Analyzing Multi-microgrid with Stochastic Uncertainties Including
Optimal PV Allocation
H. Keshtkar, J. Solanki and S. Khushalani Solanki
Department of Electrical Engineering, West Virginia University, PO Box 6070, Morgantown, WV 26505, U.S.A.
Keywords: PV Allocation, Loss Minimization, MATLAB-OpenDSS Interface, PSO, Multi-microgrid, Stochastic
Uncertainties, Stability Margin Analysis.
Abstract: This paper presents the effects of Photovoltaic (PV) location on losses of the distribution system. The optimal
location of PV is determined by using Particle Swarm Optimization (PSO) implemented in MATLAB-
OpenDSS environment. IEEE 34-node test feeder is employed to verify the feasibility and effectiveness of
the developed method. Once the optimal location is determined the challenge still remains due to the uncertain
behavior of the PV system. This effect along with other stochastic behaviors such as the uncertain output
power of loads like Plug-in Hybrid Electric Vehicles (PHEVs) due to their stochastic charging and
discharging, that of a wind generation unit due to the stochastic wind speed, and that of a solar generating
source due to the stochastic illumination intensity, add problems like frequency oscillations in a microgrid.
Hence, frequency control of a multi microgrid system is also addressed.
1 INTRODUCTION
Electric energy is produced in large power plants and
transmitted through High Voltage (HV) transmission
systems to be distributed to consumers through Low
Voltage (LV) distribution networks. Distribution
system dynamics are changing with the siting of
electricity generation closer to the loads and these are
called Distributed Generation (DG) units
(Mohammadi, 2012). These units have less
environmental impact, easy siting, high efficiency,
enhanced system reliability and security, improved
power quality, lower operating costs due to peak
shaving, and relieved transmission and distribution
congestion.
However, depending on the location of DG units,
some of the problems may be more pronounced and
hence it is important to site the DG units to optimally
exploit their potential. This paper, therefore, develops
algorithms to optimally place the distributed
generator considering the changing demand and
generation conditions over a day. With distributed
generators the distribution network can work in
isolation being separated from the feeder network to
form a micro-grid without affecting the transmission
grid’s integrity. One of the DG technologies is
Photovoltaic (PV), with penetrations increasing from
hundreds of kWs to MWs in LV network. Due to
these increasing penetrations in distribution systems,
the utilities and planning engineers are increasingly
interested in determining the best locations to place
these units (Prenc, 2013).
Much of the research work on PV allocation
assumes a constant generation making the problem
deterministic (Medina, 2006 – Shukla 2008). For
example, in (Shukla, 2008), analytical methods are
presented to determine the optimal location of PV
with constant generation to improve the power
quality. In reality however the PV output has
variations and hence an optimal location profiling the
daily irradiation and energy production is necessary
(Ackermann, 2001).
These units can be installed near load centers or at
remote nodes to avoid large power transfers. Since
distribution systems are now being operated as
microgrids that form smart cities, analysis on the
effects of these placements for microgrids is deemed
necessary (Duenas, 2014). The problem becomes
more complex with interconnection of multi-
MG/smart cities with more energy layers (Guo, 2010)
as compared to a single microgrid/smart city which is
a passive system.
In this paper we develop an optimal PV location
algorithm and analyze the effects of the PV
generations and loads like PHEVs on transient
stability of a multi-microgrid system. Since the
241
Keshtkar H., Solanki J. and Khushalani Solanki S..
Analyzing Multi-microgrid with Stochastic Uncertainties Including Optimal PV Allocation.
DOI: 10.5220/0005452902410249
In Proceedings of the 4th International Conference on Smart Cities and Green ICT Systems (SMARTGREENS-2015), pages 241-249
ISBN: 978-989-758-105-2
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
variability of a PV system can affect the findings of
the optimal location a stochastic model is utilized for
both PV generations and PHEV loads. Stochastic
modeling and Monte Carlo simulation (MCS) are
common methods to perform the stochastic optimal
planning (Duenas, 2014), which have been widely
used in OPF (Guo, 2010, Zhang, 2011), distribution
network planning (Soroudi, 2011, Zhipeng, 2011),
power market design (Sofla, 2012 – Shresta, 2008),
distribution system extension planning (Kai, 2012)
and microgrid energy management (Niknam, 2012).
Here we utilize different models for stochastic
behaviors of variable generation and demand units.
The organization of this paper is as follows:
Section 2 discusses about modelling of the
distribution and microgrid system. The stochastic
modelling of uncertain parameters in the multi-
microgrid system is also formulated in this section.
PSO algorithm for optimal allocation of PV is
discussed in Section 3. Section 4 describes the case
studies and also presents the results of simulations.
The paper is concluded in Section 5.
2 MODELING AND PROBLEM
FORMULATION
We strongly encourage authors to use this document
several microgrids with many DG units such as
diesel, wind and solar generators integrated. In
distribution systems, losses can increase operation
cost and therefore it is essential to determine the
optimal placement of these generators to minimize
the total losses of the multi-microgrid system. Here
an optimal power flow type of problem is formulated
and heuristic algorithm such as particle swarm
optimization is selected.
2.1 Optimization Algorithm
In this paper we will consider the placement of solar
generation on ⊂ nodes of multi-microgrid due to
restriction imposed by distribution network operators.
We assume that solar generation contribute majorly
to the active power of the system thereby reducing the
problem to minimization of active power losses. The
methodology proposed here is described in three
basic steps:
1) A constrained non-linear optimization problem is
formulated to minimize the real power losses.
Equality constraints related to distribution power
flow equations and inequality constraints related to
node voltage limits, generation capacity constraints
and feeder current constraints are considered.
2) An intelligent computational technique like PSO
is employed with reduced computational complexity
due to reduction in search space from to 
where are the number of PV units to be placed.
Since the order of units also matters a permutations
calculator rather than a combinations calculator is
employed.
3) PSO is combined with three phase distribution
power flow computed using backward forward sweep
algorithm while the PSO globally optimizes to find
the optimal DG placements and the distribution
power flow determines the constraints violations. In
backward forward sweep method, Kirchhoff’s
Current Law and Kirchhoff’s Voltage Law are used
to compute the bus voltage from farthest node in the
backward sweep. Then in forward sweep,
downstream bus voltage is updated starting from
source node. The procedure stops after the mismatch
of the calculated and the specified voltages at the
substation is less than a convergence tolerance.
4) If the optimal values of two consecutive iterations
are same with all constraints satisfied the algorithm is
deemed to have converged. If not, the process
continues until the criteria is satisfied.
The minimization objective function is formulated as
shown in (1).
11
ik
NN
loss
ik
FP


