You are on page 1of 9

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3085674, IEEE
Transactions on Industry Applications

A Center of Mass Determination for Optimum Placement of


Renewable Energy Sources in Microgrids
Salman Harasis, Member, IEEE, Hassan Abdelgabir, Member, IEEE, Yilmaz Sozer, Senior Member, IEEE,
Mithat Kisacikoglu, Member, IEEE, Ali Elrayyah, Member, IEEE

Abstract— The objective of this paper is to determine the curtailment minimization, enhancement of load bus voltage
location and size of the renewable energy sources (RESs) to be profile, system stability improvement, system capacity
installed in microgrid networks that have multiple distributed maximization, and cost minimization. Transient stability is
generators (DGs). Since RESs have faster dynamics and low considered a major issue in microgrid operation as microgrids
inertia, microgrid sources could encounter large disturbances have less amount of energy reserve and inertia compared to
in their power production when RESs are deployed. When the conventional power systems. Several solutions have been
disturbance magnitude exceeds a certain limit, the stable discussed in the literature to enhance the performance and
operation of the microgrid can be compromised. In this study, stability of microgrids. Optimum placement of energy storage
the impact of installing a certain RES capacity on a microgrid
devices has been considered as a potential candidate to
is numerically evaluated. A novel methodology is developed to
minimize microgrid transient energy unbalance [6, 7]. The
accurately determine the location and size of a newly installed
RES in a microgrid. The proposed methodology considers the
existence of conventional energy sources has been discussed
complete system dynamics in the process of locating and sizing to increase the reliability of microgrids [5-7]. The optimal
the installed RES. The stability of the microgrid is examined localization and capacity sizing of DGs have been addressed
using a modified microgrid load flow model. The proposed in several pieces of research considering different aspects and
method automatically checks for the optimum RES placement objectives. A fault-based approach for optimal placement of
to ensure minimal effect on microgrid stability. The energy storage devices in a microgrid is reported in [6]. The
effectiveness of the proposed methodology is verified authors introduced structure preserving energy function to
experimentally on a microgrid testbed. improve the transient stability of the investigated microgrid.
The optimum bus wherein the storage devices to be installed
Keywords— load flow analysis, microgrid, renewable energy has been decided based on the voltage drop on each bus. The
sources, stability, the center of the mass point. research carried out in [7] analyzed the transient stability of a
microgrid in the presence of a battery storage element. The
I. INTRODUCTION authors have proven that the deployment of distributed battery
Due to the technological advancements in power storage elements yields a shorter clearing time when a three-
electronics along with the recent innovations in the small- phase fault occurs compared to a larger centralized
scale autonomous generation units that utilize renewable deployment. However, the optimal sizing and the impact of
energy sources (RESs), the concept of microgrids has deploying storage elements on the dynamic power-sharing of
emerged as a future technology for distribution networks [1]. the DGs have not been discussed in [6, 7].
These autonomous distributed generation systems have Heuristic and stochastic optimization algorithms have
several benefits on utility, customers, and the environment been utilized in several studies to optimally operate
[1, 2]. Integrating distributed energy sources, increasing microgrids. The authors in [8] employed a particle swarm
system reliability, reducing line losses, decreasing depletion optimization approach to find the optimum size and location
of fossil resources, and reducing adverse environmental of DG units in a microgrid in both islanded and grid-
impacts are examples of the benefits of microgrids [3]. connected modes. The optimal sizing is performed based on
The stability of large-scale power networks becomes a the cost analysis of the microgrid considering the fuel cost
critical concern when installing RESs into microgrids that where the objective function is set to minimize the overall
contain several DGs. Due to the intermittent nature and low system cost. However, the stability and dynamic operation of
inertia of inverter-based DGs, microgrids may experience the system have not been considered in the objective function
poor voltage and frequency dynamics during disturbances [4, of the proposed optimization. Das et al. [9] investigated the
5]. The disturbances may severely affect the system stability effect of connecting a PV system to an islanded microgrid that
if a sudden change in the generation or a large load swing contains internal combustion engines and micro-turbines.
occurs. As the magnitude of the disturbance exceeds certain Multi-objective swarm optimization has been employed for
thresholds, power system instability could be induced cost optimization. The optimization has been built to
resulting in system failure which deteriorates the reliability minimize the operational cost considering only the steady-
of microgrids. Therefore, the placement process of RES state operation. In [10], the optimal operation of DGs was
should account for uncertain and intermittent production investigated in microgrids using another heuristic algorithm
characteristics of RES that affect the stability, reliability, and called the imperial competitive algorithm. The algorithm
optimal operation of the system. accounts for the uncertainty in the load and renewable power
The placement of RES in a microgrid has been discussed generators. The work relied only on the statistical modeling
in the literature to tackle several aspects, such as power of the system and the probability density functions to prove
S. Harasis is with Department of Electrical and Computer Engineering, the effectiveness of the proposed algorithm. The authors in
The University of Akron, Akron, OH 44325, and Tafila Technical University [11] performed energy management for a microgrid using the
Tafila, Jordan (email:skh60@uakron.edu), H. Abdelgabir, and Y. Sozer are bee-colony-algorithm for optimal placement of renewable
with Department of Electrical and Computer Engineering, The University of sources in microgrids. The algorithm was employed to
Akron, Akron, OH 44325 (e-mails: hsa25@zips.uakron.edu,
ys@uakron.edu). M. Kisacikoglu is with Department of Electrical and analyze the economic dispatch of distributed generators to
Computer Engineering, The University of Alabama, Tuscaloosa, AL 35487
(e-mail: mkisacik@ua.edu). A. Elrayyah is with Qatar Environment and
Energy Research Institute, Doha, Qatar (email:aelrayyah@qf.org.qa)

0093-9994 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: The University of Akron Libraries. Downloaded on June 23,2021 at 16:11:07 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3085674, IEEE
Transactions on Industry Applications

