Energy 126 (2017) 622e637

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Demand side management in a smart micro-grid in the presence of

renewable generation and demand response
G.R. Aghajani a, *, H.A. Shayanfar b, H. Shayeghi b, c
a
Department of Electrical Engineering, Science and Research Branch, Islamic Azad University, Tehran, Iran
b
Centre of Excellence for Power System Automation and Operation, Department of Electrical Engineering, Iran University of Science and Technology, Tehran,
Iran
c
Technical Engineering Department, University of Mohaghegh Ardabaili, Ardabil, Iran

a r t i c l e i n f o a b s t r a c t

Article history: In this study, a stochastic programming model is proposed to optimize the performance of a smart micro-
Received 9 September 2015 grid in a short term to minimize operating costs and emissions with renewable sources. In order to
Received in revised form achieve an accurate model, the use of a probability density function to predict the wind speed and solar
17 November 2016
irradiance is proposed. On the other hand, in order to resolve the power produced from the wind and the
Accepted 12 March 2017
solar renewable uncertainty of sources, the use of demand response programs with the participation of
residential, commercial and industrial consumers is proposed. In this paper, we recommend the use of
incentive-based payments as price offer packages in order to implement demand response programs.
Keywords:
Smart microgrid
Results of the simulation are considered in three different cases for the optimization of operational costs
Renewable generation and emissions with/without the involvement of demand response. The multi-objective particle swarm
Demand side management optimization method is utilized to solve this problem. In order to validate the proposed model, it is
Demand response program employed on a sample smart micro-grid, and the obtained numerical results clearly indicate the impact
of demand side management on reducing the effect of uncertainty induced by the predicted power
generation using wind turbines and solar cells.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction AMI system generally includes:

Future distribution systems will certainly face the increased
Smart meters with Power Line Carrier (PLC) communications
penetration of wind and solar renewable sources, which have an installed at the customer premises. The smart meter of medium
intermittent natural behavior. This may endanger the security of and large customers using General Packet Radio Service (GPRS)
the system operation [1,2]. In order to implement advanced plan- could be directly connected to the utility.
ning for Distributed Energy Resources (DERs) to ensure the eco-
To manage all smart meter measured data from each installa-
nomic and safe operation of these systems, and Advanced tion, Data Concentrators (DC) are installed in the proximity of
Measuring Infrastructure (AMI) is necessary [3e5]. AMI establishes 20 kV/400 V distribution transformers. Data concentrators
a bidirectional telecommunication between customers and elec- integrate PLC communications that exchange information with
tricity companies to provide readability, monitoring, and remote smart meters and communicate with central.
control of meters; data collection and transmission to electricity
Meter Data Management Systems (MDMSs) are mainly Meter
companies; processing and analysis of information, as well as the Data Management & Repository (MDM/R) systems in which the
implementation of energy consumption management in an received unprocessed data are collected from all meters or
attempt to ensure the reliability of the system and to guarantee the sensors then processed in order to deliver the required data to
creation of a balance between supply and demand [6e8]. distributed system operator and application systems.
To manage and control a smart microgrid, the structure of the
One of the main drawbacks in the management of renewable
resources, including wind and solar energies, is the issue of un-
* Corresponding author. Basij Sq, Ardabil, Iran. Tel./fax: þ98 4533513957. certainty in their behavior, such that before the use of solar and
E-mail address: aghajani__gholamreza@yahoo.com (G.R. Aghajani). wind energy and other renewable energies in power system,

http://dx.doi.org/10.1016/j.energy.2017.03.051
0360-5442/© 2017 Elsevier Ltd. All rights reserved.
 G.R. Aghajani et al. / Energy 126 (2017) 622e637 623

Nomenclature RCDRj ðtÞ demand response program coats of the j th load in

period t
r total number of residential consumers PGrid(t) active power bought/sold from/to the utility in period t
c total number of commercial consumers pGrid ðtÞ offered price bought/sold from/to the utility in period t
i total number of industrial consumers CDG
i;s ðtÞ running cost of the ith DG unit at the t th period in the
h efficiency of the PV system sth scenario
Si solar irradiance(kW/m2) CDR
j;s ðtÞ the cost caused by load reduction by the j th DRPs
Prs probability of scenarios during the t th period in the sth scenario
aw scale parameter Weibull distribution ENSn,s(t) amount of involuntarily load shedding in period t and
bw shape parameter Weibull distribution scenario s
Vwind wind speed(m/s) EmiDG(t) average pollution of DG units
Vm average wind speed EmiGrid(t) average pollution of Grid
PR rated power of the wind turbine CO2;i ðtÞ carbon dioxide pollutants of ith DG unit in period t (kg/
Vci cut-in speed of the wind turbine MWh)
Vr rated speed of the wind turbine SO2;i ðtÞ sulfur dioxide pollutants of ith DG unit in period t (kg/
Vco cut-in speed of the wind turbine MWh)
fv(Vwind) wind speed probability density function NOx;i ðtÞ nitrogen oxide pollutants of ith DG unit in period t (kg/
Fv(Vwind) wind speed distribution function MWh)
Pw(Vwind) power output of WECS(kW) PDemandL;s load consumption in period t and scenario s
fPw ðPw Þ probability density function for the power output of PDR,s(t) active power participated in DPRs
WECS RDG(i,t) scheduled spinning reserve provided by DG I in period
fPPV ðsiÞ solar irradiance probability density function t
FPPV ðsiÞ solar irradiance distribution function PDG(i,t,s) active output power of DG I in period t and scenario s
PPV ðsiÞ power output of PVS min
PDG;i minimum output power limit of DG i
Ph power output of HSWPS max
PDG;i maximum output power limit of DG i
fh ðPh Þ probability density function for the power output of Wess(t) battery energy storage at time t
HSWPS hcharge(hdischarge) charge(discharge) efficiency of the battery
RC(r,t) amount of load reduction planned by each residential
consumer in period t List of abbreviations
CC(c,t) amount of load reduction planned by each commercial PDF Probability Density Function
consumer in period t CDF Cumulative Distribution Function
IC(i,t) amount of load reduction planned by each industrial DRP Demand Response Program
consumer in period t DSM Demand Side Management
RCmaxt maximum load reduction proposed by each residential DER Distributed Energy Resource
consumer in period t AMI Advanced Metering Infrastructure
CCmaxt maximum load reduction proposed by each DR Demand Response
commercial consumer in period t PSO Particle Swarm Optimization
ICmax
t maximum load reduction proposed by each industrial MOPSO Multi-Objective Particle Swarm Optimization
consumer in period t WES Wind Energy System
zr;t amount of incentive payment to each residential PLC Power Line Carrier
consumer in period t DC Data Concentrators
zc;t amount of incentive payment to each commercial MDM/R Meter Data Management & Repository
consumer in period t PVS Photovoltaic System
zi;t amount of incentive payment to each industrial PVS Photovoltaic System
consumer in period t HSWPS Hybrid Solar-Wind Power System
FCost total expected cost DG Distribution Generation
FEmission total emissions VOLL Value of Lost Load
COC(t) certain operational cost function EENS Expected Energy Not Served
UOC(t) uncertain operational cost function MT Micro-Turbine
Pi(t) output power ith unit in period t WT Wind Turbine
pi ðtÞ offered price ith unit in period t PV Photovoltaic
Ii(t) on and off status of the ith DG in period t FC Fuel Cell
SUi(t) start up or shut down cost of the ith DG in period t PCC Point of Common Coupling
RCDG i ðtÞ reserve costs of the ith DG in period t GPRS General Packet Radio Service
MDMS Meter Data Management System

