Energy 210 (2020) 118602

Contents lists available at ScienceDirect

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

Optimal management of home loads with renewable energy

integration and demand response strategy
Eity Sarker a, *, Mehdi Seyedmahmoudian a, Elmira Jamei b, Ben Horan c, Alex Stojcevski a
a
School of Software and Electrical Engineering, Swinburne University of Technology, Melbourne, VIC, 3122, Australia
b
College of Engineering and Science, Victoria University, Melbourne, VIC, 3011, Australia
c
School of Engineering, Deakin University, Geelong, VIC, 3216, Australia

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

Article history: The implementation of proper energy management techniques and utilization of renewable energy re-
Received 29 March 2020 sources enhance the energy efficiency and stability of future grid systems. This research proposed a home
Received in revised form energy management model consisting of microgrid framework and demand side management (DSM)
17 July 2020
technique. To reduce peak load, peak to average, and energy cost, households’ loads were shifted on the
Accepted 8 August 2020
Available online 21 August 2020
basis of price-based tariff such as flexible and time of use tariff. Simulation was carried out using binary
particle swarm optimization algorithm in MATLAB. The microgrid was mathematically modeled, and the
impacts of DSM integrated microgrid were analysed for different households in terms of electricity cost
Keywords:
Demand side management
reduction. Simulations suggested that DSM implementation significantly reduced peak loads and
Demand response renewable resources produced trade-off. Renewable energy integration with DSM can be a promising
Load shifting technique approach for the significant reduction in total electricity cost of households by paying less for purchasing
BPSO algorithm electricity and selling the surplus electricity to the grid.
Energy tariff © 2020 Elsevier Ltd. All rights reserved.
Renewable energy generation

1. Introduction integrating distributed energy resources, such as wind turbines and

solar panels, and distributed energy storage like batteries [8,9].
The increase in energy consumption as well as the rapid growth Moreover, microgrid has distinct features, including reliability, low
of population and the lack of implementation of proper manage- investment costs, and regulations of different distributed generator
ment techniques result in an extreme spike in energy demand [1,2]. units’ output voltage and current [10]. Additionally, microgrid can
Existing electrical grid worsens the situation due to their old- be incorporated with different demand side management (DSM)
fashioned design as well as redundant and overstressed infra- frameworks and operated in grid connected and off-grid modes.
structure [3]. Recently environmental pollution gains much atten- DSM refers to the amendments of consumers’ energy con-
tion among the scientists and environmentalists because of public sumption pattern to enhance the efficiency of electrical energy
consciousness of reducing carbon emission and political pressure systems and network [11e14]. This technique modifies daily energy
[4]. About 85% of the total global energy consumption depends on consumption pattern to achieve a desired load profile [15e17].
fossil fuels [5]. The excessive usage of fossil fuels is associated with Numerous scholars studied the implementation of DSM in resi-
the release of substantial CO2 emission. The integration of renew- dential loads management [18e20]. For instance, Gottwalt et al.
able energy resources (RES) in power generation is the most carried out a simulation for shifting residential loads on the basis of
effective and feasible way to promote sustainable development and €
time of use (TOU) tariff [21]. Ozkan also developed DSM model to
reduce environmental pollution [6]. reduce peak load and energy cost for residential area [22]. Bharathi
The utilization and optimization of RES lead to the concept of et al. solved DSM optimization problems for residential loads
microgrid as a replacement for fossil energy sources [7]. Recently, management using genetic algorithm to reduce peak load [23].
the use of microgrid system has gained significant popularity in Srinivasan et al. developed a game theory based dynamic pricing
finding ways to increase the stability of energy supply by model for residential and commercial electricity sector in
Singapore [24].
Renewable energy integration or microgrid has gained signifi-
* Corresponding author. cant attention in recent years because of the economic and
E-mail addresses: [email protected], [email protected] (E. Sarker).

https://doi.org/10.1016/j.energy.2020.118602
0360-5442/© 2020 Elsevier Ltd. All rights reserved.
2 E. Sarker et al. / Energy 210 (2020) 118602