(1)
Considering the conductor current



and
current contribution from the solar generator as

, the losses of line i
k
can be expressed as in (2).
2
2
2
3. .
cos sin
3
3. .
cos sin
3
ik ik
i
k
loss loss ik
DG
iiii
i
i
ik
DG
kkkk
k
k
PIR
P
PQ
V
V
R
P
QP
V
V











(2)
The losses over a period of one day continuously
change due to changes in active power injections of
the solar generation and changes in load consumption
patterns. However, the location of PV once
determined cannot change, so an optimal location
should consider these variations in losses. The
optimization algorithm is subjected to the following
constraints.
(i) Generator Rating Constraint: Based on peak
power generation, the minimum and maximum
limits have been imposed on the generation
capacity as
SMARTGREENS2015-4thInternationalConferenceonSmartCitiesandGreenICTSystems
242
min max
iii
ggg
PPP
(3)
(ii) Voltage Constraint: The optimal siting has to
be obtained such that there are no bus voltages
limit violations.
min maxi
VVV
(4)
(iii) Power Balance Constraint: The total power
demand should be less than or equal to total
power generation.
1
1
i
i
r
dg
i
r
dg
i
PP
QQ
(5)
(iv) Feeder Current Constraint: The feeder,
current flowing through the feeder should be
less than its thermal limit.
ik
ik th
II
(6)
An unconstrained formulation considering both
objectives and constraints from (1-6) is then given in
(7). Typical operations constrain the voltage to be
around the nominal node voltages of the microgrid
whereas the lines have to be limited to their thermal
values.
11 1
2
2
1
2
2
11
()
ik i
i
ik
NN r
xlossgd
ik i
N
Nom
gd ii
i
NN
th ik
ik
F
PPP
QQ VV
II