network operators have always used storage services to manage solution to resolve this uncertainty is felt more than ever. One of
production shortages and to create a balance between production these solutions is the use of Demand Response Programs (DRPs)
and consumption. Today, with the advent of renewable energies, [9,10].
such as wind and solar energy, and the lack of certainty in their Recently, significant studies have been conducted for better
production potential, the need to provide storage and find a implementation of demand side management programs and
624 G.R. Aghajani et al. / Energy 126 (2017) 622e637

modeling their roles in creating a balance between generation and
Use probabilistic modeling of wind, solar, and wind-solar
consumption in the presence of renewable generation, considering powers as a function of output power generation to provide
their stochastic behavior. Demand response programs were used in more compliance between planning and reality.
Ref. [11] to manage the operation of a smart micro-grid with wind
Consider a probabilistic multi-objective model and use the
and solar resources, and Particle Swarm Optimization (PSO) algo- MOPSO method by considering Pareto criterion and fuzzy-based
rithm was applied to solve the proposed model so that the pollution mechanism to solve the intended problem.
emission function was not considered in modeling the micro-grid
management. Multi-objective operation planning in a smart dis- The rest of this paper is organized as follows: after the Intro-
tribution grid with wind and solar resources was evaluated in duction, in Section 2, the problem is described in detail. In Section
Ref. [12] as a probabilistic model to reduce operational costs and 3, the studied smart micro-grid is introduced. In Section 4, the
emissions; Rayleigh and beta Probability Density Functions (PDFs) proposed algorithm is presented based on Pareto criterion. In
were used for modeling variations in the wind speed and solar Section 5, the simulation and analysis of numerical results are
radiation, respectively. In this reference, simultaneous modeling of discussed, and finally, the last section reports the important con-
solar and wind power generation is not considered, and on the clusions of this study.
other hand, the definition of pollution function from elements such
as SO2 and NOx is ignored, and the 3-constraint method is used for 2. Statement of the problem
problem-solving. A similar problem was discussed by Zakariazadeh
et at. in Ref. [13], where a scenario tree approach was used to solve In this study, a probabilistic model is proposed for short-term
the problem; in this study, authors have ignored solar power energy management in order to minimize the operational cost
modeling. and emissions in a smart micro-grid. Due to the stochastic behavior
Online optimal management and modeling of a micro-grid with of wind and solar energies, their accurate prediction is not possible
poly-generation were studied in Ref. [14] the using mesh adaptive and is always associated with uncertainty error in next-day plan-
direct search method in which the uncertainty caused by renew- ning. Therefore, to provide more compliance between planning and
able generation was ignored. In Ref. [15], using demand response reality, a PDF is used to model the behaviors of wind, solar and
programs was proposed for controlling the frequency of a smart hybrid solar-wind power systems in an attempt to obtain optimal
microgrid with renewable generation. A multi-objective function results despite uncertainties. To remove the uncertainty induced by
has become a single objective function, and Mixed Integer Liner these resources, incentive-based payment demand response pro-
Programming (MILP) method is used to solve the proposed model. grams are proposed. However, it is assumed that planning for
The use of demand side management in a smart grid, consid- generation resources and consumption demands in a smart micro-
ering the wind power generation (wind farm) and the resulting grid is performed by the distribution system's operator, which has
uncertainty, was studied by Cicke et al. in Ref. [16] in order to in- the possibility of managing and controlling the grid using distri-
crease social welfare. In this research, the authors don't consider bution management systems and advanced metering infrastruc-
the use of solar power generation and do not use incentive-based ture. The following section is dedicated to the modeling and
demand response programs that can cause consumers the moti- introduction of the objective function.
vation to participate. Using a stochastic planning approach based
on the Monte Carlo method was suggested in Ref. [17] for modeling
the stochastic behavior of wind and Demand Response (DR) 2.1. Proposed model for Wind Energy System (WES)
considering the influence of wind power as an operational storage
in an energy market. Wind turbines convert wind energy into electrical energy.
This paper aims to find the optimal operation of the smart Output power from the wind turbine depends on parameters, such
microgrid with the purpose of minimizing operational costs and as wind availability, wind speed, the wind turbine's power curve,
emissions and considering the concept of DR in smart grids for and shape and size of the turbine. Probabilistic models are devel-
covering the uncertainty caused by wind and solar power genera- oped based on available historical information; since wind speed is
tion and taking into account the stochastic natural behavior. Since a stochastic variable, the meteorological data can be appropriate for
consumer's participation in these programs is considered to be
completely voluntary, an incentive-based demand response 1.4
method is proposed for implementing demand response programs.
PDF CDF
In this method, programs are considered in the form of offering
1.2
packages of price and storage of DR for residential, commercial, and
industrial consumers, where consumers can choose one of the
offered packages and participate in a demand response program 1

depending on their conditions. Rayleigh and beta PDFs are pro-

Probability

posed to model the wind and solar power generation. The proposed 0.8
multi-objective model is solved by using Multi-Objective Particle
Swarm Optimization (MOPSO) method, considering the Pareto 0.6
criterion with nonlinear sorting based on fuzzy mechanism.
In short, the main contributions of this study are:
0.4

The use of DRPs to cover the uncertainty caused by wind and

0.2
solar power generation in a smart microgrid by considering the
objective functions related to the operation costs and pollution
emissions. 0
0 0.5 1 1.5

Propose the use of offered packages of price strategy in order to Normalized Speed
implement demand response programs.
Fig. 1. Wind speed distribution model.
 G.R. Aghajani et al. / Energy 126 (2017) 622e637 625

estimating the wind energy potential of a site. According to the

2.2. Proposed model for photovoltaic system (PVS)
wind speed behavior, Rayleigh distribution is used to model wind
[18]. Rayleigh distribution is a particular form of Weibull distribu-
Photovoltaic generators are systems that convert sunlight into
tion in which the shape index is equal to 2. Therefore, assuming aw
electricity in a way that the solar system output is entirely
as the scale parameter, and bw¼2 as the shape parameter, the
dependent on the amount of sun radiation. Considering the solar
probability density and cumulative distribution functions are as
irradiance behavior, beta PDF and CDF are used to model it ac-
follows:
cording to (8) and (9) [22,23].
 2 ! 8
FV ðvwind Þ ¼ 1  exp 
vwind
(1) < Gða þ bÞ sia1 ð1  siÞb1
>
0  si  1;a  0;b  0
aw f B ðsiÞ ¼ GðaÞGðbÞ
>
:
0 otherwise
 2 !
2 vwind (8)
f V ðvwind Þ ¼ vwind exp  (2)
a2w aw
Zsi
Gða þ bÞ a1
If vm is considered the average wind speed of a specific site, the FB ðsiÞ ¼ si ð1  siÞb1 dsi (9)
scale parameter is given: GðaÞGðbÞ
0
    pffiffiffiffi
1 1 1 p where si indicates the amount of solar irradiance (kW/m2). a and b
vm ¼ a w G 1 þ ¼ aw G ¼ aw
2 2 2 2 are parameters of beta PDF that can be calculated from the mean
(3)
2 value and standard deviation of solar irradiance data and are uti-
aw ¼ pffiffiffiffivm
p lized as follows:
 