environmental benefits of renewable energy [25e28]. Additionally, energy with various power ratings. In the BPSO algorithm, a set of
the modeling of microgrid with the application of DSM can play a household devices is presented by D ¼ {d1, d2, … …,dn}.
pivotal role in reducing peak load, energy inefficiency, and opera- For each household, the scheduling vector of energy consump-
tional cost of electricity provider, thus helping to reduce carbon tion of the appliances can be presented by Dn ¼ ½d11 ; ………; dtn ,
footprints. This integrated approach can reduce the amount of where dtn is the energy consumption of appliances n, running for t
energy required to buy from the grid. Recent works were dedicated hour. The total energy consumption by appliances n in the house-
to the implementation of DSM along with the integration of hold is estimated by the following expression:
renewable energy for achieving a balance between energy gener-
ation and consumption. Quiggin et al. implemented the demand X
24
response program in a residential microgrid with the integration of En ¼ dtn (1)
solar photovoltaic, wind turbine, and energy storage [29]. This t¼1
method optimized the energy balance between supply and demand Subjected to
by reducing peak demand fluctuations by 16%, thus significantly
reducing CO2. Palma-Behnke et al. modeled a microgrid system X
i¼n
with solar photovoltaic, wind turbine, diesel generator, battery dtn ¼ ui *Xit (1a)
bank, and water supply system [30]. The DSM mechanism along i¼1
with artificial neural network was used for determining optimal
operation and reducing costs. Shen et al. carried out an incentive- where ui and Xit represent the power rating and operation time of
based demand response program in a microgrid, which consisted ith appliance, respectively. The devices were scheduled based on
of micro turbines, wind turbine, fuel cells, solar photovoltaic, their daily energy consumption. The scheduled energy consump-
storage devices, and controllable load [31]. The authors solved an tion of the appliances can be expressed as follows:
operational scheduling problem to achieve an optimal scheduling
of the batteries and diesel generators. Nunna and Doolla proposed X
ET
En ¼ dtn (2)
an intelligent energy management system with demand response
t¼ST
program to reduce peak load demand in the microgrid, and they
performed simulation by using Java Agent Development framework where ST and ET are the schedule start and end time of the appli-
[32]. To reduce electricity cost, an incentive method was recom- ances, respectively.
mended by the authors to motivate the consumers for participating The cost function CTOU , which represents the cost of energy in
in the demand response program. Philippou et al. investigated a each hour, was calculated based on flexible pricing tariff and time of
price-based DSM mechanism and performed sensitivity analysis to use (TOU) tariff provided by the utility company. In Victoria, there
reduce peak load during summer and winter [33]. are some time of use tariff are available such as TOU, flexible
The state of the art review, discussed above suggests that a wide pricing, single rate and block rate tariff based on what time the
number of literatures have focussed on the demand response using electricity is used. Flexible pricing tariff introduced three periods,
various algorithms; however studies related to different price- such as peak, off-peak, and shoulder while, TOU tariff considered
based demand response using BPSO and consideration of uncer- only peak and off-peak periods. Table 1 shows the typical TOU tariff
tainty in renewable generation integrated with demand response and flexible pricing tariff of Victoria, Australia [https://www.
are very limited in the literature. Therefore, this research work canstarblue.com.au/electricity/victoria-electricity-tariffs/]. The
proposes an integrated framework for home energy management, price of electricity at peak hour is higher as compared to price at
which includes load scheduling based on different tariff rate and off-peak hour due to the high demand at peak hour. The peak and
fulfilling the scheduled demand from renewables. The major con- off-peak hours for TOU tariff are the same for weekdays and
tributions of this research work are as follows: A DSM system is weekends. In the case of flexible tariff, no peak period for weekends
modeled based on flexible and time of use tariff rate to minimize exists. However, shoulder periods are extended from 7am to 10pm.
the households’ energy cost and the optimization problem was The time span of 24 h was divided into equal time slots, where
solved using BPSO algorithm with sensitivity analysis. A microgrid t 2T. The cost is the function of amount of energy consumed. The
was modeled considering uncertainty of renewables generation cost function (i.e., CTP and CTO) for TOU tariff is presented by the
and the effect of renewable generation on scheduled load was following equations. Here, CTP and CTO are the energy consumption
studied with surplus power selling. For this purpose, a compre- cost during peak and off-peak hours, respectively.
hensive case study was carried out for scheduling households’ load
and evaluating different situations over weekend and weekday. X
Tp
CTP ¼ EnT *MTP (3)
2. Load modeling and DSM implementation for T¼7
multiobjective optimization
X
To
2.1. Household load modeling CTO ¼ EnT *MTO (4)
T¼23
DSM implementation aims to reduce load demand during peak
hours, reduce electricity bill, and maximize the use of renewable where MTP and MTO are the costs of electricity per unit for TOU tariff
energy as well as reduce the usage of electricity from the main during the peak and off-peak periods, respectively. In the case of
distribution grid. A BPSO-based load scheduling mechanism was TOU tariff, the total cost (CTT) for 24 h duration is the summation of
applied to manage residential load demand because the swarm- cost during peak and off-peak hours, as shown in the following
based intelligence methods are found as effective methods for equation:
engineering problems [34]. In BPSO algorithm, each hour in a day
CTT ¼ CTP þ CTO (5)
was denoted by row vector. The proposed DSM technique sched-
uled households’ shiftable appliances. Each of the household de- In the case of flexible pricing tariff, the cost function (i.e., CFP CFS
vices considered in this study consumed different amounts of CFO) is given as follows. Here, CFP, CFS and CFO are the energy
 E. Sarker et al. / Energy 210 (2020) 118602 3

Table 1
Typical electricity tariff (average value) in Victoria, Australia.

Flexible TOU

Peak Off-peak Shoulder Peak Off-peak

Time 3pme9pm 10pme7am 9pme10pm and 7ame3pm 7ame11pm 11pme7am

Rate (ctAUD/kWh) 45 19.5 35.5 36.5 17

consumption cost during peak shoulder, and off-peak hours, loads, respectively. FL is the flexible pricing tariff. To schedule the
respectively. household appliances, the objective function is fed to the BPSO
algorithm on basis of electricity price. In order to optimize the
X
Tp
objective function, BPSO algorithm was executed considering iter-
CFP ¼ EnT *MFP (6) ation 30, swarm size 20, inertia weight 0.5, velocity clamping 1 and
T¼15
acceleration coefficient 2 with following steps.
Step 1: Loading of hourly power consumption data, duration of
X
Ts
appliance operation and tariff rate. Estimation of total power con-
CFS ¼ EnT *MFS (7)
sumption and total cost.
T¼21; 7
Step 2: Random initialization of the particles with the binary
value between 0 and 1. The status of load defined by xi ¼ ½x1; x2 ; …
X
To
……xn ; cxi ε ð0; 1Þ, where xi is the decision variable. The value xi ¼
CFO ¼ EnT *MFO (8)
1 represents the connection of the load and xi ¼ 0 indicates the
T¼22
disconnection of the load.
where MFP, MFS, and MFO are the per unit electricity price for flexible Step 3: Random initialization of velocity of the particle accord-
pricing tariff during peak, shoulder, and off-peak hours, respec- ing to the equation, vi ¼ ½v1; v2 ; ………vn ; cvi ε ð0; 1Þ.
tively. For flexible pricing tariff, the total cost of energy consump- Step 4: Evaluation of the objective function (equation (10)) and
tion (CFT) is the summation of cost during peak, shoulder, and off- estimation of pbest and gbest value.
peak hours and expressed as follows: Step 5: Setting of the iteration ¼ iteration þ 1:
Step 6: Updating of velocity of particle according to the
CFT ¼ CFP þ CFS þ CFO (9) following equation.
 
