Energy Reports 8 (2022) 5396–5411

Contents lists available at ScienceDirect

Energy Reports
journal homepage: www.elsevier.com/locate/egyr

Research paper

Multi-objective optimization strategy for home energy management

system including PV and battery energy storage
∗
Ziye Song, Xin Guan , Meng Cheng
School of Energy and Power Engineering, University of Shanghai for Science and Technology, Shanghai, 200093, China

article info a b s t r a c t

Article history: As an important part of the smart grid, home energy management system (HEMS) is closely related
Received 1 December 2021 to power usage, in which consumers are not only the users of the power grid, but also the suppliers
Received in revised form 27 February 2022 of distributed power. In this paper, an intelligent HEMS with three adjustable strategies is proposed
Accepted 7 April 2022
to maximize economic benefits and consumers’ comfort. Besides, a novel objective function mainly
Available online xxxx
focusing on satisfying users’ needs is developed by innovating a tri-objective model concerning
Keywords: different weight coefficients which is integrated into a new algorithm to reach a balance among the
Multi-objective home energy management running costs, peak–valley balance index and the satisfaction index. Moreover, this paper innovates
Optimal load scheduling historical weather data to predict the outcome of photovoltaic power. The benefits of HEMS with
Real time tariff photovoltaic and battery energy storage system which is beneficial for peak load shaving and grid
Peak load reduction
stability utilizing three different strategies are analyzed by two different scenarios. The simulation
Genetic algorithm
results demonstrate that the proposed model has highly cut the expenditure of electricity bill by 39.81%
and maintain the balance of the power grid by reducing peak load by 50.37%, meanwhile, the index
of users’ comfort improves about 1.6 times.
© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction The HEMS is an interface to demand response (DR) programs

used by end users, which helps end users in a time-varying
1.1. Motivation energy price environment to solve a scheduling problem that
considers the generation of renewable energy, the demand for
In the past decades, worldwide energy consumption has risen each device in the household, battery storage capacity, and grid
sharply, which has brought about a variety of devastating prob- constraints. The goal of the scheduling problem is to achieve the
lems like energy shortage and environmental pollution (Cao et al., minimum energy consumption cost in the face of uncertainty in
2016). Renewable energy which has less impact on the environ- supply and demand and electricity prices without reducing the
ment can be integrated into a smart grid to alleviate pollution and comfort of end users. It is worth mentioning that in addition
to the impact on the environment, power distribution system
global warming problems. In addition, it has huge potential and
operators also pay special attention to the need to reduce op-
relatively low operating and maintenance costs (Li et al., 2021).
erating costs and improve network reliability, so as to provide
Renewable energy resources (RESs) are growing rapidly in the
users with lower emission levels, optimal costs and sufficient
grid due to the decrease in conventional energy resources and the
reliability (He and Giesselmann, 2015). Therefore, the DR program
increase in power demand. It is worth noting that due to the di-
is considered to be a new measure to help countries control elec-
rect dependence of the production of RESs on climatic conditions,
tricity consumption, and is utilized to change the normal demand
there are many uncertainties in the operation of RESs. Therefore,
consumption by reducing or increasing the load according to the
microgrids with RESs and energy storage systems (ESSs) can be
supplier’s situation when there is a shortage or surplus of elec-
effectively integrated to store and manage the additional RESs
tricity. Needless to mention, DR programs cannot singly control
production (Tan et al., 2021). Because RESs are non-dispatchable
the electricity consumption or influence the rate of participants
and exhibit poor load following especially in residential power
to reduce the demand. In most of the cases, consumers do not like
consumptions, the successful implementation of HEMS can get
or cannot spend their time to calculate and analyze their power
through these restricts.
consumption and schedule the running time of their appliances
in order to save their money (Faruqui and Sergici, 2010). DR
∗ Corresponding author. programs have been used extensively in recent years for demand
E-mail address: [email protected] (X. Guan). curve modification. Furthermore, by modifying the load peak and

https://doi.org/10.1016/j.egyr.2022.04.023
2352-4847/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

systems, and scheduling systems, the stability of the grid is im-

Nomenclature proved and operating costs and emissions are reduced to some
extent, which is in line with sustainable city policies and im-
Full scripts proves the level of sustainable urban development (Al-Ghussain
et al., 2020).
Xi,j Operating status of appliance i at time
instant j.
total 1.2. Literature review
Xday Total state matrix of the appliance.
ti Opening time of appliance i.
The methods that seek to integrate smart home into DR pro-
Li Permitted operating time of appliance i. grams on the purpose of peak demand reduction and system load
ai , b i Upper and lower opening time limits curve modification become even more important. Therefore, in
for appliance i. recent years, these programs have played an important role in en-
PVj Power of photovoltaic panels at time ergy management programs and numerous researches have been
instant j. carried out on HEMS considering DR programs. In this regard,
Sarray Total area of photovoltaic panels. Ref. Zeng et al. (2018) innovatively proposes a low-carbon real-
η Photoelectric conversion efficiency. time electricity price (LCEP) mechanism, which effectively reflects
Tj Average photovoltaic panels tempera- the external characteristics of renewable energy, and an energy
ture at time instant j. management optimization strategy based on mixed linear integer
Ij Total irradiance at time instant j. programming is proposed to intelligently control all flexible re-
SoCj Battery state of charge (SOC) level at sources in the family. Making full use of the flexibility of demand
time instant j. side resources to integrate with the grid brings economic benefits
on the user side owing to the overall lower power price. The
Chargej , Dischargej Battery charge and discharge rate at
results substantiate that the LCEP mechanism not only provides
time instant j.
a good signal to alleviate RESs restrictions but also maximizes
kd Battery self-discharging rate.
the profits of demand side with RESs grid integration. A type
Xj Battery state of charge at time instant j. of residential complexes (RCs) integrates with distributed en-
Xjtotal Matrix of battery state of charge at each ergy resources generation system and electric vehicles (EVs) is
time slot. proposed in Ref. Barhagh et al. (2020). Operators of RC energy
γ1 ,γ2 ,γ3 The weight coefficient of the three systems equipped with local generator sets can benefit from a DR
optimization objectives. program to reduce operating costs. The problem of optimal oper-
Pprice Total real-time power price matrix. ation of a grid-connected EV energy system is examined in the
Kj Real-time power price at time instant j. presence of real-time pricing of a DR program. In this reference,
Ln Power consumption matrix of the load. a mixed-integer linear programming (MILP) model for optimal
Loadj Total load demand of consumers at time operation of energy systems integrated into RCs is presented.
instant j. The simulation results indicate that the total expected cost of
Gj Actual obtained electricity from the grid RCs is reduced by 37.31%. In Ref. Wang et al. (2021), a heuristic
at time instant j. method has been exploited for load demand management based
m′j Demand electricity obtained from the on the generated power and the appraised market-clearing price.
grid at time instant j. The model takes the utilization of renewable energy resources
and the fluctuation of the power grid into consideration and the
N The running costs of the day.
results demonstrate that demand response has highly cut the
G′j Calculated buy or sell electricity at time
expenditure of electricity bill and maintain the balance of the
instant j.
power grid. The power system load serving entities pay special
N′ Calculated running costs of the day. attention to voltage stability improvement and valley fillings
Sai The satisfaction index of users. and peak shavings of load demand profile. A novel methodol-
ta,i Actual opening period for appliance i. ogy in Ref. Gupta and Verma (2021) is proposed for residential
Satmax The total maximum satisfaction index. consumers using Battery Energy Storage Systems (BESSs) as a
σm The peak–valley balance index. key component for demand side management. The results show
mi Obtained electricity from the grid at that centralized control of power dispatch from energy storage
time instant i. devices minimizes the use of node voltages and makes a great
maverage Average power purchase in a day. contribution to load curtailment optimization and social welfare
Tarmax Overall optimization objective. maximization. A simulation model that reflects the flexibility
requirements of power system and the assignment of Time-of-
σm,max Maximum variance of real-time power
use (TOU) pricing to maximize the profits of power suppliers
purchase profile.
and the stability of the grid is described in Ref. Kumamoto et al.
(2020). The results indicate that it is profitable to adopt TOU
pricing for energy costs reduction by obtaining flexibility from
users. In Ref. Teki et al. (2021), a solar photovoltaic system with
transferring a part of the load from the peak hours to the off-peak
battery storage is modeled, which is beneficial for peak load
hours, DR programs reduce the demand during the peak hours,
shaving and grid stability utilizing particular control and energy
and subsequently increase system reliability and reduce the op-
management. This energy management contributes to integrate
eration costs (Shakeri et al., 2017). Therefore, HEMS with smart solar photovoltaic power into the grid network and reduces peak
controllers can independently respond to changing environments loads about 47.14% and the peak efficiency of the system scales
without human intervention. Generally speaking, by increasing between 90% and 94%. A two-stage stochastic programming has
the penetration of smart homes in the power distribution sys- been presented in Ref. Zeynali et al. (2020) targeting on reducing
tem and integrating renewable energy systems, energy storage the electricity procurement cost of normal residential households.
5397
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

