Energy Conversion and Management 221 (2020) 113192

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

Integrated supply–demand energy management for optimal design of off- T

grid hybrid renewable energy systems for residential electrification in arid
climates
⁎
Charafeddine Mokhtaraa, , Belkhir Negroua, Abdessalem Bouferroukb, Yufeng Yaob,
Noureddine Settoua, Mohamad Ramadanc
a
Laboratoire de Promotion et Valorisation des Ressources Sahariennes (VPRS), Kasdi Merbah Ouargla University, BP 511, 30000 4 Ouargla, Algeria
b
Department of Engineering Design and Mathematics, University of the West of England, Bristol BS16 1QY, UK
c
Energy and Thermo-Fluid Group, School of Engineering, The International University of Beirut (BIU), Beirut, Lebanon

A R T I C LE I N FO A B S T R A C T

Keywords: The growing research interest in hybrid renewable energy systems (HRESs) has been regarded as a natural and
Hybrid renewable energy system yet critical response to address the challenge of rural electrification. Based on a Bibliometric analysis performed
Energy management by authors, it was concluded that most studies simply adopted supply-side management techniques to perform
Optimal design the design optimization of such a renewable energy system. To further advance those studies, this paper presents
Rural electrification
a novel approach by integrating demand–supply management (DSM) with particle swarm optimization and
Building energy consumption
applying it to optimally design an off-grid hybrid PV-solar-diesel-battery system for the electrification of re-
Multi agent-based particle swarm optimization
sidential buildings in arid environments, using a typical dwelling in Adrar, Algeria, as a case study. The proposed
HRES is first modelled by an in-house MATLAB code based on a multi-agent system concept and then optimized
by minimizing the total net present cost (TNPC), subject to reliability level and renewable energy penetration.
After validation against the HOMER software, further techno-economic analyses including sensitivity study are
undertaken, considering different battery technologies. By integrating the proposed DSM, the results have shown
the following improvements: with RF = 100%, the energy demand and TNPC are reduced by 7% and 18%,
respectively, compared to the case of using solely supply-side management. It is found that PV-Li-ion represents
the best configuration, with TNPC of $23,427 and cost of energy (COE) of 0.23 $/kWh. However, with lower RF
values, the following reductions are achieved: energy consumption (19%) and fuel consumption or CO2 emission
(57%), respectively. In contrast, the RF is raised from 15% (without DSM) to 63% (with DSM). It is clear that the
optimal configuration consists of wind-diesel, with COE of 0.21 $/kWh, smaller than that obtained with a stand-
alone diesel generator system. The outcomes of this work can provide valuable insights into the successful design
and deployment of HRES in Algeria and surrounding regions.

1. Introduction fact that they live at far away distances from towns and cities, especially
those located in the Sahara region of the country. As connecting these
1.1. Background areas with the national electricity grid is technically difficult, diesel
generator (DG) is currently the only source used for rural electrifica-
Currently, buildings account for 40% of global electricity usage, tion. Due to the steady increase in diesel fuel prices, the high cost for
making a share of up to 30% of total CO2 emission [1]. Because of the transportation, and concerns over its degradation and environmental
development of new electric equipment and the intense use of smart effects, serving electricity to these remote areas using DG is becoming
devices, the demand for electricity has increased significantly in the less attractive socioeconomically. Within this context, the Algerian
Algerian building sector, with an annual increase rate of 8% since 2017 Government has made great efforts towards developing and using re-
[2]. Although the majority of buildings in the country are grid-con- newable energy technologies in rural regions by installing a number of
nected [3], there remain large numbers of small and rural communities PV and wind power plants [4,5]. However, only a few megawatts have
that are unable to access the national electricity grid mainly due to the been installed in the Sahara region of the country until now. This is

⁎
Corresponding author.
E-mail address: [email protected] (C. Mokhtara).

https://doi.org/10.1016/j.enconman.2020.113192
Received 16 April 2020; Received in revised form 4 July 2020; Accepted 6 July 2020
0196-8904/ © 2020 Elsevier Ltd. All rights reserved.
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

Nomenclature NOCT Nominal Operating Condition Temperature

PV Photovoltaic
AC Alternating Current RER Renewable Energy Resources
ACL Agent Communication Language Rg Ramping Factor
BS Battery Storage SA Storage Agent
CA Control Agent SOC State of Charge of the Battery
COE Cost Of Energy STC Standard (Reference) Test Conditions
CRF Capital Recovery Factor t Time Interval (One Hour)
DA Design Agent Tc Cell Temperature
DC Direct Current Tref Reference Temperature
DG Diesel Generator Va Actual Wind Speed
DSM Demand-Supply Management Vci Cut-In Wind Speed
FIPA Foundation of Intelligent Physical Agents Vco Cut-Out Wind Speed
G Solar Radiation (W/m2) Vr Rated Wind Speed
GA Generation Agent Wt Wind Turbine
HRES Hybrid Renewable Energy System Pnpv The rated power of a Pv system under Stc
HVAC Heat Ventilation and Air Conditioner PDG_R DG maximum power generation at (t)
JADE Java Agent Development Framework Ppv The output power of the Pv
KT Temperature Coefficient of Power Gref Solar radiation at Reference Conditions (E.G. 1000 W/M2)
LA Load Agent Ppv Amount of power generation from solar PV at (t)
LPSP Loss of Power Supply Probability Pdg Amount of power generation from DG at (t)
MAS Multi-Agent System Pwt Amount of power generation from WT at (t)
NASA National Aeronautics and Space Administration Pr WT rated power
NBR Number

mainly due to the dependency of renewable energy technologies on the particle swarm optimization (PSO) for achieving a compromise be-
weather and the effect of environmental factors (mainly dust wind), tween several techno-economic, environmental and socio-political ob-
high cost of installation and other techno economic barriers. In order to jectives for the optimization of HRES. Zhang et al. [19] utilised an
solve these issues, hybrid renewable energy systems are becoming more advanced heuristic technique to find an optimal configuration of a
popular and provide a reliable pathway to provide electricity for rural HRES, and later integrated a neural network (NN) weather forecasting
areas. Generally, an off-grid hybrid renewable energy system (HRES) with an optimization algorithm for the optimal sizing of an off-grid
consists of at least one renewable source, one conventional source HRES [20]. Both genetic algorithm (GA) and PSO algorithm were
(mostly diesel generator) and/or an energy storage system (such as combined in the study by Mellouk et al. [21] to optimize a HRES. The
batteries). Recently, HRESs have attracted an increased interest from economic evaluation of a micro-grid HRES was performed by Yu et al.
researchers and policymakers because of their technical and environ- [22]. Kaabeche and Bakelli [23] carried out the size optimization of a
mental benefits. The challenge in designing such a solution concerns its HRES considering various electrochemical energy storage technologies.
economic feasibility [6]. Clearly this cannot be attained without the Hamanah et al. [24] presented a new methodology to optimize of hy-
development of an energy management method, since an efficient brid PV-wind-diesel-battery energy system, in which the minimization
management system will guarantee that HRES works in an efficient and of the annual cost was the target objective considering different con-
economical way [7]. straints. Haratian et al. [25] used HOMER software to examine the
techno-economic feasibility of a stand-alone HRES. Similarly, the sizing
of a hybrid PV/diesel/battery energy system for a rural Saharan com-
1.2. Literature review and current research gap
munity in Algeria was investigated by Fodhil et al. [26], also using
HOMER software. Fodhil [27] used PSO with the ε-constraint method
Much research work has been undertaken on the design of HRES,
for the optimal design of a PV/DG/battery hybrid energy system for
using various energy management and sizing approaches [8]. The op-
rural electrification in Algeria. Meanwhile, Duman et al. [28] con-
timal sizing of a hybrid wind/photovoltaic renewable energy system
ducted a technical and financial feasibility analysis for an off-grid hy-
considering the effect of the wind turbine model was investigated by
brid system with variable climatic environments using HOMER soft-
Mehrjerdi [9]. Abo-elyousr et al. [10] used a multi-objective optimi-
ware. Anoune [29] used a TRNSYS deterministic based approach to
zation algorithm for the optimal design of HRES, considering a variety
perform the sizing of a PV-wind based hybrid energy system for supply
of alternative fuels. Tao and Javed [11] performed the sizing of a hybrid
electrical load demands in isolated areas. Jafar et al. [30] implemented
PV-wind-battery system for an island case study. Assaf [12] proposed
a sizing method that integrated energy management to optimize the
an integrated solar-based HRES to supply remote areas with low cost
size of a stand-alone HRES using a flower pollination technique. Das
and renewably. The optimal design of PV-wind-battery HRES was also
et al. [31] carried out a techno-economic analysis and optimal design of
carried out by Tudu et al. [13]. The design optimisation of an off-grid
an off-grid HRES using water cycle algorithm and moth-flame meta-
hybrid renewable energy system including thermoelectric generators
heuristic optimization techniques for a radio transmitter station in
was discussed by Rodolfo et al. [14], in which minimizing the net
India. Halabi et al. [32] carried out a performance analysis of a hybrid
present cost was the main objective. Adefaratia and Bansal [15] eval-
PV/diesel/battery system through HOMER software for a rural area
uated the reliability, economic and environmental benefits of hybrid
case study in Malaysia. Rullo et al. [33] developed a bi-level optimi-
PV-wind-electric storage-diesel systems for rural communities. Jam-
zation framework by integrating an energy management system to the
shidi and Askarzadeh [16] optimized an off-grid HRES using multi-
GA based sizing method for optimal sizing of a hybrid PV/wind energy
objective crow search algorithm (MOCSA). Moradi et al. [17] devel-
system with energy storage.
oped an energy management technique based on an advanced dynamic
Although the optimal sizing of HRES is always an important task,
programming method for the best design of a stand-alone HRES.
peak load shaving is becoming ever more necessary for limiting the
Eriksson [18] proposed a multi-objective approach by implementing a