ðtBPSO þ1Þ
Vi ¼ w:VitBPSO þ C1 :randðÞ: xtpbest:i
BPSO
 xti BPSO
 
2.2. Objective function and BPSO algorithm execution for load þ C2 :randðÞ: xtgbest:i
BPSO
 xti BPSO (11)
sceduling
where C1 and C2 are the cognitive and social parameter, i is the
The load scheduling was formulated as a minimization problem number of particle (i ¼ 1; 2; 3…:nÞ;t is the number of iteration, and
and the optimal scheduling of the household appliances was ob- w is the inertia weight. xtpbest:i
BPSO
is the particle’s individual best so-
tained by solving the minimization problem as follows. In the
minimization function, CT represents the total electricity cost, lution at time t (pbest) and xtgbest:i
BPSO
is the global best solution (gbest).
which can be either based on the flexible pricing tariff or TOU Step 7: Normalizing of new velocity according to the sigmoid
pricing tariff. The minimizing function, Zm can be expressed as function as follows:
follows:
t þ1 1
v0 i BPSO ¼ þ1
(12)
X
T
1þe
t
vi BPSO
Zm ¼ CT (10)
t¼1 Step 8: Updating of the position of the particle according to the
following expression.
Subjected to
8 Tp ðtBPSO þ1Þ ðtBPSO þ1Þ
X XTo xi ¼ xti BPSO þ Vi (13)
>
>
>
> EnT *MTP þ EnT *MTO 2TOU
>
< T¼7 Step 9: Evaluation objective function and determination of cost
T¼23
CT ¼ according to eqation 10.
>
> X
Tp X
Ts X
To
>
> Step 10: Updating of pbest, gbest and inertia weight.
>
: EnT *MFP þ EnT *MFS þ EnT *MFO 2FL
Step 11: Termination of simulation when reaches at maximum
T¼15 T¼21;7 T¼22
iteration, otherwise repeat step 4.
(10a)
3. Microgrid modeling
CTU  CTS (10b)
The proposed microgrid consists of three subsystems, namely,
XX
i¼n energy generation, residential load demand, and energy distribu-
EnT ¼ ui *Xit ; tion subsystems. The schematic diagram of a hybrid microgrid is
T i¼1 shown in Fig. 1. The wind turbine (WT) and solar photovoltaic (PV)
cT2Specific tariff duration and ci2appliance (10c) panels worked as renewable energy generators. In our proposed
model, residential load profile was used as a demand subsystem.
where CTU and CTS indicate total cost for unscheduled and scheduled The microgrid was connected to the grid through AC bus, and whole
4 E. Sarker et al. / Energy 210 (2020) 118602

Fig. 1. Conceptual design of proposed microgrid model.

microgrid system worked as a power distribution subsystem. DSM hourly wind speed [38]. In this study, the hourly wind speed data
technique was applied in the residential loads to schedule the load for the study area was collected from timeanddate.com [https://
demand. The energy generated from the renewable sources was www.timeanddate.com/weather/australia/melbourne/historic?
used to meet the scheduled load demand on an hourly basis. month¼11&amp;year]. A Weibull probability distribution function
was used for predicting the wind speed distribution according to
3.1. PV system the following equations [39e41]:

The PV system produces electrical energy from solar energy, and  k

 k1  V
the output of the PV system depends on solar irradiance, efficiency k V c

of the PV panel, and atmospheric temperature [35]. In this study, f ðVÞ ¼ * *e ; 1  k  10 (18)
c c
performance data of solar PV (%) was used to model the solar power
output behaviour based on the Beta probability distribution func-  s 1:086
tion according to the following equations [36,37]: k¼ (19)
8 V
< Gða þ bÞ *Sða1Þ *ð1  SÞðb1Þ ; 0  S  1; a  0; b  0
>
f ðSÞ ¼ GðaÞ*GðbÞ V
>
: c¼   (20)
0; otherwise
G 1 þ 1k
(14)

  where k, c, V, V and s are shape factor, scale factor, real time wind
S*ð1 þ SÞ
b ¼ ð1  SÞ* 1 (15) speed, mean and standard deviation, respectively. The output po-
s2 wer from WT can be estimated by the following equation [38,42]:
8
S*b >
> 0; V < VI and V  Vo
a¼ (16) >
>
<
1S V  VI
PWTOUT ¼ f ðVÞ* PWTR * ; VI  V < VR (21)
>
> VR  VI
where a and b are the parameters of Beta distribution function. S, S >
>
:
and s2 are performance of solar PV (%), mean and standard devia- PWTR ; VR  V < V0
tion, respectively. The output power from solar PV was estimated
by the following equation: where PWTOUT and PWTR indicate the output power from WT and
rated power of WT, respectively. VR, VI, and VO represent the rated
PVout ¼ PVr *FPV *f ðSÞ (17) wind speed, cut-in speed, and cut-out speed, respectively.

where PVout is the output power of solar PV. PVr and FPV indicate the
rated power of solar PV (kW) and performance of the solar PV (%),
respectively. The performance of the solar PV data was based on the 3.3. Revenue from renewables
Victorian state performance on the particular date [https://pv-map.
apvi.org.au/live]. The hourly power generated from the solar PV and WT is
compared with the scheduled load demand for each of the
3.2. Wind turbine households. The hourly surplus energy (i.e., excess energy from
renewables after fulfilling the demand) and energy deficit (i.e.,
The wind power system converts the wind speed into electrical energy required to purchase from grid after consuming the energy
power. The output power of the WT energy system depends on the from renewables) were estimated using the following equation:
 E. Sarker et al. / Energy 210 (2020) 118602 5