In this study, a proper analytical battery degradation cost model most commonly used objective, mainly due to its family demands
according to cost function is proposed and future battery storage and ease of utilization. Maximizing the comfort or well-being of
cost reductions on the HEMS is investigated which results in total users, optimizing processing to reduce peak demand load profiles,
economic benefits. and minimizing greenhouse gas emissions or maximizing the use
There are several recent research papers on the home energy of renewable energy are other main goals adopted. Therefore, in
management (HEM) strategies. A water filling energy distributive order to realize a compound HEMS scheduling model, compre-
algorithm-based HEMS is proposed in Ref. Rajendhar and Jeyaraj hensively considering the aspects of technology, economy and
(2020), in which minimization of total electrical energy costs environment objectives, a multi-objective optimization method
are considered as the main objective and the index of comfort, should be utilized (Costanzo et al., 2012; Qayyum et al., 2015;
the stress of the battery and the main grid are also included. Vilar and de Mattos Affonso, 2016).
The structural properties associated with the optimal control of The objective functions studied in the previously literatures
a HEMS is developed in Ref. Langer and Volling (2020). In this are mainly concentrated on coping with one, such as cost re-
model, a representative building with an air-sourced heat pump, duction, environment protection, and user comfort improvement.
a photovoltaic system, a battery storage system, and thermal This paper poses a compound integrated multi-objective opti-
storage systems is considered. Based on optimal control and mization function to take weights and integrate them into one
considering seasonal effects, the target states of charge (SOC) of objective function. In addition, one of the main contributions of
the storage systems can be derived and the solution time can be this paper is the integration of predicted real-time generated PV
effectively reduced by utilizing the approach of rolling horizon. power for scheduling, while in the previous literatures, there is
In Ref. Khemakhem et al. (2019), a double layer supervision no photovoltaic power taken into account and lacking of utilizing
strategy is proposed to flatten power load profile in a residential photovoltaic prediction data for prescheduling. Though there are
scenario with the method of implementing a demand response many works in the literature, most of the HEM works neither
algorithm and Plug-in Electric Vehicle power management, which consider a photovoltaic integration nor include the storage of sur-
contributes to flatten the fluctuations in demand and achieve plus energy. This paper proposes a combined strategy of battery
the near optimal power. An interactive building power demand energy storage integration to solve the problem that RESs are
management strategy based on genetic algorithm is adopted to non-dispatchable and exhibit poor load following especially in
facilitate the optimization of the grid in Ref. Xue et al. (2014). residential power consumptions.
In this paper, building thermal masses are also integrated into In this paper, three different strategies are presented con-
the model of the building, which contributes to a significant sidering HEM scheduling, photovoltaic integration, and battery
relief of grid power fluctuations. An article Ref. Elma and Se- energy storage integration. A mixed integer nonlinear program-
lamogullari (2015) mainly focused on reducing peak demand ming model integrated into HEMS for optimal operation of energy
and residential appliances shifting is developed by innovating systems is presented. Moreover, a real-time price-based DR pro-
the method of voltage control which is integrated into a new gram is used, and historical weather data is utilized as the input
HEM algorithm to increase the energy efficiency. A hybrid al- of weather prediction model to predict the outcome of photo-
gorithm, Genetic Binary Particle Swarm Optimization (GBPSO), voltaic power. Finally, the main contributions of this paper are
is innovatively proposed in Ref. Javaid et al. (2017), the results as follows:
show that GBPSO based HEM controller has a good performance • Providing a tri-objective and three-in-one objective opti-
on cost reduction and Peak-to-Average Ratio curtailment. In Ref. mization framework for day-ahead energy management of
Shakeri et al. (2018) a novel control algorithm for the HEMS household considering smart homes.
considering battery storage system is demonstrated to apply in • Providing three different strategies considering scheduling
any conventional houses, which achieves up to 15% of electricity of residential appliances, electricity dispatch based on PV in-
cost reduction and maximization of comfort index. In Ref. Shirazi tegration, and energy storage integrated power dispatching.
and Jadid (2015) represents a technique which can conduct opti- • Providing photovoltaic prediction data with deep learning
mal scheduling of residential energy consumption automatically for the input of the optimization scheduling.
resulting from home automation system shortage and lack of re- • Investigating the effect of different weights of objectives on
sponse to time-varying pricing of users. The results demonstrate the peak–valley balance index, the running costs and the
that the model proposed and an appropriate scheduling load con- satisfaction index.
sumption can reduce energy costs and ensure residents’ comfort • Investigating the effect of battery energy storage system on
as much as possible. A heuristic Forward–Backward Algorithm the operating time of residential appliances and peak–valley
(F-BA) in minimization of energy costs of thermal appliances balance index.
considering comfort index is proposed in Ref. Golmohamadi et al.
(2019). This paper also presents a Peak Flatten Scheme in re- 2. Problem formulation
sponse to peak demands that emerged in the power system
and optimizes the power distribution on the bus by shifting the The problem of obtaining the optimal power matching and
power load from the weak lines to alleviate congestion, which price revenue in HEMS which efficiently integrates power gener-
contributes to economic benefits and the stress of the power grid ation, energy storage option, and electrical energy consumption
relief. or load is addressed. The proposed HEMS manages the devices
by regulating the loads that cannot be arbitrary interrupted, re-
1.3. Contribution garding the same kind of electrical appliances that operates at
multiple periods as different loads, and converting thermostati-
Recently, in addition to economic goals, HEMS have pursued cally controlled loads (TCLs) into flexible loads (FLs) at the same
other goals such as reducing grid fluctuations, improving user sat- time. The opening time range of loads is mainly determined by
isfaction, reducing environmental impact, and so on. It should be comfortable temperature range for users. Whereas, users can also
noted that these goals affect and restraint each other. For exam- set the opening time freely according to actual needs, which is
ple, When the HEMS is performing optimal scheduling, the HEMS more user-friendly.
model adopts different objectives according to the perspective The HEMS problem is, therefore, to find one or a set of the best
of the decision maker. Minimization of the running costs is the (according to one or multiple objectives) strategies that decide on
5398
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