(7)
2.2 Power Flow Equations
The power balance constraints include the power flow
equations that are solved iteratively and expressed as
() () ()
() () ()
ikk
ikk
Vi aV i bIi
I
icVidIi


(8)
Figure 1: Current flowing on line connecting node i and j.
The currents are those flowing on lines connecting
nodes i and j as shown in Fig. 1 and a,b,c,d are matrix
constants for the models of different components of
the distribution network. Here loads are modelled as
constant current injections,




, and
distributed generations as negative active power
loads. The generators are modelled using (9) that
relates the voltage
N
u and the current
N
i
N
N
NN N
N
u
ui s
U
(9)
where
N
s is the nominal complex power and
N
is a
characteristic parameter of the node N . The model (9)
is called exponential model (
Price, 1993) and is widely
adopted in the literature on power flow analysis
(Haque, 1996). Notice that
N
s is the complex power
that the node would inject into the grid, if the voltage
at its point of connection were the nominal voltage
N
U .
However, under islanded conditions the
microgrids lose stability instantly and hence
frequency response is of concern. The uncertain
behavior of the wind generations, hybrid electric
vehicles and solar generation result in frequency
deviations, an analysis of which requires uncertainty
models.
2.3 Uncertainty Models
The uncertainties considered in this paper include
wind speed, solar radiation, and load disturbance and
stochastic models for them are developed here:
2.3.1 Uncertainty of Wind Generating Units
It has been observed that wind speed deviations
follow a Weibull type distribution as shown in Eq.
(10) (Guo, 2010).
1
(,,) exp
kk
kW W
pW ck
cc c

 


 

 

(10)
Where, scale factor

and shape factor 
.
, W and σ are the average value and
standard deviations of wind speed, respectively, and
Γ(•) is the gamma function. Using the know
probability distribution function the relationship
between the output power of a wind generating unit
and the wind speed can be obtained and the details are
provided in (Mohammadi, 2012).
2.3.2 Uncertainty of Load
Load uncertainties are twofold: those associated with
AnalyzingMulti-microgridwithStochasticUncertaintiesIncludingOptimalPVAllocation
243
changing loads corresponding to daily consumption
and those corresponding to transportation related
consumption. For the daily consumption, hourly
average load demand is scaled by a disturbance factor
α=1+δ
h
, where δ
h
is the hourly disturbance coefficient
and α is the disturbance factor. Both α and δ
h
follow
normal distribution with mean zero. Normal
distribution of load can perfectly model the variable
daily load in the power system P
L
=D*α is obtained
using D the hourly load data and α the disturbance
factor.
The other types of loads are the Plugin hybrid
electric vehicles. Since the behavior of consumers in
charging their PHEVs is highly behavior dependent,
three different stochastic processes for modelling
PHEV are considered here.
Brownian Motion: This is the continuous analog of
symmetric random walk distributions, where each
increment W(s+t)-W(s) is Gaussian with distribution
N (0, t) and increments over disjoint intervals are
independent. It is typically simulated as an
approximating random walk in discrete time. Here,
charging and discharging of PHEVs have been
modelled by Brownian motion with sigma of 3.
Poisson Process: This involves generation of random
events so that: (i) arrivals occur independent of each
other (ii) two or more arrivals do not occur at the same
time (iii) the arrivals occur with constant intensity.
Number of arrivals N (t) that occur from time zero up
to time t is Poisson distributed with expected value
lambda*t. The counting process N (t) is a Poisson
process. The successive times between connections
are Exponential (lambda) distribution. Here, arrival
time of PHEVs has been modeled by Poisson process
with lambda of 3.
M/M/1 Model: This is one of the random
distributions (Markov process) in the category of
Queuing systems. Discrete time intervals are
considered so PHEV arrivals to a service center occur
according to an independent sequence of a (1), a (2)
…, where a (k) is the number of arrivals during time
slot number k. Only one PHEV can be
charged/discharged per slot (single server system).
Additional PHEVs are in waiting until service is
available. Therefore the number of PHEVs in the
system at time k is given by

