scispace - formally typeset
Search or ask a question
Journal ArticleDOI

An Optimal Energy Management System for Islanded Microgrids Based on Multiperiod Artificial Bee Colony Combined With Markov Chain

TL;DR: A multiperiod artificial bee colony optimization algorithm is implemented for economic dispatch considering generation, storage, and responsive load offers and shows cost reduction, convergence speed increase, and the remarkable improvement of efficiency and accuracy under uncertain conditions.
Abstract: The optimal operation programming of electrical systems through the minimization of the production cost and the market clearing price, as well as the better utilization of renewable energy resources, has attracted the attention of many researchers. To reach this aim, energy management systems (EMSs) have been studied in many research activities. Moreover, a demand response (DR) expands customer participation to power systems and results in a paradigm shift from conventional to interactive activities in power systems due to the progress of smart grid technology. Therefore, the modeling of a consumer characteristic in the DR is becoming a very important issue in these systems. The customer information as the registration and participation information of the DR is used to provide additional indexes for evaluating the customer response, such as consumer's information based on the offer priority, the DR magnitude, the duration, and the minimum cost of energy. In this paper, a multiperiod artificial bee colony optimization algorithm is implemented for economic dispatch considering generation, storage, and responsive load offers. The better performance of the proposed algorithm is shown in comparison with the modified conventional EMS, and its effectiveness is experimentally validated over a microgrid test bed. The obtained results show cost reduction (by around 30%), convergence speed increase, and the remarkable improvement of efficiency and accuracy under uncertain conditions. An artificial neural network combined with a Markov chain (ANN-MC) approach is used to predict nondispatchable power generation and load demand considering uncertainties. Furthermore, other capabilities such as extendibility, reliability, and flexibility are examined about the proposed approach.

Summary (2 min read)

I. INTRODUCTION

  • F EXIBILITY requirements in electric power systems and presence of non-dispatchable intermittent generation leads to development of Microgrids (MG)s [1] .
  • Therefore, presenting powerful optimization algorithms to extract the best possible solution for the MGs is very important.
  • These techniques are trying to seek good (near-optimal) solutions at a reasonable computational cost without being able to guarantee either feasibility or optimality, or even in many cases to state how close to optimality a particular feasible solution is.
  • According to the advantages of this method, it is applied in the present paper for the optimization of MG operation in terms of performance, generation resources scheduling and economic power dispatch.
  • The proposed model utilizes ANN for primary predictions.

II. ALGORITHMS IMPLEMENTATION FOR EMS

  • It comprises different units, namely ANN-MC, EMS and LEM units.
  • As shown in Fig. 1 , four different algorithms are presented for implementing EMS based on LEM by using heuristic techniques or without using any optimization method.
  • Flexibility, good accuracy, speed in decision making and plug and play abilities of LEM unit, MCEMS, EMS-MINLP (EMS based on mixed integer non-linear programming) and EMS-PSO (EMS based on particle swarm optimization) algorithms are discussed in detail in the previous studies [1] , [14] , [15] .
  • Therefore, these are not addressed in the present paper and only EMS-MABC algorithm is described.

A. EMS-MABC algorithm

  • This algorithm encompasses ANN-MC, MABC and LEM units as illustrated in Fig. 1 .
  • In the proposed model, two ANNs are used for prediction.
  • Then, according to TPM, the probability of predicted value is calculated during the next step.
  • The procedure can be summarized as follows (Fig. 3 ): EQUATION Step 1 TPM calculation based on 600 data points of wind speed; Step 2 Design of ANN-1 for primary prediction by using 300 other points of wind speed data; TPM accumulation According to Fig. 3 , ANN-1 designed in the previous step is applied for the primary prediction.
  • It must be checked during optimization process if the generated population members satisfy constraints or not.

IV. PROBLEM FORMULATION

  • The problem formulation is divided into two parts which are closely connected and dependent on each other.
  • The first part is related to the prediction error of uncertainty model and the other include MG constraints.

A. Error criteria for uncertainty consideration

  • The prediction error of a model is classically defined as the difference between the measured and predicted values.
  • A horizon dependent model error e t t+∆t is given by EQUATION where X denotes non-dispatchable resources and load demand entries.
  • The most commonly used evaluation criterion is the MAPE defined as follows [24] .
  • EQUATION where ∆t and N describe the prediction horizon and number of prediction, respectively.
  • It is very important to reduce MPE because a large prediction error and consequently wrong control commands may cause an unstable condition for non-dispatchable resources.

B. MG mathematical modeling

  • The system under study is considered as an islanded MG including non-dispatchable (WT and PV in this study) and dispatchable generation resources (MT in this study) and ES supplying some responsive (EWH and DR in this study)/ non-responsive loads (NRL).
  • The objective of economic dispatch problem is in fact minimizing the total production cost while satisfying generation resources constraints.