4.1. Load profiles and scheduling of the loads

t t
ES=D ¼ PVOUT þ PWTOUT  Ent (22)
Fig. 2 shows the weekday’s average load profiles for four
where ES/D is amount of energy surplus or deficit in an hour. The households and their comparison with the load profiles after DSM
positive value of ES/D indicates energy surplus, whereas negative implementation under the flexible pricing and TOU tariff scheme.
value represents energy deficit. The monthly revenue generation by The peak operation is one of the most important tasks in power
selling surplus energy can be estimated using the following equa- system as it causes the fluctuation of power transmission from
tion, where FRAV is the average feed-in-tariff in Victoria: main grid and disturbance in reliable operation of power grid [43].
Therefore, it is necessary to smooth the demand load curves of
" !#
X
30 X
24   multiple users interconnected with power grids. The proposed
t t
RM ¼ PVOUT þ PWTOUT  Ent *FRAV model reduced the peak load demand and distributed in other time
d¼1 t¼1 (23)
periods, which provided a load curve with low fluctuation. The
t t
where PVOUT þ PWTOUT > Ent loads were shifted based on electricity price and consumers’ pref-
erence keeping the total load demand same. Fig. 2 shows that the
each of the households’ hourly electricity consumption was
different because of variation in preference and behaviour of en-
ergy consumption. Whether the consumers used the flexible pric-
4. Results and discussion ing tariff or TOU tariff scheme, they were not getting any benefit
from the tariff scheme because most of the electricity was
The performance of the proposed microgrid model was inves- consumed during peak periods. In the case of flexible pricing tariff,
tigated for a case study. The simulation was carried out in four the peak period was between 15th hour to 21st hour, whereas the
households in Victoria, Australia. The electricity consumption of peak period was between 7th hour to 23rd hour for TOU pricing
each of the households and different set of appliances are listed in tariff. In the peak period, the electricity providers charged a higher
Table 2. The start time and end time of the appliances was the users’ price for per unit of electricity compared to off-peak and shoulder
preferable time to use the appliances. In this mode, shiftable and periods. In addition, the total electricity demand in peak period was
nonshiftable loads are considered for the estimation of overall about 8.24 kWh for the household 1 under flexible pricing tariff.
energy requirement and cost. However, only shiftable loads were After the implementation of DSM, the total energy demand during
considered for load shifting simulation.

Table 2
Appliances and power consumption pattern for households.

Device Preferred time of use Hourly consumption (kW) Power rating (kW)

1h 2h 3h

Household 1
Washing machine 10ame5pm 0.25 0.25 0 0.25
Vacuum cleaner 12pme3pm 0.9 0 0 1.8
Oven 6ame9am and 12:30pme1:30pm 0.37 0.37 0 1.1
Iron 7pme9pm 0.34 0 0 1
Rice cooker 5pme7pm 0.25 0.25 0 0.5
Induction cooker 5pme8pm 0.2 0.2 0.1 0.2
Room heater 6pme11pm 0.95 0.95 0.95 1.9
Toaster 7ame9 am 0.17 0 0 0.7
Household 2
Washing machine 11ame2pm 1.25 0.625 0 1.25
Vacuum cleaner 10ame1pm 1 0 0 2
Iron 5pme10pm 0.67 0 0 2
Rice cooker 6ame9am and 7pme9pm 0.35 0.35 0 0.7
Dish washer 5pme7pm 1.8 0 0 1.8
Room heater 7pme10pm 2 1 2 2
Toaster 7ame9am 0.29 0 0 1.75
Hair dryer 7ame8am 0.36 0 0 2.2
Air Fryer 5pme7pm 0.5 0 0 1.5
Oven 6ame9am 0.25 0.25 0.25 1.5
Household 3
Washing machine 9ame1pm 0.25 0.041 0 0.25
Vacuum cleaner 10ame12pm 1 0 0 2
Blender 6ame9am and 6pme8pm 0.16 0 0 0.5
Iron 7pme10pm 0.66 0 0 2
Rice cooker 7ame9am and 6pme9pm 0.35 0.35 0 0.7
Room heater 6pme10pm 1.9 1.9 0.9 1.9
Toaster 7ame9am 0.13 0.13 0 0.8
Oven 8ame9am 0.23 0.23 0.23 1.4
Household 4
Vacuum cleaner 9ame4pm 1.2 0 0 1.2
Blender 4pme6pm 0.1 0.1 0 0.3
Iron 9pme10pm 0.8 0 0 2.4
Rice-cooker 6pme8pm 0.35 0.35 0 0.7
Double Hot Plate 6pme9pm 0.7 0.7 0 1.4
Room heater 7pme11pm 1.9 1.9 1.9 1.9
Oven 7ame10am 0.2 0.2 0.2 1.2
6 E. Sarker et al. / Energy 210 (2020) 118602

Fig. 2. Average hourly load profile of households in weekday.