Therefore, in the case of substituting aw in Probability Density mð1 þ mÞ
a¼m 1 (10)
Function (PDF) and Cumulative Distribution Function (CDF), the s2
Rayleigh model for Wind Energy System (WES) will be obtained as
a function of average wind speed according to (4) and (5). Fig. 1  
mð1 þ mÞ
demonstrates a sample of probability density and cumulative dis- b ¼ ð1  mÞ 1 (11)
s2
tribution functions.
Now, solar irradiance can be converted into solar power using
pv 2 !
p vwind wind
Eq. (12) [24].
f V ðvwind Þ ¼ exp  (4)
2 v2m 4 v2m
PPV ðsiÞ ¼ Ac $h$si (12)

pv 2 ! where PPV(si) indicates the amount of output power from PV (kW)
wind
FV ðvwind Þ ¼ 1  exp  (5) for the amount of irradiance s; Ac is surface areas of the arrays (m2);
4 vm
and h is efficiency of the PV system.
For a certain WES, the characteristic of the output power can be Therefore, in the case of using Eq. (8), the probability density
defined as below [19]: function fB(PPV) for the output power of PVS is expressed as follows:
8
80 vwind < vci >
> Gða þ bÞ 
>
> < ðAc hsiÞa1 ð1Ac hsiÞb1 if PPV 2½0;PPV ðsiÞ
>
> f PPV ðPPV Þ¼ G ð aÞ G ðb Þ
< PR ðvwind  vci Þ
>
vci  vwind < vr >
>
Pw ðvwind Þ ¼ ðvr  vci Þ (6) : 0 otherwise
>
>
> PR
> vr  vwind < vco
>
: (13)
0 vwind  vco

where vci, vr, vco, and vwind are the cut-in speed, rated speed, cut-
off speed, and actual speed of the wind turbine, respectively, and 2.3. Proposed model for hybrid solar-wind power system (HSWPS)
PR is the rated power of the turbine. The wind turbine used in this
study is of AIR403 type [20], where, PR ¼ 15kW; vci ¼ 3:5m=s; vco ¼ Power generation by hybrid system Ph is equal to total power
18m=s; vr ¼ 17:5m=s . Fig. 2 shows the power curve for this wind output from WES system plus the power output from PVS system.
turbine.
Ph ¼ Pw þPPV (14)
In this paper, the PDF fPw(Pw) for the output power of WES can be
obtained using Eqs. (4) and (6) by the application of the trans- Assuming that PW and PPV are independent in terms of perfor-
formation theorem [21] as: mance in accordance with relations (6) and (12), density function

8
> 1  ½Fv ðvco Þ  Fv ðvci Þ Pw ¼ 0
>
> 2 0
>
> !  12 3
>
< v  v   Pw
p 6 v þ ðv  v Þ
f PW ðPw Þ ¼
r ci
$  vci þ ðvr  vci Þ$
Pw
 exp4  @
ci r c PR
A 75 0 < Pw < PR (7)
>
> PR 2v 2 P p2ffiffiffivm
>
> m R
p
>
>
:
Fv ðvco Þ  Fv ðvr Þ Pw ¼ PR
626 G.R. Aghajani et al. / Energy 126 (2017) 622e637

16

14

Turbine Output Power(kw) 12

10

0
2 4 6 8 10 12 14 16 18 20 22 24
Wind Speed(m/s)
Fig. 2. Wind turbine model AIR403 power curve.

for random variable Ph as convolution between density functions

PW and PPV is defined as follows [25]: IPði; tÞ ¼ ICði; tÞ$zi;t ; ICði; tÞ  ICmax
t (18)

f h ðPh Þ ¼ f PW ðPw Þf PPV ðPPV Þ (15) where r, c, and i represent the number of residential, commercial,
and industrial consumers; RC(r,t), CC(c,t), and IC(i,t) indicate the
Since representing the continuous PDF in the mathematical
amount of load reduction planned by each residential, commercial,
form seems to be difficult, the application of Monte Carlo simula-
and industrial consumer in period t; RCtmax , CCtmax , and ICtmax
tion is used in such cases to achieve different scenarios; yet,
indicate the maximum load reduction proposed by each consumer
generating different scenarios also adds to the mathematical
in period t; zr;t , zc;t , and zi;t show the amount of incentive payment
complexities of the problem. The appropriate strategy for pre-
venting mathematical complexities is to discretize the continuous
PDF by dividing it into different intervals. Depending on the desired
accuracy, PDF function can be divided into a number of discrete
instants with different possible levels. In this case, the surface
enclosed by each interval represents the median probability of that
interval. Therefore, in this paper, the probability density function
proposed for each wind and solar system is divided into seven
ranges per hour in order to supply the requested power. As a result,
any wind and solar systems have seven scenarios to participate in
production planning every hour.

2.4. Model of demand response programs

In the present study, electricity consumers are considered to be

residential, commercial, and industrial, and the following equations
demonstrate the modeling for their behavior. Constraints show that
sum of energy reduction by each consumer at every hour should be
lower or equal to the maximum amount of its offers.

RPðr; tÞ ¼ RCðr; tÞ$zr;t ; RCðr; tÞ  RCmax

t (16)

CPðc; tÞ ¼ CCðc; tÞ$zc;t ; CCðc; tÞ  CCmax

t (17)

Fig. 3. Classification of DRPs.

 G.R. Aghajani et al. / Energy 126 (2017) 622e637 627

to each consumer in period t; and RP(r,t), CP(c,t), and IP(i,t) repre- probability of scenario Prs during the t th period and sth scenario,
sent the cost due to load reduction by residential, commercial, and which are affected by the probabilistic amounts of wind and solar
industrial consumers in period t for the proposed load reduction, parameters in each scenario. This part of the operational cost
respectively. function includes the running cost of distributed generation units,
cost of load reduction due to demand response programs, and costs
2.5. Objective functions associated with Value of Lost Load (VOLL) and Expected Energy Not
Served (EENS) for consumers.
Demand response can be categorized based on the consumer's
participation in changing their consumption behavior into two X
T X
T X
T X
S
groups of price-based and incentive-based demand response pro- Min f 1 ðXÞ ¼ FCost ðtÞ ¼ COCðtÞ þ Prs  UOCs ðtÞ
grams (Fig. 3), and each of these groups is divided into several sub- t¼1 t¼1 t¼1 s¼1
groups. More detail is presented in Refs. [9,10]. (19)
Since incentive-based responsibility programs deal with price
signals and are considered voluntary, modeling it on the basis of where Prs is the probability of scenario s. Certain and uncertain
offering packages of price, given the reduced amount of demand, is operational cost functions are defined according to (20) and (21),
represented as Fig. 4 by considering the demand response pro- respectively.
grams based on incentive payment.

X
NDG i
COCðtÞ ¼ ½Pi ðtÞpi ðtÞIi ðtÞþSUi ðtÞjIi ðtÞIi ðt  1ÞjþRCDG
i ðtÞ
i¼1
(20)
X
J
þ j ðtÞIBuy ðtÞPGridBuy ðtÞpGridBuy ðtÞISell ðtÞPGridSell ðtÞpGridSell ðtÞ
RCDR
j¼1