V. APPLICATION TO TEST GRID

  • EMS-MABC algorithm is implemented and validated experimentally over the IREC s MG.
  • All the microsources with any characteristic can easily be emulated by digital signal processing.
  • This MG has two non-dispatchable resources (PV and WT), a dispatchable resource (MT), and ES integrated with some responsive (EWH and DR) and NRL.
  • Emulators specifications are presented in the previous papers [1] .
  • Then, all optimal power set-point of each microsource will be dispatched to them at each time interval based on The ability of the proposed algorithm under several scenarios is considered for optimal scheduling and operation of resources, minimizing the generation cost as well as applying demand side management.

VI. RESULTS AND DISCUSSION

  • The results of experimental evaluation of the proposed algorithm over IREC s MG are presented.
  • During 06:00-12:00 period, MCEMS has already used ES for supplying a part of power shortage, while in EMS-MABC, ES is operated in the charging mode and continuing to reach SOC.
  • The DR constraints are expressed with various status flags, the information of other consumers and the excess power generated has been modeled to obtain the minimum total generation cost and less market clearing price.
  • This combined programming has been evaluated over a MG Testbed.
  • The proposed approach shows more decrease in the objective function than EMS-PSO algorithm while reducing computation time.

Did you find this useful? Give us your feedback

Content maybe subject to copyright    Report

Aalborg Universitet
An optimal energy management system for islanded Microgrids based on multi-period
artificial bee colony combined with Markov Chain
Marzband, Mousa; Azarinejadian, Fatemeh; Savaghebi, Mehdi; Guerrero, Josep M.
Published in:
I E E E Systems Journal
DOI (link to publication from Publisher):
10.1109/JSYST.2015.2422253
Publication date:
2017
Document Version
Early version, also known as pre-print
Link to publication from Aalborg University
Citation for published version (APA):
Marzband, M., Azarinejadian, F., Savaghebi, M., & Guerrero, J. M. (2017). An optimal energy management
system for islanded Microgrids based on multi-period artificial bee colony combined with Markov Chain. I E E E
Systems Journal, 11(3), 1712 - 1722. https://doi.org/10.1109/JSYST.2015.2422253
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
- Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
- You may not further distribute the material or use it for any profit-making activity or commercial gain
- You may freely distribute the URL identifying the publication in the public portal -
Take down policy
If you believe that this document breaches copyright please contact us at vbn@aub.aau.dk providing details, and we will remove access to
the work immediately and investigate your claim.