Table 1
Parameter settings for three strategies and two scenarios.
Parameter name Parameter value
Number of parents 200
Number of offspring 200
Number of generations 300
Gene mutation probability 0.1
Crossover probability 0.9

supply source (grid) and energy storage equipment (battery). The

HEMS is responsible for managing the hardware and the energy
flows between the home energy components. All the components
involved in this paper are addressed as follows.
Electrical load. Appliances are the main energy consumption
in the HEMS, therefore, particular regulation is of great impor-
tance to home appliances. After the appliance is turned on, the
operation cannot be interrupted until the operation is satisfied.
All equipment will not be interrupted by process, therefore, under
the premise that the required working time of the equipment is
Fig. 1. The overall schematic for managing the electrical energy in a smart home. known, as long as the opening period is determined, the working
status of the day will be determined. Mathematical formulation
of the proposed model is shown as follows:
how much energy to buy from or sell to the grid or how much
0, / [ti , ti + Li ]
{
j∈
energy to store in the battery (Zupančič et al., 2020). Xi,j = (1)
The overall schematic that illustrates the problem of managing 1, j ∈ [ti , ti + Li ]
the electrical energy in a smart home is presented in Fig. 1. The ⎡
X1,1 ··· ··· ··· X1,48
⎤
historical weather data is first retrieved and used to predict the
⎢ .. .. .. .. .. ⎥
photovoltaic power generation data which is used as an input ⎢ . . . . . ⎥
total
for the optimization procedure every half hour in the next day. Xday = ⎢ Xi,1 ··· Xi,j ··· Xi,48 ⎥ (2)
⎢ ⎥
The user or the HEMS operator then chooses a solution strategy
⎢ . .. .. .. .. ⎥
⎣ . .
. . . .
⎦
(regarding the specific working period and rated power of the
electrical appliances participating in the scheduling in the next Xn,1 ··· ··· ··· Xn,48
day, and optimization objectives) that is preferred by the user, 1 ≤ ai ≤ bi ≤ 48, ti ∈ [ai , bi ] (3)
which is then uploaded to the HEMS central unit and utilized to
manage the energy flows within the system. Finally, the corre- Here, Eqs. (1) and (2) define the operating status of a single
sponding algorithm is activated to obtain the specific schedule device and the total state matrix of the devices in a day, where
for electricity consumption in the next day and carried out. In this subscript i denotes the ith device and subscript j denotes the jth
paper, 24 h are divided into 48 time slots in a day, namely, half time interval. In this paper, 48 time slots are considered in a day
an hour for each time slot. And the day begins at 24:00 lasting which means HEM is implemented for every 30 min slot. Eqs. (3)
for 24 h, therefore, time slot at 0 corresponds to the time at ensures that the upper limit and lower limit of the ith device
24:00, time slot at 24 corresponds to the time at 12:00, and so (ai and bi , respectively) are under the time slots limit. Where ti
on. According to the RTP tariff, the electrical energy consumption denotes the opening time, and Li denotes the operating time that
charges for different time slots are given in Fig. 2. permitted.
Photovoltaics. A regression model (Muzathik, 2014) is used to
2.1. Optimization runs determine the energy generation for the given weather conditions
and solar panels with fixed installation angle. Given the total area
The optimization was performed using a workstation with an of the photovoltaic panels (Sarray ), the power output (PV j ), where
Apple Silicon M1 CPU @3.2 GHz with 16 GB of RAM. Python subscript j denotes the jth time interval, can be approximated as
(Python) was used as the programming language and the Geatpy follows:
library (Geatpy) was used as the optimization framework. The
PVj = ηSarray Ij 1 − 0.005 Tj + 25
( ( ))
(4)
data preprocessing was implemented using the NumPy (van der
Walt et al., 2011) and Pandas (McKinney) libraries. where η denotes photoelectric conversion efficiency of solar pan-
Due to the resource-intensive experiment, the parameter set- els, Ij denotes the total irradiance, where subscript j denotes the
tings were chosen based on a smaller set of preliminary runs. jth time interval, Tj denotes the average photovoltaic panels tem-
As a result, the parameter settings for two scenarios and three perature. Within the same day, the inclination and the physical
strategy optimizations specified in Tables 1 and 5, respectively, properties of the panel will not change, so η and Sarray can be
were set. Each optimization was executed 40 times in order to regarded as a constant. Whereas, the correlation coefficient of Tj
obtain representative results. is far less than 1, and only Ij and Tj are the influence variable of
PV j . There is a strong linear relationship between PV j and Ij .
2.2. Model Battery. The battery’s state of charge at the end of the jth
time interval (SoC j ) is calculated from the previous state of charge
The home energy management system, an intelligent network (SoC j−1 ), the battery charge (Chargej ), the discharge (Dischargej ),
control system based on the smart grid, comprises components and the self-discharging rate kd :
that are energy generation equipment (distributed photovoltaic
modules, wind generators), energy consumption (load), energy SoCj = kd · SoCj−1 + Chargej − Dischargej (5)
5399
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

Fig. 2. Details of RTP tariff scheme for different time slots.

