees Ul Islam

Department of Electrical and Electronic Engineering, Islamic University of Technology, Board Bazar, Gazipur 1704, Bangladesh

Abstract
Load flow analysis is a significant tool for dynamic analysis of a conventional power system which provides the steady state values
of voltage magnitudes and angles at fundamental frequency. However, due to the absence of slack bus in an islanded microgrid,
modified load flow algorithms should be adopted considering the system frequency as solution variables. This work proposes the
application of nature inspired hybrid optimization algorithms for solving the load flow problem of islanded microgrids. Several
nature-inspired algorithms such as, Genetic Algorithm (GA), Differential Evolution (DE) algorithm, Flower Pollination Algorithm
(FPA) and Grasshopper Optimization Algorithm (GOA) are separately merged with Imperialistic Competitive Algorithm (ICA) to
form four hybrid algorithms named as ICGA, ICDE, ICFPA and ICGOA and their performances are tested on a modified IEEE 37-
Bus microgrid system. To identify the algorithm with better performance among the proposed ones, a number of statistical tests
were conducted using SPSS statistical analysis software. From the statistical analysis, it was identified that on an average, ICDE
takes less number of iterations and consequently needs lesser execution time compared to others for solving the load flow problem
of islanded microgrid system.
Keywords: nature-inspired hybrid optimization, islanded microgrid, load flow analysis

1. Introduction sponsibility of slack bus among different distributed energy re-

In modern time, microgrid systems have evolved as an or-
ganized and flexible architecture comprising of distributed en-
ergy resources (DERs) which can be a potential replacement
of the aging electrical infrastructure with enhanced operabil-
ity, reliability and reduced CO2 emission to alleviate the envi-
ronmental change [1] [2] [3]. Various controlling techniques
and approaches are required by microgrid for local energy bal-
ancing and maximizing the monetary outcome [4] [5] [6] [7].
Microgrid has gained much acceptance due to its functionality
as aggregated distributed generation (DG) unit in both islanded
and grid connected mode [8]. A Photovoltaic (PV) generator,
battery, loads and precise connections were used first by Na-
tional Technical University of Athens (NTUA) for checking the
feasibility of islanded microgrid [9]. Nonetheless, solution of
the steady state operating point of islanded microgrid through
load flow analysis still remains as a challenge and hence, both
conventional and non-conventional approaches are adopted by
the researchers till date.
In [10], conventional two step load flow approach was used
considering the highest rated local bus as the slack bus. In an-
other study, conventional three phase power flow method with
sequence frame power solver was used for distributing the re-
∗ Corresponding
author.
Email addresses: saadabdullah@iut-dhaka.edu (Saad Mohammad
Abdullah), ashik123@iut-dhaka.edu (Ashik Ahmed),
quazinafees@iut-dhaka.edu (Quazi Nafees Ul Islam)
sources (DERs) [11]. But, in case of islanded microgrids, it
is not possible to assume the system frequency to be constant
due to the absence of slack bus which restricted the use of
conventional load flow techniques for islanded microgrids [12]
[13] [14]. It is essential to have regulation and stabilization of
voltage and frequency level along with real and reactive power
sharing for islanded microgrid [15] [16]. Considering steady-
state system frequency as one of the power flow variables; in
[13], a three-phase power flow method was proposed using the
Newton-trust region method considering decentralized droop
control schemes for the DERs. A modified version of the con-
ventional Gauss-Seidel method was used in [17], where load
demands were shared by the DGs keeping the system frequency
and voltages within specified limits. Modified Newton-Raphson
method was used in [18], where generator bus was formulated
as droop bus considering the absence of slack bus for an is-
landed microgrid. In most of these studies, the system model
was developed in stationary reference frame considering the
voltages and currents as phasors which only allowed steady
state analysis of the system but lacked necessary information
for linearizing a dynamic model at the system level. There are
angular differences between synchronous reference frames of
the generating units which were not determined by these meth-
ods. For a consistent system level operating point these angle
differences are critical to obtain. In [19], the system model was
developed in synchronous reference frame and a quasi-Newton
method was introduced to solve the load flow analysis consid-
ering the system frequency, reference frame angles and voltage