2
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

rapid growth in residential energy demand [34]. Peak-load shaving can 1) The majority of works on the optimization of HRESs have been fo-
also help to address the oversizing problem in HRES, and in the cused on techno-economic parameters such as the net present cost
meantime to reduce the system cost. Generally, peak load shaving is and system reliability;
defined as a way to flatten the load by either limiting the peak load or 2) The particle swarm optimization and HOMER software are the most
shifting some loads to lower demand periods [35]. From the literature, used optimization techniques;
three peak-load shaving strategies are identified; 1) the integration of 3) The solar, wind energy and battery are the most used renewable
an energy storage system; 2) vehicle to home (V2H) technology; and 3) source and storage system, respectively;
demand-side management or demand response. Tu et al. [36] carried 4) Most of the studies are focused on supply-side management to en-
out a multi-layered demand scheduling using Mixed Integer Linear sure the demands are met even during the peak periods. In contrary,
Programing for the optimization of hybrid PV/wind/DG/battery system little work has been done for the effect of peak load shaving on the
with the aim of minimizing the cost of energy (COE). In a study carried optimal design of HRES.
out by Rullo et al. [37], the optimization of a biomass-based microgrid
with demand–supply management (DSM) was performed to minimize Besides, those applied demand-side management techniques such as
the operating cost of a biomass combined heat and power microgrid, load shifting were merely feasibility studies for the non-primary load
using a load shifting algorithm based on economic linear programming (such as washing machine and water pumping). Therefore, there have
of a model predictive control. Similarly, Yang et al. [38] carried out a been no demand-side management strategies implemented until now
techno-economic and environmental optimization of a PV/battery hy- for primary loads (such as air conditioning and other HVAC systems) in
brid energy system within DSM. Sarkar et al. [39] undertook an optimal the optimisation of HRES. This can be demonstrated by the re-
design of a hybrid solar PV/wind/biogas/battery energy system for commendations of a review paper conducted recently [35], in which
ensuring higher system reliability through the use of HOMER software. the authors suggested that future works should be focused on the proper
In his work, a demand-side management (i.e. peak load shaving) application of demand-side management and the transition towards
strategy has been established. Wu et al. [40] presented a demand-side smart home technologies. Thus, an integrated management system that
management technique for optimal operation of a hybrid PV/battery enables controlling both sides (i.e. supply-side and demand-side) could
energy system. Thiaux et al. suggested a DSM strategy to find out the provide for a broader range of opportunities in optimizing the overall
best configuration of a HRES for off-grid regions. [6]. Mohseni et al. energy management system [37] and to ensure the optimal design and
proposed a novel method incorporating a demand response technique operation of HRES.
for the optimal sizing of an off-grid hybrid PV/wind turbines/battery/
electric vehicle, subject to satisfying a reliability index for meeting the 1.3. Objective and contributions
required loads [41].
From these aforementioned literature and based on an extensive In order to overcome the limitations highlighted above, in this
Bibliometric analysis of some 699 published papers on the topic over paper, the authors will propose a novel approach; i.e. a DSM strategy to
the last ten years from the Elsevier database, using the first four key- be developed and integrated with a PSO based optimization algorithm
words of this study, as shown in Fig. 1 (using VOS viewer tool), the for the energy management and optimal design of an off-grid PV/wind/
following observations can be made: diesel/battery HRES for the electrification of a residential building lo-
cated in Adrar, in the Algerian Sahara. The main contributions of this

Fig. 1. Schematic view of the Bibliometric analysis of hybrid renewable energy system. Note that the most used keywords are those located very near to red colour
and which are written in large letters, conversely, the least used keywords are those very near to blue colour and which are written in small and transparent letters.
(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

work are firstly the design optimization of an off-grid hybrid PV/wind/ distributed energy systems. There are many tools for the implementa-
diesel/battery energy system for a residential building, including a new tion of MAS, including JADE. JADE framework is an open-source agent-
DSM strategy; and secondly, the techno-economic performance analysis oriented middleware with the most widespread use of this kind. Agents
enabling the evaluation of the benefits of DSM in the enhancement of in MAS are autonomous and can use messages to interact and under-
the COE and the building energy end-user consumption. Finally, a stand the environment. Communication among agents in JADE is im-
sensitivity study will be performed to determine the gain of DSM by plemented based on FIPA-specified agent communication language
varying renewable contributions and installed battery technologies. (FIPA-ACL), which is the most important standardization activity con-
To the authors’ knowledge, this particular work presents a new ducted in the field of agent technology. JADE can be integrated with
approach that has not been applied previously. The novelty of this work MATLAB software platform. The MAS-based framework for the simu-
is related to the integrated demand–supply management strategy, lation of the suggested HRES in MATLAB has been introduced in this
which is based on the state of supply system and the operation of the air work, in which five agents are defined:
conditioner (the main load in hot dry climates), i.e. we have developed
a control system in which the set point temperature for air conditioner 1. Load Agent (LA): This agent is responsible for calculating and
is changed (between 24 °C (high-energy consumption) and 29 °C (low controlling hourly loads, and provides the required information to
energy consumption) according to the available energy from renewable other agents. It also receives climatic data from the environment.
sources. This method is more advantageous than shifting load strategies 2. Generation Agent (GA): All energy sources including WT, solar PV,
(that were mostly applied in previous work) which are, unfortunately, and diesel generator are included in the generation agent. This
in most cases not possible to apply and not viable for primary loads agent is in charge of evaluating the state of the sources of supply and
(cooling demand for example) and unacceptable by consumers. The provides information every hour on the available energy potential.
DSM strategy will be used to further prove its benefits concerning the It is responsible for adding, deleting, changing, connecting or dis-
improvements achieved on COE and building energy consumption (fuel connecting any energy source.
consumption as well). In addition, our proposed approach is seeking to 3. Storage Agent (SA): SA includes a battery bank. The SA has per-
avoid the oversizing of the HRES as we can reduce the investment and manent behaviour, because it can work in case of charge and/or
operating costs. discharge. However, this agent has some charge limits, i.e. at each
The developed optimization model is fully implemented using hour the SA supervises the state of charge of batteries and receives
MATLAB software. Thus, the simulation of the proposed HRES is carried information from GA and LA to decide when batteries could run on
out based on hourly data over a one-year period by making a direct charge or discharge mode.
comparison with the HOMER software to validate the accuracy of the 4. Design Agent (DA): DA is an independent system operator re-
proposed model. sponsible for the optimization of the HRES. This agent selects and
In Section 2, a description of the used materials and methods in this refines the size of the HRES’ components to find out the optimal
study are presented. Sections 3 and 4 contain the results obtained and configuration of HRES.
their explanations. Section 5 summarizes the findings and provides 5. Control Agent (CA): CA is the supervisor of the system and is solely
prospects of proposed approach. concerned with the monitoring of the different information ac-
cording to expectations from/to other agents.
2. System modelling and configuration
The following sub-sections describe the mathematical models of the
In this work, a new approach is proposed for the energy manage- five agents above and the information flow chart between them.
ment and the size optimization of a HRES to provide the electricity
needs for an off-grid residential building in Adrar, south of Algeria. The 2.1. Load agent
studied hybrid system is shown in Fig. 2, including PV solar panels,
small wind turbines (WT), DG, battery storage, and a power converter. 2.1.1. Building description
Decentralized controlling systems such as multi-agent systems As this study is concerned with the effective design of HRES to sa-
(MAS) are more and more prominent with the growing demands in tisfy the electricity needs of households in remote parts of Algeria, for

Fig. 2. Schematic view of the investigated hybrid renewable energy system.