HEMS include running costs, CO2 emission, maximum peak load,

1, Charging aggregated energy consumption and battery life expectancy. The
⎧
⎪
⎨ objectives considered in this paper are customer satisfaction, grid
Xj = 0, Not working (6) balance and the running costs.
⎪
⎩
−1, Discharging The running costs (N) comprise the cost of buying and profit
of selling the electric energy during each time interval. G′j is ob-
Xjtotal = [X1 , X2 , . . . , X48 ] (7) tained by multiplying the values less than 0 in Gj by 0.5 (surplus
electric energy sells for half of the real-time power purchase
Here, Eqs. (6) defines the battery states of charge and dis-
price). The running costs of the day is calculated in Eqs. (12):
charge. During each time interval, the battery has three different
states and each state runs independently. Further, the energy can N = G′j × Pprice (12)
be transferred between the grid and the battery. Eqs. (7) indicates
Introduce the upper limit of electricity consumption per unit
the charge and discharge states of the battery for 48 time slots.
time (limit), and penalize the strategies that do not meet the
The smart grid sends the price signal to the HEMS. Pprice
upper limit of electricity consumption. The actual running costs
denotes the total real-time power price matrix in Eqs. (8), and
of the day (N ′ ) are defined in Eqs. (13):
real-time power price at time instant j is denoted with Kj . The
Limit
power consumption matrix of the load is denoted with Ln in
N,
⎧
Eqs. (9). Consumers aggregate load demand at time instant j is ⎪ Electricity consumption at all times is less than
denoted with Loadj that is calculated from the multiply between
⎪
⎪
or equal to limit
⎪
total
⎨
Ln and Xday in Eqs. (10): N′ =
⎪ N + 99, Electricity consumption at a certain moment is
Pprice = K0 , . . . , Kj , . . . , K47
[ ] ⎪
(8) ⎪
⎪
⎩
greater thanlimit
Ln = [P1 , . . . , Pn ] (9) (13)
total
Loadj = Ln · Xday (10) The satisfaction index (Sat) refers to the impact on users
caused by the change of residential load operating time before
The HEMS prioritizes the use of photovoltaic modules to sup-
and after optimization. The more operating time changes, the
ply the power to the customers when the strategy is deployed.
lower comfort of residents. It comprises the allowable open-
The smart grid receives the electric energy from the grid when
ing period of the device and the actual opening period of the
the photovoltaic power is insufficient to meet the needs of users. device:
Conversely, the HEMS sells the remaining electric energy to the ⎧
grid at the real-time power sale price(half of the real-time power ⎨ bi − ta,i , t ∈ [a , b ]
a ,i i i
purchase price) when there is a surplus of photovoltaic power. Sai = b i − ai (14)
0, else
⎩
Eqs. (11) indicates the actual amount of electricity that needs to
be obtained from the grid, where m′j denotes the electric energy where bi and ai denote the upper and lower limits of the opening
that needs to be obtained from the grid at time instant j, when period of the device respectively, and subscript i indicates ith de-
m′j < 0 indicates that the HEMS has surplus photovoltaic power vice. ta,i denotes the actual opening period of the device. Eqs. (15)
sold to the grid during this period: defines the total maximum satisfaction index model:
Gj = Loadj − PVj = m′1 , . . . , m′j , . . . , m′48
[ ] n
(11) ∑
Satmax = Sai (15)
2.3. Objectives i=1

Here, the device is turned on in the allowable opening period

Usually, the electricity bill of the day is set to be the only at the earliest time when ta,i equals ai and the value of the
optimization objective to seek the greatest economic benefits satisfaction index is 1. In contrast, the value of the satisfaction
for users (Zhou et al., 2014; Missaoui et al., 2014). The future index is 0 when the device is turned on outside the allowable
humanized home energy management system should also take opening period, to which the minimum value of the satisfaction
the needs of users into account in all aspects. At the same time, index reaches.
the impact on the environment and the fluctuation to the grid The peak–valley balance index (σm ) is used to measure the
have to be taken into consideration. Often-used objectives in degree of grid fluctuation which is significant to shave the peak
5400
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

and fill the valley to the load of power purchasing profile in The system architecture of the proposed HEM model is de-
the contemporary urban power grid system. With the growing picted in Fig. 3. Smart meter is an advanced metering device,
demand for electricity user side, power generators also increase which is based on advanced communication, computer and mea-
in order to fully meet the electricity demand of the load side. surement technology to collect, analyze and manage electrical
If there is a mismatch between power supply and demand, it energy information data. In addition to the functions of power
will cause huge power equipment investment and operational consumption recording and power transmission, smart meters
redundancy. The demand-side optimization algorithm balances can be used to measure some physical quantities, such as tem-
the load profile, which can effectively reduce power fluctuations perature, humidity, and light intensity. At the same time, it also
in the grid, realize the flexibility of electric load and improve the collects the specific energy consumption of each residential appli-
flexibility of energy regulation. The peak–valley balance index is ance. Taking the access of distributed photovoltaic modules into
defined in Eqs. (16): account, smart meters also have bidirectional multi-rate metering
functions and user-side control functions. The central control
Σi48
=1 (mi − maverage )
2
platform is an intelligent computing framework built into the
σm = (16)
48 user interaction terminal, which aims to manage and optimize
where σm denotes the variance of real-time power purchase energy utilizing proposal. It also receives the information from
profile, mi denotes the amount of electricity that needs to be the smart meters and adopts the scheduling mechanism which
obtained from the grid, subscript i indicates ith time, and maverage arranges the work sequence of the smart residential appliances
denotes average power purchase in 48 periods of the day. The participating in the scheduling through optimized algorithm cal-
lower σm indicates better peak–valley balance. culations. The battery energy storage system (BESS) in the home
Further, when discussing the balance of the power profile energy management system can store photovoltaic power that
of home energy management system with storage batteries, the cannot be consumed in real time, and improve the utilization of
reference should be the power purchase profile. renewable energy; on the other hand, it can adjust the charging
and discharging strategy to buy electricity during the low electric-
2.4. Multiple objective optimization model ity demand period and use electricity during the peak period to
alleviate the regional electricity load and make economic profits
for users.
Multiple objective optimization model is proposed to evalu-
As a part of HEMS, smart residential appliances can be inte-
ate the performance of the strategies according to the specific
grated with smart meters, and central control platform to com-
objectives that mentioned above. In this model, all optimization
plement smart residential functions. The residential appliances
objectives are normalized, the weight of which is determined by
are classified into three groups based on their inherent operating
the weight coefficient. The objective function Tar max is defined
characteristics. Those are rigid loads (RGLs), flexible loads (FLs)
in Eqs. (17):
and temperature-controlled loads (TCLs). RGLs such as lights,
Sat σm N and refrigerators, etc. have fixed time and power consumption
Tarmax = γ1 − γ2 − γ3 (17)
Satmax σm,max Nmax patterns in conformity with consumers’ needs. The scheduling of
RGLs will bring a great negative impact on the user experience,
where Satmax denotes the maximum value of satisfaction index
whereas which have influence on the maximum limits for resi-
which reaches to the peak when all flexible load devices are
dential power and have to be counted in the real-time electricity
turned on at the time when users are most satisfied, σm,max
consumption. So, the RGLs are considered in the optimal load
denotes maximum variance of real-time power purchase profile,
scheduling. FLs such as dishwashers, washing machines, televi-
Nmax is the maximum value of the running costs. γ1 ,γ2 ,γ3 denote
sions etc. have continuous power consumption patterns but these
the weight coefficient of the three optimization objectives respec-
can be shifted to preferable time slots. Once the flexible loads are
tively, the value of which is between 0 and 1, and the sum of
turned on, they will maintain stable operation at the rated power
them is 1. It can be set freely according to actual needs when
and cannot be stopped until the required working time is reached.
conducting different optimization steps.
Therefore, FLs can be scheduled by controlling the opening time.
TCLs such as air conditioners, electric water heaters, etc. have
3. Strategies for home energy management system flexible power consumption patterns but these can be influenced
by consumers’ needs. According to residential consumers’ daily
With the rapid development of power grid technology, the requirements, they can fix their specific operating time periods
concept of the Internet of Things (IoT) has been integrated into regarding to suitable temperature ranges. Therefore, TCLs can be
modern buildings to meet the electricity demand of residents in transformed into FLs to improve the satisfaction index at the
terms of comfort, safety, and efficiency (Amin and Wollenberg, sacrifice of increasing the running costs.
2005). The IoT promotes the deployment of smart sensors and The proposed HEM model is validated in this section with
other advanced metering devices, making it possible to remotely three different strategies studies and two different scenarios for
interact with electricity consumption information and to monitor each strategy. In each strategy, the simulation results of HEM
and drive residential appliances. From the perspective of the de- model are compared with normal model without optimization.
mand side, HEMS conforms to the intelligent process of the power In this paper, a multi-objective mixed integer nonlinear (MINLP)
grid. It can interact with residential equipment and public utility problem method is performed using Python with genetic algo-
equipment and adjust the equipment operation schedule to re- rithm to solve the proposed HEM model.
spond to external information. Meanwhile, HEMS has achieved
the access to renewable energy power generation, allowing users 3.1. Optimization algorithm
to use photovoltaic power to meet their own daily electricity
demands or sell electricity for profit. HEMS is mainly composed In this paper, the genetic algorithm framework based on
of smart meters, smart appliances, user interaction terminals, Python is selected for optimization calculation. Geatpy is a high-
central control platforms, distributed photovoltaic modules and performance Python genetic algorithm library and open evo-
battery energy storage systems, and implements its functions lutionary algorithm framework. It provides library functions of
through these equipment. many implemented genetic and evolutionary algorithm related
5401
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