2
decrease the running cost. Another stochastic-based technique the center of mass (COM) point as RES from the selected
has been reported in [12] where evolutionary programming location has a similar impact on each DG. The work proposed
has been utilized for optimal placement of renewable DGs in in this paper considers the placement of RESs based on the
a radial distribution system. The technique considered the actual dynamic response of each source and circuit
steady-state expected power loss and voltage drop on the component in the microgrid. The output power dynamics of
buses as a criterion for optimum placement. In [13], heuristic all sources and the physical circuit configuration are
optimization framework, called gravitational search considered to determine the optimal location of a new RES.
algorithm, is proposed for a grid connected PV system to Table I compares the proposed methodology with the
minimize energy losses in the power system. The authors
state-of-art-optimization techniques. The developed table
examined the methodology under uncertain and variable
shows the approaches discussed in the literature with the
loading conditions. The algorithm determined the optimal
placement of the PV system based on the expected status and considered objectives and operating modes. In summary, the
charging time of plug-in electric vehicle battery. In [8-13], the objective and the contributions of the paper are to:
optimizations considered steady-state operation only. The • develop a novel methodology to optimally locate RES
transients and the effect of the renewable generators on in microgrids considering their dynamic operations;
system stability have not been discussed, and the effect of • provide an accurate solution for microgrid dynamics
large-scale integration of PV sources has not been addressed. by considering the circuit physical and control
In [14], the researchers considered the problem of low-inertia parameters;
zone in a real power system equipped with virtual inertia • evaluate the stability of the microgrid using their
sources interfaced with RESs. small-signal models and eigenvalues analysis;
An H2 norm-based optimization approach is used to study • develop a deterministic approach to precisely find the
the problem of optimal placement and tuning of the installed center of the mass point of microgrids; and
units that yields better power and frequency excursion • evaluate the output power-sharing ratio between DGs
dynamics. The work reported in [15] dealt with the optimal in steady-state and transient operating conditions.
location and size of several DGs based on the power loss The paper is organized as follows: Section II presents the
sensitivities and Kalman filter algorithm. The authors load flow analysis (LFA) model for microgrids. In Section III,
developed an optimal locator index to determine the optimal the proposed COM point determination method is presented.
locations of newly installed DG. The developed index Section IV provides a case study on a seven-node microgrid
depends mainly on the weighting factors that are determined network. Section V shows the experimental validation of the
according to user preferences. The optimization is tested proposed strategy. The discussion of the effectiveness of the
under a small step change in the output power of the DG that proposed work is provided in Section VI. The conclusion and
is equal to a fraction of the rated network capacity. Moreover, summary are presented in Section VII.
the dynamic effect on the other DGs has not been discussed.
Another methodology is developed in [16] for optimal II. LOAD FLOW ANALYSIS OF MICROGRID
placement and sizing of wind and PV in distribution networks The standard LFA for conventional power systems is a
under uncertain load profile and renewable generation. The tool used to calculate the voltages at each bus and the power
optimization aims to minimize the cost of the network by flows in the network when the parameters of network
decreasing the imported power from the utility grid and the admittances, generators, and loads are given [22]. For each
overall losses of the system. In [17], the size and placement bus, voltage magnitude, 𝑉𝑉𝑖𝑖 , voltage phase, 𝛿𝛿𝑖𝑖 , real power, 𝑃𝑃𝑖𝑖 ,
of renewable energy resources are performed based on long- and reactive power, 𝑄𝑄𝑖𝑖 can be determined. Depending on
term steady-state optimization for power loss minimization whether the bus is a swing bus, load bus, or voltage-controlled
without incorporating the system dynamics. bus, two of these variables are known and the other two are
The study reported in [18] presents the impact of a unknowns. A number of equations equal to the number of
renewable-based microgrid on low voltage weak distribution unknowns are developed for each bus. The solution is
networks. The existence of extreme conditions in typical obtained by solving the equations numerically.
operation is tested by applying changes in average voltage In [23], a new load-flow analysis (LFA) model is
values and line losses. The authors in [19] proposed a method proposed for islanded microgrids. The authors consider the
to determine the optimal location and size of a newly installed droop settings of all the integrated sources and their roles in
DG based on a predetermined level of system reliability. The microgrid management. In this study, the LFA model
authors introduce an energy index of reliability to calculate proposed in [23] is adopted to obtain the optimum location
the expected energy loss in the system and thereby to and size of the RES to be installed in a microgrid. The impact
determine the required size of the DG at a specific location. of installing certain RES capacity at specific nodes on the
In [20], the authors developed a sensitivity-based stability of the microgrid is analyzed. The modified load flow
optimization method to determine the optimal location of DG. analysis model is adopted to relate the variation of RES
The sensitivity of Lagrange multipliers is used to determine production to the operation of other DGs. Furthermore, the
the ranges of the effectiveness of the optimal solutions. In proposed method searches for the optimum placement of the
[21], the authors decided the best candidate places based on installed RES to minimize the impact of the dynamics on
the sensitivity analysis of different objective functions microgrid stability.
calculating Lagrange multipliers of the solutions considering In this section, the modified LFA model for droop-
different load buses. controlled microgrids is considered to find the optimum
The target of the proposed optimum placement of RES is placement and sizing of the RES to be installed. It is important
to make all DGs installed in a microgrid see an equal level of to mention that microgrid LFA (MLFA) is a tool that
disturbance based on the microgrid equivalent circuit. considers the droop settings of all sources without a need for
Therefore, the dynamic effect of installing new RES at a a swing bus to regulate the voltage. As in the standard load
specific node can be equally distributed among all DGs. This, flow, a set of equations is developed for each node depending
in turn, reduces the impact of RES transients on system on its type. Therefore, according to the MLFA, there are three
stability. The optimum placement point of RES is defined as types of nodes in the microgrid network: source node,

0093-9994 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: The University of Akron Libraries. Downloaded on June 23,2021 at 16:11:07 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3085674, IEEE
Transactions on Industry Applications

3
TABLE I
STATE-OF-THE-ART COMPARISON TABLE
*Dynamic Hardware
Ref Approach Application Objective Function
Operation Validation
[6] Voltage post fault status Grid connected system Minimizing the fault clearing time  
Islanded and grid-
[8] PSO Minimizing the cost  
connected modes
Minimizing the cost and
[9] Multi-objective PSO Islanded microgrid  
maximizing the PV power
[10] Imperial competitive algorithm Grid connected Minimizing the cost  
Bee-colony-algorithm and Markov
[11] Islanded microgrid Minimizing the cost  
chain
Power loss and voltage drop
[12] Evolutionary programming Radial system  
minimization
[13] Gravitational search algorithm Grid connected Minimizing the energy losses  
[14] H2 based optimization algorithm Grid connected Minimizing the frequency excursion  
[15] Kalman filter based optimization Grid connected Minimizing the power loss  
2-stage MATLAB-based MinMax
[16] Grid connected Minimizing the cost  
optimization
[17] PSO Grid connected Minimizing the power loss and cost  
[19] Simulated annealing Islanded microgrid Maximizing the system reliability  
Grid connected/
Lagrange multipliers-based
[20]-[21] different IEEE bus Minimizing the cost  
optimization
systems
Minimizing the transient power
Proposed COM Islanded Microgrid  
oscillation
*Dynamic operation (: steady state operation only, : transient analysis is included)