4
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

that reason, a typical residential building is selected as a case study, temperature for cooling for very hot locations, where the outside
which is known as ‘F2’ according to Algerian standards of construction temperature can exceed 60 °C).
[42]. The building includes a bedroom, a sitting room, a corridor, a The global electricity demand of the building including both loads is
kitchen, and a bathroom. The whole floor area is 64 m2. The height of then assumed for both desired temperatures in cooling mode. With a
the building is 3 m. In Fig. 3, both 2D floor plan view and 3D building 24 °C set point temperature, the global demand is found to be
model (using a design builder software) are presented. In addition, the 8698 kWh/year. However, for a desired set point temperature of 29 °C,
characteristics of the building are provided in Table 1. the overall demand of the building is found to be 6948 kWh/year.

2.1.2. Geographic location and climatic data 2.2. Generation agent

The building is simulated under the climate of Adrar, Sahara of
Algeria, where the temperature and relative humidity can reach 50 °C Generation agent includes all produced sources, including renew-
and 5% in summer. The hourly climatic data for Adrar, over an entire able and non-renewable sources. The mathematical modelling of this
year (as an average of 10 years, from 1990 to 2010), are extracted from agent and their components are provided as follows.
Meteonorm software. The hourly ambient temperature and wind speed
in Adrar are presented in Figs. 4 and 5, respectively. In addition, the 2.2.1. Solar PV
daily average solar irradiation (G) and clearness index (CI) at each Solar PV panel is a device that converts solar energy into electric
month of the year are given in Fig. 6. energy. The energy delivered by the PV could be quantified based on
the solar radiation and the air temperature as given in Eq. (1) [44]:
2.1.3. Load estimation G NOCT − 20 ⎤
The load profile of the studied building is divided into two parts. Ppv = PNpv × × ⎡1 + Kt × ⎛ ⎡Tamb +
⎜ × G − Tref ⎞ ⎤ ⎟

Gref ⎢
⎣ ⎝ ⎣ 800 ⎦ ⎠⎥⎦
The first part includes all different electric devices, except for the
(1)
cooling chiller. Hence, the second part includes the load profile of space
cooling which is evaluated by Energy Plus software [43]. where Ppv is the rated power of the PV system. G and Gref represent solar
The primary concern is the hourly electricity demand for the radiation at time t and at standard conditions, respectively. Tamb and Tref
building’s appliances, which includes a refrigerator, a lighting system, are the ambient temperature at time t and at standard conditions re-
and a TV. This load is evaluated based on the number and the rated spectively. Kt is the temperature coefficient of power and its value de-
power of each device, and the usage time of occupants. Table 2 gives a pends on the PV panel’s technology.
summary of the building appliances and their characteristics. The
hourly demand for electric appliances per day on average over the year 2.2.2. Wind turbine (WT)
is provided in Fig. 7. The wind turbine is used to produce electricity from the kinetic
The secondary load includes the demand for heating, ventilation, energy available in the wind flow. The power output of the WT Pwt can
and air-conditioning (HVAC) systems which are the major energy be computed using Eqs. (2)–(4), as follows [45]:
consumption devices in the building and thus become ideal candidates
Pwt = 0 if V < Vcutin or V > Vcutout (2)
for substantial reductions in energy demand [1]. In hot dry climates as
in Adrar, the required demand for space cooling is assumed more than Pr Vcutin3
half of the total electricity demand. To evaluate the energy demand for Pwt = V 3 ⎛ 3
⎜
3
⎟
⎞−⎛⎜
3 − Vcut 3
⎞ × Pr if V > Vcutin and
⎟

⎝ Vr − Vcut in ⎠ ⎝ Vr in ⎠
space cooling of the investigated building, here both Energy Plus and
Design Builder software are used. Only the bedroom and the sitting V < Vr (3)
room have an air conditioner each. Thus, the conditioned floor area is
Pwt = Pr if V > Vr and V < Vcutout (4)
32 m2. Fig. 8 shows the hourly demand for space cooling in case of
24 °C (that correspondents the desired temperature for high level of where Pr is the rated power of the WT. V is the wind speed at the
thermal comfort) and 29 °C set point temperature (a suggested desired studied location (for each hour). The Vcutin, Vcutout, Vr represent cut in,

Fig. 3. The layout of the building.

5
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

Table 1
Building components and characteristics of walls.
Element Construction (outside to inside) (m) U-value (W/m2K)

Ground floor Concrete (0.1), Extruded Polystyrene (0.04), Concrete (0.1), Ceramic floor tiles (0.02) 0.573
Roof Bitumen (0.01), Cement (0.01), Concrete Block (0.2), Cement (0.01) 2.994
Internal Wall Cement (0.01), Brick (0.1), Cement (0.01) 2.079
External Wall Cement (0.01), Brick (0.1), Air gap (0.01), Brick (0.1), Cement (0.01) 1.39
Glazing / 1.978
Door / 2.823

Fig. 4. The hourly ambient temperature at Adrar region, Algeria.

Fig. 5. Hourly wind speed at Adrar region, Algeria.

cut out and rated wind speed of WT respectively (normally provided by dependent on its output power at each time step as expressed by Eq. (5)
the manufacturer). [46]:

Fcons = a∙PDG + b∙PDG _r (5)

2.2.3. Diesel generator (DG)
DG is used in a hybrid energy system to meet the load demand in where PDG (t) is power generated by DG (kW) at each hour (t), Fcons is
case the total available power generated from renewable energy and fuel consumption (L/h), PDG _r is the rated power of DG generated at
batteries is not sufficient. The amount of consumed fuel by DG is each hour (t), a and b are constants (L/kW), which represent the

6
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

Fig. 6. Daily average solar radiation and clearness index in Adrar, Algeria.

Table 2 coefficients of fuel consumption, with standard values of 0.08415 and

Building appliances and their characteristics. 0.246, respectively.
Element Rated power (W) Number Daily use (hours)

Refrigerator 320 1 24 2.2.4. Converter

Light 40 5 18 The power converter is a device that converts the electrical energy
TV 100 1 8
from AC into DC or vice versa. The rated power of a converter depends
on the peak load. The efficiency of the converter is evaluated by Eq. (6)
[27].

Fig. 7. Profile of electricity demand for appliances per day.

7
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

Fig. 8. Energy demand for space cooling during summer period.

Poutput 2.4. Design agent (DA)

ηcnv =
Pinput (6)
DA is the agent that performs the optimization of the HRES. In this
where Poutput and Pinput are the output and the input power from/to study, DA is incorporated into the PSO algorithm, which works well to
converter, respectively. minimize the objective function and thus obtain the best HRES size. The
objective function to be minimized is the cost of energy of the HRES,
2.3. Storage agent concerning the critical value of loss of power supply probability (LPSP).
The critical value of LPSP is set at 1%, which represents the allowable
In this study, the storage agent solely includes batteries, which are loss of power for the entire yearly demand. Further, fuel consumption
easy to install and possess high energy efficiency. Excess electricity can be minimized to achieve the increased share of renewable energy.
from renewable sources is used to start charging the battery, while it
can be used in the event of energy shortages. The state of charge of a 2.5. Problem formulation
battery in the discharge and the charge settings could be determined by
Eqs. (7) and (8), respectively [46]. In this study, total net present cost (TNPC) is another objective
function to be minimised as defined by the objective function (f) in Eq.
E (t ) (10), in which five decision variables (i.e. capacity of PV panels, bat-
Eb (t + 1) = Eb (t ) × (1 − σ ) − ⎜⎛ l − Eg (t )⎟⎞ × ηBD
⎝ ηcnv ⎠ (7) teries, diesel generator, converter and number of wind turbines) must
be optimally sized, subject to constraints defined in Eqs. (11)–(13):
E (t )
Eb (t + 1) = Eb (t ) × (1 − σ ) + ⎜⎛Eg (t ) − l ⎟⎞ × ηBC f = min∙TNPC (PPV _r , NWT _r , PDG _r , Ebmax , PCnv _r ) (10)
⎝ ηcnv ⎠ (8)
0 ≤ LPSP ≤ LPSP max (11)
where El (t) and Eg (t) are the energy demand and generated power,
respectively. ηBD and ηBC represent the discharge and charge efficiencies RF min ≤ RF ≤ 1 (12)
of the battery. Parameter σ is the self-discharge of the battery, which is
max max max max
set to be zero in this study. ηcnv is efficiency of the converter. At each 0 ≤ PPVr , NWTr , PDGr , Ebmax , PCnvr ≤ PPVr
, NWT , PDG r
, BS max , PCnv
time step t, the state of charge of a battery Eb (t )is constrained by the (13)
minimum and the maximum capacities of the storage Ebmin and Ebmax as
specified in Eq. (9): More details of the mathematical formulae of the objective function,
constraints and techno-economic and environment evaluating criteria
Ebmin ≤ Eb (t ) ≤ Ebmax Where Ebmin = (1 − DOD) × Ebmax (9) are provided in the following subsections.