1
An optimal energy management system for islanded
Microgrids based on multi-period artificial bee
colony combined with Markov Chain
Mousa Marzband, Fatemeh Azarinejadian, Mehdi Savaghebi, Member, IEEE, and Josep M. Guerrero, Fellow,
IEEE
Abstract—Optimal operation programming of electrical
systems through minimization of production cost and market
clearing price (MCP) as well as better utilization of renewable
energy resources has attracted the attention of many researchers.
To reach this aim, energy management systems (EMS) has
been studied in many research activities. Moreover, demand
response (DR) expands customer participation to power
systems and results in a paradigm shift from conventional
to interactive activities in power systems due to the progress
of smart grid technology. Therefore, modelling of consumer
characteristic in DR is becoming so important issues in
these systems. The customer information as the registration
and participation information of DR is used to provide
additional indices for evaluating customer response, such
as consumer
0
s information based on the offer priority, DR
magnitude, duration, and minimum cost of energy (COE).
In this paper, a multi-period artificial bee colony (MABC)
optimization algorithm is implemented for economic dispatch
considering generation, storage and responsive load offers. The
better performance of the proposed algorithm is shown in
comparison with the modified conventional energy management
system (MCEMS) and its effectiveness is validated experimentally
over a Microgrid (MG) Testbed. The obtained results show
cost reduction (by around 30%), convergence speed increase
as well as remarkable improvement of efficiency and accuracy
under uncertain conditions. An artificial neural network (ANN)
combined with Markov-chain (MC) (ANN-MC) approach is used
to predict non-dispatchable power generation and load demand
considering uncertainties. Furthermore, other capabilities such
as extendibility, reliability and flexibility are examined about the
proposed approach.
Index Terms—artificial bee colony, demand response,
microgrid, Optimum energy management, optimum scheduling
of DG, responsive load demand, uncertainty.
This work was supported by the Energy Technology Development and
Demonstration Program (EUDP) through the Sino-Danish Project “Microgrid
Technology Research and Demonstration” (meter.et.aau.dk).
M. Marzband is (email: mousa.marzband@manchester.ac.uk) with School
of Electrical and Electronic Engineering, Faculty of Engineering and Physical
Sciences, Electrical Energy and Power Systems Group, The University of
Manchester, Ferranti Building, Manchester, M13 9PL, United kingdom.
M. Marzband (email: m.marzband@liau.ac.ir) and F. Azarinejadian (email:
niloofar.azari@yahoo.com) are with Department of Electrical Engineering,
Lahijan Branch, Islamic Azad University, Lahijan, Guilan, Iran.
M. Savaghebi and J. M. Guerrero are with the Department of Energy
Technology, Aalborg University, DK-9220 Aalborg East, Denmark (email:
mes@et.aau.dk , joz@et.aau.dk).
NOMENCLATURE
Acronyms
ABC artificial bee colony
ANN artificial neural network
DR demand response
DSM demand side management
EGP excess generated power
EMS energy management system
ES energy storage
ES+ ES during charging mode
ES- ES during discharging mode
EWH electric water heater
MG microgrid
LEM local energy market
MABC multi-period ABC
MCEMS modified conventional EMS
MC Markov-chain
MCP market clearing price
MLP multi-layer perceptron
MINLP mixed integer non-linear programming
MPE maximum prediction error
MT micro-turbine
NRL non-responsive load
PSO particle swarm optimization
PV photovoltaic
RLD responsive load demand
SOC state-of-charge
TPM transition probability matrix
UP undelivered power
WT wind turbine
Variables
π
A
the supply bids by A (e/kWh)
A {WT, PV, MT, ES-, ES+, UP, EGP, & EWH}
λ
MCP
t
MCP at each time t in MCEMS (e/kWh)
λ
0MCP
t
MCP at each time t in EMS-MABC (e/kWh)
P
A
t
available power of A in MCEMS (kW)
P
0A
t
available power of A in EMS-MABC (kW)
˜
P
A
t
real power set-points of A in MCEMS (kW)
˜
P
0A
t
real power set-points of A in EMS-MABC (kW)
P
A
t
available power of A (kW)
P
n
t
non-responsive load (NRL) demand (kW)
SOC
t
battery SOC in MCEMS (%)
SOC
0
t
battery SOC in EMS-MABC (%)
P , P limit of power (kW)
E, E limit of energy (kWh)
this document downloaded from www.microgrids.et.aau.dk is the preprintversion of the paper:
M. Marzband, F. Azarinejadian, M. Savaghebi, and J. M. Guerrero, "An optimal energy management system for islanded
Microgrids based on multi-period artificial bee colony combined with Markov Chain", IEEE Systems Journal, 2015.