coupling node, and load node. For the source node, the source of the network as shown in Fig. 1(b). The third type is load
is connected as shown in Fig. 1(a) where four different node where the load is connected as pictured in Fig. 1(c).
variables are unknowns: voltage magnitude 𝑉𝑉𝑠𝑠𝑠𝑠 , voltage For both coupling and load nodes, there are only two
phase 𝛿𝛿𝑖𝑖 , real power 𝑃𝑃𝑖𝑖 , and reactive power 𝑄𝑄𝑖𝑖 . With the help unknowns associated: the node voltage magnitude and its
of voltage and frequency droop relations, the unknowns can phase angle, which can be obtained by solving (4) and (5) for
be found by solving the power and voltage equations given in the coupling node and load node, respectively. The
(1) through (5) with the frequency and voltage droop relations parameters, 𝑍𝑍𝑐𝑐𝑛𝑛 is the impedance of the line between node n
are 𝑓𝑓𝑠𝑠𝑠𝑠 = 𝑓𝑓𝑟𝑟𝑟𝑟𝑟𝑟,𝑖𝑖 − 𝐾𝐾𝑝𝑝𝑝𝑝 𝑃𝑃𝑖𝑖 and 𝑉𝑉𝑠𝑠𝑠𝑠 = 𝑉𝑉𝑟𝑟𝑟𝑟𝑟𝑟,𝑖𝑖 − 𝐾𝐾𝑞𝑞𝑞𝑞 𝑄𝑄𝑖𝑖 . and node n+1, and 𝑍𝑍𝐿𝐿 is the load impedance for the
designated load node. Equations (1)-(5) represents the
𝑉𝑉𝑠𝑠𝑠𝑠 𝑒𝑒 −𝑗𝑗𝛿𝛿𝑖𝑖 − 𝑉𝑉𝑛𝑛 𝑒𝑒 −𝑗𝑗𝛿𝛿𝑛𝑛 complete MLFA model. Based on the power flow modeling,
𝑃𝑃𝑖𝑖 + 𝑗𝑗𝑄𝑄𝑖𝑖 − 𝑉𝑉𝑠𝑠𝑠𝑠 𝑒𝑒 𝑗𝑗𝛿𝛿𝑖𝑖 � �=0 (1)
𝑍𝑍𝑠𝑠𝑠𝑠 ∗ a system of 𝑁𝑁𝑠𝑠 sources and 𝑁𝑁𝐿𝐿 loads will have a (6NS+2NL-1)
number of equations. Solving these equations numerically
𝑉𝑉𝑠𝑠𝑠𝑠 − �𝑉𝑉𝑟𝑟𝑟𝑟𝑟𝑟,𝑖𝑖 − �𝐾𝐾𝑞𝑞𝑞𝑞 𝑄𝑄𝑖𝑖 �� = 0 (2) using Gauss-Seidel or Newton Raphson methods provides the
LFA solutions for the microgrid network.
�𝐾𝐾𝑝𝑝𝑝𝑝𝑝𝑝 𝑃𝑃𝑟𝑟𝑟𝑟 � − �𝐾𝐾𝑝𝑝𝑝𝑝 𝑃𝑃𝑖𝑖 � = 0 (3)
III. CENTER OF MASS DETERMINATION FOR MICROGRID
𝑉𝑉𝑛𝑛 𝑒𝑒 𝑗𝑗𝛿𝛿𝑛𝑛 − 𝑉𝑉𝑛𝑛+1 𝑒𝑒 𝑗𝑗𝛿𝛿𝑛𝑛+1 𝑉𝑉𝑛𝑛 𝑒𝑒 𝑗𝑗𝛿𝛿𝑛𝑛 − 𝑉𝑉𝑠𝑠𝑠𝑠 𝑒𝑒 𝑗𝑗𝛿𝛿𝑖𝑖 The objective of the proposed method is to determine the
+
𝑍𝑍𝑐𝑐𝑐𝑐+1 𝑍𝑍𝑠𝑠𝑠𝑠 (4) location and capacity of a source to be installed. A microgrid
𝑉𝑉𝑛𝑛 𝑒𝑒 𝑗𝑗𝛿𝛿𝑛𝑛 − 𝑉𝑉𝑛𝑛−1 𝑒𝑒 𝑗𝑗𝛿𝛿𝑛𝑛−1 configuration, shown in Fig. 2, is considered where several
+ +⋯=0 conventional generators are connected. Node-x is the variable
𝑍𝑍𝑐𝑐𝑐𝑐
node, and its location is determined by the proposed
|𝑉𝑉𝐿𝐿 |2 𝑗𝑗𝛿𝛿𝐿𝐿
𝑉𝑉𝐿𝐿−1 𝑒𝑒 −𝑗𝑗𝛿𝛿𝐿𝐿−1 − 𝑉𝑉𝐿𝐿 𝑒𝑒 −𝑗𝑗𝛿𝛿𝐿𝐿 algorithm.
∗ − 𝑉𝑉𝐿𝐿 𝑒𝑒 �� � The location, size, and production variability of the RESs
𝑍𝑍𝐿𝐿 𝑍𝑍𝑐𝑐𝑐𝑐 ∗
𝑉𝑉𝐿𝐿+1 𝑒𝑒 −𝑗𝑗𝛿𝛿𝐿𝐿+1 − 𝑉𝑉𝐿𝐿 𝑒𝑒 −𝑗𝑗𝛿𝛿𝐿𝐿 (5) are the key factors to determine whether the system can
+� ∗ �+⋯� maintain the stability or not. The impact of installing certain
𝑍𝑍𝑐𝑐𝑐𝑐+1 RES capacity at specific nodes such as node-x of Fig. 2 on the
=0 stability and operation of the system is evaluated using two
The phase angle of one of the sources must be taken as a parts: (i) microgrid load flow analysis (MLFA) tool that
reference and therefore, the node of the reference source has considers the droop settings of all sources; and (ii)
only three unknowns which indicates that (3) can be canceled ddetermining the impact of a step-change in the RES
for that node. The second node type is the coupling node production at the terminals of each generator
where the source coupling impedance is connected to the rest
DG3

DG1 Load1

Load2 DG4

Node x
DG2
(a) (b) (c)
Fig. 1. The types of microgrid nodes: (a) the source node; (b) the
coupling node; and (c) the load node. Fig. 2. A microgrid structure with several DGs.

0093-9994 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: The University of Akron Libraries. Downloaded on June 23,2021 at 16:11:07 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3085674, IEEE
Transactions on Industry Applications