In this study, a multi-objective stochastic programming model

X
NDG X
J
will consider operation costs and pollution emissions in the pres- UOCs ðtÞ ¼ CDG CDR
i;s ðtÞ þ j;s ðtÞþENSs ðtÞ  VOLLðtÞ (21)
ence of distributed generation resources and wind and solar sour- i¼1 j¼1
ces. Moreover, DRPs based on incentive-based payment will be
used to remove the uncertainty caused by the random behavior of where Pi(t) and pi(t) indicate the amount of output power and
wind and solar resources in a 24-h planning period. In these pro- offered price for the ith unit during the t th period; binary Ii(t)
grams, demands are considered as responsive residential, com- represents the on and off mode of the ith DG during the t th period;
mercial, and industrial demand. Fig. 5 shows a clear picture of the SUi(t) shows running and shutting down costs for the ith unit
optimization model. during the t th period; RCiDG ðtÞ and RCjDR ðtÞ are reserve costs of the
ith DG and demand response programs for the j th load during the t
2.5.1. Operational cost function th period; PGrid-Buy(t) and PGrid-sell(t) represent the amount of ex-
Here, the operation function is divided into two parts: certain change power with utility in period t; pGrid-Buy(t) and pGrid-sell(t)
operational costs - including the fixed running and start-up costs of indicate the offered price for exchange power with utility in the
Distribution Generations (DGs), spinning and non-spinning reserve open electricity market during the t th period; Ci;s DG ðtÞ and C DR ðtÞ
j;s
costs provided by DGs, as well as demand response programs - and show the running cost of the ith DG unit and the cost due to load
costs of power, which is bought/sold from/to the utility; and un- reduction provided by the j th DRPs during the t th period in the sth
certain operational costs - by considering and realizing the scenario; and ENSs(t) and VOLL(t) are the Expected Energy Not
Served (EENS) in the sth scenario at t th period and Value of Lost
Load (VOLL) at t th period, respectively. In Eq. (19),
X T ¼ ½X1 ; X2 ; :::; XT  is the variables state vector that includes active
power produced by each DG, charge and discharge power of the
battery, and the active power exchanging with the upstream grid.
Regarding the operational costs in the proposed model, it is
assumed that DGs, along with the DRs, are the spinning reserve
providers for compensating the uncertainty caused by wind and
solar, renewable generations. Therefore, the DG that is considered
in the off mode at its ground state is turned on to provide reserve
power; thus, fixed and start-up costs of DGs with the possibility
equal to unity are included in the first part of the operational cost
function. In other words, during the occurrence of scenario in on/off
real-time modes, the DG remained constant during the planning,
and output power was only carried out according to the pro-
grammed values for the day ahead. Hence, in the real-time occur-
rence of each scenario, the reverse payment would not change and,
as a result, the reserve cost is considered with the possibility of one
Fig. 4. DRP's offer package. in the first parts of the objective function.
628 G.R. Aghajani et al. / Energy 126 (2017) 622e637

Optimization Model

Stochastic model:
Customers: WECS and PVS
Industrial, Commercial, Residential
Participants Objective Function Operating Cost:
Sources: MOPSO Cost function/constraints
DG,MT,FC,PV,WT,Batt,Grid
Environmental Cost and Emission

Output Power:

DG, MT, FC, PV, WT, batt, Grid, Customers

Fig. 5. Optimization model structure.

2.5.2. Pollution emissions function 2.6.2. Reserve and DG power constraint

Pollution emissions function includes functions such as the Maximum and minimum power generations by each unit are
amount of pollution caused by DG units and the amount of pollu- constrained and can be expressed as follows:
tion caused by the grid at the time of purchase. The pollutants
include CO2, SO2, and NOx, and the mathematical model of pollution Pmin max
DG;i :Iði; tÞ  PDG ði; t; sÞ  PDG;i $Iði; tÞ ci; t; s (27)
emissions function can be obtained as follows:
RDG ði; tÞ  PDG ði; t; sÞ  PDG ði; t; 0Þ c i; t; s (28)
X
T X
T 
Min f 2 ðXÞ ¼ FEmission ðtÞ ¼ ½EmiDG ðtÞ þ EmiGrid ðtÞ
t¼1 t¼1
(22) 2.6.3. Battery constraints
The average pollution caused by renewable DG units can be Since there are limitations of charging and discharging in stor-
calculated as follows: age devices during each time interval, the following limitations and
equations can be expressed for a battery type:
NDG 
X 
EmiDG ðtÞ ¼ EDG DG DG DG
CO2 ðiÞ þ ESO2 ðiÞ þ ENOx ðtÞ  Pi ðtÞ (23) Wess ðtÞ ¼ Wess ðt  1Þ þ hcharge ðtÞPcharge ðtÞDt$Icharge
i¼1 1
 Pdischarge ðtÞDt$Idischarge Idischarge ðtÞ þ Icharge ðtÞ
where DG ðiÞ,
ECO DG ðiÞ,
ESO
and DG ðiÞ
ENO
indicate the amount of CO2, SO2,
hdischarge
2 2 x
and NOx, pollution caused by the ith DG, respectively, that kg/MWh  1Wess;min  Wess ðtÞ  Wess;max Pcharge ðtÞ
is its measurement unit. Similarly, pollution caused by the grid at
 Pcharge;max ; Pdischarge ðtÞ  Pdischarge;max
the time of energy purchase can be written as follows:
(29)

EmiGrid ðtÞ ¼ EGrid Grid Grid
CO2 þ ESO2 þ ENOx Þ  PGrid ðtÞ (24)
where, Wess(t) and Wess (t-1) separately; shows the amount of
energy stored within the battery at t and t-1 time; Pcharge (Pdischarge)
is the amount of allowed charge/discharge during a defined period
of time (Dt), hcharge (hdischarge) is battery efficiency during charging/
2.6. Problem constraints discharging; Wess,min and Wess,max are lowest and the highest
amounts of energy stored in the battery; and Pchrge,max (Pdischarge,-
The typical smart microgrid is assumed to operate with the max) is maximum battery charge/discharge during each time in-
following constraints. terval Dt.

2.6.1. Power balance constraint 2.7. Typical smart microgrid system

The total power generated by DGs purchased from the utility;
and load reduction caused by demand response programs in each A microgrid usually includes a set of DGs, energy reserves, and
interval and scenario must be equal to the total demand loads. load systems that can be operated independently or in conjunction
with the area's main electrical grid [26,27]. Development of micro-
X
NDG X
Ns grids is a part of a smart grid concept; due to the advantages of
PDG;i;s ðtÞ þ PGrid;s ðtÞ ¼ PDemandl;s ðtÞ  PDR;s ðtÞ (25) micro-grids, such as reduced energy costs and improved reliability
i¼1 l¼1 and system security, it is obvious that there are common objectives
among micro-grids and smart grids [28]. Also, advantages such as
where PDemand l,s is the amount of L th demand level during the t th green technology development and the use of demand response
period and sth scenario. In Eq. (26), PDR,s(t) is the amount of active programs in micro-grids depend on the use of smart grid tech-
power participation in demand response programs and can be nologies. As observed in Fig. 6, the smart micro-grid studied here
described as follows: has three types of consumers: residential, commercial, and indus-
X X X trial, along with power generation resources such as Micro-Turbine
PDR;s ðtÞ ¼ RCðr; t; sÞ þ CCðc; t; sÞ þ ICði; t; sÞ (26) (MT), Wind Turbine (WT), Photovoltaic (PV) cell, Fuel Cell (FC), and
r c i
battery and diesel generators; therefore, this grid has the capacity
to exchange energy with the utility. The installation characteristics
 G.R. Aghajani et al. / Energy 126 (2017) 622e637 629

Fig. 6. Typical smart micro-grid system.

Table 1
Bids and emissions coefficient of the DG sources.

Unit Type Bid (Vct/kWh) Start-up/Shut-down Cost (Vct) CO2 (kg/MWh) SO2 (kg/MWh) NOx (kg/MWh) Pmin ðkWÞ Pmax ðkWÞ