the peak period was reduced to 6.41 kWh due to the shifting of shoulder rate is charged most of the day (7ame10pm). Similar to
loads from peak to off-peak and shoulder hours. Therefore, almost weekday, peak and off-peak rates were charged in the weekend for
22.2% (1.83 kWh) of loads in peak periods was shifted from peak TOU pricing tariff. Hence, DSM was applied to shift the loads from
periods to off-peak and shoulder periods, where the per unit peak periods. However, no notable load shifting was obtained for all
electricity price was lower than that during peak periods. Similarly, the cases because of the similar reasons, as described earlier. When
the implementation of DSM shifted approximately 18.2%, 23.4%, DSM was applied, the individual peak load was reduced signifi-
and 20.6% of loads in peak periods from peak periods to off-peak cantly and distributed throughout peak periods. However, the in-
and shoulder periods for households 2, 3, and 4, respectively. dividual peak load remained in the same hour for all the
Additionally, the individual peak load for all the households was households, as illustrated in Fig. 3.
also reduced after DSM implementation, as shown in Fig. 2. In the
case of TOU pricing tariff, the energy providers offer only peak and 4.2. Performance gain in terms of energy and cost
off-peak hour pricing scheme. Most of the hours of the day
(7ame11pm) are considered peak period, where consumers usually Fig. 4 shows the effect of swarm size and iteration on optimal
prefer to use their loads. Therefore, no significant load shifting was solution for the household 1 under flexible tariff. The number of
observed in the case of TOU pricing tariff due to the users’ prefer- iteration depends on the complexity of the optimization problem
ence of load utilization, nature of loads, and extended hours of peak and affects the probability of achieving optimal solution as well as
periods. However, the implementation of DSM reduced the indi- the computation time. As can be seen from Fig. 4, the optimal so-
vidual peak load and distributed the reduced amount of load in the lution was obtained between 15 and 22 iteration for swarm size
entire peak periods for all the households, as presented in Fig. 2. 5e30 with inertia weight 0.5, velocity clamping 1 and acceleration
Fig. 3 displays the weekend’s average hourly electricity con- coefficient 2. In previous study, Guerdia suggested 10e50 iterations
sumption for the households. The average electricity consumption for optimization problems with medium complexity and 200e400
in the weekend for all the households was higher than in weekdays. iterations for optimization problems with high complexity,
In the case of flexible tariff, DSM was not applied for weekend respectively [44]. The increases in swarm size from 5 to 20
because of the absence of peak period for weekend. Moreover, a remarkably reduced the iteration number for obtaining optimal
 E. Sarker et al. / Energy 210 (2020) 118602 7

Fig. 3. Average hourly load profile of households during weekend.

number of iteration was considered 20 and 30 for all the cases.

Fig. 5 depicts the effects of DSM implementation on the
reduction of electricity cost in weekday for all the households
considered. The DSM allows the consumers to shift the load from
peak hours to off-peak and shoulder hours as discussed earlier,
causing substantial electricity cost (Fig. 5). The reduction of hourly
electricity cost depends on the type of selected appliances and the
flexibility of the operation time, which were provided by the con-
sumers for load scheduling. The hourly electricity cost presented in
Fig. 5 shows that the electricity cost in peak periods was about
65.1% of the total cost for household 1, which was reduced to
approximately 50.7% after DSM application. Thus, the total elec-
tricity consumption was significantly reduced during peak periods.
Similarly, the electricity cost in peak periods was reduced from
59.7%, 52.7% and 51.7%e51.6%, 42.3% and 43.2% of the total cost
under flexible pricing tariff for households 1, 2, and 3, respectively.
In the case of TOU pricing tariff, a slight reduction in electricity cost
during peak periods was observed for all the households.
Fig. 6 shows the hourly electricity cost reduction during week-
Fig. 4. Performance of BPSO at different swarm size. end after load scheduling, particularly under TOU pricing tariff. In
the case of flexible pricing tariff, no load was scheduled. Fig. 6
shows that the pattern of hourly electricity cost of unscheduled
solution; however no significant difference was observed in further
loads under flexible pricing tariff was almost similar to that of
increase from 20 to 30 (Fig. 5). Therefore, the swarm size and
8 E. Sarker et al. / Energy 210 (2020) 118602

Fig. 5. Hourly cost curves for average load of households in weekday.

unscheduled loads under TOU pricing tariff. When TOU tariff rate is 4.3. Renewable energy integration
considered, most of the household appliances were operated dur-
ing the peak periods. The cost of electricity consumed in peak pe- We considered residential microgrid, which can be connected to
riods under TOU tariff was about 94.9%, 94.3%, 97.9%, and 97.5% of the main grid to import and export power for each of the house-
total electricity cost for households 1, 2, 3, and 4, respectively. holds. The proposed microgrid model contains a 3 kW solar panel
However, with the application of DSM, the cost of electricity in peak and a 1.5 kW wind turbine. The hourly power output from solar PV
periods for the households was reduced to 92.1%, 93.5%, 95.9%, and was estimated using the Victorian hourly performance data of solar
95.5% of total electricity cost, respectively. PV. According to the technical specifications of the 1.5 kW wind
The user always wants to reduce their electricity cost and pre- turbine widely used in Australia [https://www.
fers to maintain a balance of load with a low peak to average ratio australianwindandsolar.com/aws-hc-wind-turbines], the rated,
(PAR). In the present case study, the application of DSM showed a cut-in, and cut-out speeds are 10.5 m/s, 2.7 m/s, and 12.1 m/s,
significant reduction in electricity cost and PAR. Table 3 shows the respectively. Hence, the hourly wind power output was predicted
comparison between the flexible pricing and TOU pricing tariff based on the hourly wind speed. However, the accurate prediction
scheme for load scheduling in terms of peak load reduction, total of solar and wind power is not possible due to their stochastic
electricity cost reduction, and PAR reduction. In weekdays, the PAR behaviour and associated with uncertainty. Therefore, the optimal
and electricity cost reduction were 12.8%e27.1% and 5.4%e6.6% power output from solar and wind was estimated using probability
under flexible pricing tariff, whereas those for TOU tariff were distribution function, which indicates the frequency of wind speed
12.8%e20.4% and 2.0%e2.9%, respectively, due to the application of (Fig. 7(A)) and frequency of PV performance (Fig. 7(B)). The highest
DSM. In the weekend, the load scheduling under TOU tariff reduced relative frequency of 60% was observed for 5.4 m/s in the case of
the PAR and electricity cost by 14.9%e20.1% and 1.0%e2.9%, weekday wind speed distribution, while 7.2 m/s showed the
respectively. highest frequency of 41% for weekend distribution. The Beta dis-
tribution of PV performance for both weekday and weekend
 E. Sarker et al. / Energy 210 (2020) 118602 9

Fig. 6. Hourly cost curves for average load of households in weekend.