2
SOC maximum SOC (%)
SOC minimum SOC (%)
t time step
I. INTRODUCTION
F
EXIBILITY requirements in electric power systems
and presence of non-dispatchable intermittent generation
leads to development of Microgrids (MG)s [1]. An MG
can be defined as a small power system consisting of
power converter-based generation, energy storage devices,
small classic synchronous generation and various types of
loads. This configuration with a proper control could provide
lots of advantages to consumers such as better power
quality, higher reliability, more flexibility, and less operation
and generation cost [2]–[5]. Adequate amount of demand
side delivery in MGs has significant importance due to
limitations of using non-dispatchable resources [6], [7]. The
problem of demand-supply mismatch exists in these systems
if energy generation resources are not adequate to supply
the requested load and no proper EMS is employed. An
EMS makes optimal use of available DGs while ensuring
the flexibility, reliability and quality of the supply. However,
it may also fail to produce the load demand if total
demand is more than the total generated power. Under such
scenarios, utilizing backup systems such as energy storage
(ES), diesel generators or applying demand response (DR)
helps to reduce the demand-supply mismatch [8], [9]. At
present, ES can be implemented only in small scale and for
a short-time supply. Moreover, DR mechanism may leads
to reduction of the fluctuations resulting from random and
unwanted requests which may help to provide peak shaving
[10], [11]. The combined operation of ES and DR with
DG technologies provide more reliability for MG operation
[8], [9], [12]. Hence, intelligent control systems must be
developed to accommodate ES and DR in MGs in order to
supply consumers as required [6], [13]. Optimal management
of MG generation units requires exact determination of
constraints to describe the operation problem considering the
output power generation with the least possible generation
cost [14]. These are often represented as a large scale,
non-convex, nonlinear, mixed-integer problems. Therefore,
presenting powerful optimization algorithms to extract the best
possible solution for the MGs is very important. Deterministic
optimization methods are highly dependent on the system and
their definition is very difficult for large complex systems.
In solving optimization problems with a high-dimensional
search space particularly in UC and ED problems, the
deterministic and stochastic optimization algorithms do not
provide a suitable solution because the search space increases
exponentially with the problem size, therefore solving these
problems using exact techniques (such as [15]) is not practical.
On the other hand, these problems can be solved with
the non-deterministic polynomial-hard (NP-hard) problem.
Heuristic algorithms such as genetic algorithm [16], particle
swarm optimization [17], ant colony optimization [18] and bee
colony optimization [19] are some optimization methods used
for unit commitment within MGs [14]. Some algorithms give a
better solution for some particular problems than others. These
techniques are trying to seek good (near-optimal) solutions at a
reasonable computational cost without being able to guarantee
either feasibility or optimality, or even in many cases to state
how close to optimality a particular feasible solution is. In
addition, most of these approaches have a stochastic behaviour.
Therefore, it is made effort to present a deterministic heuristic
search algorithm based on a swarm meta-heuristic algorithm.
In [1], the design of an energy management system (EMS)
is developed in order to obtain the best purchasing price
in day-ahead market (DAM), as well as to maximize the
utilization of existing DER and study the system stability
is reported. However, no optimization approach was used in
that work. Furthermore, the research work presented in this
paper is a continuation of the work by the authors [15],
where a framework for combining stochastic optimization,
non-dispatchable resources/load demand uncertainties, and
local optimization is needed.
Amongst them, special attention is paid to the optimization
algorithms based on artificial bee colony (ABC) for solving
optimization problems due to the population-based search
capability, simplicity of implementation, adequate convergence
speed and robustness [14], [20]–[23]. According to the
advantages of this method, it is applied in the present paper
for the optimization of MG operation in terms of performance,
generation resources scheduling and economic power dispatch.
For increasing effectiveness and usability in MG applications,
an algorithm based on multi-period ABC (MABC) is proposed
in this paper for solving energy management problems over
a real MG for a day-ahead period. It is noteworthy that the
proposed algorithm can also find the global optimal point in
the multi-dimensional and great search space.
Another approach proposed in this paper is based on
modeling the uncertainty in load demands and the generation
of renewable resources. A model is presented for very
short-term prediction by using artificial neural network (ANN),
Markov-chain (MC) and linear regression. The proposed
model utilizes ANN for primary predictions. Then, the
second-order MC is applied to determine transition probability
matrix (TPM) for primary prediction. Finally, a linear
regression is used between the primary predictions and
probability values obtained by MC for the final prediction.
The MC is applied to modify the predicted values according
to long-term pattern of the resource data. Applying ANN
without using statistical models, increases the number of input
variables for both training and utilization [24]. Further, two
limitations on the use of ANN models also exist that seriously
affect the prediction performance, namely, over-training and
extrapolation [25]. Over-training occurs when the capacity of
the ANN for training is too great, because too many training
iterations is allowed. For extrapolation, the advantages of the
ANNs have not been determined when they are required to
perform estimation beyond available experimental data [25].
Both of these ANN imperfections are taken into account in
the model proposed in this paper.
The contributions of the paper are as follows:
1) development of an intelligent algorithm based on
ABC within a real MG towards supporting real time
applications;