Preprint submitted to Elsevier March 24, 2019

 Fig. 1. Block diagram of control strategy of droop controlled inverter for individual DER
magnitudes as the load flow variables. Most of these load flow
techniques use gradient-based algorithms which requires evalu-
ation of derivatives for a series of complex equations. Gradient-
based techniques often fail to obtain a global solution as there
is possibility of a local minima solution being considered as a
global minima solution [20].
Nature-inspired optimization algorithms can be good alter-
native to the gradient-based techniques in obtaining a global so-
lution. Multi-solution based evolutionary algorithms have even
better possibility of avoiding a local optimum by exploring a
larger portion of the search space [21]. For droop-controlled is-
landed microgrid, a load flow algorithm was introduced where
particle swarm optimization (PSO) technique was used to deter-
mine the droop parameters [14]. Hybridization among different
evolutionary algorithms can result in better exploration of the
search space. In [22], hybridized ICGA algorithm was used for
load flow analysis by incorporating imperialist competitive al-
gorithm (ICA) with the multi-solution based genetic algorithm
(GA). Fairly good performance was obtained in the aforemen-
tioned work; however, the system modeling was done in sta-
tionary reference frame and only one hybridization algorithm
was used. Comparison among different hybrid algorithms were
not considred and hence there is still room for further research.
In this work, application of hybridized nature-inspired opti-
mization algorithms was proposed to solve the load-flow prob-
lem of droop controlled islanded microgrids in the synchronous
reference frame. To come up with a fairly optimal result; sev-
eral multi solution-based optimizers, such as, genetic algorithm
(GA), differential evolution (DE) algortithm, flower pollina-
tion algorithm (FPA) and grasshopper optimization algorithm
(GOA) were adopted and each of them were separately com-
2
bined with ICA to obtain four hybrid algorithms. Statistical
analysis was conducted to compare the performance of each of
the hybridized algorithms to obtain the optimal solution. The
rest of the paper is organized as follows. In Section 2, the devel-
opment of the mathematical model of the droop controlled mi-
crogrid system is reviewed based on different literatures. Sec-
tion 3 contains discussions regarding the load flow approach
and the proposed hybrid algorithms. Comparison among the
proposed hybrid algorithms along with the load flow results ob-
tained for a case study system is carried out in Section 4. In the
end, the concluding remarks are added in Section 5.
2. Mathematical Model of Microgrid
Multiple distributed energy resources (DER) are aggregated
in a microgrid system. In most of the cases the DERs provide
DC output. Thus, before connecting to a bus, the DERs are in-
terfaced with inverters for the conversion from DC to AC. As
a result, developing mathematical model of the inverter along
with its associated controllers is important for the analysis of
microgrid systems. The discussion in this section describes
the dynamic model of droop controlled inverters along with the
necessary load and line equations to develop the complete mi-
crogrid model. The modeling technique described in this sec-
tion is based on the studies carried out in [19] [23] [24].
2.1. Inverter Model
2.1.1. Reference Frame Transformation
The inverter is coupled to the inverter bus through an LCL
filter as shown in Fig. (1). The three phase capacitor voltage
and inductor currents of the LCL filter are transformed into
synchronous reference frame through abc to dq transformation.
This is accomplished using the following transformation voq∗ = Vn − nQ (10)

cos(θ) cos(θ − 2π ) cos(θ + 2π )  

  where, ωn represents the nominal frequency set point and Vn
 
vq   3 3  v  represents the nominal set point of the q − axis output volt-
2π   a 