() ( 1) () 1, 2
(1)()1
(1) 0
nk nk ak k
nk ak
n


(11)
This recursion defines a Markov chain n (k), k 1. So
M/M/1 is a single server buffer model in continuous
time. Considering PHEV arrivals as Poisson process
with intensity λ, an exponentially distributed random
mean service time
1
is employed for each PHEV.
The resulting system size N (t) for t 0, is a Markov
process in continuous time which evolves as follows:
Starting from N (0) = n_0, wait an exponential time
with intensity λ+μ (intensity λ if n_0=0), then charge
with probability /  and discharge with
probability /. Here, number of PHEVs in the
system (PHEV load size) has been modelled by
M/M/1 model with λ of 1.5 and μ of 0.8.
2.3.3 Uncertainty of PV
The uncertainty of solar radiation is mainly because
of the stochastic weather conditions. In this paper,
cleanness index is used to model the uncertainties of
weather condition. The relationship between the
cleanness index and the solar irradiation can be
obtained from (Srisaen, 2006). The distribution
function of cleanness index can be expressed as.
()
( ) exp( )
tu t
tt
tu
kk
P
kC k
k
(12)
Where, k
t
indicates the mean value of cleanness
index, k
tu
is the 0.864 theoretically, 





where 2 17.519 exp
1.3118
1062exp
5.0426
/

and 

/


.
The PV generation varies with the solar
irradiation which varies according to the cleanness
distribution. The relationship between the solar
irradiation and the PV output power can be obtained
from (Huang, 2006).
2.3.4 Load Frequency Control (LFC)
Load Frequency Control (LFC) has been
implemented in the second part of the simulations for
a multi-microgrid system. LFC in microgrids with
nominal frequency of 50 Hz is designed to maintain
frequency within 49.9 and 50.1 in normal condition
by controlling tie-line flows and generator load
sharing (Kroposki, 2008). The control strategy should
damp the frequency oscillation in steady state and
minimize them in transient state while maintaining
stability.
Microgrids considered in this paper has hybrid
power generation consisting of wind generators,
photovoltaic, diesel generators. Power supplied to the
load Ps is the sum of output power from wind turbine
generators P
w
, diesel generators P
g
, photovoltaic
generation P
pv
, total loss power P
Loss
and output of
PHEV P
phev
given by
s
wg pvlossphev
P
PPP P P

(13)
SMARTGREENS2015-4thInternationalConferenceonSmartCitiesandGreenICTSystems
244
Table 1: Total daily loss for different PV locations.
PV Bus 808 814 816 828 832 834 840 848 860 890
Total daily Loss (MWh) 3.8129 3.6524 3.6513 3.6182 3.4299 3.4457 3.4300 3.4294 3.4295 3.8688
LFC system in this paper uses this power flow
balance equation for adjusting the frequency of the
microgrid. Modeling of different parts of the system
is discussed in the rest of this section.
Modeling of Wind Turbine Generator (WTG). The
wind turbine is characterized by non-dimensional
curves of power coefficient C
p
as a function of both
tip speed ratio λ and blade pitch angle β. The tip speed
ratio, which is defined as the ratio of the speed at the
blade tip to the wind speed, can be expressed by
blade blade
W
R
V
(14)
Figure 2: Characteristic curve of output mechanical power
versus wind speed of the studied WTGs.
where R
blade
(= 23.5 m) is the radius of blades and
ω
blade
(=3.14 rad/s) is the rotational speed of blades.
The expression for approximating C
p
as a function of
λ and β is given by
(3)
(0.44 0.0167 )sin 0.0184( 3)
15 0.3
p
C