Fig. 3. The architecture of home energy management system.

Table 2
The implementation of the optimization algorithm for the optimization of the HEMS strategy.

operators, such as population initialization, selection, crossover, algorithm template used in this optimization work for simulation
mutation, reinsertion, multi population migration, multi-objective calculation.
optimization, etc. It also provides an open evolutionary algorithm
framework to realize diversified evolutionary algorithms. 3.2. Demand response (DR) strategy
According to the above algorithm steps, the chromosome fit-
ness function code is designed in Geatpy, as shown in Table 2. Demand response strategy which is the fundamental structure
The lower the electricity costs, the higher the fitness: is first integrated into the optimal strategy. Operating parameters
According to the fitness function, input various parameters of residential appliances in this strategy are tabulated in Table 3.
and variables, write the specific problem description script, im- Under the circumstance without optimal load scheduling, all elec-
port it into the main framework script of genetic algorithm, and trical appliances start to operate in the earliest constrained time
select SOEA_ EGA_ Templet (single objective genetic algorithm period, and the value of real-time power consumption is shown
template with elite individual reservation) is the evolutionary in Fig. 4. Most of the power consumption behavior of users is
5402
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

Table 3
Operating parameters of residential appliances.
Serial number Residential appliances Opening period Rated power/kW Duration/h
1 Water Heater (WH) 17:00–22:00 3 2
2 Air Conditioner 1 (AC1) 19:00–21:00 1.5 3
3 Air Conditioner 2 (AC2) 11:00–12:00 2 2
4 Air Conditioner 3 (AC3) 17:00–21:00 2 3
5 Rice Cooker 1 (RC1) 11:00–11:30 1 0.5
6 Rice Cooker 2 (RC2) 17:00–17:30 1 0.5
7 Microwave Oven 1 (Oven1) 11:00–11:30 2 0.5
8 Microwave Oven 2 (Oven2) 17:00–17:30 2 0.5
9 Electric Vehicle (EV) 0:00–4:00 2.4 3.5
10 Sweeping Robot (SR) 8:00–23:00 0.7 0.5
11 Dishwasher 1 (DW1) 13:00–16:30 1 0.5
12 Dishwasher 2 (DW2) 19:00–23:00 1 0.5
13 Washing Machine (WM) 7:00–22:30 0.3 0.5
14 Illumination 1 (Light1) 6:00 0.03 2
15 Illumination 2 (Light2) 18:00 0.03 6
16 Refrigerator 0:00 0.6 24

Fig. 4. Real-time power consumption of the day without scheduling.

Fig. 5. Real-time consumption of the day with and without scheduling in strategy(A).

concentrated in the three time periods of 11:00–12:30, 17:00– 3.3. Photovoltaic (PV) integrated strategy
21:30 and 0:00–3:00 in strategy (A), namely, power consumption
at noon, daily power consumption in the evening and the charge It will have a great impact on power consumption during the
of electric vehicles in the early morning. The running costs of day time after integrating photovoltaic modules. The daily power
strategy (A) are 6.44$ calculated. consumption profile and the usage of photovoltaic power without
The power consumption of each time slot with optimal load and with optimal load scheduling are depicted in Figs. 7 and
scheduling which compared with strategy (A) is given in Fig. 5. 8 respectively. Fig. 7 shows that all photovoltaic power is used
The optimal load scheduling of all residential appliances with completely in real time during the period form 11:00 to 12:30,
strategy (A) are shown in Fig. 6. Compared with strategy (A), which alleviates the electricity load during the noon peak period
the charging time of electric vehicles in the early morning after to a certain extent, but the peak during the noon period still
scheduling is postponed to 3:00–6:00, and the power consump- exists. From Fig. 8, it is clear that the power consumption profile
tion during the noon period is postponed to the low electricity during the noon period is smoother, and the peak of the power
price period and the distribution is more even. The peak of power purchase profile during the noon period is also shaved compared
consumption at 11:00 in strategy (A) is shaved and most of the with Fig. 7. The power consumption during the noon period is
power consumption in the evening is switched to the time slots postponed as evenly as possible to the time period when there
after 21:00 which is staggered with the evening electricity price is a photovoltaic power surplus, and the photovoltaic utilization
peak. rate is also increased after scheduling.
5403
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

Fig. 6. The optimal load scheduling of all residential appliances in strategy(A).

Fig. 7. The daily power consumption and the usage of photovoltaic power without scheduling in strategy(B).

Fig. 8. The daily power consumption and the usage of photovoltaic power with scheduling in strategy(B).

When calculating the benefits of photovoltaics, the cost of Compared with two schemes in strategy A, the optimization
photovoltaic power generation needs to be considered. The peak algorithm proposed in this paper has an effective benefit on costs
value of photovoltaic power generation in this paper is 2.3 kW. reduction, and the daily electricity expense decreases from 6.44$
Taking into account the costs of solar panels, installation ma- to 5.06$ after optimization scheduling (a reduction of 21.43%).
terials, supporting instruments, and maintenance fees, the cost Compared with scheme 1 and scheme 3, although the installation
of photovoltaic power generation is about 2$/W, namely, the cost of the household photovoltaic power generation system is
total cost of the photovoltaic system is 4600$. The life of the high, it can be found that the user’s daily power generation cost is
photovoltaic system is about 20 years, and the calculation shows reduced from 6.44$ to 5.308$ (a reduction of 17.57%). This shows
that the daily cost of the photovoltaic system is 0.63$. that the introduction of photovoltaic power generation in the
The optimal load scheduling of all residential appliances in home energy management system will pay equipment costs, but
strategy (B) are shown in Fig. 9. The running costs and photo- will still bring obvious practical economic benefits. At the same
voltaic utilization rate in strategies above are tabulated in Table 4. time, it can be found in comparison with the top three schemes.
5404
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

Fig. 9. The optimal load scheduling of all residential appliances in strategy(B).