4
In the first part, for a particular deployment and location
of the newly installed RES, the MLFA model provided in (1)-
(5) is first modified in such a way that the equation of the real
part of (1) can be neglected since the RES source node has a
known real power. Thus, the modified MLFA would have
6𝑁𝑁𝑠𝑠 + 2𝑁𝑁𝐿𝐿 − 2 equations and could be represented as:
𝛧𝛧(𝑃𝑃1 , 𝑄𝑄1 , 𝑉𝑉𝑠𝑠1 , 𝛿𝛿1 , … ) = 0 (6)
In the second part, the state-space model of the microgrid
including the RES is developed. Since the main concern of
the proposed method is to observe the impact of installing a
new DG on the production of the other DGs, only the power- Fig. 3. Block diagram of the proposed method.
sharing dynamics are considered here, knowing that the
dynamics of voltages and currents are much faster and have a given power capacity on a certain node in the microgrid can
a negligible effect on the system stability [22, 24, 25]. be evaluated in terms of the risk for instability. Furthermore,
Therefore, only three state variables for each source are the proposed methodology automatically checks for different
considered. The phase angle of 𝑖𝑖 𝑡𝑡ℎ source can be given based placements of the new DG until it finds a specific point that
on the droop relation as 𝜃𝜃𝚤𝚤̇ = 𝑓𝑓𝑠𝑠𝑠𝑠 = 𝑓𝑓𝑟𝑟𝑟𝑟𝑟𝑟,𝑖𝑖 − 𝑚𝑚𝑃𝑃𝑃𝑃 𝑃𝑃𝑖𝑖𝑖𝑖 in which serves as a COM of the microgrid where disturbances at the
generator's terminals would be proportional to their capacity
𝑃𝑃𝑖𝑖𝑖𝑖 is the real power after averaging the instantaneous power
and their controllers’ robustness. According to the relative
𝑃𝑃𝑖𝑖 , so the phase-shift between the 𝑖𝑖 𝑡𝑡ℎ source and reference transient power-sharing among the sources, when the
source (𝑟𝑟𝑟𝑟) can be obtained by: placement is performed on the COM point, microgrid stability
𝛿𝛿𝚤𝚤̇ = 𝑓𝑓𝑠𝑠𝑠𝑠𝑠𝑠 − 𝑓𝑓𝑠𝑠𝑠𝑠 = 𝐾𝐾𝑝𝑝𝑝𝑝𝑝𝑝 𝑃𝑃𝑟𝑟𝑟𝑟 − 𝐾𝐾𝑝𝑝𝑝𝑝 𝑃𝑃𝑖𝑖 (7) is expected to be enhanced as all generators experience the
The instantaneous real and reactive power are low pass same level of perturbation and no source is brought closer to
filtered using a first-order filter as given in (8) and (9), the instability region than the others.
respectively where 𝜔𝜔𝑐𝑐 is the cut-off frequency of the first The proposed methodology is summarized in Fig. 3. The
methodology is divided into two stages: (i) the transfer
order low pass filter [25, 26].
function between the power-sharing of the newly installed
𝑃𝑃𝚤𝚤𝚤𝚤̇ = −𝜔𝜔𝑐𝑐 𝑃𝑃𝑖𝑖𝑖𝑖 + 𝜔𝜔𝑐𝑐 𝑃𝑃𝑖𝑖 (8) RES and the other DGs is generated; (ii) the response of the
̇
𝑄𝑄𝚤𝚤𝚤𝚤 = −𝜔𝜔𝑐𝑐 𝑄𝑄𝑖𝑖𝑖𝑖 + 𝜔𝜔𝑐𝑐 𝑄𝑄𝑖𝑖 (9) transfer functions obtained in the first stage is analyzed. Each
case from 1 to X represents a specific location and size of the
Equations (7)-(9) represent the state-space model of the
installed RES. The total number of cases is ∑𝑚𝑚 𝑛𝑛
1 𝐶𝐶 ∑1 𝑅𝑅 where
droop-controlled microgrid system and can be expressed as:
C is the number of source nodes (Ns), and R is the size of the
𝑥𝑥̇ = 𝐴𝐴𝐴𝐴 + 𝐵𝐵1 𝑢𝑢 + 𝐵𝐵2 𝑃𝑃𝑅𝑅𝑅𝑅𝑅𝑅 (10) RES. The steps of the first stage are explained in the previous
𝑇𝑇
analysis where the second performs the following steps: (i)
where 𝑥𝑥 = �𝛿𝛿1 , 𝑃𝑃1𝑓𝑓 , 𝑄𝑄1𝑓𝑓 , … , 𝛿𝛿𝑅𝑅𝑅𝑅𝑅𝑅 , 𝑃𝑃𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 , 𝑄𝑄𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 , … , � is the obtain the transfer functions between power-sharing of each
state vector; and 𝑢𝑢 = �𝑃𝑃1 , 𝑄𝑄1, , … , 𝑄𝑄𝑅𝑅𝑅𝑅𝑅𝑅 , … , 𝑃𝑃𝑖𝑖 , 𝑄𝑄𝑖𝑖, … � is the
𝑇𝑇 DG to the new DER for a particular placement; (ii) obtain the
step response of the transfer functions obtained in stage-1 and
dependent inputs. The inputs can be expressed in terms of the
the associated overshoot of each transfer function; (iii)
state variables, and 𝑃𝑃𝑅𝑅𝑅𝑅𝑅𝑅 is the independent input of the
calculate the standard deviation (S) for the obtained
system which is the new integrated DG. Applying small
perturbation method to the system in (10) and Taylor’s series percentage overshoot values, where 𝑆𝑆 = �∑𝑖𝑖(𝑥𝑥𝑖𝑖 − 𝑥𝑥̅ )2 ÷ 𝑛𝑛 ,
expansion to (6), result in the linearized state-space model of 𝑥𝑥̅ is the mean value of the peak times, and 𝑛𝑛 is the number of
the system as: peak times (equivalent to the number of DGs in the
microgrid); (iv) check for different possible placements and
𝛥𝛥𝛥𝛥̇ = 𝐴𝐴 𝛥𝛥𝛥𝛥 + 𝐵𝐵1 𝛥𝛥𝛥𝛥 + 𝐵𝐵2 𝛥𝛥𝛥𝛥 (11)
repeat from step 1; and if it is the last possible placement,
𝜕𝜕𝜕𝜕(𝑥𝑥, 𝑢𝑢, 𝑃𝑃) 𝜕𝜕𝜕𝜕(𝑥𝑥, 𝑢𝑢, 𝑃𝑃) 𝜕𝜕𝜕𝜕(𝑥𝑥, 𝑢𝑢, 𝑃𝑃)
𝛥𝛥𝛥𝛥 + 𝛥𝛥𝛥𝛥 + 𝛥𝛥𝛥𝛥 = 0 (12) compare the standard deviation (𝑆𝑆) of all placements and the
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 one with the lowest 𝑆𝑆 value to obtain the optimum placement.
𝛥𝛥𝛥𝛥, the threshold condition can be expressed as in (13) and If the calculated standard deviation is zero, the placement is
substituted into (11) as in (14): considered a COM as the other DGs experience an equal level
−1 of disturbance.
𝜕𝜕𝜕𝜕(𝑥𝑥, 𝑢𝑢, 𝑃𝑃) 𝜕𝜕𝜕𝜕(𝑥𝑥, 𝑢𝑢, 𝑃𝑃)
𝛥𝛥𝛥𝛥 = − � � � 𝛥𝛥𝛥𝛥 The standard deviation is used as a criterion to determine
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 the sizing and the best location according to the equivalent
(13)
𝜕𝜕𝜕𝜕(𝑥𝑥, 𝑢𝑢, 𝑃𝑃) circuit of the system seen from the RES connection points.
+ 𝛥𝛥𝛥𝛥� The best size is determined empirically by sweeping different
𝜕𝜕𝜕𝜕
𝛥𝛥𝛥𝛥̇ values of the installed RES droop coefficient until the
−1 standard deviation conditions are satisfied. The flowchart in
𝜕𝜕𝜕𝜕(𝑥𝑥, 𝑢𝑢, 𝑃𝑃) 𝜕𝜕𝜕𝜕(𝑥𝑥, 𝑢𝑢, 𝑃𝑃)
= �𝐴𝐴 − 𝐵𝐵1 � � � �� 𝛥𝛥𝛥𝛥 Fig. 4 describes the procedure of the process.
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
−1 (14) IV. CASE STUDY
𝜕𝜕𝜕𝜕(𝑥𝑥, 𝑢𝑢, 𝑃𝑃) 𝜕𝜕𝜕𝜕(𝑥𝑥, 𝑢𝑢, 𝑃𝑃)
+ �𝐵𝐵2 − 𝐵𝐵1 � � � �� 𝛥𝛥𝛥𝛥 A seven node microgrid with two conventional DGs and
𝜕𝜕𝑢𝑢 𝜕𝜕𝜕𝜕
= 𝐴𝐴𝑛𝑛𝑛𝑛𝑛𝑛 𝛥𝛥𝛥𝛥 + 𝐵𝐵𝑛𝑛𝑛𝑛𝑛𝑛 𝛥𝛥𝛥𝛥 = 0 four loads distributed across different nodes is considered in
this study. The RES will be installed at one of the three source
Based on the modeling carried out, the stability of the
nodes. Fig. 5 represents the actual microgrid, where RES is
system is evaluated using the eigenvalues analysis and the connected to one of the possible nodes (i.e., source node 6).
transfer functions between the newly installed DG and the The other possible nodes are source nodes 2 and 4.
other sources. Therefore, the integration of specific DG with

0093-9994 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: The University of Akron Libraries. Downloaded on June 23,2021 at 16:11:07 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3085674, IEEE
Transactions on Industry Applications

5
Initial size of RES TABLE III
{size=fun(KpRES)} DROOP COEFFICIENTS
Parameter Symbol Value
Run the proposed 𝑘𝑘𝑝𝑝1 2e-4 rad/W.sec
algorithm (Fig. 3) Frequency 𝐾𝐾𝑝𝑝2 3e-4 rad/W.sec
droop
coefficients
𝐾𝐾𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 0.7e rad/W.sec (2.0 kW)
-4
Sweep Kp RES