1 Diesel 0.586 0.15 890 0.0045 0.23 30 300

2 MT 0.457 0.96 720 0.0036 0.1 6 30
3 FC 0.294 1.65 460 0.003 0.0075 3 30
4 PV 2.584 0 0 0 0 0 25
5 WT 1.073 0 0 0 0 0 15
6 Bat 0.38 0 10 0.0002 0.001 30 30
7 Grid e 0 950 0.5 2.1 30 30

are presented in Table 1, which includes DG price offers, cost of In this study, the concept of Pareto optimization is applied to the
starting up and shutting down of units, amount of greenhouse gas basic principles of PSO algorithm [30] (developing the algorithm of
emissions caused by DGs and utility, as well as minimum and Multi-Objective Particle Swarm Optimization (MOPSO) [31]),
maximum power generations [29]. simultaneous minimization of operational costs and emission
functions with renewable generation, and DG and DR which are
carried out. Application of the algorithm to the problem considered
3. MOPSO algorithm in this study can be accomplished with the following steps.
Step 1: Defining the input data.
Since multi-objective optimization problems include multiple These data are related to smart micro-grid technical specifica-
conflicting objective functions, equality constraints and inequalities tions and include production capacity, proposed power price, and
must be optimized simultaneously. the operational and emission costs of DGs. In addition, the data
include the mean values and variance of wind speed and solar
Min FðXÞ ¼  ½ f1 ðXÞ f2 ðXÞ … fn ðXÞ T irradiance on the next day, and the requested demand from the
gi ðXÞ < 0 i ¼ 1; 2; :::; Nueq (30)
Subject to : daily load curve.
hi ðXÞ ¼ 0 i ¼ 1; 2; :::; Neq
Step 2: Obtaining the amount of wind and solar power from
proposed equations.
where F(X) is a vector containing objective functions and X is a
Step 3: Generating an initial population as XT¼[X1, X2, …, XT].
vector containing optimization variables, fi(X) is the objective
Step 4: Applying a power dispatch algorithm to the generated
function ith; gi(X) and hi(X) are constraints of equality and
population and calculation of fitness function according to Eq. (19)
inequality; and n is the number of objective functions. For a multi-
or (22).
objective optimization problem either X or Y solutions can be one of
Step 5: Identifying non-dominated solutions.
the two possible solutions. One will dominate another, or none is
Step 6: Separating non-dominated solutions and storing them in
dominated by any of the other solutions. Therefore, in an optimi-
an archive.
zation problem, one solution X will dominate Y if the following two
Step 7: Selecting the best particle from the non-dominated
conditions are met:
response archive as a leader.
The process of selecting a leader is as follows:
c j2f1; 2; :::; ng; f j ðXÞ  f j ðYÞ
(31) The explored search space is divided into equal parts and a
d k2f1; 2; :::; ng; f k ðXÞ < f k ðYÞ
probability distribution is distributed to each part. Finally, the best
Therefore, the Pareto set solutions can be obtained through non- particle is selected as the leader using the roulette wheel method.
dominated solutions (desired answers) on search space. Finally, the Step 8: Updating the new velocity and position for each particle.
answer is obtained among non-dominated solutions stored in the Step 9: Updating the best position for each particle.
archive.
630 G.R. Aghajani et al. / Energy 126 (2017) 622e637

Fig. 7. Flowchart of MOPSO algorithm.

For updating the best position, the particle's new position is calculated using optimization results for each objective function. mki
compared with its previous position. is in the range of 0e1 such that mki ¼ 0 indicates incompatibility of
8 the solution with the objectives of designer, while mki ¼ 1 repre-
> Pbest;i ðtÞ Pbest;i ðtÞ3Xi ðt þ 1Þ sents full compatibility.
>
<
Xi ðt þ 1Þ Xi ðt þ 1Þ3Pbest;i ðtÞ Fig. 7 shows the flowchart of the proposed algorithm used for
Pbest;i ðt þ 1Þ ¼ (32)
> select randomly
> solving the optimization problem.
:

Pbest;i ðtÞorXi ðt þ 1Þ otherwise

4. Simulation and analysis of numerical results
Step 10: Adding the current non-dominated solutions to the
archive.
The smart microgrid connected to the utility in Fig. 6 has three
Step 11: Removing the dominated solutions from the archive.
residential, commercial, and industrial consumers whose load de-
Step 12: If the number of members in the archive exceeds the
mand from the system and daily load curve are shown in Fig. 8 [32].
determined capacity, extra members would be removed.
It is considered a test system, so that the total energy consumption
Step 13: Evaluating the criterion for program termination.
during the day is equal to 4034 kWh. The real-time market price of
If the maximum number of repetitions is established, the opti-
APX is shown in Fig. 9 [33].
mization process would stop; otherwise, the current population
The spinning reserve cost due to DGs is considered to be 20% of
would replace the previous population, and the algorithm would
the highest marginal cost of energy generation [34]. Hourly wind
return to Step 7.
speed data, which are taken from the weather forecast website
Step 14: Selecting the best interactive solution.
(Willy Online Ply Ltd.) [35], are shown in Fig. 10. The solar cell
In order to choose a better solution from among the obtained
considered here is a 25 kW SOLAREX MSX, composed of solar
optimal responses, the fuzzy decision-making function with a
panels of 10✕2.5 kW with h¼18.6% and s¼10m2 [36]; Fig. 11 shows
membership function is considered, in which the exact number of
the average hourly solar irradiance [37].
variables can be located; where mki represents the optimality
It is assumed that the power coefficient of the wind turbine and
amount of the objective function i in among Pareto optimal
PV system is equal to 1, while other DGs and loads locally
response k, which is calculated as follows:
compensate for the required reactive power through capacitor
8 placement in the related buses. The value of lost load is considered
>
> 1 f i  f min
>
>
i to be equal to 1.33 V/kWh [38]. In a typical system, including a
>
< f max f battery with a capacity of 30kWh, minimum and maximum charges
mki ¼ i i
f max
i < f i < f min
i (33)
> f max f min
> are considered to be 10% and 100% of the battery capacity,
>
> i i
>
: respectively, with a charge and discharge efficiency of 94% [39,40].
0 f i  f max
i The offered packages provided for demand response programs are
shown in Table 2. For their implementation, it is assumed that 40%
where fimax and fimin are the upper and lower limits of the objective of consumers participate in demand response programs [41].
function i, respectively. In the proposed method, these values are
 G.R. Aghajani et al. / Energy 126 (2017) 622e637 631

120
Residential Industrial Commercial

100

Demand(% of Maximum) 80

60

40

20

0
2 4 6 8 10 12 14 16 18 20 22 24
Time(h)
Fig. 8. Daily load curves for different consumers.

To evaluate the effect of planning for energy level, reserve, and

5 DR in the operational costs and pollution emissions function, and to
4.5
resolve the uncertainty caused by wind and solar resources, the
problem is considered in three different conditions:
Real Time Market Price(Ect/kWh)

4

Case 1: Considering operational cost and emission functions
3.5
without demand response
3
Case 2: Considering operational cost and emission functions
with demand response
2.5
Case 3: Simultaneously considering multi-objective functions of
2
operational costs and emissions

1.5 In all the cases, power generation units are supposed to have the
capability of participating in the smart microgrid depending on
1
their technical and economic characteristics and exchange of en-
0.5 ergy with the utility through a Point of Common Coupling (PCC) in
the case of excessive generation and demand. In order to evaluate
0 the effects of the proposed model, it has been implemented in
0 5 10 15 20 25
Time(h) MATLAB software on a PC (2.6 MHz with 4 GB of RAM).
Case 1: Operational cost and emission functions without demand
Fig. 9. The real-time market prices of APX. response

12

10
Wind Speed(m/s)

2 4 6 8 10 12 14 16 18 20 22 24
Time(h)
Fig. 10. Hourly wind speed forecast.
632 G.R. Aghajani et al. / Energy 126 (2017) 622e637

when energy prices are high, the utility purchases energy from the
1 smart microgrid in which the power consumption is provided by
DGs with the priority of offered price. The results of Table 4 show
that, in most operational periods, due to the high pollution of
utility, the utility purchases power from the smart microgrid in
solarirradiance(kW/m )