Table 4 Table 5
Simulation results of HEM optimization model for four different schemes in two The general battery storage system specifications.
strategies. S. no. Parameter Rating
Strategy Scheme With or With or The Utilization rate 1 Battery capacity 7.2 kWh
number number without without running of photovoltaics 2 Charging efficiency 85%
photo- scheduling costs (%) 3 Discharging efficiency 95%
voltaics ($/day) 4 SOC limit 5%–95%
1 Without Without 6.440 0 5 Charging power limit 0 kW–3 kW
A
2 Without With 5.060 0 6 Discharging power limit 0 kW–3 kW
3 With Without 5.308 56.03
B
4 With With 3.876 66.97

the battery life is 4500 cycles, and the daily cost of the battery is
0.1$.
In scheme 4, when the photovoltaic system is integrated, the The overall schematic that illustrates the specific implementa-
optimized scheduling algorithm will further reduce the power tion mode under the framework of genetic algorithm is presented
cost of users and maximize the saving of users, and the proportion in Fig. 10. The charge and discharge states of the battery for 48
of the running costs is reduced by 39.81% compared with scheme time slots as the novel chromosomes is first retrieved and used
1. It is clear that the optimization scheduling algorithm is used as an input for the optimization procedure. The result of the
to optimize the opening time when integrating the photovoltaic optimization is the fitness. The battery energy storage strategy
system, which will reduce the running costs and maximize the designed in this section is:
savings for the users.
1. Generate the novel chromosomes to be the charge and
3.4. Battery energy storage (BES) integrated strategy discharge states vector of the battery for 48 time slots Xjtotal .
2. Initialize the values of SOC and k, and read Xj according to
As is depicted in Fig. 8, the peak of power consumption is the time sequence to confirm the strategy of charge and
concentrated in the evening, and the peak of photovoltaic power discharge state in this period.
generation is at noon, therefore, the peak of photovoltaic power 3. If Xj is equal to 1, the strategy of discharge is adopted. In
can only meet the demands of peak power consumption at noon, the case of photovoltaic surplus, the photovoltaic power
it is hard to increase the overall photovoltaic utilization rate by supply is given priority to the equipment. If there is no
the measure of scheduling. Considering the issues above compre- photovoltaic surplus power, the battery will be considered
hensively, this section integrates battery energy storage system to supply power to the equipment. The shortfall will be
into the HEMS based on the opening period obtained from the made up by the purchase of electricity from the grid.
optimization results of the previous section in strategy (B), which 4. If Xj is equal to 0, the battery does not work and the pho-
stores the surplus power in time during the peak period of pho- tovoltaic power supply is given priority to the equipment.
tovoltaic power generation at noon and uses it during the peak The shortfall will be made up by the purchase of electricity
period at night to increase the utilization rate of photovoltaics. from the grid.
And the general battery storage system specifications are given 5. Add penalty function to filter and eliminate unreasonable
in Table 5. schemes.
The lithium iron phosphate battery banks are usually preferred
for the home energy storage system applications because of its Further, photovoltaic power has priority over the grid when con-
high battery capacity, excellent performance, life cycle character- sidering the power supply to charge the battery. Additionally,
istics, easy availability, and environment friendly. In this study, when considering the power supply to the equipment, the pri-
the battery storage capacities of all residential households are ority of photovoltaic power is greater than that of the battery.
assumed to be the same which is 7.2 kWh. And the charging ef- And both of them have priority over the grid. Finally, benefiting
ficiency and discharging efficiency are 85% and 95% respectively. from the strategy the photovoltaic power generated in real-time
The SOC limit of the battery will be constrained between 5% and is consumed as much as possible at any period.
95% to ensure the battery safety. Also Charging power limit and The daily power consumption profile and the usage of photo-
discharging power limit are constrained under 3 kW both. In voltaic power of strategy with battery integration are depicted
this way, it can not only ensure that the remaining photovoltaic in Fig. 11. It is clear that the photovoltaic power that is not
power generation can be fully absorbed in real time, but also the consumed in real-time is completely stored by the battery and
battery has the ability to ensure all RGLs and some FLs of users in realizes the full utilization of the power. Due to the abundant
the emergency of external power failure. The battery cost is 500$, output of photovoltaics, except for 11:00, there is almost no
5405
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

Fig. 10. The specific implementation mode under the framework of genetic algorithm.

Fig. 11. The daily power consumption and the usage of photovoltaic power of strategy with battery integration.

electricity purchase from 8:00 to 16:00, which shows that pho- economic benefits. Humanized home energy management system
tovoltaic power combined with battery supply fully meets the should also take the demands of users in all aspects into account
real-time power demand during the day. The strategy makes in the future. First of all, the satisfaction index is considered, the
full utilization of low electricity price period at dawn for power HEMS chooses different opening time slots within the allowable
purchase, electric vehicle charging, and battery use. The advan- operating period of the equipment will make a difference. In
tages of the battery have been fully demonstrated after 18:00 addition, the peak and valley balance of the power purchase
in the evening. The power purchase profile is lower than the profile is considered, with reference to the stability of the grid and
power consumption profile for most of the time except 21:00, power consumption safety, the fluctuation of the power purchase
indicating that the proposed HEM model with battery energy profile should be minimized. From the previous section, how-
storage systems enhances the energy management benefits. The ever, the introduction of suitable battery energy storage system
battery SOC levels in strategy (C) are shown in Fig. 12. can consume all the photovoltaics, so we will not discuss the
photovoltaic utilization rate in this section.
4. Simulation results The proposed HEM model is validated in this section with one
case study and two different scenarios for this case. This case
Most of the previous studies on the optimization of the home analyzes two mathematical models of optimization objectives,
energy management system, the daily electricity bill is set to and a multi-objective optimization model that meets multiple
a single objective optimization for users seeking to maximize needs.
5406
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

Fig. 12. The battery SOC levels in strategy(C).