  2  2π age. The droop constants m and n are calculated from the given
vd  =  sin(θ) sin(θ − ) sin(θ + )  vb  (1)
vo 3  3 3  v range of frequency and voltage magnitude.
 1 1 1  c
2 2 2 ωmax − ωmin
m= (11)
where, vq and vd are respectively the q − axis and d − axis Pmax
components of the filter capacitor voltage and va , vb and vc are Voq,max − Voq,min
voltages in the stationary reference frame. The reference frame n= (12)
Qmax
transformation of the filter inductor current, il and the output
current, io can be obtained using similar relationships. In (1), 2.1.4. Voltage Controllers
θ represents the transformation angle which is calculated by a The voltage controller compares between the reference and
phase locked loop (PLL). Details of this transformation tech- measured values of frequency and voltage, and generates the
nique is given in [25]. reference values of the output filter inductor currents through a
2.1.2. Phase Locked Loop (PLL) pair of PI controller. The voltage controller equations are
A dq−based PLL is used to measure the phase and fre- ϕ̇d = ω∗ − ωPLL (13)
quency. In the PLL, a proportional-integral (PI) controller is
used to force the d − axis component of the capacitor voltage to ∗
ild = k pv,d (ω∗ − ωPLL ) + kiv,d ϕd (14)
become 0. This results in the steady-state voltage magnitude to
be equal to its q − axis component. The PLL equations are ϕ̇q = voq∗ − voq (15)
∗
φ̇PLL = −vod (2) ilq = k pv,q (voq∗ − voq ) + kiv,q ϕq (16)
where, ϕd and ϕq represent the integrator states of the voltage
ωPLL = −k p,PLL vod + ki,PLL φPLL (3) controllers. The proportional and integral gains of the respec-
θ̇ = ωPLL (4) tive d − axis and q − axis controllers are represented by k pv,d ,
k pv,q , kiv,d and kiv,q .
where, φPLL is the integrator state of the PI controller. k p,PLL and
ki,PLL are respectively the proportional and integral gain, ωPLL 2.1.5. Current Controllers
is the calculated frequency and θ is the transformation angle. ∗
The reference values of filter inductor current (ildq ) are com-
2.1.3. Power Controller pared with the measured filter inductor current using the current
In the power controller, the instantaneous active (p) and re- controllers. Voltage references are provided as outputs by these
active (q) power outputs are calculated first from the capacitor current controllers, which are used to generate switching sig-
voltage and output current. Then, the droop controller gener- nals for the inverter. Two PI controllers are used for this pur-
ates the voltage magnitude and frequency references based on pose. The cross-coupling terms appearing due to the reference
the active and reactive power values. The instantaneous active frame transformation are also eliminated by these controllers.
(p) and reactive (q) power outputs are given by The current controller equations are
∗
3 γ̇d = ild − ild (17)
p= (vd iod + vq ioq ) (5)
2 ∗ ∗
vid = k pc,d (ild − ild ) + kic,d γd − ωn L f ilq (18)
3
q = (vq iod − vd ioq ) (6) γ̇q =
∗
ilq − ilq (19)
2
Average active (P) and reactive (Q) power values are calculated ∗ ∗
viq = k pc,q (ilq − ilq ) + kic,q γq + ωn L f ild (20)
by passing the instantaneous power outputs through a first order
low pass filter (LPF). The filter equations are where, γd and γq are the integrator state of the current con-
trollers. k pc,d , k pc,q , kic,d and kic,q represent the proportional and
Ṗ = ωc p − ωc P (7) integral gains of the d − axis and q − axis controllers respec-
tively.
Q̇ = ωc q − ωc Q (8)
2.1.6. LCL Filter Equations
where, ωc is the cut-off frequency of the low pass filter.
The inverter output is connected to the microgrid through an
The P − ω and Q − V droop equations are used to generate LC filter and coupling inductor. The filter inductor (L f ), filter
the frequency reference, ω∗ and q − axis voltage magnitude ref- capacitor (C f ) and coupling inductor (Lc ) collectively form the
erence, v∗oq respectively. The equations are LCL filter. The parasitic resistance of these components are
also considered for the inverter model as shown in Fig. (1).
ω∗ = ωn − mP (9) The filter dynamics are governed by the following equations
3
 1
i̇ld = (−r f ild + vid − vod ) + ω∗ ilq (21)
Lf
1
i̇lq = (−r f ilq + viq − voq ) − ω∗ ild (22)
Lf
1
i̇od = (−rc iod + vod − vbd ) + ω∗ ioq (23)
Lc
1
i̇oq = (−rc ioq + voq − vbq ) − ω∗ iod (24)
Lc
1
v̇od = (ild − iod ) + ω∗ voq + Rd (i̇ld − i̇od ) (25)
Cf
1
v̇oq = (ilq − ioq ) − ω∗ vod + Rd (i̇lq − i̇oq ) (26)
Cf
Fig. 2. Line configuration between two buses
In (23) and (24), vbd and vbq represent the bus voltages at the
grid side of the coupling inductor.
2.1.7. Local to Global Reference Frame Transformation
Each inverter model is developed in its own local reference
frame. For modeling a microgrid system with several inverters,
it is necessary to translate the values defined in the local refer-
ence frame of an inverter to a common reference frame called
the global reference frame. This transformation can be achieved
by " # " #
Fq f
= R(δ) q (27)
Fd fd
" #
cosδ −sinδ
R(δ) = (28)
sinδ cosδ
where δ is the angular difference between local and global ref-
erence frame. In (27), lowercase letter is used to indicate local
reference frame and uppercase letter is used to indicate global
reference frame. The angle δ is defined by

δ̇ = ω − ωPLL (29) Fig. 3. Line and load currents at a particular bus

where, ω is the frequency of the global reference frame and

ωPLL is the frequency measured by PLL of a particular inverter. Line currents between two adjacent buses i and j connected
Often, the reference frame of the first inverter in the system is through a transmission line can be described by
chosen as the global reference frame. In this work, we set ω1 = 1
ωPLL and δ1 = 0, which implies δ̇1 = 0. For other inverters, I˙line,di j = (Vbdi − Vbd j − Rlinei j Iline,di j ) + ωIline,qi j (32)
Llinei j
ω 6= ωPLL and δ has to be calculated following (29).
1
2.2. Equations for Load and Line I˙line,qi j = (Vbqi − Vbq j − Rlinei j Iline,qi j ) − ωIline,di j (33)
Llinei j
To complete modeling the entire microgrid model, it is nec-
essary to formulate the state equations for load and line in the where, 0 ≤ i < j ≤ N. N represents the total number of buses
global reference frame. Loads can be of constant impedance in the system.
type which is basically a combination of resistors and inductors 2.3. Bus Voltage Equations
(RL loads) as depicted in Fig. (2). The equations of RL load The virtual resistance method can be used to find the ex-
connected to the ith bus can be described by pression of the bus voltages in the global reference frame. To
1 determine the bus voltage at the ith bus using virtual resistance
I˙load,di = (Vbdi − Rloadi Iload,di ) + ωIload,qi (30) method, a high resistance connection is considered between bus
Lloadi
i and ground. This high resistance actually represents open cir-
1 cuit. Ideally this resistance should be infinite, but typically a
I˙load,qi = (Vbqi − Rloadi Iload,qi ) − ωIload,di (31) large value of resistance is considered for modeling purpose.
Lloadi