0.8
2

most of the periods. Results in Fig. 12 indicate that, since wind and
solar power are devoid of any pollution most of the time, these
0.6 resources reach their maximum power generation by considering
the pollution emission function.
However, since the offered price of these resources is higher
0.4 than that of other power generation resources, they cannot receive
much attention when considering the optimal operational cost.
Results reported in Ref. [29] verify this conclusion.
0.2 Case 2: Operational cost and emissions functions with demand
response
In this step, operational costs and emissions are separately
0 minimized with the involvement of DR. The optimal allocation of
0 5 10 15 20 25
power generation of units for minimizing operational costs and
Time(h)
emissions is shown in Tables 5 and 6, respectively.
Fig. 11. Hourly solar irradiance forecast. Comparison of the results presented in Tables 3 and 5 show that,
in the case of using demand response programs, wind power
generation is reduced from 8.51 kW to 7.87 kW, and solar power
Table 2 generation is reduced from 4.82 kW to 3.34 kW. On the other hand,
Price-quantity offer package for DRPs.
comparison of results obtained from the optimization of emission
Quantity (kW) function with/without DR indicates that use of these programs
Price (Vct/kWh) reduces the wind and solar power generations from 50.61 kW to
47.39 kW, and from 91.39 kW to 89.50 kW, respectively. Fig. 13 shows
DRP1 0e5 5e10 10e50 50e70
0.06 0.13 0.26 0.36 the amount of power generation by wind turbine and solar cell
DRP2 0e5 5e20 20e30 30e60 considering the operational cost and pollution emissions with the
0.04 0.07 0.28 0.43 involvement of demand response.
According to Fig. 14, it can be said that, considering the pollution
emissions, the use of DRPs reduces the production capacity of wind
In this case, the operational costs and emission are separately turbine and solar cell, and also shifts the demand from peak periods
minimized without considering the DR. The optimal allocation of to off-peak periods. In this case, when the customers participate in
the power generation of the units for minimizing the operational the DR program and accept a reduction of their consumption at a
costs and emissions is shown in Tables 3 and 4, respectively. The specific hour, it allows the system operator to reduce the scheduled
results of Table 3 suggest that in the early hours, when the price of power of generating units.
energy is low, the battery starts to be charged, and from 9 to 16, Case 3: Simultaneous consideration of multi-objective operational

Table 3
Energy resources scheduling for operation cost objective function without DR.

Hour units

DG (kW) MT (kW) FC(kW) WT (kW) PV(kW) Batt (kW) Utility (kW)

1 30.000 7.9514 7.283 0.178 0.000 11.587 30.000

2 32.287 12.912 27.010 0.178 0.000 19.688 26.299
3 45.130 7.189 20.433 0.069 0.000 19.786 21.963
4 38.075 6.000 24.565 0.000 0.000 30.000 29.359
5 30.000 8.835 13.957 0.403 0.000 14.207 26.011
6 37.393 8.068 23.610 0.091 0.000 21.981 23.818
7 30.000 12.489 19.287 0.016 0.000 0.705 22.912
8 33.650 9.813 26.953 0.063 0.075 26.505 22.939
9 104.789 28.719 21.201 0.178 0.112 30.000 30.000
10 234.763 6.000 3.213 0.000 0.000 15.975 30.000
11 211.313 8.245 19.145 0.000 0.851 15.444 30.000
12 283.868 6.000 5.132 0.000 0.000 30.000 30.000
13 272.073 6.000 8.400 0.054 1.695 28.255 29.966
14 218.992 13.747 24.529 0.473 0.342 21.758 29.843
15 213.897 14.123 28.932 0.714 0.859 27.265 29.791
16 226.268 13.190 6.804 0.300 0.338 24.099 30.000
17 114.225 29.999 29.974 0.000 0.550 30.000 25.252
18 95.459 27.799 30.000 1.741 0.000 30.000 30.000
19 105.819 29.999 30.000 1.302 0.000 30.000 29.879
20 122.196 14.555 29.017 0.000 0.000 26.876 27.355
21 155.728 28.322 28.442 1.300 0.000 30.000 23.792
22 79.629 28.946 26.685 0.555 0.000 29.860 29.323
23 54.699 29.097 25.325 0.717 0.000 25.161 30.000
24 31.509 10.515 26.946 0.172 0.000 28.682 25.174
 G.R. Aghajani et al. / Energy 126 (2017) 622e637 633

Table 4
Energy resources scheduling for emission objective function without DR.

Hour units

DG (kW) MT (kW) FC(kW) WT (kW) PV(kW) Batt (kW) Utility (kW)

1 37.654 17.999 29.918 1.428 0.000 30.000 30.000

2 30.000 23.862 24.679 0.426 0.000 27.435 27.403
3 30.000 13.837 26.968 1.428 0.000 30.000 27.233
4 30.105 6.678 30.000 0.801 0.000 30.000 29.584
5 30.000 6.000 27.317 1.785 0.000 29.897 30.000
6 30.000 8.457 30.000 0.4146 0.000 30.000 27.871
7 32.727 16.279 30.000 1.501 0.000 30.000 26.508
8 50.717 30.000 29.678 1.305 0.197 29.731 21.628
9 59.526 30.000 30.000 1.630 3.727 30.000 0.116
10 111.604 29.726 28.456 1.809 7.525 29.507 10.627
11 125.775 30.000 30.000 8.775 10.449 29.999 20.000
12 152.640 30.000 29.999 10.410 11.950 29.999 30.000
13 142.210 30.000 29.988 3.915 23.899 29.984 29.997
14 164.908 30.000 30.000 2.345 21.050 30.000 28.304
15 175.367 30.000 30.000 1.780 7.875 30.000 19.027
16 161.097 29.987 29.019 1.284 4.225 30.000 14.607
17 158.091 29.869 27.953 1.593 0.493 30.000 18.000
18 153.215 30.000 30.000 1.785 0.000 30.000 30.000
19 115.726 30.000 28.812 1.275 0.000 30.000 21.187
20 108.116 29.540 29.986 1.785 0.000 29.469 21.103
21 98.827 30.000 30.000 1.300 0.000 29.873 30.000
22 73.699 30.000 30.000 1.300 0.000 30.000 29.999
23 44.515 29.573 29.999 0.915 0.000 30.000 29.997
24 38.253 29.227 30.000 0.615 0.000 30.000 5.096

Forecast 60 Forecast
25
WT Powr for Operation Cost without DR PV Powr for Operation Cost without DR
WT Power for Emission Cost without DR PV Power for Emission Cost without DR
50
20
WT Power(kW)

PV Power(kW)

40
15

30

10
20

5
10

0 0
0 5 10 15 20 25 0 5 10 15 20 25
Time(h) Time(h)
(a) (b)
Fig. 12. Output power (a) Wind turbine (b) Solar cell, considering operating cost and emission without DR.

and emission cost functions simultaneous minimization of these two functions when demand
In case 3, the optimal power allocation of the units is carried out response programs are implemented.
for the simultaneous minimization of operational costs and emis- The results of Fig. 16 show that maximum wind and solar power
sions as two inconsistent functions with/without the involvement generation is related to the case in which pollution emission is
of DR. According to Fig. 15, since the objectives of operational cost taken into account; therefore, it is possible to establish a balance
and emissions costs are opposite, moving from initial points on between them by simultaneous optimization.
curves toward the endpoints along the Pareto path is equal to the Results are shown in Table 7 for a better comparison of output
change in the operation behavior from low cost and more pollution wind power and solar cell power from the perspective of operation
to higher cost and low pollution, where the optimal operation point costs and emissions with/without the presence of demand re-
can be determined by fuzzy mechanisms. sponses. Results show that among the proposed scenarios, the state
The results of Fig. 15 indicate that, in the case of using demand of considering the operation cost is the best state to resolve the
response programs, it is possible to improve the optimal operation uncertainty resulted from solar and wind resources.
point such that the operational cost and pollution emissions are By comparing the results of simulations calculation of requested
reduced by 21% and 14%, respectively. Fig. 16 shows the amount of power in different cases and the amount of requested energy (4034
wind and solar power generation with the minimization of oper- kWh), it can be observed that the amount of energy not supplied is
ational cost function and emissions function, as well as with the negligible and does not have much impact on the results.
634 G.R. Aghajani et al. / Energy 126 (2017) 622e637

Table 5
Energy resources scheduling for operation cost objective function with DR.