Obtain the transfer functions between the 𝐾𝐾𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 0.8e-4 rad/W.sec (1.75 kW)
output power of the sources [Hx(s)] Voltage droop
𝑘𝑘𝑞𝑞1 , 𝐾𝐾𝑞𝑞2 , 𝐾𝐾𝑞𝑞𝑞𝑞𝑞𝑞𝑞𝑞 1.0e-4 V/Var
coefficients
Calculate the value of “S”
TABLE IV.
From the obtained Hx(s)
EIGENVALUES OF THE MICROGRID
𝜆𝜆1 , 𝜆𝜆2 = −18.8075 ± 𝑗𝑗81.59 𝜆𝜆3 , 𝜆𝜆4 = −18.8075 ± 𝑗𝑗49.59
No Is the
Yes 𝜆𝜆5 = −45.4176 𝜆𝜆6 = −36.6354
candidate solution is
COM? 𝜆𝜆7 = −40.4149 𝜆𝜆8 = −37.7785
End TABLE V.
Fig. 4. The overall procedure of the proposed algorithm. THE CORRESPONDING NODAL VOLTAGES
Node Voltage
Vs2 δ2 Vn4 Vn5 Vn6 Vn7 DG1 (Vs1) 119.9688 ∠0o
LS2 Zc5 Zc6
DG2 Zc4 DG2 (Vs2) 119.2014 ∠-0.05o
RES (VRES) 119.1397∠-0.06o
ZL3

ZL4

Vn3 Vn1
LRES

119.195∠-0.05o
ZL2

Vn2 Vn2 119.161∠ -0.07o


δRES

RES

Vn1 Vn3 118.199∠-0.04o


VRES

Zc1 Ls1 Vn4 118.153∠ -0.03o


ZL1

Vn5 117.545∠ -0.08o


δ1

Vn6 117.223 ∠ -0.06o


DG1
Vs1

Vn7 117.149∠ -0.07o


Fig. 5. Microgrid with a new integrated RES. 80

Table II provides the physical parameters of the


60

40

microgrid, and Table III contains the frequency and voltage 20

droop coefficients of the sources. The RES is considered to 0


Imag

have the highest rating among the sources. Therefore, the -20

frequency droop coefficient of RES, i.e., 𝐾𝐾𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 , is designed to -40

be the smallest compared to the other coefficients. The droop -60

is designed small to examine the effect of large-scale -80

integration of RES on microgrid stability.


-45 -40 -35 -30 -25 -20 -15
Real

First, the case when a 2.0 kW RES is placed at node 6 Fig. 6. Low-frequency dominant modes mapping of the microgrid.
shown in Fig. 5 is taken into consideration. The MLFA model
The results show that the step response of H1 (s) and
is developed for the system configuration, and the solution
provides the voltages and phase angles at each node as well H2 (s), the case where the RES is installed at node 2, has an
as the real and reactive power of each source. Then, the approximately similar response. The percentage of overshoot
solution of MLFA is taken along with the state-space model values are 10.4% and 9.65%, respectively with less than 1.0%
of the microgrid provided in (10) to conduct the stability difference in the output power. This indicates that DG1 and
analysis using eigenvalues analysis. The obtained eigenvalues DG2 experienced a comparable level of perturbation, and they
of the study system are shown in Table IV, and the mapping undergo similar transients when RES is integrated at node 2.
of microgrid low-frequency eigenvalues is depicted in Fig. 6. For instance, in the case where RES is installed at node 4, the
The corresponding nodal voltages are given in Table V. percentage overshoot of H1 (s) and H2 (s) are 11.2% and
The sensitivity analysis of the microgrid eigenvalues 8.02%, respectively, which are pretty different than each
performed in [27] has confirmed that the slowest mode, 𝜆𝜆1 is other. When RES is integrated at node 6, the percentage
associated with the real powers of the sources in the system. overshoot of H1 (s) and H2 (s) are 11.8% and 5.2%,
The placement at node 2 and 4 cases are also checked in order respectively, which represents the worst-case scenario as the
to obtain the COM point. For each case, two transfer functions power overshoot of DG1 is twice the overshoot of DG2. This
are obtained as: would cause a transient overcurrent to DG1. Accordingly, for
𝑃𝑃1𝑓𝑓 (𝑠𝑠) 𝑃𝑃2𝑓𝑓 (𝑠𝑠) the optimal placement of RES and reliable configuration of
𝐻𝐻1 (𝑠𝑠) = , 𝐻𝐻2 (𝑠𝑠) = (15) the microgrid, the RES is to be installed at node 2.
𝑃𝑃𝑅𝑅𝑅𝑅𝑅𝑅 (𝑠𝑠) 𝑃𝑃𝑅𝑅𝑅𝑅𝑅𝑅 (𝑠𝑠)
Table VI provides the calculated standard deviation of the
According to (15), the impact of RES power dynamics on percentage overshoot values for each case (node 2, node 4,
the other sources can be determined. The response of the and node 6). %𝑂𝑂𝑂𝑂1 is the percentage overshoot of the transfer
transfer functions are compared by installing the RES at function H1(s), and %𝑂𝑂𝑂𝑂2 is the percentage overshoot of the
different locations.
TABLE II transfer function H2(s). Fig. 7 through Fig. 9 show the step
THE PHYSICAL PARAMETERS OF THE SYSTEM response of 𝐻𝐻1 (𝑠𝑠) and 𝐻𝐻2 (𝑠𝑠) for the cases when RES is
Parameter Symbol Value installed at node 2, node 4, and node 6, respectively. It is
[𝑍𝑍𝑐𝑐1 , 𝑍𝑍𝑐𝑐2 , 𝑍𝑍𝑐𝑐3 ] [0.02, 0.04, 0.034] 𝛺𝛺 shown from the figures that the overshoot values are close to
Line impedance each other in Fig. 7 and largely different in Fig. 8 and Fig. 9.
[𝑍𝑍𝑐𝑐4 , 𝑍𝑍𝑐𝑐5 , 𝑍𝑍𝑐𝑐6 ] [0.026, 0.023, 0.03 ] 𝛺𝛺
Therefore, node 2 is most likely to serves as the optimal point
Source Inductance [𝐿𝐿𝑆𝑆1 , 𝐿𝐿𝑆𝑆2 , 𝐿𝐿𝑅𝑅𝑅𝑅𝑅𝑅 ] [2.54, 3.18, 2.2] mH
as the standard deviation of the overshoot values is the
Load impedance [𝑍𝑍𝐿𝐿1 , 𝑍𝑍𝐿𝐿2 , 𝑍𝑍𝐿𝐿3 , 𝑍𝑍𝐿𝐿4 ] [12 + 𝑗𝑗1, 8, 10 + 𝑗𝑗1.1, 9]𝛺𝛺 minimum among the nodes in the microgrid.