Table 3 basis. The households can sell the excess energy (surplus) from
Summary of the load shifting results (reduction of PAR, PL and Cost in percentages renewables after fulfilling their demand (i.e., positive value in
after DSM implementation).
Fig. 9). The negative value indicates the hourly amount of energy
Consumers Flexible price tariff TOU tariff TOU tariff that require to buy from the grid, whereas zero (i.e., base line in
(Weekday) (Weekday) (Weekend) Fig. 9) suggests that the generation from renewables is equal to the
PAR PL Cost PAR PL Cost PAR PL Cost households’ electricity demand. In the case of weekday, energy
Household 1 27.1 27.1 6.6 16.1 16.1 2.7 19.0 19.0 2.9 deficit was observed during 6ame9am and 18pme22pm for all the
Household 2 15.9 15.9 6.3 20.4 20.4 2.3 20.1 20.1 1.0 households under flexible and TOU pricing tariff due to the use of
Household 3 20.1 20.1 5.4 12.8 12.8 2.9 14.9 14.9 2.2 more loads and less generation from renewables. The higher sur-
Household 4 12.8 12.8 5.9 14.9 14.9 2.0 16.0 16.0 2.1 plus energy was obtained during the mid-day for both tariff cases
PAR- Peak to average ratio; PL- Peak load. associated with the higher renewable energy generation and less
energy consumption in those periods, as shown in Fig. 9(A) and (B).
On the contrary, the energy deficit for weekend occurred during
exhibited almost same pattern. Fig. 8 presents the average hourly 9ame10am and 15pme23pm. Similar to weekday, weekend also
predicted power generation from the renewable energy sources for showed maximum energy surplus during mid-day. Therefore,
weekday and weekend. The hourly power output for weekday and approximately 2.49e6.92 kW of surplus energy was obtained per
weekend showed almost the similar trend. The higher renewable day for the studied households by the incorporation of DSM and
output predicted during the mid-day for both weekday and renewable energy resources. Accordingly, the estimated monthly
weekend was due to the higher PV output, which was linked to the surplus energy was in the range between 85 kW and 122 kW which
higher intensity of solar irradiance during those periods. can be sold into grid to generate revenue.
Fig. 9 illustrates the average hourly energy surplus and deficit
with the integration of renewable generation for each of the
households after load shifting. The amount of energy that the 4.4. Trade-off from DSM integrated with microgrid
households can sell to the market and the amount of energy that
the households need to buy from the grid was estimated hourly Fig. 10 shows the monthly electricity cost and benefits gained
10 E. Sarker et al. / Energy 210 (2020) 118602

Fig. 7. Probability distribution function (A) Weibull distribution for wind power (B) Beta distribution for solar power.

Fig. 8. Average power output from renewable energy sources (A) Weekday (B) Weekend.

from the DSM implementation and renewable integration. Each of 7% of the electricity cost without DSM, which can compensate the
the households generated revenue by selling surplus energy, and users’ remaining electricity cost after DSM implementation and
the value was estimated based on the average Victorian feed-in- renewable integration. Therefore, considering all the scenarios i.e.,
tariff of 10 ctAUD/kWh [https://www.canstarblue.com.au/ load shifting, power generation from renewables and revenue from
electricity/victoria-electricity-tariffs/]. Fig. 10 (A) shows that the surplus energy, the households considered for the current case
monthly electricity cost was reduced by 5.5%e6.1% for flexible study need to spend from only 18e26% of the total electricity cost
pricing tariff when some of the total household’s loads were shifted without DSM.
from peak hours to off-peak and shoulder hours based on the
electricity price and consumers’ preference. On the contrary, the 5. Conclusions and recommendations
maximum electricity cost reduction in the case of TOU pricing tariff
was only just above 2.5%, as shown in Fig. 10 (B). The current work aims to minimize the monetary expenses of
In the case of TOU pricing tariff, an insignificant amount of load electricity consumption in households by shifting the loads to the
was shifted from peak hours to off-peak hours due to users’ pref- times with lower electricity price and utilizing energy from re-
erence and nature of loads, which were linked to a small reduction newables. A residential household load management model was
in electricity cost. Analysis of flexible and TOU pricing tariff shows proposed based on the price-based demand response along with
that users’ monthly electricity cost was reduced significantly for the integration of renewable energy resources. Simulation and case
both tariff cases (Fig. 10). For instance, over 76% of base electricity studies were conducted in different households to analyze the
cost (i.e., electricity cost without DSM) reduction was obtained for effectiveness of the proposed model (i.e., DSM integration with
household 4 under flexible pricing tariff due to the combined ef- renewables) under flexible and TOU pricing schemes using BPSO
fects of load shifting and renewable integration. Additionally, the algorithm in MATLAB. Results showed the potential benefits of DSM
monthly revenue generation from surplus energy was more than implementation in reducing peak load and electricity cost. As a
 E. Sarker et al. / Energy 210 (2020) 118602 11

Fig. 9. Hourly energy surplus and deficit for each of the households after load shifting (A) flexible pricing weekday, (B) TOU weekday, (C) flexible pricing weekend, and (D) TOU
weekend.

Fig. 10. Monthly cost analysis under different scenarios (A) flexible pricing tariff (B) TOU pricing tariff.
12 E. Sarker et al. / Energy 210 (2020) 118602