3
2) presentation of an algorithm based on artificial neural
network (ANN) combined with Marcov chain to consider
system uncertainties;
a) prediction of wind speed in a very short-time (adequate
for real-time optimization);
b) reduction of prediction error and uncertainty of
predictions;
c) significant reduction in calculation time which is
considered very critical in real-time applications.
3) experimental implementation of the proposed smart
algorithm demonstrating some benefits including flexible
multi-device support, fast development with a running
time proper for real-time applications.
II. ALGORITHMS IMPLEMENTATION FOR EMS
The EMS proposed in this paper is depicted in Fig. 1.
It comprises different units, namely ANN-MC, EMS and
LEM units. As shown in Fig. 1, four different algorithms
are presented for implementing EMS based on LEM by using
heuristic techniques or without using any optimization method.
Flexibility, good accuracy, speed in decision making and
plug and play abilities of LEM unit, MCEMS, EMS-MINLP
(EMS based on mixed integer non-linear programming) and
EMS-PSO (EMS based on particle swarm optimization)
algorithms are discussed in detail in the previous studies [1],
[14], [15]. Therefore, these are not addressed in the present
paper and only EMS-MABC algorithm is described.
PVWTn
ttt
P,P,P
PVWTn
ttt
P,P,P
Fig. 1: Proposed algorithm for implementing EMS
A. EMS-MABC algorithm
This algorithm encompasses ANN-MC, MABC and LEM
units as illustrated in Fig. 1. Since LEM unit is explained in
detail in [1] and [15], only ANN-MC and MABC units are
discussed bellow.
1) ANN-MC unit: In this study, MC method is applied
for obtaining long-term trends in wind speed data. Thus,
a simple ANN structure with the minimum number of
input variables and data regulations is required for training
and the over-training problem can be solved with the
proposed structure. As the MC method keeps the signals
long-term behavior in the memory, the error obtained from
the extrapolated prediction is also reduced. As another solution
for extrapolation problem, the artificial samples covering the
entire range are drawn as much as possible based on the
existing knowledge about the proposed problem then used for
ANN initializing to ensure that most of the future prediction
involves interpolation. The outline of the proposed model is
shown in Fig. 2. A set of wind speed data 2.5s in a 175min
period is used to improve the model accuracy for predicting
wind speed up to 7.5s ahead (total of 4200 wind speed
data). In Fig. 2, TPM is transition probabilities matrix for the
primary prediction, forward neighborhood indices (FNI)s and
Backward neighborhood indices (BNIs) are two upper/lower
states and their corresponding probabilities, respectively. v
i
tk
is the real speed data at time t-k, and ˆv
tk|t
is the predicted
wind speed data for t+k and i states an index of the model
i
th
vector used in ANN-1. Also, v
t1
, v
t2
, · · · , v
tn
are
considered as wind real speed data which are used for forming
TPM by MC. In the proposed model, two ANNs are used for
prediction. The first ANN (ANN-1) is applied for primary
prediction and short-term obtaining of wind speed signal,
where 10 real-time speed data from t to t-10 are used as input
variables. Primary prediction can take place by ANN-1 for
different time horizon. For training, 30 sets of data with 10
measured wind speed in each set are selected. After primary
prediction, the provided TPS for the values and four other
indices with primary predictions are fed as input variables
to the second ANN (ANN-2). At the end, the implemented
model based on two ANNs and MC method can be utilized
for predicting different time horizons. Multi-layer perceptron
(MLP) is used for ANN-1 which includes an input layer, a
hidden layer and an output layer. In the output layer, only
one neuron is used as ˆv
tk|t
in which k is the time step
and ˆv is the anticipated wind speed at time t+k (calculated
at time t). Because the number of neurons in each layer
have an effect on the speed and network stability, sensitivity
analysis shows that the structure of ANN-1 with the least
mean absolute percentage error (MAPE) is equal to 5, 2
and 1 neurons for input, hidden and output layers with 30
training vectors and 0.01-0.08 learning rates. Based on the
wind speed data histogram, wind speed states have become
compatible with the 1m/s upper and lower limit difference of
the wind speed for reaching high accuracy at an acceptable
time. Based on the state matrix, it is possible to find the
number of transitions from the two previous states during wind
speed data sequence to the next state at time t+k. Finally,
TPM is calculated. TPM is formed by using 600 wind speed
data and the calculated matrix is used as primary prediction
values (Fig. 2). At the beginning, Markov state is calculated for
the primary prediction values by ANN-1 for one step ahead.
Then, according to TPM, the probability of predicted value
is calculated during the next step. This process is carried out
for all of the primary predictions. It must be noted that the
prediction values of the previous step are generated by ANN-1.
For the final prediction, MLP has been used for ANN-2. The
number of neurons in the input layer is selected by considering
the calculation of time and error (maximum prediction error
(MPE) and MAPE). Since ANN-2 has six input and one output