0093-9994 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: The University of Akron Libraries. Downloaded on June 23,2021 at 16:11:07 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3085674, IEEE
Transactions on Industry Applications

6
TABLE VI be the optimum placement of RES but not a COM. To make
STANDARD DEVIATIONS OF THE PERCENTAGE OVERSHOOT VALUES
RES at RES at RES at
this point a COM, the standard deviation should be zero. To
Parameter achieve this, deploying 1.75 kW RES drives the standard
node 2 node 4 node 6
Output power percentage overshoot of
10.4 11.2 11.8
deviation (S) to zero and makes node 2 serve as a COM point
H1: (%OS1) of the microgrid where no source is brought closer to the
Output power percentage overshoot of instability region than the others. Fig. 10 shows the step
9.65 8.02 5.2
H2: (%OS2)
response of this case where DG1 and DG2 experience the
Maximum transient power difference:
0.75 3.78 6.6 same amount of overshoot.
|%OS1 − %OS2 |
Standard deviation of the overshoot
values: 0.375 1.59 3.3
S {%OS1, %OS2}
The relative difference of S S2 4.24S2 8.8S2

Fig. 10. The step response when 1.75 kW RES is installed at COM
(node 2).
To validate the performance advantages of the proposed
methodology over the methods discussed in the literature, the
Fig. 7. The step response when 2.0 kW RES is installed at node 2. proposed optimization is applied on a 9-bus system
considered in [6]. The line parameters are taken from
Table A.2 (appendix section of [6]). The rated power of the
sources is 10.0 kW, the droop coefficients are (kp=1.0e-5),
and the loads are 4 kW, 3 kW, and 6 kW. The voltage
magnitudes and phase angles are taken based on the data
given in Table X. According to the network configuration,
there are six different candidate buses where the new source
can be integrated (i.e., bus 1-6). Table VII illustrates the
optimal placement of the newly installed source for the
system considered in [6] based on the proposed optimization.
Fig. 8. The step response when 2.0 kW RES is installed at node 4. It can be seen from Table VII that bus-2 serves as an optimal
location whereas bus-4 and -6 are the worst candidate buses.
Table VIII compares the results of the optimization
technique reported in [6] and the proposed methodology. It
is clear from the results that there is a high agreement
between the two methodologies as both of them consider the
dynamic operation of the system. According to [6], bus-3 is
the most optimal location and bus-2 is the second optimal
location compared to the rest of the buses. According to the
proposed optimization, buses 2 and-3 are the first and second
optimal locations compared to the rest of the buses,
Fig. 9. The step response when 2.0 kW RES is installed at node 6. respectively. According to both methodologies, the results
Based on the results shown in Table VI, node 2 is the also show that the difference between bus-2 and bus-3
optimum node to deploy the RES, node 4 is the second placements is the minimum among the other buses. This
optimum node, and node 6 is the worst node for the RES to confirms the applicability and significance of the proposed
methodology.
be installed at. It is clear that the step response of 𝐻𝐻1 (𝑠𝑠) and
It is important to mention that as circuit parameters of the
𝐻𝐻2 (s) have 0.35 and 0.233 steady-state values, respectively.
system are known, the proposed methodology can be applied
These values represent the ratio of the frequency droop
to any system with a larger number of buses and yields
coefficients of the sources [23, 24] which confirms the
accurate results compared to other stochastic-based
validation of the modeling carried out in Section II and III.
optimization techniques. Furthermore, even with a large
The steady-state values of the power-sharing are given as:
system, only a few nodes are usually tested as candidates for
𝑃𝑃1𝑓𝑓 (𝑠𝑠) 𝐾𝐾𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 the placement process. This, in turn, helps to narrow the
lim = (16)
𝑠𝑠→0 𝑃𝑃𝑅𝑅𝑅𝑅𝑅𝑅 (𝑠𝑠) 𝑘𝑘𝑝𝑝1 options and maintain the efficiency of the proposed
𝑃𝑃2𝑓𝑓 (𝑠𝑠) 𝐾𝐾𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 methodology. However, any change in system parameters
lim = (17) would change the dynamics and thereby the obtained values
𝑠𝑠→0 𝑃𝑃𝑅𝑅𝑅𝑅𝑅𝑅 (𝑠𝑠) 𝑘𝑘𝑝𝑝2
of the standard deviation. Consequently, the accuracy of the
Although the system is an 8th order system, the step proposed algorithm is affected by the correct knowledge of
response exhibits a low order response characteristic which the circuit parameters. As a result, any change in the circuit
indicates that the system has dominant modes, and the system components of the system would change the placement results
order could be reduced. The results show that node 2 is the and may impact the system stability.
optimum point to deploy the RES but it does not serve as
COM as the standard deviation is 0.375. Node 2 is found to

0093-9994 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: The University of Akron Libraries. Downloaded on June 23,2021 at 16:11:07 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3085674, IEEE
Transactions on Industry Applications

7
TABLE VII.
COMPARISON OF STANDARD DEVIATION VALUES OF THE SYSTEM
STUDIED IN [6]
RES at RES at RES at RES at RES at RES at
Parameter
bus-1 bus-2 bus-3 bus-4 bus-5 bus-6
Output power
percentage (a) Microgrid testbed and DG1 components: On the figure: (a) IGBT-
5.21 11.52 13.18 23.25 10.69 8.63 based inverter, (b) DSP board, (c) Interfacing control board, (d) current
overshoot of
H1: (%OS1) sensor, (e) L-filter, and (f) JTAG emulator.
Vn7
Output power
percentage
20.24 10.53 13.99 4.77 11.85 22.95 DG2
overshoot of Vn6
ZL4
sw Vn4
H2: (%OS2) Vn5

Output power PDCS-2 Inv-2 Ls2


Zc4 Zc5
RES

percentage sw
8.25 11.78 12.25 11.39 5.25 5.43
overshoot of Vn3
Zc3 ZL3
H3: (%OS3) LRES Inv-RES PDCS-RES
PACL
Standard ZL2
Vn2 DG1
deviation of the sw
overshoot Vn1
values: 6.488 0.538 0.710 ≈7.645 2.872 ≈7.645
S {%OS1, Ls1 Inv-1 PDCS-1
ZL1
%OS2, %OS3}
PACL
TABLE VIII
COMPARISON BETWEEN THE RESULTS OBTAINED IN [6] AND THE (b) The configuration of the experimental microgrid
PROPOSED ALGORITHM Fig. 11. Experimental setup and microgrid configuration.
According to the
Order According to [6] Proposed 2000

methodology
1500
1st optimal location 3 2
%OS =12.1%
2nd optimal location 2 3 Output Power (W)
1000
DG1

3rd optimal location 5 5 500

4th optimal location 1 1


%OS =11.8%
5th optimal location 4 4, 6 0 DG2

6th optimal location 6 - -500


DG1 DG2 RES

200 300 400 500 600 700 800 900 1000


Time (ms)

V. EXPERIMENTAL RESULTS Fig. 12. The output power of the DGs when RES is installed at node 2.
To validate the effectiveness of the proposed 2000

methodology, an experimental test has been conducted on a 1500

microgrid testbed that contains three different sources. The %OS


DG1
=10.3%

microgrid testbed and the configuration are shown in Fig. 11.


1000
Output Power (W)

Programmable DC supplies (PDCSs) are used to supply the 500

inverters. The loads are combinations of resistive and 0 %OS =13.2%

inductive loads where static load banks and programmable


DG2

DG1 DG2 RES

AC loads (PACLs) are used. The microgrid dynamics depend -500


200 300 400 500 600 700 800 900 1000

on the frequency-power droop control, where the power


Time(ms)

Fig. 13. The output power of the DGs when RES is installed at node 4.
dynamics are slow. Therefore, the structure of the 2000

experimental setup should not impact the arrived results.