4
 Bus voltage expressions are typically dependent on any in- The elements of the state matrix, Amg and input matrix, Bmg is
coming inverter output current, incoming or outgoing line cur- defined by equations (2), (7), (8), (13), (15), (17), (19), (21)
rents and currents flowing through the connected load as shown to (26), (29) and (30) to (33). The input vector, umg can be
in Fig. (3). The voltage expression at bus i can be expressed as represented in terms of the bus voltages as
h iT
i−1
X N
X umg = Vbd1 Vbq1 ... . . . Vbd N VbqN (44)
Vbdi = rn (Iodi − Iload,di + Iline,d ji − Iline,dik ) (34)
j=1 k=i+1 3. Load Flow Analysis
In the conventional load flow analysis the voltage and fre-
i−1 N
X X quency of the slack bus is constant. But, in case of islanded
Vbqi = rn (Ioqi − Iload,qi + Iline,q ji − Iline,qik ) (35)
microgrid the concept of slack bus is not applicable as the sys-
j=1 k=i+1
tem frequency is variable. As a result, for an islanded micro-
where, j ≤ i < k ≤ N. grid with droop controlled inverter; the system frequency has to
2.4. Overall Microgrid Model be considered as one of the load flow variables along with the
voltage magnitudes and reference angles contributed by each
The equations of the inverter, load and line model described
inverter in the system. The state variable, x can be described in
so far can be used to represent the overall microgrid model. The
terms of the load flow variables as
state vector of a droop controlled inverter connected to the ith
bus can be formulated as h
x = ω δ2 . . . . . . δK voq1 . . . . . . voqK
i
(45)
xinvi = [δi Pi Qi ϕdi ϕqi γdi γqi
(36) where, ω , δ and voq represents the system frequency, reference
ildi ilqi vdi vqi iodi ioqi φPLLi ]T angle and voltage magnitude respectively and K represents the
If there are total p number of inverters connected to the sys- total number of inverters in the system. The constraints of the
tem, then the combined state vector of all the inverters can be objective problem can be defined as
represented as
h i ωmin ≤ ω ≤ ωmax
xinv = xinv1 xinv2 . . . . . . xinv p (37)
δmin ≤ δ ≤ δmax
Considering the equations of the load model described from
vmin max
oq ≤ voq ≤ voq
(30) to (31), the state vector of a load connected at the ith bus is
h iT 3.1. Problem Formulation
xloadi = Iload,di Iload,qi (38)
The objective of the load flow analysis is to minimize the
If there are total N number of buses in the microgid system and sum of absolute mismatch values of active and reactive power
one specific RL load is connected to each bus, then the com- of the inverters. The objective function can be written as
bined state vector of all the loads will be    
X K  XK 
f (x) =  ∆Pi  +  ∆Qi  (46)
xload = [Iload,d1 Iload,q1 . . . . . . Iload,d N Iload,qN ]T (39)
 i=1   i=1 

The equations of line current from (32) to (33) can be used to For a droop controlled inverter, the power mismatch equa-
represent the states of a line between buses i and j as tions are the difference between the inverter output power cal-
h iT culated at the global reference frame and the reference values
xlinei j = Iline,di j Iline,qi j (40) set by the droop controllers. For the ith inverter the active and
reactive power mismatch equations are
The overall state vector of all the lines can be represented as
3 (ωn − ω)
xinvi = [Iline,d1 j Iline,q1 j . . . . . . Iline,dkN Iline,qkN ]T (41) ∆Pi = (Vodi Iodi + Voqi Ioqi ) − (47)
2 mi
where, 1 < j ≤ k ≤ N (Vn − voqi )
3
∆Qi = (Voqi Iodi − Vodi Ioqi ) − (48)
2 ni
Based on the combined state vectors of the droop controlled
inverters, loads and lines; the states of the overall microgrid To determine the power mismatch values, a set of equations
model can be described as has to be solved. Solving these equations include the calcula-
h i tion of bus voltages and inverter output currents. For simplicity,
xmg = xinv xload xline (42) inverter output voltage across the capacitor can be considered as
a voltage source behind its coupling impedance as shown in Fig.
State-space model of the whole microgrid system will have the
(4). The process of determining the power mismatch values is
following form
described in the following steps.
ẋmg = Amg xmg + Bmg umg (43)
5

Fig. 4. Steady-state equivalent circuit of inverter model at bus i
Fig. 5. Norton equivalent circuit of steady-state inverter model at bus i
Step 1: First of all, following equations (27) and (28); the
d − axis and q − axis components of the inverter output volt-
age is transformed in the global referemce frame using the
reference angle (δ). Then, the output voltage of the ith in-
verter in terms of a complex quantity can be calculated as
Vodi = sin(δi )voqi
Voqi = cos(δi )voqi
Voi = Vodi + jVoqi
Step 2: Before determining the bus voltages it is required
to calculate the current injected to a particular bus. For this
study, only constant impedance loads are considered. So, the
inverter are the only sources to inject current to their respec-
tive buses. The current injected by the inverters can be easily
calculated by transforming the circuit shown in Fig. (4) to
its Norton equivalent as shown in Fig. (5). Then, the current
injected by the ith inverter is calculated as
Voi
Iin ji = IS C i =
Zc (ω)
Step 3: The bus voltages can now be calculated from the
injected currents as
Vb = Zbus (ω)Iin j (53)
For islanded microgrids the bus impedance matrix is a func-
tion of frequency and it has to be updated at each iteration.
For an N-bus system the vector of injected currents at each
bus is given by

Iin j p due to inverter at bus p (p ≤ N)