4.1. Case study the peak–valley balance index is better optimized when imple-
menting the HEM strategy with battery energy storage system
The main aim of this case study is to analyze the peak load integrated. Therefore, the satisfaction index and the running costs
management and cost management for each scenario separately. are optimized merely when scheduling the opening periods at
The battery energy storage system is integrated in this case the first step, and the peak–valley balance index and the running
to obtain best multi-objective optimization results. Scenario 1 costs are optimized merely when the BES strategy is integrated.
represents the balance of power consumption profile and power When scheduling the opening periods of all residential appliances
purchase profile under the circumstance of peak–valley balance in the first step, coefficients (γ1 = 0.66, γ2 = 0, γ3 = 0.34) in
index as the single optimization objective. Multi-objective opti- Eqs. (17) are assigned. The optimal load scheduling of all residen-
mization model according to the running costs, the satisfaction tial appliances with multiple optimization objectives in Scenario
index and peak–valley balance index is considered in Scenario 2. 2 are shown in Fig. 16. The daily power consumption profile and
Scenario 1 tries to minimize the power consumption profile and the usage of photovoltaic power without battery installation in
the power purchase profile by shifting FLs and TCLs from high the first step are depicted in Fig. 17. The opening time periods
load demand periods to low load demand periods. Consequently, of optimization with battery integration in the second step is on
the running costs of this scenario might increase. Scenario 2 is the basis of optimal load scheduling in the first step. Coefficients
a compromised approach which covers three optimization objec- (γ1 = 0, γ2 = 0.66, γ3 = 0.34) are assigned with battery
tives compared to Scenario 1. integration in the second step. The daily power consumption
profile and the usage of photovoltaic power with multi-objective
4.2. Scenario 1 optimization and battery integration are presented in Fig. 18.
The battery SOC levels of Scenario 2 are shown in Fig. 19. As is
When scheduling the opening periods of all residential appli- depicted in Fig. 18, compared with Strategy (C), the fluctuation
ances in the first step, coefficients (γ1 = γ3 = 0, γ2 = 1) in of the power purchase profile is more moderate and the opening
Eqs. (17) are assigned. The optimal load scheduling of all residen- periods of residential appliances have been scheduled, which are
tial appliances with single optimization objective in Scenario 1 is more satisfactory.
shown in Fig. 13. The opening time periods of optimization with
battery integrated in the second step are on the basis of optimal
4.4. Discussion
load scheduling in the first step. The daily power consumption
profile and the usage of photovoltaic power with peak–valley
balance objective in Scenario 1 are depicted in Fig. 14. It is clear The main aim of this paper is to strike a balance among the
that the multiple peaks of the power purchase profile have been objectives of the proposed HEM model simultaneously which
reduced in the early morning. The battery SOC levels in Scenario is tabulated in Table 6. It seems that optimization with a sin-
1 are depicted in Fig. 15. In Scenario 1, the battery has only gle objective might cause losses to other objectives that have
two charge and discharge cycles, therefore, the power purchase not been integrated. For example, when the peak–valley balance
profile in the evening is flatter, and there are no charge and index is used as the single optimization objective, the running
discharge cycles in the early morning when the electric vehicle is costs are 5.802$, which is two times more compared with it that
being charged at a rated power. It is clear that the stationarity of is used as the optimization objective. Considering that there is
power purchase profile has been improved in the early morning a wide range of demands for power consumption strategies in
in Scenario 1. Scenario 1 in case study with single optimization practical applications, it is necessary to establish a multi-objective
objective according to peak–valley balance index deals with the optimization model.
optimization problem of the fluctuation of the grid, this approach There is a certain inverse proportional relationship between
may not meet the optimal aggregated objective function values. the running costs and the variance of the power purchase pro-
That is why authors have implemented the proposed HEM model file. This is because when pursuing low running costs, power
in another scenario which is explained as follows. consumption will be concentrated in the low electricity price
period, resulting in a sharp increase in power consumption dur-
4.3. Scenario 2 ing this period. The integration of battery energy storage and
photovoltaic systems can alleviate the problem to a certain ex-
In this scenario, a multi-objective optimization model accord- tent. The multi-objective model of scenario 2 emphasizes the
ing to the running costs, the satisfaction index and the peak– peak–valley balance index, so the running costs are 78.5% of the
valley balance index is considered. The satisfaction index can maximum value, and the variance is only 40% of the maximum
only be optimized when scheduling the opening periods, and value. It proves that the model responds more sensitively to
5407
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

Fig. 13. The optimal load scheduling of all residential appliances with single optimization objective in Scenario 1.

Fig. 14. The daily power consumption profile and the usage of photovoltaic power with peak–valley balance objective.

Fig. 15. The battery SOC levels in Scenario 1.

Fig. 16. The optimal load scheduling of all residential appliances with multiple optimization objectives in Scenario 2.

5408
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

Fig. 17. The daily power consumption profile and the usage of photovoltaic power without battery integration in Scenario 2.

Fig. 18. The daily power consumption profile and the usage of photovoltaic power with multi-objective optimization and battery integration in Scenario 2.

Fig. 19. The battery SOC levels in Scenario 2.

Table 6
Simulation results of HEM optimization mode for three different conditions.
Strategies Optimization The peak–valley The running The satisfaction
objectives balance index costs ($/day) index
Strategy C N 4.03 2.840 0.30
Scenario 1 σm 1.58 5.802 0.44
Scenario 2 σm , N, and Sat 2.00 5.054 0.79

changes in the peak–valley balance index and the weight coef- the satisfaction index of which are less than 0.5 either. Therefore,
ficient of the running costs. when scheduling the opening periods of residential appliances, it
Most of the predecessors neglected the satisfaction index is necessary to assign a larger weight for the satisfaction index.
when they were optimizing. After single objective optimization The results show that the optimization benefit is more obvious
of the running costs and peak–valley balance index respectively, that the value of satisfaction index in Scenario 2 is increased by
5409
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

163.33% on the basis of Strategy (C). From the overall analysis, Declaration of competing interest
the proposed HEM model in Scenario 2 is the most practical case
compared to all other scenarios and strategies. The authors declare that they have no known competing finan-
cial interests or personal relationships that could have appeared
5. Conclusion to influence the work reported in this paper.