where DOD is the depth of discharge of the battery, which depends on

the battery’s technology. 2.5.1. Total net present cost (TNPC)
As batteries are not ‘clean’ energy storage systems, having a high TNPC is widely used in the design of HRES. It can be evaluated
cost of investment and a short life span [47], it is thus mandatory to based on Eqs. (11)–(18) [48].
select the best battery technology to reduce their capacity as much as TNPC ($) = (Ccap + CO & M + Crep + Cfuel ) (14)
possible. In this context, two battery technologies are considered,
namely Lead-acid and Li-ion, both having different technical–economic where Ccap, Crep, CO&M, Cfuel are the costs of investment, replacement,
characteristics. operation and maintenance and the cost of fuel, respectively.

8
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

Ccap ($)=(PPV _r ·CPV + PWT _r ·CWT + PDG _r ·CDG + Ebmax ·CBS + PCnv _r ·CCnv 8760
Fuel consumption (L /year ) = ∑ fuelcons (t ),
(15) t=1 (24)
T
1 CO2 (kg /year ) = EF∙Fuel consumption (25)
CO & M ($) = 0.02·C Acap· ∑
(1 + i) k (16)
k=1 where EF is the emission factor of the DG, which depends on the type of
diesel engine fuel properties. The value of this factor is in a range of
Crep ($) 2.4–2.8 kg/lit [50]. This study takes a lower value of 2.4 kg/lit.
T T
⎛ 1 1
= ⎜Ebmax ·CBS · ∑ + PCnv ·CCnv· ∑ 2.6. Particle swarm optimization (PSO) algorithm
k = 10 (1 + i) k k = 12 (1 + i) k
⎝
T
1 ⎞ PSO is considered to be the most frequently used method in the field
+ PDG _r ·CDG· ∑
k = a,2a, …… . < T (1 + i) k ⎟ (17)
of artificial intelligence [51]. This meta-heuristic optimization tech-
⎠
nique was firstly introduced by Kennedy and Eberhart in 1995 [27]. In
T PSO, a set of particles (also known as swarm), defined by their positions
1
Cfuel ($) = fuelCons ·Cfuel· ∑ and velocity vectors, move through the search space seeking to reach
k=1
(1 + i) k (18) their local optimum and global optimum for the swarm. Therefore,
where PNpv, PWT_r, PDG_r, PCnv, Ebmax represent the rated power and the there are two optimum values to determine each particle’s position. The
capacity of PV, WT, DG, converter, and battery, respectively. CPV, CWT, first is the optimum value that has been achieved by each particle (i.e.
CDG, CCnv, CBS represent the investment and replacement costs of PV, local maximum) and the second is the optimum value obtained for the
WT, DG, converter, and battery, respectively. Here, fuelCons is fuel entire population (i.e. global maximum). The ultimate goal is to achieve
consumption. the optimal solution for the objective function. The PSO algorithm
solves the defined problem according to the following steps:
2.5.2. Loss of power supply probability (LPSP)
LPSP is widely used to evaluate the system reliability. Generally, 2.6.1. Step 1: initialization
LPSP is the ratio of the total energy deficit over the total energy demand
during one year of operation. LPSP represents the rate of dissatisfaction (1) Load the hourly solar irradiation, ambient temperature, wind speed
of the load and its value can be determined by Eq. (19) [11]: and building load for an entire year.
(2) Load the components characteristics and economic parameters as
T
∑t = 1 El (t ) − PPV (t ) − PWT (t ) − PDG (t ) − (Eb (t ) − Ebmin ) presented in Table 3.
LPSP (%) = T
∑t = 1 Pload (t ) (3) Set the constraints limits for LPSP and RF, and set the decision
variables limits (including the minimum and the maximum values
(19) for each variable) as given by Table 4, and set a random value for
them.
2.5.3. Renewable fraction (RF)
RF is introduced to set the minimum value of renewable energy 2.6.2. Step 2: initialization of PSO parameters
contribution in the overall load served by the HRES. RF can be eval- To determineω, C1, and C2 , the following coefficients must be cal-
uated using Eq. (20) [49]. culated using Eqs. (26) and (27) [18]:

∑ PDG 2K
RF (%) = 1 − φ=
∑ PPV + PWT + PDG (20) 2−∅− ∅2 − 4∅ (26)
∅ = ∅1 + ∅2 ≥ 4 (27)
2.5.4. Cost of energy (COE) where φ is the constriction coefficient. Typically, ∅1 and ∅2 are set to
COE is obtained by dividing the total cost of the system (including 2.05. The constant k is generally less than unity. Hence, ω, C1, and C2
costs related to the initial investment and all costs associated with the are calculated using Eqs. (28)–(30):
operation of the system components) by the annual energy served as ω=φ (28)
given by Eq. (21).
C1 = φ∙∅1 (29)
CAnualized
COE ($/ kWh) =
Eserved (21) C2 = φ∙∅2 (30)

where CAnualized and CRF are the total annualized cost and capital re- Further settings of the PSO and simulation parameters are given in
covery factor, respectively, which are defined by Eqs. (22) and (23) as Table 5 [53].
follows.
2.6.3. Step 3: determine particles’ positions and fitness
CAnnualized ($) = TNPC∙CRF (22)

[i (i + 1)T ] (1) Initialization of particle’ velocity and position for all the popula-
CRF = tion,
[(i + 1)T − 1] (23)
(2) Evaluation of the fitness of each particle, and evaluate the optimal
where T is the project lifetime, and i is the real interest rate. value for the initial population (i.e. the minimum TNPC).

2.5.5. Fuel consumption and CO2 emission 2.6.4. Step 4: estimate the new positions
As most remote areas suffer from the high cost of fuel associated
with its transport, it is also important to reduce fuel consumption. (1) Updating the optimum individual and global fitness and locations,
Besides, CO2 emissions are also evaluated based on the amount of and
consumed fuel. Eqs. (24) and (25) are used to evaluate the consumed (2) Updating the velocity and location of each particle.
fuel and generated CO2 by the HRES [50].

9
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

Table 3 indicated velocity towards a position. The termC1 r1 (Pki − Xki ) is the
HRES’ components and economic parameters. learning parameter. C2 r2 (Pkg − Xki ) is so-called the social component, Pi
Generation source Parameters Specification
is the optimal individual particle position and Pg is the optimal global
position, C1 and C2 are personal (cognitive) and global (social) learning
Solar PV Nominal power (kW) 1 coefficients, respectively; r1 and r2 are random numbers (their values
Capital cost ($/kW) 1600 are between 0 and 1), ω is a coefficient.
O&M cost (% of Capital cost) 2
The procedure is reproduced until all pre-defined conditions have
Temperature coefficient of power −0.41
(%/°C) been met, such as the number of iterations or pre-defined goal fitness
Lifetime (Year) 20 levels.
Wind turbine Nominal power (kW) 2.1
Capital cost ($/kW) 3000 [52] 2.6.5. Step 5: output of obtained results
O&M cost (% of Capital cost) 2% After the simulation is finished, key results are either plotted or
Cut-in speed (m/s) 3
exported in tabular forms for presentation and further analysis shown
Cut-out speed (m/s) 20
Rated speed (m/s) 11
later.
Lifetime (Year) 20