4
variables, the number of neurons in each layer must be located
in the range of variables and the best structure for ANN-2 with
the least MAPE is estimated as 3,0 and 1 neuron for the input,
hidden and output layers with 10 training vector and learning
rate between 0.01-0.05. In Fig. 3, the proposed model is shown
for ANN-MC unit.
1
i
t
ANN-1 for
primary
prediction
(m=10)
2
i
t
v
i
tm
v
1
t
2
t
v
tn
v
+
tkt
v
(
)
01
:
TPM
(
)
01
:
FNIs
(
)
10
:
BNIs
Fig. 2: General outline for the proposed model in ANN-MC
unit
Step 1:
MC transition probability matrix formation
Step 2:
ANN-1 design for primary prediction
[600,1]
[30,10]
Step 3:
ANN-1 test and ANN-2 design
[10,10]
TPM
FNI
BNI
Ack
Step 4:
Final test
[3200,10]
[3200,3]
[3200,1]
Fig. 3: Flowchart of four stages for implementation of the
proposed model for ANN-MC unit
Annual pattern of non-dispatchable power generation and
load demand are captured by ANN. Then, second-order MC
is applied to calculate TPM. Basically in TPM formation,
the non-dispatchable power generation and load demand
time series are converted to power states which contain the
generated and consumed powers among certain values. Based
on state matrix, it is possible to find the number of transition
from two preceding states in the sequence of power data
to another state at time. Firstly, Markov state for primary
values predicted by ANN-1 is calculated for one step ahead.
Then, according to TPM, the probability of predicted value
is calculated in the next step. This process is carried out for
all primary predictions. It should be noted that the predicted
values are produced in the previous step by ANN-1. For longer
prediction horizon, transition probabilities are necessary for
steps ahead. In these cases, the above TPM is multiplied in
ANN-1 according to the number of time steps in the future. It
is difficult to determine the relationships between the primary
prediction and the coefficients obtained from MC. Since ANNs
can encode complex and non-linear relations, ANN-2 is used
to capture the relationships between the primary prediction
and obtained probabilities. The transition probability of the
predicted values state and ANN-1 output are fed to ANN-2
in order to achieve higher prediction accuracy under uncertain
conditions in comparison with primary predicted values. The
procedure can be summarized as follows (Fig. 3):
Step 1
TPM calculation based on 600 data points of wind speed;
Step 2
Design of ANN-1 for primary prediction by using 300
other points of wind speed data;
Calculation of MAPE and MPE
Step 3
Implementation of MC model for testing ANN-1 and
designing ANN-2 (another 100 data set);
finding non-dispatchable and load demand data state
evaluation of different states transition
TPM calculation
TPM accumulation
According to Fig. 3, ANN-1 designed in the previous
step is applied for the primary prediction. Then, the TPM
calculated in step 1 is used to calculate the required
coefficients. ANN-2 provides six input variables of
primary wind speed prediction, their transition probability
values, FNI-1 and FNI-2 of the current predicted states
and BNI-1 and BNI-2 of previous predicted states.
Step 4
Design of ANN-2 for secondary prediction by using
ANN-1 and TPM.
Both ANNs and TPM obtained are applied steps for the
final prediction.
All the above steps must be applied for different prediction
time horizons.
2) MABC unit: The flowchart of MABC unit is shown
in Fig. 4. The highlighted areas in this Figure are the
modifications made to ABC algorithm in order to adapt it in
MG application. Each response of the optimization problem
has D variables. In this paper, D = 7 is considered including
WT (P
W T
t
), PV (P
P V
t
), MT (P
MT
t
), charging and discharging
power ES (P
ES+
t
and P
ES
t
), EWH (P
EW H
t
) and DR (P
DR
t
)
variables. The proposed algorithm is trying to find the optimal
values for the design variables that minimize the objective
function. Therefore, X
i
t
is defined as X
i
t
= x
i,1
t
, x
i,2
t
, · · · , x
i,7
t
vector. The elements are x
i,1
t
= P
i,W T
t
, x
i,2
t
= P
i,P V
t
, x
i,3
t
=
P
i,MT
t
, x
i,4
t
= P
i,ES+
t
, x
i,5
t
= P
i,ES
t
, x
i,6
t
= P
i,EW H
t
and
x
i,7
t
= P
i,DR
t
. These variables are divided into two categories
of dependent (P
i,MT
t
, P
i,ES+
t
, P
i,ES
t
, P
i,EW H
t
and P
i,DR
t
)
and independent (P
i,W T
t
and P
i,P V
t
) variables. Since WT
and PV are non-dispatchable resources which are affected
by weather conditions, MT and ES powers can be varied
depending on the power generated by WT and PV and energy
consumed by load. To begin, independent variables must
be made considering ANN-MC unit output. It is necessary
to involve target population members in program planning,
implementation and evaluation of objective function. It must
be checked during optimization process if the generated
population members satisfy constraints or not. Then, by using
valid values for these independent variables and associated
constraints, dependent variables can be generated randomly.
Furthermore, after selecting a food source, the onlooker bee
generates a new food source. MABC unit is illustrated by a

Citations
More filters
Journal ArticleDOI
TL;DR: A comparative and critical analysis on decision making strategies and their solution methods for microgrid energy management systems are presented and various uncertainty quantification methods are summarized.

617 citations

Journal ArticleDOI
TL;DR: In this paper, a load restoration optimization model is proposed to coordinate topology reconfiguration and microgrid formation while satisfying a variety of operational constraints, which exploits benefits of operational flexibility provided by grid modernization to enable more critical load pickup.

232 citations

Journal ArticleDOI
TL;DR: In this article, an algorithm for energy management system (EMS) based on multi-layer ant colony optimization (EMS-MACO) is presented to find energy scheduling in microgrid (MG).

223 citations

Journal ArticleDOI
TL;DR: In this article, the authors investigated several sustainable hybrid renewable systems for electricity production in Iran and concluded that the hybrid configuration composed of photovoltaic (PV), wind turbine, diesel generator and battery produced the best outcome with an energy cost of 0.151$/kWh and 15.6% return on investment.