The multi-objective home energy management system with References

three different strategies according to HEM scheduling, photo-
Al-Ghussain, L., Samu, R., Taylan, O., 2020. Sizing renewable energy systems with
voltaic integration, and battery energy storage integration is pro- energy storage systems in microgrids for maximum cost-efficient utilization
posed in this paper for residential consumers. The HEM problem of renewable energy resources. Sustainable Cities Soc. 55, 102059.
with a novel three-in-one objective function is modeled as a Amin, S.M., Wollenberg, B.F., 2005. Toward a smart grid: power delivery for the
multi-objective non mixed integer linear problem, and genetic 21st century. IEEE Power Energy Mag. 3, 34–41.
Barhagh, S., Abapour, M., Mohammadi-Ivatloo, B., 2020. Optimal scheduling of
algorithm is used to solve the complex problem. In this paper, a electric vehicles and photovoltaic systems in residential complexes under
real-time pricing tariff scheme is utilized to engage in the HEMS real-time pricing mechanism. J. Cleaner Prod. 246, 119041.
scheduling model and which prices are given in different time Cao, X., Dai, X., Liu, J., 2016. Building energy-consumption status worldwide and
periods. The proposed problem is implemented to simulate two the state-of-the-art technologies for zero-energy buildings during the past
decade. Energy Build. 128, 198–213.
scenarios selected by users in a practical system. Costanzo, G.T., Zhu, G., Anjos, M.F., 2012. A system architecture for autonomous
With the method of genetic algorithm framework, the home demand side load management in smart buildings. IEEE Trans. Smart Grid
energy management system with three different strategies inte- 3, 2157–2165.
gration is scheduled and calculated. Finally, photovoltaic equip- Elma, O., Selamogullari, U.S., 2015. A new home energy management algorithm
with voltage control in a smart home environment. Energy 91, 720–731.
ment and optimization strategy have brought good economic
Faruqui, A., Sergici, S., 2010. Household response to dynamic pricing of
benefits, compared with the initial scheme, the running cost of electricity: a survey of 15 experiments. J. Regul. Econ. 38, 193–225.
which is reduced by 39.81%. The battery integration strategy can 2021. Geatpy. https://www.geatpy.com/. [Accessed Dec 2021].
recover the excess photovoltaic power in time, and can purchase Golmohamadi, H., Keypour, R., Bak-Jensen, B., 2019. Optimization of household
energy consumption towards day-ahead retail electricity price in home
power in the low-price period. And the benefits of the proposed
energy management systems. Sustainable Cities Soc. 47, 101468.
model with the integration of battery energy storage system Gupta, P., Verma, Y.P., 2021. Voltage profile improvement using demand side
are greatly enhanced by the way of optimizing the charge and management in distribution networks under frequency linked pricing regime.
discharge sequence of the battery based on the optimization of Appl. Energy 295, 117053.
equipment start-up time, which has decreased the cost of the He, M., Giesselmann, M., 2015. Reliability-constrained self-organization and
energy management towards a resilient microgrid cluster. In: 2015 IEEE
electricity bill. The results show that the operation cost of the Power & Energy Society Innovative Smart Grid Technologies Conference
battery integration strategy is reduced by 26.73% based on the (ISGT). pp. 1–5.
photovoltaic integration strategy. From the overall analysis, the Javaid, N., Naseem, M., Rasheed, M.B., 2017. A new heuristically optimized home
proposed HEM model with multi-weight optimization objectives energy management controller for smart grid. Sustainable Cities Soc. 34,
211–227.
can take the case of three objectives into account at the same Khemakhem, S., Rekik, M., Krichen, L., 2019. Double layer home energy super-
time, furthermore, it has obviously enhanced the flexibility of vision strategies based on demand response and plug-in electric vehicle
user selection compared with single objective optimization. control for flattening power load curves in a smart grid. Energy 167,
The future work to the current study can be done in the 312–324.
Kumamoto, T., Aki, H., Ishida, M., 2020. Provision of grid flexibility by distributed
following areas:
energy resources in residential dwellings using time-of-use pricing. Sustain.
(a) The current household energy management system still Energy Grids Netw. 23, 100385.
needs users to set their preferences and power demand in ad- Langer, L., Volling, T., 2020. An optimal home energy management system for
vance, and the interactions are cumbersome. With the develop- modulating heat pumps and photovoltaic systems. Appl. Energy 278, 115661.
Li, C., Yao, Y., Zhao, R., Xie, K., 2021. Optimization model for time-varying
ment of science and technology, the intelligent interactive ter-
settlement of renewable energy consumption considering accommodation
minal of the future HEMS will automatically sense users’ power difficulty and supply–demand interaction. Int. J. Electr. Power Energy Syst.
consumption habits, and can flexibly respond to emergencies 125, 106469.
(power failure, sudden weather change, etc.), so that users can McKinney, W., 2010. Data structures for statistical computing in Python. In:
really feel convenient. Proceedings of the 9th Python in Science Conference. pp. 51–56.
Missaoui, R., Joumaa, H., Ploix, S., 2014. Managing energy smart homes according
(b) In the future, the smart perception systems can adjust the to energy prices: analysis of a building energy management system. Energy
weights by analyzing the behavior of the residents automatically Build. 71, 155–167.
with help of big data and artificial intelligence, and it can regulate Muzathik, A.M., 2014. Photovoltaic modules operating temperature estimation
a certain range of residents by establishing a virtual community. using a simple correlation. Int. J. Energy Eng. 4, 151.
2019. Python. https://www.python.org/. [Accessed May 2019].
(c) In order to simplify the research, this paper divides a day
Qayyum, F.A., Naeem, M., Khwaja, A.S., 2015. Appliance scheduling optimization
into 48 periods, and the start-up and operation of all equipment in smart home networks. IEEE Access 3, 2176–2190.
are in half an hour, which is ideal. The unit period of equipment Rajendhar, P., Jeyaraj, B.E., 2020. A water filling energy distributive algorithm
operation can be further subdivided to improve the practical based HEMS in coordination with PEV. Sustain. Energy Grids Netw. 22,
100360.
significance of the optimization strategy on the premise that
Shakeri, M., Shayestegan, M., Abunima, H., 2017. An intelligent system archi-
computational capacity allows. tecture in home energy management systems (HEMS) for efficient demand
response in smart grid. Energy Build. 138, 154–164.
Shakeri, M., Shayestegan, M., Reza, S.M.S., 2018. Implementation of a novel home
CRediT authorship contribution statement
energy management system (HEMS) architecture with solar photovoltaic
system as supplementary source. Renew. Energy 125, 108–120.
Ziye Song: Conceptualization, Methodology, Writing – origi- Shirazi, E., Jadid, S., 2015. Optimal residential appliance scheduling under
nal draft, Data curation, Writing – review & editing, Software, dynamic pricing scheme via HEMDAS. Energy Build. 93, 40–49.
Tan, K.M., Babu, T.S., Ramachandaramurthy, V.K., Kasinathan, P., Solanki, S.G.,
Formal analysis. Xin Guan: Writing – review & editing, Investi-
Raveendran, S.K., 2021. Empowering smart grid: A comprehensive re-
gation, Supervision. Meng Cheng: Software, Validation, Software, view of energy storage technology and application with renewable energy
Investigation. integration. J. Energy Storage 39, 102591.

5410
Z. Song, X. Guan and M. Cheng Energy Reports 8 (2022) 5396–5411

Teki, V.K., Maharana, M.K., Panigrahi, C.K., 2021. Study on home energy man- Zeng, W., Von Appen, J., Selzam, P., 2018. Active residential load management
agement system with battery storage for peak load shaving. Mater. Today: based on dynamic real time electricity price of carbon emission. Energy
Proc. 39, 1945–1949. Procedia 152, 1027–1032.
van der Walt, S., Colbert, S.C., Varoquaux, G., 2011. The NumPy array: a structure Zeynali, S., Rostami, N., Ahmadian, A., et al., 2020. Two-stage stochastic home
for efficient numerical computation. Comput. Sci. Eng. 13, 22. energy management strategy considering electric vehicle and battery energy
Vilar, D.B., de Mattos Affonso, C., 2016. Residential energy management system storage system: An ANN-based scenario generation methodology. Sustain.
with photovoltaic generation using simulated annealing. In: 2016 13th Energy Technol. Assess. 39, 100722.
International Conference on the European Energy Market (EEM). pp. 1–6. Zhou, S., Wu, Z., Li, J., 2014. Real-time energy control approach for smart
Wang, C., Zhang, Z., Abedinia, O., 2021. Modeling and analysis of a microgrid home energy management system. Electr. Power Compon. Syst. 42,
considering the uncertainty in renewable energy resources, energy storage 315–326.
systems and demand management in electrical retail market. J. Energy Zupančič, J., Filipič, B., Gams, M., 2020. Genetic-programming-based multi-
Storage 33, 102111. objective optimization of strategies for home energy-management systems.
Xue, X., Wang, S., Sun, Y., 2014. An interactive building power demand man- Energy 203, 117769.
agement strategy for facilitating smart grid optimization. Appl. Energy 116,
297–310.

5411