Iin j = 

(54)
0
 otherwise
Step 4: After determining the bus voltages, the output cur-
rent of the ith inverter can be determined by
Voi − Vbi
I oi = (55)
Zc (ω)
The d − axis and q − axis components of the inverter output
current in the global reference frame is given by
Iodi = <{Ioi } (56)
Ioqi = ={Ioi } (57)
Step 5: The equations from (49) to (57) are sufficient to cal-
culate the active and reactive power mismatch values for each
inverter by solving the equations described in (47) and (48).
Thus, the values of the power mismatch equations can be
used to evaluate the objective function as indicated in equation
(46).
3.2. Algorithm for Load Flow Analysis
Gradient based algorithms are often difficult to apply in op-
timization problems where the derivation process of the prob-
lem is complex and these algorithms also suffer from local min-
ima entrapment. These drawbacks can be compensated by the
use of non-gradient based evolutionary algorithms and by com-
bining them to form hybrid algorithms can lead to better global
optimization. Considering this fact; in this study, several hybrid
(49) algorithms based on nature-inspired optimization algorithms have
been designed to perform the load flow analysis of islanded mi-
(50) crogrids. These hybrid algorithms have been designed keeping
(51) the imperialist competitive algorithm (ICA) as the main frame.
For better exploration of the search space, four other evolu-
tionary algorithms namely; genetic algorithm (GA), differential
evolution (DE), flower pollination algorithm (FPA) and grasshop-
per optimization algorithm (GOA) were separately combined
with ICA to obtain four hybrid algorithms ICGA, ICDE, ICFPA
and ICGOA. Detailed explanation of each evolutionary algo-
rithm can be found in [20] [21] [26]. The flowchart of the pro-
posed algorithm is given in Fig. (6). The process of applying
the proposed hybrid algorithms for the load flow analysis is de-
scribed in the following steps.
(52)
Step 1: Initialization of the system data of islanded micro-
grid.
6
Fig. 6. Flowchart of the algorithms used for load flow analysis

7
Step 2: Generation of initial population for the state vari-
ables as stated in (45). In case of ICA, the population indi-
viduals are called countries. For this study, the total number
of countries was set to 100.
Step 3: For each country, solve equations (49) to (57) and
calculate the values of active and reactive power mismatch
for each inverter using equations (47) and (48). Then, find
the value of the objective function using equation (46) for
each country in the population.
Step 4: Next step is to generate the empires based on the
fitness values of the countries. First, the countries are sorted
according to their objective function values. Then, depend-
ing on the fitness values of the countries; the empires are gen-
erated by selecting the imperialists and assigning colonies to
them. For, this study among the 100 countries, 5 were chosen
as imperialists and rest of them were assigned as colonies to
the imperialists.
Step 5: The positions of the colonies are then moved towards
the position of imperialist by a process called assimilation.
Step 6: In this step, the positions of some of the colonies are
modified randomly by doing revolution.
Step 7: If there is a colony which has a lower fitness value
than the imperialist, then their positions are interchanged.
This process is referred as intra-empire competition.
Step 8: The hybridization process of ICA with GA, DE, FPA
and GOA is carried out in this step. Four distinct hybrid algo-
rithms are obtained by following the four cases as indicated
in Fig. (6).
Case 1; ICGA is accomplished by following case 1,
where the updated empires from the previous step is set
as parents in GA. Then, the positions of the imperialists
and colonies are updated through crossover, mutation
and selection process [20].
Case 2; The resulting algorithm will be ICDE if case 2
is followed. Here, the empires are assigned as agents of
the DE algorithm. The positions of the countries (im-
perialists and colonies) are updated through mutation,
crossover and selection process of DE [20].
Case 3; The empires are assigned as population of flow-
ers in FPA. The population is then modified by mimick-
ing either global or local pollination process depending
upon a probability switch [20]. If this case is adopted
the resulting hybrid algorithm will be ICFPA.
Case 4; In order to attain ICGOA, in this case the em-
pires are set as the positions of grasshoppers in GOA.
the positions are updated by simulating the swarming
behavior of grasshoppers through a mathematical model
[21].
Selecting any one of these four cases in this step is the prime
difference among the four hybrid algorithms. The rest of the
steps are similar for each algorithm.
8
Fig. 7. Single-line diagram of the modified IEEE 37-bus system
Step 9: Intra-empire competition is performed again. The
objective function values of the imperialists and their respec-
tive colonies are monitored and the imperialist is replaced
with a colony if the colony attains a lower objective function
value.
Step 10: In this step, first of all the total fitness of each
empire is calculated. Then, in imperialistic competition, the
weakest colonies are identified and are given to the empires
which have the most likelihood to possess them. If an empire
ends up with no colonies then that empire will be eliminated.
Step 11: The solution set which is providing the best fitness
value will be identified in this step.
Step 12: If the stopping conditions are satisfied, then the
whole process will be terminated. Otherwise, the calcula-
tions will be repeated from step 5. For this study, the opti-
mization process will terminate if any one of the following
two stopping criterion is satisfied.
1. If the value of the best fitness is less than a pre-specified
threshold (ε) value which is set to 10−5 for this study.
2. If the number of iteration is less than a pre-specified
value of maximum number of iterations. For this study,
the maximum number of iteration was set to 50.
4. Case Study
In section 2, the dynamic model of microgrid system was
discussed. The proposed load flow algorithms were outlined in
section 2. In order to validate the applicability of these algo-
rithms for load flow analysis, the modified IEEE 37-bus system
was considered as a case study system. This section contains
the discussions on load flow analysis of the modified IEEE 37
bus system using the proposed algorithms. This section also in-
cludes a comparative study on the results obtained through each
algorithm to identify the best one.
The standard IEEE 37-bus system is modified by connect-
ing seven inverters at different bus locations as indicated in [27].
The single line diagram of the modified 37-bus system is shown
in Fig. (7). The seven inverters are connected at buses 15, 18,
22, 24, 29, 33 and 34 as indicated by the green dots in Fig.
(7). Only constant impedance loads are considered in this case
 Table 1
Inverter bus locations, power ratings and droop co-efficients [19]