Hour units

DG (kW) MT (kW) FC(kW) WT (kW) PV(kW) Batt (kW) Utility (kW)

1 37.694 12.202 8.359 0.243 0.000 0.473 23.029

2 36.791 12.318 13.077 0.000 0.000 16.677 23.490
3 32.129 15.173 13.239 0.283 0.000 19.486 23.662
4 31.252 9.991 21.058 0.178 0.000 29.479 30.000
5 30.178 7.711 3.304 0.000 0.000 29.998 3.804
6 32.588 6.000 7.898 0.046 0.000 3.391 17.859
7 36.320 7.426 25.822 0.459 0.000 20.765 29.736
8 37.811 13.354 29.065 0.030 0.003 26.825 7.911
9 82.384 10.887 22.688 0.178 0.931 27.931 30.000
10 163.871 22.707 17.409 0.000 0.672 18.340 30.000
11 239.305 6.475 5.723 0.000 0.000 1.503 30.000
12 205.704 8.094 15.044 0.048 0.377 4.268 29.999
13 223.543 6.000 27.248 0.005 0.279 7.202 29.875
14 267.051 6.000 5.700 0.000 0.000 3.751 30.000
15 217.144 25.307 27.230 0.000 0.000 11.251 29.932
16 175.002 12.794 26.700 0.131 0.531 5.823 29.981
17 128.741 27.860 29.831 1.785 0.550 29.066 7.166
18 47.588 30.000 29.854 0.000 0.000 27.804 29.753
19 55.726 29.981 30.000 1.302 0.000 29.990 30.000
20 91.621 29.253 28.888 1.747 0.000 28.817 29.674
21 131.616 25.949 24.737 0.769 0.000 26.608 29.679
22 101.889 16.946 19.488 0.238 0.000 25.349 26.089
23 30.000 14.163 27.543 0.183 0.000 23.249 29.862
24 31.719 15.337 30.000 0.246 0.000 22.667 28.365

Table 6
Energy resources scheduling for emission objective function with DR.

Hour units

DG (kW) MT (kW) FC(kW) WT (kW) PV(kW) Batt (kW) Utility (kW)

1 30.000 6.000 11.962 0.178 0.000 28.859 30.000

2 30.704 10.370 25.188 0.107 0.000 29.056 26.425
3 30.000 11.401 27.743 0.000 0.000 30.000 29.143
4 30.000 6.000 3.000 0.000 0.000 19.000 30.000
5 30.000 6.000 3.000 8.58E-19 0.000 6.000 30.000
6 30.013 6.010 3.051 0.005 0.000 21.862 29.941
7 32.110 9.840 29.792 1.205 0.000 30.000 28.956
8 30.000 10.442 23.967 0.000 0.161 30.000 24.570
9 35.074 30.000 29.692 1.606 3.249 30.000 15.378
10 39.394 29.719 28.609 3.037 7.525 29.878 19.836
11 80.775 30.000 30.000 8.775 10.450 30.000 30.000
12 106.976 29.999 29.791 10.409 11.335 29.994 11.493
13 42.214 30.000 29.986 3.915 23.900 29.999 29.985
14 129.915 29.997 30.000 2.199 21.050 29.969 1.868
15 126.938 29.918 30.000 1.781 7.127 29.991 20.243
16 71.943 30.000 30.000 1.276 4.158 28.990 24.633
17 57.665 30.000 29.999 1.785 0.550 29.999 30.000
18 67.661 30.000 30.000 1.649 0.000 30.000 5.689
19 102.590 30.000 30.000 1.221 0.000 30.000 23.188
20 96.809 30.000 29.999 1.770 0.000 29.582 18.161
21 58.733 29.999 30.000 1.294 0.000 30.000 29.974
22 36.309 30.000 30.000 1.300 0.000 30.000 17.390
23 51.233 29.999 29.998 0.757 0.000 30.000 18.013
24 33.639 30.000 30.000 0.123 0.000 29.578 10.341

Results are given in Table 7 for a better comparison of solar and of an optimization function with two inconsistent objectives. The
wind generation capacities from the perspective of operating costs total operational cost of the microgrid, and pollution caused by
and emissions with/without DR. We see that among the proposed pollutants were considered in three different conditions. Moreover,
modes, taking into account operating costs is the best case to a probabilistic programming method was used to model the sto-
resolve the uncertainty derived from wind and solar resources. chastic behavior of the wind and solar cells' power generation. For
better performance of the smart microgrid, the possibility of energy
exchange with the utility was assumed.
5. Conclusion In order to manage consumption, it is assumed that consumers
can participate in DRPs based on incentive payment. The proposed
In this study, a probabilistic programming was implemented for price package and the amount of demand reduction are used to run
the smart microgrid, by considering the DR as the compensation for these programs.
uncertainty caused by wind and solar power generation in the form
 G.R. Aghajani et al. / Energy 126 (2017) 622e637 635

Forecast 60 Forecast
25 PV Powr for Operation Cost with DR
WT Powr for Operation Cost with DR
WT Power for Emission Cost with DR PV Power for Emission Cost with DR
50
20
WT Power(kW)

PV Power(kW)
40

15
30

10
20

5
10

0 0
0 5 10 15 20 25 0 5 10 15 20 25
Time(h) Time(h)

(a) (b)
Fig. 13. Output power (a) Wind turbine (b) Solar cell, considering operating cost and emission with DR.

300 The MOPSO method, based on fuzzy techniques, was used to

BeforeDR AfterDR solve the proposed model and achieve an optimal response.
The results of the simulations showed that if consumers
250
participate in DRPs, there is a possibility for reducing the opera-
tional costs and emissions. Among the studied cases, simultaneous
consideration of operational costs and pollution emissions, with
200
the involvement of DR, produced the best results by reducing
operational cost and pollution emissions by 21% and 14%, respec-
150 tively. In addition, results of the simulation showed that consid-
ering the pollution function as the main objective, it increased the
operational costs such that maximum use of demand response
100 programs occurred in this case, compared with the operating
condition. Another important result is providing a model with a
simple structure. As this model shows, if consumers participate in
50
demand response, in addition to covering the production shortages
caused by the uncertainty derived from wind and solar power, this
0
will result in a reduction of operating costs and the system's overall
0 5 10 15 20 25 pollution.
Time(h)

Fig. 14. Load demand before and after DR implementation.

X: 1585
2080 Y: 2070
2800
X: 1759
Y: 2751 2060
2750

Min Operating Cost 2040 Min Operating Cost

2700
2020
Emission(Kg)

2650 2000

2600 1980
Best Solution With DR
Best Solution Without DR 1960
2550

Min Emission 1940 Min Emission

2500
X: 1638
1920 Y: 1937
X: 2119
2450 X: 1911 Y: 2420 X: 1782
Y: 2465 1900 Y: 1887
2400
1750 1800 1850 1900 1950 2000 2050 2100 2150 1880
1550 1600 1650 1700 1750 1800
Operating Cost(Ect)
Operating Cost(Ect)
(a) (b)
Fig. 15. Pareto criterion for operating costs and pollution (a) without DR (b) with DR.
636 G.R. Aghajani et al. / Energy 126 (2017) 622e637

70
Cost Optimization
Cost Optimization
20 Emission optimization 60 Emission optimization
Multi-Objective Optimization
Multi-Objective Optimization
50
15
WT Power(kW)

PV Power(kW)
40

10 30

20
5

10

0 0
0 5 10 15 20 25 0 5 10 15 20 25
Time(h) Time(h)
(a) (b)
Fig. 16. Power generation (a) Wind turbine (b) Solar cell.

Table 7
Comparison of solar and wind generation capacity from the perspective of operating costs and emissions without/with DR.