Diesel generator Rated power (kW) 1

2.7. Control agent (CA)
Capital cost ($/kW) 800 [36]
O&M cost (% of Capital cost) 2% CA is in charge of monitoring the components of the HRES. This
Fuel price ($/L) 0.2 agent follows an energy management strategy to ensure a balance be-
Lifetime (hour) 30,000
tween the demand and the supply sides and to ensure a compromise
Lead-acid (L-acid) Battery Battery capacity (kWh) 1 between the objective functions (cost, reliability, etc.) and to find the
Capital cost ($/kWh) 180
optimum size of HRES. As stated earlier, a DSM approach has been built
O&M cost (% of Capital cost) 2
DOD (%) 60
and integrated into the optimization model. The DSM plan was pro-
Discharge efficiency 90% posed to reduce the energy usage of buildings and thus to avoid the
Charge efficiency 100% oversizing of the HRES. Two loads are introduced in the optimization,
Lifetime (Year) 5 the first is evaluated with a set point temperature in cooling mode at
Lithium-ion (Li-ion) Battery Battery capacity (kWh) 1 24 °C; and the second is evaluated at a set point temperature of 29 °C.
Capital cost ($/kWh) 300 Hence, the main steps of the size optimization of the suggested HRES
O&M cost (% of Capital cost) 2
are as follows:
DOD (%) 80
Discharge efficiency 90%
Charge efficiency 100% Case 1: Renewable sources (both solar and wind) supply sufficient
Lifetime (Year) 10 electricity and any excess electricity will be used to charge the
Converter Capital cost ($/kW) 700 [52] battery.
O&M cost (% of Capital cost) 2 Case 2: Same as Case 1 except that the unused generated energy (as
Efficiency (%) 95 it is greater than the need for building consumption while the bat-
Lifetime (Year) 10
tery is fully charged) is transferred to the dump load.
Economic parameters Project lifetime (Year) 20 Case 3: Renewable resources fail to provide sufficient energy to meet
i (interest rate) (%) 5
the first load (at 24 °C). Therefore, the control agent requests to use
the second load (at 29 °C). If the system is still unable to provide the
required load, then in this scenario the priority is to use the stored
Table 4
Decision variables and constraints limits.
energy in the batteries rather than operating the diesel generator.
Thus, DG still remains as the last option.
Parameters PPV NWT BS PDG Converter LPSP RF Case 4: The electricity provided from renewable sources is not suf-
[kW] [Nbr] [kWh] [kW] [kW] [%] [%]
ficient to meet the required load and the battery bank is drained. In
Min bound 0 0 0 1 0 0 0 this case, DG is turned on to deliver the power. When the over-
Max bound 9 2 40 5 10 2 100 all system fails to meet the demand, the loss of power supply is
therefore evaluated.

Table 5 The sequence diagram that illustrates the different interactions be-
Parameters of the PSO. tween the five agents, is presented in Fig. 9. Besides, a flowchart of the
Parameter Value proposed optimization model including the developed DSM is presented
in Fig. 10.
Number of iterations 100
Population Size 25
The dimension of the search variables 5 3. Simulation using in-house code
k 0.75
constriction coefficients (both) 2.05 In this study, a MATLAB code has been developed and implemented
to optimize the proposed HRES. First, the modelling of five agents is
performed. Each agent is defined as a function file. The design agent
The location and the velocity of each particle in the swarm can be includes two sub-agents, one for implementing the particle swarm op-
adjusted using Eqs. (31) and (32) [50]: timization algorithm, which is used to solve the multi-objective pro-
Xki + 1 = Xki + vki + 1 (31) blem, and another one is created for evaluating the objective function
and constraints. In the control agent, the energy management strategy
where X is particle position and v is particle velocity in iteration k: is implemented including the proposed DSM strategy. In the generator
agent, mathematical equations for energy sources are developed to
vki + 1 = [ωvki + C1 r1 (Pki − Xki ) + C2 r2 (Pkg − Xki )] (32)
evaluate the produced energy at each time step. The storage agent
where Vki is named the inertia, which drives the particle to move at an consists of two sub-agents, namely charge and discharge agents. In

10
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

Fig. 9. The sequence diagram of MAS for HRES.

these two sub-agents, the state of charge of the battery bank is eval- as follows.
uated based on their modelling equations. The communication between
these agents is defined by call functions. The required inputs data for 3.1. First scenario
starting the simulation are imported simultaneously to MATLAB soft-
ware as vectors with a duration of 8760 h (i.e. number of hours per In this scenario, the optimal sizing of the HRES is performed without
year). After MATLAB completes the simulation, the results are exported introducing the suggested DSM strategy in the optimisation. However, a
for data post-processing and analysis. Finally, to validate the developed supply-side management strategy is applied in order to control the
PSO-based code, HOMER Pro is used simultaneously to simulate the operation of the HRES. Here, the annual electricity demand of the
proposed HRES for a baseline case simulation (i.e. without DSM), as the studied building is evaluated for a desired set point temperature of
HOMER software is unable to do the demand management. 24 °C (which represents the temperature for high thermal comfort re-
The proposed HRES is optimized with two different scenarios ac- quirements in cooling mode). This scenario is also simulated by HOMER
cording to the battery technology (L-acid and Li-ion), where a sensi- software to verify the accuracy of the developed MATLAB code.
tivity analysis is performed by imposing different values for RFminimum,
ranging from 0% to 100%. Furthermore, two simulation scenarios are 3.2. Second scenario
investigated based on the proposed DSM strategy which are integrated
with the PSO-based algorithm. The description of these two scenarios is In this scenario, the suggested DSM (that includes peak load shaving

Fig. 10. Flowchart of the proposed optimization model.

11
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

strategy and supply-side management) is integrated into the developed cost, as well as the primary and dumped loads, simultaneously. This
PSO-based MATLAB code, in which two different loads for the same approach has clear benefits compared to most previous works that
building are imported to MATLAB software to perform the simulations. merely carried out the optimal sizing of HRESs by focusing on the re-
The first load incorporated the evaluated demand for cooling at 24 °C duction of the system cost while the satisfaction of the load demand
(i.e. a standard desired temperature of cooling mode). The second load even in case of an overload demand were likely leading to the high
is obtained by increasing the energy demand for cooling in case of investment cost and the large dumped load.
setting the desired temperature at 29 °C. After applying the proposed
DSM strategy, the operating energy demand for the building is ob- 4.2. HRES’s design with a high renewable fraction
tained, in which at each time step (hour), the energy demand for the
building takes one value from the two candidate loads (either from the In the first simulation case, RFminimum = 100% is imposed in the
first load or the second load) depending on the available energy pro- optimization, in which the proposed HRES is simulated with both L-acid
duced from renewable sources. When the energy generated from re- and Li-ion battery characteristics.
newable sources is greater than the value of power from the first load, After the simulation, the results of optimal design for the HRES with
in this case, the operation demand for the building takes the value of L-acid batteries are given in Tables 7 and 8, respectively.
power from the first load; otherwise, it takes the value from the second Similarly, the results of optimal design for the HRES with Li-ion
load. After finishing 8760 h, the yearly operating load for the building battery characteristics are summarized in Tables 9 and 10, respectively.
is obtained, and based on this, the objective functions are evaluated. Furthermore, the contributions of the HRES’s components in case of
using L-acid and Li-ion batteries are presented in Figs. 12 and 13, re-
4. Results and discussion spectively.
The hourly energy demand of the building for both scenarios using
The design and techno-economic evaluation of the proposed HRES L-acid and Li-ion batteries are presented in Figs. 14 and 15, respec-
are carried out using the developed optimisation approach. After that, tively.
the obtained results with the developed MATLAB code and by HOMER From the obtained results and as shown in Figs. 14 and 15, for high
software are first compared for validation purposes. Then the results for RF values (100%), the energy demand for the building can be reduced
all investigated scenarios are presented and discussed. by 7% for both cases (i.e. using L-acid or Li-ion batteries) while ap-
plying the suggested DSM. Moreover, the amount of electricity excess or
4.1. Validation of the developed model dumped load is also reduced compared to the case without DSM. Re-
ducing the dumped load is an important factor, as the investment cost
Here, the developed MATLAB code is validated according to a of renewable-based HRESs is still very high for customers in the de-
baseline scenario with condition of RF = 100%. Fig. 11 shows the veloping countries. Besides that, for both investigated scenarios and
convergence of the developed PSO-based code, in case of using (a) L- using both battery technologies, the obtained optimal HRES config-
acid batteries and (b) Li-ion batteries, respectively and for the two in- urations consist of PV and battery storage. However, applying the DSM,
vestigated scenarios (with and without applying the suggested DSM). the capacity of battery storage can be largely reduced by 25% and the
The comparisons between the obtained results by the MATLAB code capacity of PV modules can be slightly reduced, which leads to an
and HOMER software for a baseline scenario (with RF = 100% and overall 18% reduction on both TNPC. Of this, the HRES that includes
without DSM) are presented in Table 6. PV and L-acid has a COE of 0.29 $/kWh (without DSM) and 0.26 $/kWh
It can be seen that the obtained results by the developed model are (with DSM) which makes this configuration only attractive when con-
in good agreement with those obtained by HOMER software, with small sidering renewability preferences. However, for PV/battery (Li-ion), the
and acceptable differences. Furthermore, the calculation time required COE can be reduced from 0.26 $/kWh (without DSM) to 0.23 $/kWh
for obtaining results within the developed code is lower compared to (with DSM), which makes this configuration more cost-effective than
the HOMER simulation. Moreover, the implemented PSO-based code is using a stand-alone DG system. Moreover, this configuration is more
converged rapidly, same as those in ref [53,54]. Hence, the accuracy of interesting than using solely DG, considering the effect on the en-
the implemented MATLAB code has been validated and the simulation vironment and the local government policy that encourages solutions
of the HRES can be continued for the other scenarios. As mentioned, a for saving fossil fuels. Hence, it can be concluded that within high re-
DSM strategy is implemented and tested under a case study of a typical newable contribution (RF = 100%), PV/battery (Li-ion) using the
residential building, and the aim is to reduce the system investment suggested DSM technique is the optimum solution for the electrification