(15)
The output mechanical power of the studied WTGs is
3
1
2
WrpW
P
AC V
(16)
where, ρ (= 1.25 kg/m3 ) is the air density and A
r
(=
1735 m2) is the swept area of blades. The
characteristic curve of output mechanical power
versus wind speed of the studied WTGs in this paper
is shown in Fig. 2.
3 PARTICLE SWARM
OPTIMIZATION (PSO)
The locations of the PV systems are optimized by
Particle Swarm Optimization (PSO) Algorithm to
minimize the total losses in the system presented in
(7). PSO is a multi-agent search approach, which
traces its evolution to the motion of a flock of birds
searching for food. It uses a number of particles that
are called a swarm. Each particle traverses the search
space searching for the global minimum (or
maximum). In a PSO system, particles fly within a
multidimensional search space. During flight, each
particle sets its position based on its own experience
and the experience of neighboring particles. Hence, it
makes use of the best position encountered by itself
and its neighbors. Similarly, the swarm direction and
speed of a particle is determined by the history
experience obtained by itself as well as a set of its
neighboring particles (Babaei, 2009).
Each particle is a representative of PV locations
that are variables that affect the total losses in each
iteration. Let us consider p and s as particle position
and flight speed in a search space, respectively. The
best position of a particle in each step is recorded and
represented as P
best
. The best particle’s index among
all the particles in the group is considered as G
best
. The
convergence of PSO is ensured by use of a
constriction function. Finally, the modified velocity
and position of each particle can be calculated as
shown in (17) and (18):
11 2
*( * . ()*( ) * ()*( ))
dd bestd bestd
s
k v ac rand P P ac rand G P

(
17)
11ddd
PPs
(
18)
Here d is the index of iteration, P
d
is the current
particle’s position at the d-th iteration, s
d
is the
particle’s speed of at d-th iteration, γ is inertia weight
factor, ac
1
and ac
2
are acceleration constants, rand() is
a uniform random value in the range [0,1], and k is
the constriction factor which is a function of ac
1
and
ac
2
according to (19):
2
2
|2 4 |
k
ac ac ac

(19)
Where ac=ac
1
+ac
2
and ac>4. Appropriate choice of
inertia weight, γ, makes a balance between global and
local explorations. In general, γ is calculated
according to (20) (Das, 2006):
max min
max
max
iter
iter