i Bus Pmax (kW) Qmax (kVAR) mi −1 ni −1

1 15 15 15 2387.3 1250
2 18 8 8 1273.2 666.7
3 22 10 10 1591.5 833.3
4 24 15 15 2387.3 1250
5 29 8 8 1273.2 666.7
6 33 10 10 1591.5 833.3
7 34 15 15 2387.3 1250

study. The branch and load parameters are considered to be the

same as used in [27]. For all the inverters, a nominal voltage
of vn = 170V was chosen and nominal frequency was set to
ωn = 2π60 rad/s. The maximum power ratings and the droop
co-efficients for each inverter is given in Table 1. Fig. 8. Convergence graph for the best results of each algorithm

Next, load flow analysis was performed by applying ICGA,

ICDE, ICFPA and ICGOA algorithms separately to the modi-
fied IEEE 37-bus system. The simulations were performed us-
ing a personal computer with a processor of intel core i7-8550
at 1.8GHz and with a installed RAM of 8GB. Each algorithm
was executed for 30 times. For each run, the number of itera-
tions to reach the stopping criterion and the overall execution
time was recorded. These data are summarized in Table 2 in or-
der to compare the performance of each algorithm. As it can be
seen from Table 2, the average number of iterations required by
ICDE is 27, which is lower compared to the other algorithms.
On an average, ICDE requires 95.512 sec to complete the opti-
mization process for load flow analysis, which is much less than
the other three algorithms. The convergence graph of each al-
gorithm considering their best and worst results is respectively
shown in Fig. 8 and Fig. 9. From both the graphs, it can be seen
that all the algorithms attain very high fitness values in the ini- Fig. 9. Convergence graph for the worst results of each algorithm
tial iterations and gradually after completing several iterations
the fitness values obtain convergence. From Fig. 8 and Fig. 9, it hypothesis (H1 ) assumes that the means of the data sets are not
is also evident that for both the cases; ICDE is providing faster equal. From both Table 3 and Table 4, it can be seen that, the
convergence compared to the other algorithms. p − values of t − tests for all the pairs are smaller than 0.05 ex-
cept the pair of ICFPA and ICGOA, which indicates that these
To further validate the results, SPSS statistics software was two algorithms does not possess significant differences among
used to perform statistical analysis of the obtained data. To themselves. Based on the studies done so far, it can be con-
demonstrate the uniqueness of each algorithm, independent sam- cluded that ICDE is significantly unique from ICGA, ICFPA
ples t − tests were performed to compare means of each algo- and ICGOA. As indicated in Table 2; in terms of the average
rithm. In this study, data from two algorithms were defined values of the number of iterations and the execution time re-
as grouping variables at a time. Table 3 and Table 4 show the quired to complete the load flow analysis, ICDE exhibit better
t−test results based on the required number of iterations and the accuracy than the other algorithms. The results of the load flow
execution time as the comparison variable respectively. When- analysis of the modified IEEE 37-bus system using ICDE is
ever independent samples t − test is performed in SPSS, the given in Table 5. Among the 30 independent runs the best result
software also generates a corresponding F − test results which is tabulated here. The p.u. value of the steady state frequency is
determine whether the data sample of two groups have equal 0.9962. The inverter bus locations and per unit (p.u.) values of
variances or not. If the p − value of the F − test is greater the inverter output voltages along with the values of the active
than the significance level of 0.05, then the group variances are and reactive powers generated by each inverter is tabulated in
considered to be equal. Otherwise, equal variances can not be Table 6. It has been observed that all bus voltages lie within 5%
assumed. For the t − test, the null hypothesis (H0 ) assumes that of the rated bus voltage.
the mean values of the data sets are equal and the alternative

9
 Table 2
Iterations and Time required by ICGA, ICDE, ICFPA and ICGOA algorithms

Best Result Average Result Worst Result

Parameter
ICGA ICDE ICFPA ICGOA ICGA ICDE ICFPA ICGOA ICGA ICDE ICFPA ICGOA
No. of Iter. 23 23 26 29 30 27 32 42 40 30 39 50
Run Time 80.215 79.659 90.597 101.586 105.528 95.512 146.213 159.265 140.287 120.108 305.142 569.403
(sec)