The new RES has been deployed at the source nodes (node 1500

2, node 4, and node 6) and the output power of the other DGs 1000
%OS
DG1
=9.9%
Output Power (W)

are analyzed. The output power of the DGs when the RES
source has been installed at node 2, node 4, and node 6 are
500

shown in Fig. 12-14. It can be seen from the figures that DG1
%OS =19.2%
DG2
0

and DG2 were sharing the full load according to their droop -500
DG1 DG2 RES

coefficients before RES is installed. After RES is installed, 200 300 400 500 600
Time (ms)
700 800 900 1000

the power-sharing of DG1 and DG2 are largely reduced as Fig. 14. The output power of the DGs when RES is installed at node 6.
RES has a larger capacity compared to other DGs. The The results obtained in the experiment study confirm the
transient power overshoot is calculated for each DG in all validity of the proposed optimization. This shows that the
cases to be used to find the value of the standard deviation. experimental results agree to a large extent with the
Therefore, the optimum placement point can be determined. mathematical modeling and the transfer function-based
Comparing the power overshoot results of all cases analysis performed on the considered microgrid.
depicted in Fig. 12-14, it can be seen that DG1 and DG2 Table IX analyzes and compares the experimental results,
experienced very close transients when RES is installed at quantitatively. The percentage overshoot and standard
node 2. The transient power difference between DG1 and deviation values of the placement cases are provided. The
DG2 (∆𝑃𝑃𝐷𝐷𝐺𝐺1𝑚𝑚𝑚𝑚𝑚𝑚 − ∆𝑃𝑃𝐷𝐷𝐺𝐺2𝑚𝑚𝑚𝑚𝑚𝑚 ) in this case, is about 29 W. The table results confirm that node 2 is the most optimal point to
transient power difference when RES is installed at node 4 is serve as a COM with a small standard deviation of 0.11
81 W. On the other hand, it is 153 W when RES is installed compared to 1.03 for node 4 and 3.29 node 6 placement.
at node 6. This is considered a larger difference compared to In the proposed work, the sensitivity of the parameter
node 2 placement. change can also be inferred from Table IX. According to the

0093-9994 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: The University of Akron Libraries. Downloaded on June 23,2021 at 16:11:07 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3085674, IEEE
Transactions on Industry Applications

8
microgrid parameters given in Table II, the standard deviation
at node 2 should be zero with the integration of 1.75 kW RES.
However, due to the difference between the actual circuit
parameters and the model-based analysis (built based on
Table I data), the standard deviation is 0.11 instead of zero.
This kind of marginal discrepancy may appear due to
measurement error in the system parameters.
Fig. 15 shows the power ratio between DG1 and DG2 when
RES is placed at the three source nodes. The power-sharing
ratio of the sources follows the equations given in (16) and
(17) during the steady-state and deviates during the transient.
The transient depends on the control dynamics and the
equivalent circuit of the microgrid seen by the RES at each
node. Fig. 15(a) shows the case when RES is installed at node
2 (COM point) where the transient power-sharing slightly Fig. 16. The corresponding voltage and current waveforms at the COM
changes after RES insertion and both DGs experience similar point (1: the node voltage, 2: DG1 current, 3: DG2 current, 4: RES current).
transient. Fig. 15(b) shows the case when RES is installed at
node 4 where worse transient power-sharing occurs and both VI. DISCUSSION
DGs experienced different transient power. Fig. 15(c) shows As stabilizing islanded microgrid operation with the
the worst case when RES is installed at node 6. In this case, existence of RESs becomes a crucial issue, the procedure
the DGs transient power difference is the largest. Therefore, performed in this paper aims to maintain system stability by
this node is the worst node to install RES as it may destabilize accurately finding the optimum place and size of the
the microgrid if RES is switched on or off during normal integrated RES. The methodology developed in this paper can
operating conditions. be used as an accurate tool for improving microgrid stability
The corresponding instantaneous voltage and current under the high penetration of renewables. As the parameter
waveforms of the DGs when RES is installed at the COM values of the lines, sources, and loads can be easily known in
(node 2) are shown in Fig. 16. The figure shows the microgrid small-scale microgrids, implementing the proposed
has stable voltage and current profiles before and after methodology provides a clear answer about the stability of
integrating the RES source that reduces the power-sharing of microgrids when power generation changes abruptly in
the existing DGs in the microgrid. microgrid. Therefore, the effect of the uncertainty and
variability of renewable sources on microgrid stability can be
TABLE IX.
THE EXPERIMENTAL RESULT VALUES OF THE STANDARD DEVIATION
carefully addressed and studied. Moreover, plug and play
RES at RES at RES at capability of microgrid sources can be practically enabled and
Parameter realized by considering the proposed methodology.
node 2 node 4 node 6
Output power percentage overshoot of
12.1 10.3 9.9
The optimum placement is crucial when RES is switched
H1: (%OS1) off. In this case, the power overshoot causes transient
Output power percentage overshoot of
11.8 13.2 19.2 overcurrent and overvoltage that may overload DGs and
H2: (%OS2) trigger the stability of the microgrid. As the obtained transfer
Maximum transient power difference: functions between the output power of the integrated DGs can
0.3 2.9 9.3
|%OS1 − %OS2 | be utilized to design the filters, the droops, and tune the
Standard deviation of the overshoot controllers of the DGs for reliable and stable operation of
values: 0.11 1.03 3.29 microgrids.
S {%OS1, %OS2} It is worth to mention that combining a stochastic-based
The relative difference of S S2 9.3S2 29.9S2 optimization with the proposed approach would increase the
capability of the proposed methodology under uncertain
parameter values. Therefore, technical, and non-technical
1 1
constraints can be included in the optimization scheme. In
0.8 0.8
addition to that, as the implemented control strategy affects
0.6 0.6
the dynamics of the microgrid, different control strategies can
Power Ratio

Power Ratio

RES Integration

be applied in microgrid depending on the service provided.


RES Integration
0.4 P /P 0.4
DG1 DG2 P /P

Depending on whether the source is grid feeding, grid


DG1 DG2

0.2 0.2

0
300 400 500 600 700 800 900
0
300 400 500 600 700 800 900
following, or grid supporting, the control strategy can be
Time (ms) Time (ms)
designed. As a result, the dynamics are influenced by the
(a) Placement at Node 2 (b) Placement at Node 4
1 control type and control parameters. The work done here
0.8 considers that all sources are equipped with droop control and
0.6
operate as grid-forming sources. This is considered as the
system under study is an islanded microgrid and the sources
Power Ratio

RES Integration
0.4

are required to stabilize the microgrid voltage. The


P /P
DG1 DG2

consequences of the above-mentioned issues on the optimal


0.2

placement methodology will be investigated in future work.