(20)
AnalyzingMulti-microgridwithStochasticUncertaintiesIncludingOptimalPVAllocation
245
Where iter
max
is the maximum number of iterations,
and iter is the number of the iterations up to current
stage. The iterations continue until it reaches the
iter
max
or the difference between the losses calculated
by best particles of the last two iterations is less than
a predefined threshold.
4 CASE STUDIES AND
SIMULATION RESULTS
The optimal PV locations are determined using the
formulations discussed and the PSO technique. It is
shown that losses are minimized under varying daily
load consumptions. The PSO method described in
section 3 was implemented in Matlab programming
language and the unbalanced power flow solution is
obtained using OpenDSS. With the placement of PVs
at these optimal locations a frequency stability
analysis of a multi-microgrid system is performed
under stochastic load and generation behavior.
4.1 Optimal PV Allocation
As a preliminary analysis a 0.5 MW of PV generation
is considered with an unknown optimal location that
would result in minimum line losses. IEEE 34 node
benchmark distribution system as shown in Fig 4 is
considered that is inherently unbalanced with three
phase cables and conductors and three phase, two
phase and single phase loads. The characteristic
curves of the PV are as shown in Fig. 3 (a)-(d).
Figure 3: PV Characteristic curves; (a) Irradiation-time, (b)
Temperature-time, (c) P
mpp
-temperature, (d) Efficiency-
power.
Initially losses are evaluated with single PV
integration at node 848 of the IEEE 34 node system.
Losses for an entire day are plotted as shown in Fig.
5 as the PV generation and loads vary throughout the
day.
The active power losses are low at night and in
early morning time periods, when loading is less.
Figure 4: IEEE 34 node test feeder configuration.
Figure 5: Daily losses with PV placement at node 848.
However, when the loading increases and active
power generations from PV source increases, the
losses increase peaking from 15:00 – 16:00 hours.
Table 1 summarizes the total daily losses with PV
placements at different nodes. It is seen the best
location is node 848 which is located further from the
main grid and close to high-demand consumers.
Similar results are obtained using the developed PSO
algorithm and it is seen that losses are reduced by
29% as compared to no PV installation and 13% as
compared to worst PV installation.
Table 2: Total daily loss for best and worst multi-PV
locations.
Best PV locations Worst PV locations
PV Nodes 832, 848, 860 808, 814, 890
Total daily Loss
(MWh)
3.5620 3.8408
4.2 Optimal Multi-PV Allocation
Optimal locations for three PVs are obtained for IEEE
34 node test feeder by optimization algorithm (PSO)
to achieve the minimum total daily losses. As seen in
Table 2 the optimal node locations are 832, 848 and
860. They show 8% improvement in losses as
compared to the worst locations found heuristically.
4.3 Stochastic Uncertainties in
Multi-microgrid
With the determined optimal PV locations frequency
0 5 10 15 20 25
0
0.2
0.4
0.6
0.8
1
Time (hours)
Irradiation
(a) Irradiation-tim e
0 5 10 15 20 25
25
30
35
40
45
50
55
60
Time (hours)
Temperature
(b) Temperature-time
0 10 20 30 40 50 60 70 80 90 100
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
Temperature ('C)
Pmpp
(c) Pmpp-tem perture
0.1 0.2 0.3 0.4 0.5 0. 6 0.7 0. 8 0.9 1
0.86
0.88
0.9
0.92
0.94
0.96
0.98
Power (pu)
Efficiency
(d) Efficiency-output power
800
806 808
812
814
810
802
850
818
824
826
816
820
822
828 830
854 856
852
832
888
890
838
862
840
836860
834
842
844
846
848
864
858
0 5 10 15 20 25
0.16
0.17
0.18
0.19
0.2
0.21
0.22
Time (h)
Total Loss
Loss-time
SMARTGREENS2015-4thInternationalConferenceonSmartCitiesandGreenICTSystems
246
stability is evaluated considering the frequency
response models and uncertainty modelling in section
2. Moreover for multiple microgrids several islands
may occur simultaneously as a result of multiple
contingencies in the network. PSO is adopted here to
improve the frequency response by optimizing the
parameters of various controllers. The details of
controller design utilizing PSO for speed control are
available in our prior work (Keshtkar, 2014).
Consider two similar microgrids connected
through a tie line as shown in Fig. 6. This hybrid
system comprises of several RES such as Wind
Turbine Generator (WTG) and PV including Diesel
Engine Generator (DEG) as DG that contributes to
the inertia of the microgrid system. The hybrid system
also includes PHEVs and other residential and small
industrial loads. The PV systems are located
optimally using the algorithm proposed earlier in the
paper. The controllable source in this microgrid is
Diesel Engine Generator whose control parameters
are optimized by PSO to minimize the frequency
deviations due to the disturbances. The tie line
deviations of the multi-microgrid system are also
considered in the objective function for optimizing
controller parameters. The Area Control Error (ACE)
for each microgrid considering ΔP
tie
the tie-line
power flow deviation, Δf
i
the frequency deviation of
each microgrid weighted by β is as shown in (21).
i tie i
A
CE P f