Cases Wind power Solar power Wind power forecast Solar power forecast Cover percentage of wind Cover percentage of solar
(kW) (kW) (kW) (kW) power power

Operating costs without DR 8.51 4.82 57.15 91.47

Operating costs with DR 7.87 3.34 57.15 91.47 7.5% 30.7%
Emissions without DR 50.605 91.39 57.15 91.47
Emissions with DR 47.39 89.50 57.15 91.47 6.3% 2.1%
Simultaneous optimization 23.12 90.21 57.15 91.47
without DR
Simultaneous optimization 21.97 78.27 57.15 91.47 4.9% 13.2%
without DR

Operating costs with DR are written in bold.

References programming approach. Energy Convers Manag 2014;78:151e64.

[14] Mohamed FA, Koivo HN. System modelling and online optimal management
of microgrid using mesh adaptive direct search. Int J Electr Power & Energy
[1] Sun Q, Ge X, Liu L, Xu X, Zhang Y, Niu R, et al. Review of smart grid
Syst 2010;32(5):398e407.
comprehensive assessment systems. Energy Procedia 2011;12:219e29.
[15] Rezaei N, Kalantar M. Smart microgrid hierarchical frequency control ancillary
[2] Mazlumi K. Presenting a new method based on branch placement for optimal
service provision based on virtual inertia concept: an integrated demand
placement of phasor measurement units. J Operation Automation Power Eng
response and droop controlled distributed generation framework. Energy
2014;2(2):113e20.
Convers Manag 2015;92:287e301.
[3] Colak I, Kabalci E, Fulli G, Lazarou S. A survey on the contributions of power
[16] Cicek N, Delic H. Demand response management for smart grids with wind
electronics to smart grid systems. Renew Sustain Energy Rev 2015;47:
power. IEEE transactions on. Sustain Energy 2015;6(2):625e34.
562e79.
pez G, Moreno J, Amarís H, Salazar F. Paving the road toward Smart Grids [17] Afshar K, Gazafroudi AS. Application of stochastic programming to determine
[4] Lo
operating reserves with considering wind and load uncertainties. J Operation
through large-scale advanced metering infrastructures. Electr Power Syst Res
Automation Power Eng 2013;1(2):96e109.
2015;120:194e205.
[18] Rahimi E, Rabiee A, Aghaei J, Muttaqi KM, Nezhad AE. On the management of
[5] A survey on advanced metering infrastructure and its application in Smart
wind power intermittency. Renew Sustain Energy Rev 2013;28:643e53.
Grids. In: Mohassel RR, Fung AS, Mohammadi F, Raahemifar K, editors. 2014
[19] Gen B. Reliability and cost/worth evaluation of generating systems utilizing
IEEE 27th canadian conference on. Electrical and Computer Engineering
wind and solar energy. 2005.
(CCECE); 2014.
[20] Owner’s manual of the AIR403 wind turbine made by Southwest Wind power
[6] Bouhafs F, Mackay M, Merabti M. Communication challenges and solutions in
Inc. www.nooutage.com/pdf/swwp_air403_landman.pdf.
the smart grid. Springer; 2014.
[21] Abouzahr I, Ramakumar R. An approach to assess the performance of utility-
[7] Verma PK, Sharma H. Application of cloud computing In Smart grid: a review.
interactive wind electric conversion systems. Energy Conversion. IEEE Trans
International conference of advance research and innovation (ICARI-2015);
1991;6(4):627e38.
404e409.
[22] Atwa Y, El-Saadany E, Salama M, Seethapathy R. Optimal renewable resources
[8] Vardakas JS, Zorba N, Verikoukis CV. Performance evaluation of power de-
mix for distribution system energy loss minimization.,IEEE Transactions on.
mand scheduling scenarios in a smart grid environment. Appl Energy
Power Syst 2010;25(1):360e70.
2015;142:164e78.
[23] Ettoumi FY, Mefti A, Adane A, Bouroubi M. Statistical analysis of solar mea-
[9] Albadi MH, El-Saadany E. A summary of demand response in electricity
surements in Algeria using beta distributions. Renew Energy 2002;26(1):
markets. Electr Power Syst Res 2008;78(11):1989e96.
47e67.
[10] Khan AA, Razzaq S, Khan A, Khursheed F. HEMSs and enabled demand
[24] Deshmukh M, Deshmukh S. Modeling of hybrid renewable energy systems.
response in electricity market: an overview. Renew Sustain Energy Rev
Renew Sustain Energy Rev 2008;12(1):235e49.
2015;42:773e85.
[25] Tina G, Gagliano S, Raiti S. Hybrid solar/wind power system probabilistic
[11] Alavi SA, Ahmadian A, Aliakbar-Golkar M. Optimal probabilistic energy
modelling for long-term performance assessment. Sol Energy 2006;80(5):
management in a typical micro-grid based-on robust optimization and point
578e88.
estimate method. Energy Convers Manag 2015;95:314e25.
[26] Chowdhury S, Crossley P. Microgrids and active distribution networks. The
[12] Zakariazadeh A, Jadid S, Siano P. Stochastic multi-objective operational
Institution of Engineering and Technology; 2009.
planning of smart distribution systems considering demand response pro-
[27] Mazzola S, Astolfi M, Macchi E. A detailed model for the optimal management
grams. Electr Power Syst Res 2014;111:156e68.
of a multigood microgrid. Appl Energy 2015;154:862e73.
[13] Zakariazadeh A, Jadid S, Siano P. Economic-environmental energy and reserve
[28] Kanchev H, Lu D, Colas F, Lazarov V, Francois B. Energy management and
scheduling of smart distribution systems: a multiobjective mathematical
 G.R. Aghajani et al. / Energy 126 (2017) 622e637 637

operational planning of a microgrid with a PV-based active generator for part I: formulation. IEEE Transactions on. Power Syst 2005;20(4):1818e26.
smart grid applications. IEEE Transactions on. Ind Electron 2011;58(10): [35] Willy Weather. Available online: http://wind.willyweather.com.au.
4583e92. [36] The Solar Power Group Company. Accessed at, http://thesolarpowergroup.
[29] Moghaddam AA, Seifi A, Niknam T, Pahlavani MRA. Multi-objective operation com.au.
management of a renewable MG (micro-grid) with back-up micro-turbine/ [37] Reconstruction and Short-term Forecast of the Solar Irradiance. Accessed at,
fuel cell/battery hybrid power source. Energy 2011;36(11):6490e507. http://lpc2e.cnrs-orleans.fr/~soteria/.
[30] Kennedy J. Particle swarm optimization. Encyclopedia of machine learning. [38] Bouhouras AS, Labridis DP, Bakirtzis AG. Cost/worth assessment of reliability
Springer; 2010. p. 760e6. improvement in distribution networks by means of artificial intelligence. Int J
[31] MOPSO: a proposal for multiple objective particle swarm optimization. In: Electr Power & Energy Syst 2010;32(5):530e8.
Coello Coello CA, Lechuga MS, editors. Evolutionary computation, 2002 CEC'02 [39] Chen C, Duan S, Cai T, Liu B, Hu G. Smart energy management system for
proceedings of the 2002 congress on; 2002. optimal microgrid economic operation. IET Renew power Gener 2011;5(3):
[32] A benchmark low voltage microgrid network. In: Papathanassiou S, 258e67.
Hatziargyriou N, Strunz K, editors. Proceedings of the CIGRE Symposium: [40] Clement-Nyns K, Haesen E, Driesen J. The impact of charging plug-in hybrid
power systems with dispersed generation; 2005. electric vehicles on a residential distribution grid. IEEE Transactions Power
[33] Apx power spot exchange. Accessted at, https://www.apxgroup.com/trading- Syst 2010;25(1):371e80.
clearing/apx-power-uk/. [41] Staff F. Assessment of demand response and advanced metering. Docket AD-
[34] Bouffard F, Galiana FD, Conejo AJ. Market-clearing with stochastic security- 06-2-000. Federal Energy Regulatory Commission; 2006.