result of demand response program and renewable energy inte- [11] Behrangrad M. A review of demand side management business models in the
electricity market. Renew Sustain Energy Rev 2015;47:270e83. https://
gration, the consumers used a minimal amount of electricity from
doi.org/10.1016/j.rser.2015.03.033.
grid, and they could sell surplus energy to the grid. This practice [12] Siano P. Demand response and smart grids - a survey. Renew Sustain Energy
greatly impacts the reduction of the households’ monthly elec- Rev 2014;30:461e78. https://doi.org/10.1016/j.rser.2013.10.022.
tricity cost. However, the investigation on the integrated home [13] Stavrakas V, Flamos A. A modular high-resolution demand-side management
model to quantify benefits of demand-flexibility in the residential sector.
energy management model is associated with some limitations, Energy Convers Manag 2020;205:112339. https://doi.org/10.1016/
which can be a scope for future research. For instance, the capital j.enconman.2019.112339.
cost associated with the renewable generation has not been [14] Jalali MM, Kazemi A. Demand side management in a smart grid with multiple
electricity suppliers. Energy 2015;81:766e76. https://doi.org/10.1016/
considered in this case. Environmental sustainability assessment of j.energy.2015.01.027.
this integrated model can be of future research interest. Apart from [15] Ferruzzi G, Graditi G, Rossi F, Russo A. Optimal operation of a residential
these, the sensitivity of the model was performed considering only microgrid: the role of demand side management. Intell Ind Syst 2015;1:
61e82. https://doi.org/10.1007/s40903-015-0012-y.
swarm size and iteration number; however the effects of acceler- [16] Sheen JN. Economic profitability analysis of demand side management pro-
ation coefficients (i.e., cognitive and social parameter), velocity gram. Energy Convers Manag 2005;46:2919e35. https://doi.org/10.1016/
clamping-limit and inertia weight can be further investigated. j.enconman.2005.02.005.
[17] Niu J, Tian Z, Zhu J, Yue L. Implementation of a price-driven demand response
in a distributed energy system with multi-energy flexibility measures. Energy
Authorship Contribution Statement Convers Manag 2020;208:112575. https://doi.org/10.1016/
j.enconman.2020.112575.
[18] Tascikaraoglu A, Boynuegri AR, Uzunoglu M. A demand side management
Eity Sarker: Conceptualization, Methodology, Validation, strategy based on forecasting of residential renewable sources : a smart home
Investigation, Writing - Original draft, Writing - Review & system in Turkey. Energy Build 2014;80:309e20. https://doi.org/10.1016/
editing. Mehdi Seyedmahmoudian: Conceptualization, Resources, j.enbuild.2014.05.042.
[19] Sarker E, Halder P, Seyedmahmoudian M, Jamei E, Horan B, Mekhilef S, et al.
Supervision, Writing - Review & editing. Elmira Jamei: Resources, Progress on the demand side management in smart grid and optimization
Writing - review & editing. Ben Horan: Resources, Writing - review approaches. Int J Energy Res 2020. https://doi.org/10.1002/er.5631. er.5631.
& editing. Alex Stojcevski: Resources, Supervision, Writing - Review [20] Sarker E, Mehdi Mekhilef SS, Stojcevski A. Optimal scheduling of appliances in
smart grid environment using BPSO algorithm. Energy Proc 2019:1e4.
& editing. [21] Gottwalt S, Ketter W, Block C, Collins J, Weinhardt C. Demand side
management-A simulation of household behavior under variable prices. En-
Declaration of competing interest ergy Pol 2011;39:8163e74. https://doi.org/10.1016/j.enpol.2011.10.016.
€
[22] Ozkan HA. Appliance based control for home power management systems.
Energy 2016;114:693e707. https://doi.org/10.1016/j.energy.2016.08.016.
The authors declare that they have no known competing [23] Bharathi C, Rekha D, Vijayakumar V. Genetic algorithm based demand side
financial interests or personal relationships that could have management for smart grid. Wireless Pers Commun 2017;93:481e502.
https://doi.org/10.1007/s11277-017-3959-z.
appeared to influence the work reported in this paper. [24] Srinivasan D, Rajgarhia S, Radhakrishnan BM, Sharma A, Khincha HP. Game-
Theory based dynamic pricing strategies for demand side management in
smart grids. Energy 2017;126:132e43. https://doi.org/10.1016/
Acknowledgement
j.energy.2016.11.142.
[25] Singh S, Singh M, Chandra S. Feasibility study of an islanded microgrid in rural
This work was supported by the School of Software and Elec- area consisting of PV, wind, biomass and battery energy storage system. En-
ergy Convers Manag 2016;128:178e90. https://doi.org/10.1016/
trical Engineering, Swinburne University of Technology, Mel-
j.enconman.2016.09.046.
bourne, Australia. The first author is indebted to the Swinburne [26] Quang P, Scully T, Breen M, Murphy MD. Facilitating high levels of wind
University of Technology for the TFS scholarship. penetration in a smart grid through the optimal utilization of battery storage
in microgrids: an analysis of the trade- offs between economic performance
and wind generation facilitation. Energy Convers Manag 2020;206:112354.
References https://doi.org/10.1016/j.enconman.2019.112354.
[27] Kaygusuz A. Closed loop elastic demand control by dynamic energy pricing in
[1] Halder PK, Paul N, Joardder MUH, Sarker M. Energy scarcity and potential of smart grids. Energy 2019;176:596e603. https://doi.org/10.1016/
renewable energy in Bangladesh. Renew Sustain Energy Rev 2015;51: j.energy.2019.04.036.
1636e49. https://doi.org/10.1016/j.rser.2015.07.069. [28] Li B, Roche R. Optimal scheduling of multiple multi-energy supply microgrids
[2] Halder PK, Paul N, Beg MRA. Prospect of pongamia pinnata (karanja) in considering future prediction impacts based on model predictive control.
Bangladesh: a sustainable source of liquid fuel. J Renew Energy 2014. https:// Energy 2020;197:117180. https://doi.org/10.1016/j.energy.2020.117180.
doi.org/10.1155/2014/647324. [29] Quiggin D, Cornell S, Tierney M, Buswell R. A simulation and optimisation
[3] Tazi K, Abbou FM, Chaka AB, Abdi F. Modeling and simulation of a residential study: towards a decentralised microgrid, using real world fluctuation data.
microgrid supplied with PV/batteries in connected/disconnected modes d Energy 2012;41:549e59. https://doi.org/10.1016/j.energy.2012.02.007.
case of Morocco. J Renew Sustain Energy 2017;9:1e15. https://doi.org/ [30] Palma-Behnke R, Benavides C, Aranda E, Llanos J, S aez D. Energy management
10.1063/1.4979355. system for a renewable based microgrid with a demand side management
[4] Zunnurain I, Maruf M, Rahman M, Shafiullah G. Implementation of advanced mechanism. IEEE Symp. Comput. Intell. Appl. Smart Grid 2011:1e8. https://
demand side management for microgrid incorporating demand response and doi.org/10.1109/CIASG.2011.5953338.
home energy management system. Infrastructure 2018;3:50. https://doi.org/ [31] Shen J, Jiang C, Liu Y, Qian J. A microgrid energy management system with
10.3390/infrastructures3040050. demand response for providing grid peak shaving. Elec Power Compon Syst
[5] Halder P, Azad K, Shah S, Sarker E. Prospects and technological advancement 2016;44:843e52. https://doi.org/10.1080/15325008.2016.1138344.
of cellulosic bioethanol ecofuel production. In: Azad K, editor. Adv. Eco-fuels a [32] Kumar Nunna HSVS, Doolla S. Energy management in microgrids using de-
sustain. Environ. Elsevier Science & Technology; 2019. p. 211e36. mand response and distributed storage - a multiagent approach. IEEE Trans
[6] Osorio GJ, Shafie-khah M, Lotfi M, Ferreira-Silva BJM, Catala ~o JPS. Demand- Power Deliv 2013;28:939e47. https://doi.org/10.1109/TPWRD.2013.2239665.
side management of smart distribution grids incorporating renewable energy [33] Philippou Nikolas, Venizelou V, Makrides G, Hadjipanayi M, Efthymiou V,
sources. Energies 2019;12:143. https://doi.org/10.3390/en12010143. Georghiou GE. PV integration and price-based Demand Side Management:
[7] Nazir R, Laksono HD, Waldi EP, Ekaputra E, Coveria P. Renewable energy optimum time-of-use tariffs. In: Proc 33th European Solar Energy Conference
sources optimization: a micro-grid model design. Energy Procedia 2014;52: and Exhibition; 2016.
316e27. https://doi.org/10.1016/j.egypro.2014.07.083. [34] kai Feng Z, Liu S, jing Niu W, shan Li S, jun Wu H, yang Wang J. Ecological
[8] Longe OM, Ouahada K, Rimer S, Harutyunyan AN, Ferreira HC. Distributed operation of cascade hydropower reservoirs by elite-guide gravitational
demand side management with battery storage for smart home energy search algorithm with Le vy flight local search and mutation. J Hydrol
scheduling. Sustain 2017;9. https://doi.org/10.3390/su9010120. 2020;581:124425. https://doi.org/10.1016/j.jhydrol.2019.124425.
[9] Alagoz BB, Kaygusuz A, Karabiber A. A user-mode distributed energy man- [35] Gonzalez de Durana J, Barambones O. Technology-free microgrid modeling
agement architecture for smart grid applications. Energy 2012;44:167e77. with application to demand side management. Appl Energy 2018;219:
https://doi.org/10.1016/j.energy.2012.06.051. 165e78. https://doi.org/10.1016/j.apenergy.2018.03.024.
[10] Olivares DE, Mehrizi-Sani A, Etemadi AH, Can ~ izares CA, Iravani R, Kazerani M, [36] Atwa YM, El-Saadany EF, Salama MMA, Seethapathy R. Optimal renewable
et al. Trends in microgrid control. IEEE Trans Smart Grid 2014;5:1905e19. resources mix for distribution system energy loss minimization. IEEE Trans
https://doi.org/10.1109/TSG.2013.2295514. Power Syst 2010;25:360e70. https://doi.org/10.1109/TPWRS.2009.2030276.
 E. Sarker et al. / Energy 210 (2020) 118602 13