(21)
Figure 6: Configuration of the modeled multi-microgrid.
Three different stochastic behaviors of the PHEVs
discussed in section 2 are modeled along with
stochastic uncertainties of load, wind and solar
generations. Frequency response of one of the
microgrids of the multi-microgrid system is obtained
as shown in Fig. 7. Also Fig. 8 shows the magnified
frequency responses of the system in presence of
different stochastic uncertainties in the multi-
microgrid power system.
It is seen that stochastic modeling is essential to
study the transient stability of the system and that
some uncertainties can cause the frequency to
severely deviate from the nominal value. For
example, stochastic behavior of PHEVs creates
significant overshoots in the frequency response as
shown in Fig. 8 that can cause the microgrid system
to be unstable. It is observed thus that a simultaneous
modeling of stochastic behaviors is essential to design
and test reliable and robust controllers.
Figure 7: Frequency response of one of the microgrids with
different stochastic modelling.
Figure 8: Frequency response of one of the microgrids for
stability margin analysis.
5 CONCLUSIONS
In this paper a method for determining the optimal
placement of a PV system in distribution network
based on daily power consumption/production
fluctuations is described to minimize the total daily
losses. The PSO optimization algorithm shows fast
and accurate performance in calculating the optimal
position of a PV system. Therefore, it can also serve
0 50 100 150 200 250 300
-0.01
-0.005
0
0.005
0.01
0.015
time
df-time
Brownian
Poisson
3D Brownian
M/G/Infinity
M/M/1
200 210 220 230 240 250
-8
-6
-4
-2
0
2
4
6
x 10
-3
time
df
(a) df-tim e
199 200 201 202 203 204 205
-8
-6
-4
-2
0
2
4
x 10
-3
time
df
(b) df-tim e
Brownian
Poisson
3D Brownian
M/ G/In finit y
M/M/1
AnalyzingMulti-microgridwithStochasticUncertaintiesIncludingOptimalPVAllocation
247
as a tool in calculating the optimal placement of any
number and kind of DG units with a specific daily
production curve such as wind turbine systems, fuel
cells, microturbines with a goal of optimizing
distribution power system performance.
A transient response analysis of the system with
optimal PV locations and stochastic modeling of
loads and generation is obtained and control
parameters are designed using PSO. It is observed
that simultaneous stochasticity modeling of all
components should be considered for designing
robust controllers.
ACKNOWLEDGEMENTS
The authors would like to acknowledge partial
funding support from NSF#1351201 CAREER grant
for this research work.
REFERENCES
Mohammadi M., Hosseinian S.H., and Gharehpetian G.B.,
2012. Optimization of Hybrid Solar Energy
Sources/Wind Turbine Systems Integrated to Utility
Grids as Microgrid (MG) Under Pool/Bilateral/Hybrid
Electricity Market Using PSO. Solar Energy, vol. 86,
pp. 112–125.
Prenc R., Škrlec D., Komen V. 2013. Optimal PV System
Placement in a Distribution Network on the Basis of
Daily Power Consumption and Production Fluctuation
EuroCon, Zagreb, Croatia.
Medina A., Hernandez J.C., and F. Jurado, 2006. Optimal
Placement and Sizing Procedure for PV Systems on
Radial Distribution Systems. International Conference
on Power System Technology.
Singh R. K., Goswami S. K. 2009. A Genetic Algorithm
Based Approach for Optimal Allocation of Distributed
Generations in Power Systems for Voltage Sensitive
Loads. ARPN Journal of Engineering and Applied
Sciences.
Prenc R., V. Komen, N. Bogunovi, 2012. GIS-based
Determination of Optimal Accommodation of
Embedded Generation on the MV Network.
International Journal of Communications Antenna and
Propagation.
Costa P. M., M. A. Matos, 2004. Loss Allocation in
Distribution Networks with Embedded Generation.