Fig. 11. The convergence of PSO-based MATLAB code for HRES with (a) L-acid batteries; (b) Li-ion batteries.

12
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

Table 6
Validation of the developed optimization model with HOMER.
Optimizer PV (kW) WT (Nbr) DG (kW) L-acid (kWh) Li-ion (kWh) Converter (kW) TNPC ($) LPSP (%) COE ($) Initial cost ($) Excess (kWh/year)

HOMER 8.61 0 0 / 15 3.32 27,361 2.1 0.25 20,603 8478

PSO 8.38 / 15 4.45 28,314 2 0.26 21,018 6494
HOMER 8.96 0 0 20 / 3.44 31,110 2.1 0.29 20,337 8801
PSO 8.42 20 / 4.45 31,379 2 0.29 20,187 6569

Table 7
HRES’s design within L-acid battery with RF = 100%
RFminimum DSM PV WT DG L-acid Converter TNPC RF LPSP
(%) (kW) (Nbr) (kW) (kWh) (kW) ($) (%) (%)

100 NO 8.42 0 0 20 4.45 31,379 100 2

YES 6.96 0 0 16 4.18 25,717 100 2
Only DG 0 0 4 0 0 26,664 0 0.05

Table 8
HRES’s design within L-acid battery with RF = 100% (Continue).
RFminimum DSM COE Fuel CO2 Excess load_24 Load_29 Demand
(%) ($/kWh) (L/year) (Kg/year) (kWh) (kWh) (kWh) (kWh)

100 NO 0.29 0 0 6569 8698 6948 8698

YES 0.26 0 0 4648 8698 6948 8050
Only DG 0.245 4993 11,983 0 8698 6948 8698

Table 9
HRES’s design within Li-ion battery with RF = 100%
RFminimum DSM PV WT DG Li-ion Converter TNPC RF LPSP
(%) (kW) (Nbr) (kW) (kWh) (kW) ($) (%) (%)

100 NO 8.38 0 0 15 4.45 28,314 100 2

YES 7 0 0 11 4.18 23,427 100 2
0 NO 0.68 0 3 0 0.59 19,422 15 1.39
YES 0.21 1 2 0 2.17 18,383 66 1.43
Only DG 0 0 4 0 0 26,664 0 0.05

Table 10
HRES’s design within Li-ion battery with RF = 100% (Continue).
RFminimum DSM COE Fuel CO2 Excess load_24 Load_29 Demand
(%) ($/kWh) (L/year) (Kg/year) (kWh) (kWh) (kWh) (kWh)

100 NO 0.26 0 0 6494 8698 6948 8698

YES 0.23 0 0 4725 8698 6948 8055
0 NO 0.18 3904 9370 7 8698 6948 8698
YES 0.21 1667 4001 2598 8698 6948 7052
Only DG 0.245 4993 11,983 0 8698 8698 6948

of the studied building. turbine and DG. In this case, the building energy consumption is de-
creased by 19% and the contribution of DG is reduced extensively,
4.3. HRES’s design with a low renewable fraction where a RF equal to 63% and a COE of 0.21 $/kWh are reached.
However, for both scenarios (with and without DSM), it can be seen
In this case, the minimum RF value is imposed at 0% for the opti- that no battery storage is required as the wind turbine can work in the
mization. Tables 11 and 12 provide the results for this simulation case, daytime as well as in the night time. It is concluded that with low RF
using the characteristics of both battery technologies (i.e. L-acid or L- values, the COE for both scenarios is lower than in the case of using a
ion). stand-alone DG system. Nevertheless, while applying the DSM, the re-
With RFminimum = 0%, the contribution of the HRES’s components newable energy contribution is increased noticeably. Therefore, wind/
and the hourly energy demand of the building is presented in Figs. 16 DG HRES is the best configuration for the case study building with low
and 17, respectively. renewable fraction values.
By imposing a minimum RF = 0% and without applying the pro-
posed DSM technique, PV/diesel is found to be the best configuration. 4.4. Comparing results and summary
In this case, a low renewable fraction is reached (15%), in which the
majority of load demand is supplied by DG and with slight contribution The results indicate the fact that the TNPC and COE of the HRES can
from PV modules. However, after introducing the DSM strategy in the be increased while increasing the contribution of renewable energy.
optimization, the optimal HRES configuration includes a small wind Conversely, CO2 emission and fuel consumption can also be reduced.

13
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

Fig. 12. Contribution of HRES’s components for HRES with L-acid batteries (RF = 100%).

Fig. 13. Contribution of HRES’s components for HRES with Li-ion batteries (RF = 100%).

Hence, the design of the best HRES that presents optimality from HRES is smaller than the value of the COE obtained with using a stand-
techno-economic and environment perspectives is one of the main alone DG system that is merely 0.245$/kWh.
reasons for performing sensitivity analysis on renewable fraction and Comparing the obtained results, the following improvements can be
component characteristics. made by introducing the suggested DSM: i.e. with RF = 100%, the
With RFminimum = 0%, it is found that the COE of the obtained energy demand, TNPC and electricity excess are decreased by 7%, 18%

14
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

Fig. 14. Load power for HRES with L-acid batteries (RF = 100%).

Fig. 15. Load power for HRES with Li-ion batteries (RF = 100%).

and 29%, respectively. In such a case, the best solution includes PV/ These results have demonstrated that the investment cost of small-scale
battery (Li-ion), with the TNPC and COE being 23,427 $ and 0.23 $/ WT is still very high compared to PV modules.
kWh, respectively. However, by imposing 0% as minimum RF, the en- The findings of this work have proved the effectiveness of the sug-
ergy demand is reduced by 19% and fuel consumption and CO2 emis- gested approach, as it is able to address the demand–supply mismatch
sions are minimized by 57%. Conversely, the RF is raised from 15% problem, limit the oversizing of components and reduce the building
(using solely Supply-side management) to 63% (DSM) where the op- energy demand. Further extending this model to include additional
timal configuration consists of PV/wind/diesel, with COE of 0.21$/ influencing criteria, such as thermal comfort requirements and occu-
kWh, which is smaller than that using a stand-alone DG. Moreover, for pant preferences would broaden the optimization potential of the pro-
both simulation cases, small-scale WTs do not provide viable solution posed method. We anticipate additional changes to be made, howe-
without applying the proposed DSM strategy, despite that the present ver these are outside the scope of this paper and subject for future
case study has already considered the windiest location in the country. studies.