[37] Zakariazadeh A, Jadid S, Siano P. Stochastic multi-objective operational [41] Khan PA, Halder PK, Rahman S. Wind energy potential estimation for different
planning of smart distribution systems considering demand response pro- regions of Bangladesh. Int J Renew Sustain Energy 2014;3:47e52. https://
grams. Elec Power Syst Res 2014;111:156e68. https://doi.org/10.1016/ doi.org/10.11648/j.ijrse.20140303.11.
j.epsr.2014.02.021. [42] Roy S. Market constrained optimal planning for wind energy conversion
[38] Akram U, Khalid M, Shafiq S. Optimal Sizing of a Wind/Solar/Battery Hybrid systems over multiple installation sites. IEEE Trans Energy Convers 2002;17:
Grid-connected Microgrid Optimal sizing of a wind/solar/battery hybrid grid- 124e9. https://doi.org/10.1109/60.986449.
connected microgrid system. IET Renew Power Gener Res 2018;12:72e80. [43] kai Feng Z, jing Niu W, Cheng X, yang Wang J, Wang S, guo Song Z. An effective
https://doi.org/10.1049/iet-rpg.2017.0010. three-stage hybrid optimization method for source-network-load power
[39] Shaban AH, Resen AK, Bassil N. Weibull parameters evaluation by different generation of cascade hydropower reservoirs serving multiple interconnected
methods for windmills farms. Energy Rep 2019;6:188e99. https://doi.org/ power grids. J Clean Prod 2020;246:119035. https://doi.org/10.1016/
10.1016/j.egyr.2019.10.037. j.jclepro.2019.119035.
[40] Azad K, Rasul M, Halder P, Sutariya J. Assessment of wind energy prospect by [44] Ben Guedria N. Improved accelerated PSO algorithm for mechanical engi-
weibull distribution for prospective wind sites in Australia. Energy Procedia, neering optimization problems. Appl Soft Comput J 2016;40:455e67. https://
vol. 160. Elsevier Ltd; 2019. p. 348e55. https://doi.org/10.1016/ doi.org/10.1016/j.asoc.2015.10.048.
j.egypro.2019.02.167.