IEEE Transaction on Power System.
Srisaen N., and A. Sangswang, 2006. Effects of PV Grid-
Connected System Location on a Distribution System.
IEEE Asia Pacific Conference on Circuits and Systems.
Méndez V. H., J. Rivier, J. I. de la Fuente, T. Gómez, J.
Arceluz, J. Marín, 2002. Impact of Distributed
Generation on Distribution Losses. In Proc.
Mediterranean Power, Athens, Greece.
Shukla T.N., S.P. Singh and K.B. Naik, 2008. Allocation of
Optimal Distributed Generation using GA for
Minimum System Losses. Fifteenth National Power
Systems Conference (NPSC), IIT Bombay.
Ackermann T., G. Andersson, L. Söder, 2001. Distributed
generation: a definition. Electric Power Systems
Research, 2001.
Duenas P., Reneses J., Barquin J., 2014. Dealing with
multi-factor uncertainty in electricity markets by
combining Monte Carlo simulation with spatial
interpolation techniques. IET Gener. Transm. Distrib.
Guo L., Liu W., B. Jiao, B. Hong, C. Wang, 2010. Multi-
objective stochastic optimal planning method for stand-
alone microgrid system. IET Gener. Transm.
Distribution.
Zhang, H., Li, P., 2011. Chance constrained programming
for optimal power flow under uncertainty. IEEE Trans.
Power System.
Soroudi, A., Caire, R., Hadjsaid, N., Ehsan, M., 2011.
Probabilistic dynamic multi-objective model for
renewable and non-renewable distributed generation
planning. IET Gener. Transm. Distribution.
Zhipeng, L., Fushuan, W., Ledwich, G., 2011. Optimal
siting and sizing of distributed generators in distribution
systems considering uncertainties. IEEE Trans. Power
Delivery.
Sofla M. A., King R., 2012. Control Method for Multi-
Microgrid Systems in Smart Grid Environment -
Stability, Optimization and Smart Demand
participation. IEEE PES Innovative Smart Grid
Technologies (ISGT).
Shrestha, G.B., Pokharel, B.K., Lie, T.T., 2008.
Management of price uncertainty in short-term
generation planning. IET Gener. Transm. Distribution.
Kai Z., Agalgaonkar A.K., Muttaqi K.M., 2012.
Distribution system planning with incorporating DG
reactive capability and system uncertainties. IEEE
Trans. Sustain. Energy.
Niknam, T., Golestaneh F., Malekpour A., 2012.
Probabilistic energy and operation management of a
microgrid containing wind/photovoltaic/fuel cell
generation and energy storage devices based on point
estimate method and self-adaptive gravitational search
algorithm. Energy.
Price, W. W., Chiang, H. D., Clark, H. K., Concordia, C.,
Lee, D. C., Hsu, J. C., Vaahedi, E., 1993. Load
representation for dynamic performance analysis. IEEE
Transactions on Power Systems, 8(2), 472-482.
Haque M. H., 1996. Load flow solution of distribution
systems with voltage dependent load models. Elect.
Pow. Syst. Res., vol. 36, pp. 151–156.
Huang Y., Peng F.Z., Wang J., 2006. Z-Source Inverter for
Residential Photovoltaic Systems. IEEE Trans. Power
Delivery, vol.21, no.6.
Kroposki B., Basso T., and DeBlasio R., 2008. Microgrid
Standards and Technologies. IEEE Power and Energy
Society General Meeting, pp. 1-4.
Babaei E., Galvani S. and Ahmadi Jirdehi M., 2009. Design
of Robust Power System Stabilizer Based on PSO.
IEEE Symposium on Industrial Electronics and Appli-
SMARTGREENS2015-4thInternationalConferenceonSmartCitiesandGreenICTSystems
248
cations, Malaysia.
Das T.K., Venayagamoorthy G.K., 2006. Optimal Design
of Power System Stabilizers Using a Small Population
Based PSO. IEEE Power Engineering Society General
Meeting.
Keshtkar, H.; Mohammadi, F.D.; Ghorbani, J.; Solanki, J.;
Feliachi, A., 2014. Proposing an improved optimal
LQR controller for frequency regulation of a smart
microgrid in case of cyber intrusions. IEEE 27th
Canadian Conference on Electrical and Computer
Engineering (CCECE).
AnalyzingMulti-microgridwithStochasticUncertaintiesIncludingOptimalPVAllocation
249