Table 11
HRES’s design results with RFminimum = 0%
RFminimum DSM PV WT DG BS Converter TNPC RF LPSP
(%) (kW) (Nbr) (kW) (kWh) (kW) ($) (%) (%)

0 NO 0.68 0 3 0 0.59 19,422 15 1.39

YES 0 1 2 0 2.17 18,471 63 1.53
Only DG 0 0 4 0 0 26,664 0 0.05

15
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

Table 12
HRES’s design results with RFminimum = 0% (Continue).
RFminimum DSM COE Fuel CO2 Excess load_24 Load_29 Demand
(%) ($/kWh) (L/year) (Kg/year) (kWh) (kWh) (kWh) (kWh)

0 NO 0.18 3904 9370 7 8698 6948 8698

YES 0.21 1757 4217 2135 8698 6948 7028
Only DG 0.245 4993 11,983 0 8698 8698 6948

5. Conclusions beyond.

This paper has presented a novel approach for managing both the CRediT authorship contribution statement
demand and supply sides through integrating a demand–supply man-
agement technique with a PSO-based algorithm for reducing building Charafeddine Mokhtara: Data curation, Investigation,
energy consumption and thus meeting the load demand at the lowest Methodology, Resources, Software, Validation, Visualization, Writing-
possible costs. The novelty of this work is related to the integrated original draft. Belkhir Negrou: Supervision, Methodology,
demand–supply management strategy, which is based on the state of Conceptualization. Abdessalem Bouferrouk: Methodology,
supply system and the operation of the air conditioner, i.e. we have Conceptualization, Writing-review & editing. Yufeng Yao:
developed a control system in which the set point temperature for air Methodology, Conceptualization, Writing-review & editing.
conditioner is changed (between 24 °C and 29 °C according to the Noureddine Settou: Methodology, Conceptualization. Mohamad
available energy from renewable sources. Ramadan: Methodology, Writing-review & editing.
From the obtained results, the following improvements are achieved
by integrating the proposed DSM. Fist, with high renewable contribu- Declaration of Competing Interest
tion, the energy demand and TNPC are decreased by 7% and 18%,
respectively, for which, PV/battery (Li-ion) is found to be the best
The authors declare that they have no known competing financial
configuration, with TNPC and COE of 23,427 $ and 0.23$/kWh, re-
interests or personal relationships that could have appeared to influ-
spectively. Second, with low renewable usage, the following reductions
ence the work reported in this paper.
are observed including the building energy demand (19%) and both
fuel consumption and CO2 emissions (57%). In contrast, the RF is raised
Acknowledgment
from 15% (without DSM) to 63% (with DSM), where the optimal con-
figuration consists of wind/diesel, with COE 0.21 $/kWh, being smaller
The first author would like to thank the research team of VPRS la-
than that using a stand-alone DG.
boratory and the University of Kasdi Merbah Ouargla, Algeria, for their
The findings of this work have proved the effectiveness of the sug-
support. Many thanks also go to DGRSDT of the Ministry for Scientific
gested approach for the chosen case study in terms of location and
Research, Algeria. He also wishes to thank the Engineering Modelling
layout. Future work can be followed, e.g. for creating a generalized
and Simulation research group at the Department of Engineering,
model feasible for other locations and layouts in the country and
Design and Mathematics, University of the West of England, Bristol, UK

Fig. 16. Contribution of HRES’s components for HRES with RFminimum = 0%.

16
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

Fig. 17. Energy demand for HRES with RFminimum = 0%.

for hosting his short study abroad during a research visit. [16] Jamshidi M, Askarzadeh A. Techno-economic analysis and size optimization of an
off-grid hybrid photovoltaic, fuel cell and diesel generator system. Sustainable
Cities Soc 2019;44:310–20. https://doi.org/10.1016/j.scs.2018.10.021.
References [17] Moradi H, Esfahanian M, Abtahi A, Zilouchian A. Optimization and energy man-
agement of a standalone hybrid microgrid in the presence of battery storage system.
[1] Ahmad MW, Mourshed M, Yuce B, Rezgui Y. Computational intelligence techniques Energy 2018. https://doi.org/10.1016/j.energy.2018.01.016.
for HVAC systems: a review. Build Simul 2016:359–98. https://doi.org/10.1007/ [18] Eriksson EL V, Gray EM. Optimization of renewable hybrid energy systems – a
s12273-016-0285-4. multi-objective approach. Renewable Energy 2018. https://doi.org/10.1016/j.
[2] Mokhtara C, Negrou B, Settou N, Gouareh A, Settou B, CMA. Decision-making and renene.2018.10.053.
optimal design of off-grid hybrid renewable energy system for electrification of [19] Zhang W, Maleki A, Rosen MA, Liu J. Optimization with a simulated annealing
mobile buildings in Algeria: case study of drilling camps in Adrar. Alger J Env Sci algorithm of a hybrid system for renewable energy including battery and hydrogen
Technol 2020. storage. Energy 2018;163:191–207. https://doi.org/10.1016/j.energy.2018.08.
[3] L’Energie M de. Bilan des réalisation du sector de l’Energie. < http://www.energy. 112.
gov.dz/francais/uploads/MAJ_2018/Stat/Bilan_des_Realisations_du_secteur_2017_ [20] Zhang W, Maleki A, Rosen MA, Liu J. Sizing a stand-alone solar-wind-hydrogen
%C3%A9dition_2018.pdf > ; 2018. energy system using weather forecasting and a hybrid search optimization algo-
[4] Government A. Renewable Energy and Energy Efficiency Development Plan 2015- rithm. Energy Convers Manage 2019;180:609–21. https://doi.org/10.1016/j.
2030. IEA. < https://www.iea.org/policiesandmeasures/renewableenergy/? enconman.2018.08.102.
country=ALGERIA > ; 2015. [21] Mellouk L, Ghazi M, Aaroud A, Boulmalf M, Benhaddou D, Zine-dine K. Design and
[5] Haddoum S, Bennour H, Zaïd TA. Algerian Energy Policy: Perspectives , Barriers , energy management optimization for hybrid renewable energy system-case study:
and Missed Opportunities. vol. 1700134. < https://doi.org/10.1002/gch2. Laayoune region. Renewable Energy 2019;139:621–34. https://doi.org/10.1016/j.
201700134 > ; 2018. renene.2019.02.066.
[6] Thiaux Y, Dang TT, Schmerber L, Multon B, Ben Ahmed H, Bacha S, et al. Demand- [22] Yu H, Zhang C, Deng Z, Bian H, Sun C, Jia C. Economic optimization for config-
side management strategy in stand-alone hybrid photovoltaic systems with real- uration and sizing of micro integrated energy systems. J Mod Power Syst Clean
time simulation of stochastic electricity consumption behavior. Appl Energy Energy 2018;6:330–41. https://doi.org/10.1007/s40565-017-0291-2.
2019;253:113530https://doi.org/10.1016/j.apenergy.2019.113530. [23] Kaabeche A, Bakelli Y. Renewable hybrid system size optimization considering
[7] Hossain MA, Pota HR, Squartini S, Zaman F, Guerrero JM. Energy scheduling of various electrochemical energy storage technologies. Energy Convers Manage
community microgrid with battery cost using particle swarm optimisation. Appl 2019;193:162–75. https://doi.org/10.1016/j.enconman.2019.04.064.
Energy 2019;254:113723https://doi.org/10.1016/j.apenergy.2019.113723. [24] Hamanah WM, Abido MA, Alhems LM. Optimum sizing of hybrid PV, wind, battery
[8] Faccio M, Gamberi M, Bortolini M, Nedaei M. State-of-art review of the optimization and diesel system using lightning search algorithm. Arab J Sci Eng 2019. https://
methods to design the configuration of hybrid renewable energy systems (HRESs). doi.org/10.1007/s13369-019-04292-w.
Front Energy 2018;12:591–622. https://doi.org/10.1007/s11708-018-0567-x. [25] Haratian M, Tabibi P, Sadeghi M, Vaseghi B, Poustdouz A. A renewable energy
[9] Mehrjerdi H. Modeling, integration, and optimal selection of the turbine technology solution for stand-alone power generation: a case study of KhshU Site-Iran.
in the hybrid wind-photovoltaic renewable energy system design. ENERGY Convers Renewable Energy 2018;125:926–35. https://doi.org/10.1016/j.renene.2018.02.
Manage 2020;205:112350https://doi.org/10.1016/j.enconman.2019.112350. 078.
[10] Abo-elyousr FK, Nozhy AN. Bi-objective economic feasibility of hybrid micro-grid [26] Fodhil F, Hamidat A, Nadjemi O. Energy control strategy analysis of hybrid power
systems with multiple fuel options for islanded areas in Egypt. Renewable Energy generation system for rural Saharan community in Algeria. Lect Notes Networks
2018. https://doi.org/10.1016/j.renene.2018.05.066. Syst 2018;35:108–20. https://doi.org/10.1007/978-3-319-73192-6_12.
[11] Ma T, Javed MS. Integrated sizing of hybrid PV-wind-battery system for remote [27] Fodhil F, Hamidat A, Nadjemi O. Potential, optimization and sensitivity analysis of
island considering the saturation of each renewable energy resource. Energy photovoltaic-diesel-battery hybrid energy system for rural electri fi cation in
Convers Manage 2019;182:178–90. https://doi.org/10.1016/j.enconman.2018.12. Algeria. Energy 2019;169:613–24. https://doi.org/10.1016/j.energy.2018.12.049.
059. [28] Duman AC, Güler Ö. Techno-economic analysis of off-grid PV/wind/fuel cell hybrid
[12] Assaf J, Shabani B. A novel hybrid renewable solar energy solution for continuous system combinations with a comparison of regularly and seasonally occupied
heat and power supply to standalone-alone applications with ultimate reliability households. Sustainable Cities Soc 2018;42:107–26. https://doi.org/10.1016/j.scs.
and cost effectiveness. Renewable Energy 2019;138:509–20. https://doi.org/10. 2018.06.029.
1016/j.renene.2019.01.099. [29] Anoune K, Laknizi A, Bouya M, Astito A, Ben Abdellah A. Sizing a PV-Wind based
[13] Tudu B, Mandal KK, Chakraborty N. Optimal design and development of PV-wind- hybrid system using deterministic approach. Energy Convers Manage
battery based nano-grid system: a field-on-laboratory demonstration. Front Energy 2018;169:137–48. https://doi.org/10.1016/j.enconman.2018.05.034.
2019;13:269–83. https://doi.org/10.1007/s11708-018-0573-z. [30] Jafar M, Moghaddam H, Kalam A, Arabi S. Optimal sizing and energy management
[14] Dufo-López R, Champier D, Gibout S, Lujano-Rojas JM, Domínguez-Navarro JA. of stand-alone hybrid photovoltaic/wind system based on hydrogen storage con-
Optimisation of off-grid hybrid renewable systems with thermoelectric generator. sidering LOEE and LOLE reliability indices using fl ower pollination algorithm.
Energy Convers Manage 2019;196:1051–67. https://doi.org/10.1016/j.enconman. Renewable Energy 2019;135:1412–34. https://doi.org/10.1016/j.renene.2018.09.
2019.06.057. 078.
[15] Adefarati T, Bansal RC. Reliability, economic and environmental analysis of a mi- [31] Das M, Anil M, Singh K, Biswas A. Techno-economic optimization of an off-grid
crogrid system in the presence of renewable energy resources. Appl Energy hybrid renewable energy system using metaheuristic optimization approaches –
2019;236:1089–114. https://doi.org/10.1016/j.apenergy.2018.12.050. case of a radio transmitter station in India. Energy Convers Manage
2019;185:339–52. https://doi.org/10.1016/j.enconman.2019.01.107.