Table 3
Results of t-test based on number of Iterations

F-test t-test for equality of means

Method
F Sig. Mean Difference t df Sig.(2-tailed)
ICDE-ICFPA 12.819 0.001 -5.033 -7.560 40.905 0.000
ICDE-ICGA 21.065 0.000 -2.567 -3.067 36.191 0.004
ICDE-ICGOA 77.330 0.000 -15.033 -10.687 31.387 0.000
ICFPA-ICGA 1.992 0.164 2.467 2.483 58 0.016
ICFPA-ICGOA 33.548 0.000 -10.000 -6.644 39.740 0.000
ICGA-ICGOA 18.166 0.000 -12.467 -7.849 46.147 0.000

Table 4
Results of t-test based on computational time

F-test t-test for equality of means

Method
F Sig. Mean Difference t df Sig.(2-tailed)
ICDE-ICFPA 38.033 0.000 -50.701 -3.941 29.805 0.000
ICDE-ICGA 8.777 0.004 -10.016 -3.137 44.320 0.003
ICDE-ICGOA 5.209 0.026 -63.754 -4.259 29.592 0.000
ICFPA-ICGA 28.486 0.000 40.685 3.110 31.810 0.004
ICFPA-ICGOA 1.375 0.246 -13.052 -0.665 58 0.509
ICGA-ICGOA 3.360 .072 -53.738 -3.545 58 0.001

Table 5 5. Conclusion
Load Flow Results obtained by ICDE In this paper, application of nature-inspired hybrid optimiza-
tion algorithms was demonstrated for the efficient solution of
Load Flow Calculated Load Flow Calculated load flow problem of islanded microgrids. For solving the load
Variable Value Variable Value flow problem, an objective function was formulated based on
ω (rad/s) 375.5576 voq1 (V) 165.472 the absolute summation of errors in the real and reactive power
δ2 (deg) -1.3682 voq2 (V) 168.236 generations from the inverter based microgrid sources and the
δ3 (deg) -2.9203 voq3 (V) 169.621 minimization was solved using hybrid optimization techniques;
δ4 (deg) -0.3786 voq4 (V) 169.717 namely ICGA, ICDE, ICFPA and ICGOA. The hybridization
δ5 (deg) 1.0963 voq5 (V) 166.986
was performed with a view to improving the global searching
δ6 (deg) 2.2311 voq6 (V) 169.932
δ7 (deg) 2.4701 voq7 (V) 168.741
capability by enhanced exploration of the search space. A case
study was conducted on the modified IEEE 37-bus system con-
Table 6 taining seven droop controlled inverters. The performance of
p.u. output voltage and generated powers at each inverter the proposed load flow algorithms were compared through a
series of statistical tests. Based on the statistical test ICDE was
Output Generated power found to exhibit better performance than the other algorithms
i Bus
voltage Active Reactive in terms of the required number of iterations and the execu-
(p.u.) power power tion time. Therefore, ICDE can be regarded as a prospective
(kW) (kVAR) alternative to the conventional load flow techniques of islanded
1 15 0.9734 1.856 0.685 microgrids.
2 18 0.9896 3.477 0.357
3 22 0.9978 6.595 -1.424
References
4 24 0.9983 3.235 1.330 [1] A. Hirsch, Y. Parag, J. Guerrero, Microgrids: A review of technologies,
5 29 0.9823 0.905 0.803 key drivers, and outstanding issues, Renewable and Sustainable Energy
6 33 0.9996 1.685 3.493 Reviews 90 (2018) 402–411. doi:10.1016/j.rser.2018.03.040.
7 34 0.9926 0.729 5.899 [2] R. M. Cuzner, G. Venkataramanan, The status of dc micro-grid pro-
tection, in: 2008 IEEE Industry Applications Society Annual Meeting,