0
300 400 500 600 700 800 900
Time (ms)

(c) Placement at Node 6


Fig. 15. The instantaneous output power ratio between DG1 and DG2 when VII. CONCLUSION
the RES source is installed at different nodes.
In this paper, the impact of installing certain RES capacity
at specific nodes on the stability of islanded microgrids is
evaluated. A modified load flow analysis model for droop-
controlled microgrids has been proposed to obtain the
optimum placement of the renewable energy source in

0093-9994 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: The University of Akron Libraries. Downloaded on June 23,2021 at 16:11:07 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3085674, IEEE
Transactions on Industry Applications

9
microgrids. The proposed methodology evaluates the impact [18] M. A. Zehir, A. Batman, , M. A. Sonmez, A. Font, D. Tsiamitros, D.
of installing a large renewable source at certain nodes by Stimoniaris, E. Dialynas, “Impact of Renewable Based Microgrid
Supply/Demand Profiles on Low Voltage Distribution Networks,”
analyzing the impact of the newly installed renewable source Energy Procedia, 103, 231-236, 2016.
output power on the other integrated DGs. [19] M. R. Vallem and J. Mitra, "Siting and sizing of distributed generation
The study shows a novel deterministic approach that for optimal microgrid architecture," Proceedings of the 37th Annual
provides an accurate analysis of the effect of source switching North American Power Symposium, 2005., Ames, IA, USA, 2005.
on the stability of microgrids. The study also analyzes the [20] M. Gautam, N. Bhusal, M. Benidris, C. Singh and J. Mitra, "A
Sensitivity-based Approach for Optimal Siting of Distributed Energy
relation between the integrated DGs power-sharing dynamics. Resources," 2020 International Conference on Probabilistic Methods
The proposed method checks for different placements until it Applied to Power Systems (PMAPS), Liege, Belgium, 2020.
finds the COM point in which if RES is installed, the power [21] M. Benidris, Y. Tian, S. Sulaeman and J. Mitra, "Optimal location and
system stability is expected to be enhanced and maintained size of distributed energy resources using sensitivity analysis-based
approaches," 2016 North American Power Symposium (NAPS),
under dynamic operating conditions. The impact of installing Denver, CO, 2016.
RES to a microgrid system at different nodes has been [22] H. Saadat, “Power System Analysis”. New York, USA: McGraw-Hill,
evaluated and the proposed method has been used to obtain 1999.
the COM point. [23] A. Elrayyah, Y. Sozer, and M. Elbuluk, “A novel load-flow analysis
for stable and optimized microgrid operation,” IEEE Trans. Power
REFERENCES Del., vol. 29, no. 4, August 2014.
[24] N. Pogaku, M. Prodanovic, and T. C. Green,” Modelling, analysis and
[1] Z. Miao and A. Domijan, L. Fan, “Investigation of microgrids with testing of autonomous operation of an inverter-based microgrid, ”
both inverter interfaced and direct AC-connected distributed energy IEEE Trans. on Power Electron., vol. 22, no. 4, March 2007.
resources,” IEEE Trans. Power Del., vol. 26, no. 3, 2011. [25] S. Harasis, Y. Sozer, M. Elbuluk and H. Abu-Rub, " Improved
[2] S. Tiwari, and P. K. Goel, “Design and control of micro-grid fed by Transient Power Sharing of Droop Controlled Islanded Microgrids,"
renewable energy generating sources,” IEEE Trans. Ind. Appl. vol. 99, 2020 IEEE Applied Power Electronics Conference and Exposition
pp. 1-9, 2018. (APEC), New Orleans, LA, USA, 2020.
[3] P. Chiradeja, and R. Ramakumar, “An approach to quantify the [26] H. Abdelgabir, A. Elrayyah, and Y. Sozer, "A center of mass
technical benefits of distributed generation,” IEEE Trans. Energy determination for the optimum placement and deployement of the
Convers., vol. 19, no. 4, pp. 754-773, 2004. renewable energy sources for microgrid," 2018 IEEE Energy
[4] N. Soni, S. Doolla, and M. C. Chandorkar, “Improvement of transient Conversion Conference and Expo (ECCE), Portland, OR, 2018.
response in microgrids using virtual inertia,” IEEE Trans. Power Del. [27] H. Abdelgabir, A. R. Boynuegri, A. Elrayyah, and Y. Sozer, "A
vol. 28, no. 3, 2013. complete small signal modelling and adaptive stability analysis of
[5] M. D. McDonald, R. A. Walling, R. D’Aquilla, and J. G. Gleary, nonlinear droop-controlled microgrids," 2018 IEEE Applied Power
“Effect of distributed generation on regional voltage stability,” IEEE Electronics Conference and Exposition (APEC), San Antonio, TX,
PES Transmiss. Distr. Conf. Expo., pp. 1-6, 2012. 2018.
[6] Q. Sun, B. Huang, D. Li, D. Ma and Y. Zhang, "Optimal Placement of
Energy Storage Devices in Microgrids via Structure Preserving Energy
Function," IEEE Trans. Ind. Informat., vol. 12, no. 3, pp. 1166-1179,
2016.
[7] S. Arif, and Tareq Aziz, "Study of Transient Stability with Battery
Energy Storage Systems in Renewable Integrated Islanded
Microgrid," 2017 IEEE International WIE Conference on Electrical
and Computer Engineering, 2017.
[8] F. Tooryan, E. R. Collins, A. Ahmadi and S. S. Rangarajan,
"Distributed generators optimal sizing and placement in a microgrid
using PSO," 2017 IEEE 6th International Conference on Renewable
Energy Research and Applications (ICRERA), San Diego, CA, 2017.
[9] B.K. Das, Y.M. Al-Abdeli, G. Kothapalli, “Effect of load following
strategies, hardware, and thermal load distribution on stand-alone
hybrid CCHP systems,” Applied Energy, 2018.
[10] N. Nikmehr, S. Najafi-Ravadanegh, “Optimal operation of distributed
generations in micro-grids under uncertainties in load and renewable
power generation using heuristic algorithm,” IET Renew. Power
Generation, 2015.
[11] M. Marzband, F. Azarinejadian, M. Savaghebi and J. M. Guerrero,
"An Optimal Energy Management System for Islanded Microgrids
Based on Multiperiod Artificial Bee Colony Combined With Markov
Chain," in IEEE Systems Journal, vol. 11, no. 3, pp. 1712-1722, 2017.
[12] D. K. Khatod, V. Pant and J. Sharma, "Evolutionary programming
based optimal placement of renewable distributed generators," in IEEE
Transactions on Power Systems, vol. 28, no. 2, pp. 683-695, 2013.
[13] A. Ali, D. Raisz, K. Mahmoud and M. Lehtonen, "Optimal Placement
and Sizing of Uncertain PVs Considering Stochastic Nature of PEVs,"
in IEEE Transactions on Sustainable Energy, vol. 11, no. 3, pp. 1647-
1656, July 2020.
[14] B. K. Poolla, D. Grob and F. Dörfler, "Placement and Implementation
of Grid-Forming and Grid-Following Virtual Inertia and Fast
Frequency Response," in IEEE Transactions on Power Systems, vol.
34, no. 4, pp. 3035-3046, July 2019.
[15] S. H. Lee and J. Park, "Optimal Placement and Sizing of Multiple DGs
in a Practical Distribution System by Considering Power Loss,"
in IEEE Transactions on Industry Applications, vol. 49, no. 5, pp.
2262-2270, Sept.-Oct. 2013.
[16] M. Bazrafshan, N. Gatsis and E. Dall’Anese, "Placement and Sizing
of Inverter-Based Renewable Systems in Multi-Phase Distribution
Networks," in IEEE Transactions on Power Systems, vol. 34, no. 2,
pp. 918-930, March 2019.
[17] S. M. Hakimi, A. Hasankhani, M. Shafie-khah and J. P. S. Catalão,
"Optimal sizing and siting of smart microgrid components under high
renewables penetration considering demand response," in IET
Renewable Power Generation, vol. 13, no. 10, pp. 1809-1822, 2019.

0093-9994 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: The University of Akron Libraries. Downloaded on June 23,2021 at 16:11:07 UTC from IEEE Xplore. Restrictions apply.

You might also like