17
C. Mokhtara, et al. Energy Conversion and Management 221 (2020) 113192

[32] Halabi LM, Mekhilef S, Olatomiwa L, Hazelton J. Performance analysis of hybrid criteria decision-making. Energy Procedia 2019;162:171–80. https://doi.org/10.
PV/diesel/battery system using HOMER: a case study Sabah, Malaysia. Energy 1016/j.egypro.2019.04.019.
Convers Manage 2017;144:322–39. https://doi.org/10.1016/j.enconman.2017.04. [43] EnergyPlus n.d. < https://energyplus.net > (accessed June 11, 2020).
070. [44] El-bidairi KS, Duc H, Jayasinghe SDG, Mahmoud TS, Penesis I. A hybrid energy
[33] Rullo P, Braccia L, Luppi P, Zumoffen D, Feroldi D. Integration of sizing and energy management and battery size optimization for standalone microgrids: a case study
management based on economic predictive control for standalone hybrid renewable for Flinders Island, Australia. Energy Convers Manage 2018;175:192–212. https://
energy systems. Renewable Energy 2019;140:436–51. https://doi.org/10.1016/j. doi.org/10.1016/j.enconman.2018.08.076.
renene.2019.03.074. [45] Mandal S, Das BK, Hoque N. Optimum sizing of a stand-alone hybrid energy system
[34] Anvari-moghaddam A, Rahimi-kian A, Mirian MS, Guerrero JM. A multi-agent for rural electrification in Bangladesh. J Clean Prod 2018;200:12–27. https://doi.
based energy management solution for integrated buildings and microgrid system. org/10.1016/j.jclepro.2018.07.257.
Appl Energy 2017;203:41–56. https://doi.org/10.1016/j.apenergy.2017.06.007. [46] Mohamed MA, Eltamaly AM, Alolah AI, Hatata AY, Mohamed MA, Eltamaly AM,
[35] Uddin M, Romlie MF, Abdullah MF, Abd Halim S, Abu Bakar AH, Chia Kwang T. A et al. A novel framework-based cuckoo search algorithm for sizing and optimization
review on peak load shaving strategies. Renewable Sustainable Energy Rev of grid-independent hybrid renewable energy systems. Int J Green Energy
2018;82:3323–32. https://doi.org/10.1016/j.rser.2017.10.056. 2018;00:1–15. https://doi.org/10.1080/15435075.2018.1533837.
[36] Tu T, Rajarathnam GP, Vassallo AM. Optimization of a stand-alone photovoltaic e [47] Mousavi N, Kothapalli G, Habibi D, Das CK, Baniasadi A. A novel photovoltaic-
wind e diesel e battery system with multi-layered demand scheduling. Renewable pumped hydro storage microgrid applicable to rural areas. Appl Energy
Energy 2019;131:333–47. https://doi.org/10.1016/j.renene.2018.07.029. 2020;262:114284https://doi.org/10.1016/j.apenergy.2019.114284.
[37] Zheng Y, Jenkins BM, Kornbluth K, Kendall A, Træholt C. Optimization of a bio- [48] Das BK, Zaman F. Performance analysis of a PV/Diesel hybrid system for a remote
mass-integrated renewable energy microgrid with demand side management under area in Bangladesh: effects of dispatch strategies, batteries, and generator selection.
uncertainty. Appl Energy 2018;230:836–44. https://doi.org/10.1016/j.apenergy. Energy 2019;169:263–76. https://doi.org/10.1016/j.energy.2018.12.014.
2018.09.015. [49] Pro H. Renewable fraction n.d. < https://www.homerenergy.com/products/pro/
[38] Yang F, Xia X. Techno-economic and environmental optimization of a household docs/latest/_renewable_fraction.html > (accessed June 12, 2020).
photovoltaic-battery hybrid power system within demand side management. [50] Shara M, Elmekkawy TY. Multi-objective optimal design of hybrid renewable en-
Renewable Energy 2017. https://doi.org/10.1016/j.renene.2017.02.054. ergy systems using PSO-simulation based approach. Renewable Energy 2014:68.
[39] Sarkar T, Bhattacharjee A, Samanta H, Bhattacharya K, Saha H. Optimal design and https://doi.org/10.1016/j.renene.2014.01.011.
implementation of solar PV-wind-biogas-VRFB storage integrated smart hybrid [51] Twaha S, Ramli MAM. A review of optimization approaches for hybrid distributed
microgrid for ensuring zero loss of power supply probability. Energy Convers energy generation systems: off-grid and grid-connected systems. Sustainable Cities
Manage 2019;191:102–18. https://doi.org/10.1016/j.enconman.2019.04.025. Soc 2018;41:320–31. https://doi.org/10.1016/j.scs.2018.05.027.
[40] Wu Z, Tazvinga H, Xia X. Demand side management of photovoltaic-battery hybrid [52] Mohamed MA, Eltamaly AM, Alolah AI. Sizing and techno-economic analysis of
system. Appl Energy 2015;148:294–304. https://doi.org/10.1016/j.apenergy.2015. stand-alone hybrid photovoltaic/wind/diesel/battery power generation systems.
03.109. < https://doi.org/10.1063/1.4938154 > ; 2015.
[41] Mohseni S, Brent AC, Burmester D. A demand response-centred approach to the [53] Samy MM, Barakat S, Ramadan HS. A flower pollination optimization algorithm for
long-term equipment capacity planning of grid-independent micro-grids optimized an off-grid PV-Fuel cell hybrid renewable system. Int J Hydrogen Energy
by the moth-flame optimization algorithm. Energy Convers Manage 2019;44:2141–52. https://doi.org/10.1016/j.ijhydene.2018.05.127.
2019;200:112105https://doi.org/10.1016/j.enconman.2019.112105. [54] Samy MM, Barakat S, Ramadan HS. Techno-economic analysis for rustic elec-
[42] Mokhtara C, Negrou B, Settou N, Gouareh Abderrahmane BS. Pathways to plus- trification in Egypt using multi-source renewable energy based on PV/wind/FC. Int
energy buildings in Algeria: design optimization method based on GIS and multi- J Hydrogen Energy 2019. https://doi.org/10.1016/j.ijhydene.2019.04.038.

18