203 citations

Journal ArticleDOI
TL;DR: In this article, a combined sizing and energy management methodology, formulated as a leader-follower problem, is proposed to select the optimal size for the microgrid components, which is solved using a genetic algorithm.

186 citations

References
More filters
Journal ArticleDOI
TL;DR: To address the intrinsically stochastic availability of renewable energy sources (RES), a novel power scheduling approach is introduced that involves the actual renewable energy as well as the energy traded with the main grid, so that the supply-demand balance is maintained.
Abstract: Due to its reduced communication overhead and robustness to failures, distributed energy management is of paramount importance in smart grids, especially in microgrids, which feature distributed generation (DG) and distributed storage (DS). Distributed economic dispatch for a microgrid with high renewable energy penetration and demand-side management operating in grid-connected mode is considered in this paper. To address the intrinsically stochastic availability of renewable energy sources (RES), a novel power scheduling approach is introduced. The approach involves the actual renewable energy as well as the energy traded with the main grid, so that the supply-demand balance is maintained. The optimal scheduling strategy minimizes the microgrid net cost, which includes DG and DS costs, utility of dispatchable loads, and worst-case transaction cost stemming from the uncertainty in RES. Leveraging the dual decomposition, the optimization problem formulated is solved in a distributed fashion by the local controllers of DG, DS, and dispatchable loads. Numerical results are reported to corroborate the effectiveness of the novel approach.

718 citations


"An Optimal Energy Management System..." refers background in this paper

  • ...Hence, intelligent control systems must be developed to accommodate the ES and the DR in MGs in order to supply consumers as required [6], [13]....

    [...]

  • ...An adequate amount of demand-side delivery in MGs has significant importance due to the limitations of using nondispatchable resources [6], [7]....

    [...]

Journal ArticleDOI
TL;DR: The modeling, control design, and stability analysis of parallel-connected three-phase VSIs are derived and a hierarchical control scheme for the paralleled VSI system control architecture is developed.
Abstract: Power-electronics-based microgrids (MGs) consist of a number of voltage source inverters (VSIs) operating in parallel. In this paper, the modeling, control design, and stability analysis of parallel-connected three-phase VSIs are derived. The proposed voltage and current inner control loops and the mathematical models of the VSIs are based on the stationary reference frame. A hierarchical control scheme for the paralleled VSI system is developed comprising two levels. The primary control includes the droop method and the virtual impedance loops, in order to share active and reactive powers. The secondary control restores the frequency and amplitude deviations produced by the primary control. Also, a synchronization algorithm is presented in order to connect the MG to the grid. Experimental results are provided to validate the performance and robustness of the parallel VSI system control architecture.

610 citations

Journal ArticleDOI
TL;DR: In this article, an optimization model including battery life loss cost, operation and maintenance cost, fuel cost, and environmental cost is established to obtain a set of optimal parameters of operation strategy considering the lifetime characteristics of lead-acid batteries, a multiobjective optimization to minimize power generation cost and to maximize the useful life of leadacid batteries has been achieved via the NSGA-II.
Abstract: Standalone microgrids with renewable sources and battery storage play an important role in solving power supply problems in remote areas such as islands To achieve reliable and economic operations of a standalone microgrid, in addition to the consideration of utilization of renewable resources, the lifetime characteristics of a battery energy storage system also need to be fully investigated In this paper, in order to realize the economic operation of a recently developed standalone microgrid on Dongfushan Island in China, an optimization model including battery life loss cost, operation and maintenance cost, fuel cost, and environmental cost is established to obtain a set of optimal parameters of operation strategy Considering the lifetime characteristics of lead-acid batteries, a multiobjective optimization to minimize power generation cost and to maximize the useful life of lead-acid batteries has been achieved via the nondominated sorting genetic algorithm (NSGA-II) The results show that the proposed method can optimize the system operations under different scenarios and help users obtain the optimal operation schemes of the actual microgrid system

416 citations


"An Optimal Energy Management System..." refers methods in this paper

  • ...Heuristic algorithms such as a genetic algorithm [16], PSO [17], ant colony optimization [18], and bee colony optimization [19] are some optimization methods used for the UC within MGs [14]....

    [...]

Journal ArticleDOI
TL;DR: The results show the effectiveness of the proposed control structure in compensating the voltage unbalance in an islanded microgrid.
Abstract: The concept of microgrid hierarchical control is presented recently. In this paper, a hierarchical scheme is proposed which includes primary and secondary control levels. The primary level comprises distributed generators (DGs) local controllers. The local controllers mainly consist of power, voltage and current controllers, and virtual impedance control loop. The central secondary controller is designed to manage the compensation of voltage unbalance at the point of common coupling (PCC) in an islanded microgrid. Unbalance compensation is achieved by sending proper control signals to the DGs local controllers. The design procedure of the control system is discussed in detail and the simulation results are presented. The results show the effectiveness of the proposed control structure in compensating the voltage unbalance.