10
 IEEE, 2008, pp. 1–8. doi:10.1109/08IAS.2008.382. Electrical Energy Systems 26 (2016) 2727–2743. doi:10.1002/etep.2231.
[3] T. L. Vandoorn, J. C. Vasquez, J. De Kooning, J. M. Guerrero, L. Vande- [23] N. Pogaku, M. Prodanovic, T. C. Green, Modeling, analysis
velde, Microgrids: Hierarchical control and an overview of the control and testing of autonomous operation of an inverter-based micro-
and reserve management strategies, IEEE industrial electronics magazine grid, IEEE Transactions on power electronics 22 (2007) 613–625.
7 (2013) 42–55. doi:10.1109/MIE.2013.2279306. doi:10.1109/TPEL.2006.890003.
[4] J. M. Guerrero, P. C. Loh, T.-L. Lee, M. Chandorkar, Advanced control [24] M. Rasheduzzaman, J. A. Mueller, J. W. Kimball, An accurate small-
architectures for intelligent microgrids—part ii: Power quality, energy signal model of inverter-dominated islanded microgrids using dq refer-
storage, and ac/dc microgrids, IEEE Transactions on Industrial Electron- ence frame, IEEE Journal of Emerging and Selected Topics in Power
ics 60 (2013) 1263–1270. doi:10.1109/TIE.2012.2196889. Electronics 2 (2014) 1070–1080. doi:10.1109/JESTPE.2014.2338131.
[5] F. Katiraei, R. Iravani, N. Hatziargyriou, A. Dimeas, Microgrids [25] P. Krause, O. Wasynczuk, S. D. Sudhoff, S. Pekarek, Analysis of electric
management, IEEE power and energy magazine 6 (2008) 54–65. machinery and drive systems, volume 75, John Wiley & Sons, 2013.
doi:10.1109/MPE.2008.918702. [26] E. Atashpaz-Gargari, C. Lucas, Imperialist competitive algorithm: an
[6] A. Madureira, J. Pereira, N. Gil, J. P. Lopes, G. Korres, N. Hatziar- algorithm for optimization inspired by imperialistic competition, in: 2007
gyriou, Advanced control and management functionalities for multi- IEEE congress on evolutionary computation, IEEE, 2007, pp. 4661–4667.
microgrids, European Transactions on Electrical Power 21 (2011) 1159– doi:10.1109/CEC.2007.4425083.
1177. doi:10.1002/etep.407. [27] L. Luo, S. V. Dhople, Spatiotemporal model reduction of inverter-based
[7] M. A. Hossain, H. R. Pota, M. J. Hossain, F. Blaabjerg, Evolution of islanded microgrids, IEEE Transactions on Energy Conversion 29 (2014)
microgrids with converter-interfaced generations: challenges and oppor- 823–832. doi:10.1109/TEC.2014.2348716.
tunities, International Journal of Electrical Power & Energy Systems 109
(2019) 160–186. doi:10.1016/j.ijepes.2019.01.038.
[8] F. Katiraei, M. R. Iravani, P. W. Lehn, Micro-grid autonomous operation
during and subsequent to islanding process, IEEE Transactions on power
delivery 20 (2005) 248–257. doi:10.1109/PES.2004.1373266.
[9] D. Georgakis, S. Papathanassiou, N. Hatziargyriou, A. Engler, C. Hardt,
Operation of a prototype microgrid system based on micro-sources
quipped with fast-acting power electronics interfaces, in: 2004
IEEE 35th Annual Power Electronics Specialists Conference (IEEE
Cat. No. 04CH37551), volume 4, IEEE, 2004, pp. 2521–2526.
doi:10.1109/PESC.2004.1355225.
[10] H. Nikkhajoei, R. Iravani, Steady-state model and power flow analysis of
electronically-coupled distributed resource units, IEEE Transactions on
Power Delivery 22 (2007) 721–728. doi:10.1109/TPWRD.2006.881604.
[11] M. Z. Kamh, R. Iravani, A sequence frame-based distributed
slack bus model for energy management of active distribution net-
works, IEEE Transactions on smart Grid 3 (2012) 828–836.
doi:10.1109/TSG.2012.2188915.
[12] Z. Shuai, Y. Sun, Z. J. Shen, W. Tian, C. Tu, Y. Li, X. Yin, Microgrid
stability: Classification and a review, Renewable and Sustainable Energy
Reviews 58 (2016) 167–179. doi:10.1016/j.rser.2015.12.201.
[13] M. M. A. Abdelaziz, H. E. Farag, E. F. El-Saadany, Y. A.-R. I. Mohamed,
A novel and generalized three-phase power flow algorithm for islanded
microgrids using a newton trust region method, IEEE Transactions on
Power Systems 28 (2013) 190–201. doi:10.1109/TPWRS.2012.2195785.
[14] A. Elrayyah, Y. Sozer, M. E. Elbuluk, A novel load-flow analysis for
stable and optimized microgrid operation, IEEE Transactions on Power
Delivery 29 (2014) 1709–1717. doi:10.1109/TPWRD.2014.2307279.
[15] T. Senjyu, Y. Miyazato, A. Yona, N. Urasaki, T. Funabashi, Opti-
mal distribution voltage control and coordination with distributed gen-
eration, IEEE Transactions on power delivery 23 (2008) 1236–1242.
doi:10.1109/TPWRD.2007.908816.
[16] J. He, Y. W. Li, An enhanced microgrid load demand sharing strat-
egy, IEEE Transactions on Power Electronics 27 (2012) 3984–3995.
doi:10.1109/TPEL.2012.2190099.
[17] F. Mumtaz, M. Syed, M. Al Hosani, H. Zeineldin, A simple and
accurate approach to solve the power flow for balanced islanded mi-
crogrids, in: 2015 IEEE 15th International Conference on Environ-
ment and Electrical Engineering (EEEIC), IEEE, 2015, pp. 1852–1856.
doi:10.1109/EEEIC.2015.7165454.
[18] F. Mumtaz, M. Syed, M. Al Hosani, H. Zeineldin, A novel approach to
solve power flow for islanded microgrids using modified newton raph-
son with droop control of dg, IEEE transactions on sustainable energy 7
(2016) 493–503. doi:10.1109/TSTE.2015.2502482.
[19] J. A. Mueller, J. W. Kimball, An efficient method of determining operat-
ing points of droop-controlled microgrids, IEEE Transactions on Energy
Conversion 32 (2017) 1432–1446. doi:10.1109/TEC.2017.2719580.
[20] X.-S. Yang, Nature-inspired optimization algorithms, Elsevier, 2014.
[21] S. Saremi, S. Mirjalili, A. Lewis, Grasshopper optimisation algorithm:
theory and application, Advances in Engineering Software 105 (2017)
30–47. doi:10.1016/j.advengsoft.2017.01.004.
[22] M. Abedini, A novel algorithm for load flow analysis in island microgrids
using an improved evolutionary algorithm, International Transactions on

11