407 citations


Additional excerpts

  • ...operation and generation cost [2]–[5]....

    [...]

Journal ArticleDOI
TL;DR: In this paper, the optimal control of the microgrid's energy storage devices is addressed, where stored energy is controlled to balance power generation of renewable sources to optimize overall power consumption at the micro-grid point of common coupling.
Abstract: Energy storage may improve power management in microgrids that include renewable energy sources. The storage devices match energy generation to consumption, facilitating a smooth and robust energy balance within the microgrid. This paper addresses the optimal control of the microgrid's energy storage devices. Stored energy is controlled to balance power generation of renewable sources to optimize overall power consumption at the microgrid point of common coupling. Recent works emphasize constraints imposed by the storage device itself, such as limited capacity and internal losses. However, these works assume flat, highly simplified network models, which overlook the physical connectivity. This work proposes an optimal power flow solution that considers the entire system: the storage device limits, voltages limits, currents limits, and power limits. The power network may be arbitrarily complex, and the proposed solver obtains a globally optimal solution.

378 citations


"An Optimal Energy Management System..." refers background in this paper

  • ...An adequate amount of demand-side delivery in MGs has significant importance due to the limitations of using nondispatchable resources [6], [7]....

    [...]

Frequently Asked Questions (17)
Q1. What are the contributions in "An optimal energy management system for islanded microgrids based on multi-period artificial bee colony combined with markov chain" ?

To reach this aim, energy management systems ( EMS ) has been studied in many research activities. The customer information as the registration and participation information of DR is used to provide additional indices for evaluating customer response, such as consumer′s information based on the offer priority, DR magnitude, duration, and minimum cost of energy ( COE ). In this paper, a multi-period artificial bee colony ( MABC ) optimization algorithm is implemented for economic dispatch considering generation, storage and responsive load offers. Furthermore, other capabilities such as extendibility, reliability and flexibility are examined about the proposed approach. 

It is very important to reduce MPE because a large prediction error and consequently wrong control commands may cause an unstable condition for non-dispatchable resources. 

The ability of the proposed algorithm under several scenarios is considered for optimal scheduling and operation of resources, minimizing the generation cost as well as applying demand side management. 

As the MC method keeps the signalslong-term behavior in the memory, the error obtained from the extrapolated prediction is also reduced. 

by proper selection of MT, ES is operated in the charging mode in EMS-MABC and at the end of this time interval, SOC is about 80%. 

In this paper, to improve exploitation process of classic ABC, a different probability function modifying searching mechanism has been applied to the original ABC algorithm. 

by using valid values for these independent variables and associated constraints, dependent variables can be generated randomly. 

The number of neurons in the input layer is selected by considering the calculation of time and error (maximum prediction error (MPE) and MAPE). 

these modifications are based on reducing the colony size; maintaining the perturbation scheme; and using a rank selection strategy for maintaining diversity. 

good accuracy, speed in decision making and plug and play abilities of LEM unit, MCEMS, EMS-MINLP (EMS based on mixed integer non-linear programming) and EMS-PSO (EMS based on particle swarm optimization) algorithms are discussed in detail in the previous studies [1], [14], [15]. 

Update the best solution acquired so farend while Return optimal power set-pointsend forproposed MABC, binary numbers 1 and 0 are used to indicate the status of generating units ON/OFF whereas the economic dispatch is solved using the real coded ABC. 

Improving strategy throughput by constraint based management in MABC whenever the commitment status for each time interval is generated randomly or by the modification of employed/onlooker bee′s position, dispatchable constraint must be checked as follows:Step 1: If dispatchable resources constraints are met, then go to Step 3. 

The outline of the proposed model is shown in Fig. 2. A set of wind speed data 2.5s in a 175min period is used to improve the model accuracy for predicting wind speed up to 7.5s ahead (total of 4200 wind speed data). 

The minimum value of λMCPt and λ ′MCP t are respectively 0.2 e/kWh and 0.13 e/kWh which are obtained for both algorithms during 00:00-06:00. 

Since WT and PV are non-dispatchable resources which are affected by weather conditions, MT and ES powers can be varied depending on the power generated by WT and PV and energy consumed by load. 

The DR constraints are expressed with various status flags, the information of other consumers and the excess power generated has been modeled to obtain the minimum total generation cost and less market clearing price. 

These variables are divided into two categories of dependent (P i,MTt , P i,ES+ t , P i,ES− t , P i,EWH t and P i,DR t ) and independent (P i,WTt and P i,PV t ) variables.