International Journal of Power Electronics and Drive Systems (IJPEDS)

Vol. 15, No. 2, June 2024, pp. 1235~1246

ISSN: 2088-8694, DOI: 10.11591/ijpeds.v15.i2.pp1235-1246  1235

Multi-objective algorithm for hybrid microgrid energy

management based on multi-agent system

Ilham Tyass1, Abdelouahad Bellat1, Abdelhadi Raihani2, Khalifa Mansouri1

1
Signals, Distributed Systems and Artificial Intelligence (SSDIA) Laboratory, ENSET Mohammedia,
Hassan II University of Casablanca, Casablanca, Morocco
2
Electrical Engineering and Intelligent Systems (EEIS) Laboratory, ENSET Mohammedia,
Hassan II University of Casablanca, Casablanca, Morocco

Article Info ABSTRACT

Article history: In the dynamic landscape of renewable energies, microgrid systems emerge
as a promising avenue for fostering sustainable local energy generation.
Received Aug 20, 2023 However, the effective management of energy resources holds the key to
Revised Dec 21, 2023 unlocking their full potential. This study assumes the task of creating a multi-
Accepted Jan 5, 2024 objective optimization algorithm for microgrid energy management. At its
core, the algorithm places a premium on seamlessly integrating renewable
energy sources and orchestrating efficient storage coordination. Leveraging
Keywords: the prowess of a multi-agent system, it allocates and utilizes energy resources.
Through the combination of renewable sources, storage mechanisms, and
Hybrid microgrid variable loads, the algorithm promotes energy efficiency and ensures a steady
Multi-agent system power supply. This transformative solution is underscored by the algorithm's
Optimization algorithm remarkable performance in practical simulations and validations across
Renewable energy diverse microgrid scenarios, offering a prevue into the future of sustainable
Storage management energy utilization.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Ilham Tyass
Electrical Engineering and Intelligent Systems (EEIS) Laboratory, ENSET Mohammedia
Hassan II University of Casablanca
Casablanca, Morocco
Email: tyass1@yahoo.fr

NOMENCLATURE
DDC : Demand for DC loads PSG : Power supplied from grid
DAC : Demand for AC loads Pexcess : Wind and PV power excess
PPV : Photovoltaic energy production BATmax : Maximal battery capacity
Pw : Wind energy production CAPmax : Maximal capacitor capacity
PDC : Power supplied to DC loads BOC : Battery state of charge
PAC : Power supplied to AC loads SOC : Supercapacitor state of charge
PSPV : Power supplied by PV Fth : Rapid fluctuation threshold
PSW : Power supplied wind power LCmin : Minimum charge level of the supercapacitor
PSB : Power supplied by battery LBmin : Minimum charge level of the battery
PSS : Power supplied by supercapacitor DFth : Rapid fluctuation threshold of the demand

1. INTRODUCTION
In the realm of renewable energies, microgrid systems offer a promising avenue for local and
sustainable energy generation. Effectively harnessing energy resources constitutes a major challenge within

Journal homepage: http://ijpeds.iaescore.com

1236  ISSN: 2088-8694

this context. The intricate interplay among renewable energy sources, storage mechanisms, and consumption
patterns demands sophisticated strategies to ensure uninterrupted power supply while optimizing available
resources. The literature has explored various approaches to microgrid operations within energy management
paradigms [1]–[3]. However, the potential of artificial intelligence techniques has notably emerged, particularly
in systems exhibiting behaviors akin to microgrids [4]. Artificial intelligence and multi-agent systems have
demonstrated exceptional performance in domains such as network management, intelligent platform
interfaces, and database administration. For instance, multi-agent system or MAS are utilized to meet network
energy demands, adjust power based on surplus and shortage information, and choose from various options,
including coordination with power grids, battery storage systems, and controllable distributed generation plants
[5]. Similarly, an intelligent bidding tactic employing a continuous double auction was implemented, enabling
customer engagement in demand response initiatives [6]. In research [7], a multi-agent control mechanism was
introduced for buildings, where agents operate according to a newly suggested comfort metric. Similarly,
another building control system was introduced [8], emphasizing the management of energy consumption.
These examples highlight the relevance and effectiveness of MAS in complex energy management. Energy
management through MAS for implementing a hybrid system at high altitude is discussed in research [9], based
on local information, to ensure efficient and stable system operation, distributed generation sources within
microgrids are regulated by an energy management system. Some studies have also addressed Enhancing
microgrid dependability via distributed power control for distributed energy resources, utilizing network and
MAS along with communication delay technologies to efficiently oversee distributed demand [10].
Our investigation is rooted in the assumption that each agent resides within a multi-agent system,
closely linked to specific microgrid elements. A distinctive feature of our approach lies in merging the MAS
framework with an optimization algorithm. This fusion enhances the integration and efficient utilization of
renewable energies while optimizing storage, employing batteries and supercapacitors within a hybrid
architecture. This novel synergy optimizes energy distribution, ensuring efficient and balanced power supply.
The primary aim of our work is to develop a multi-objective optimization algorithm for microgrid energy
management. This algorithm prioritizes renewable energy integration and efficient coordination of storage
between batteries and supercapacitors.
The structure of this article entails examining fundamental aspects and requirements in section 1 that
frame our energy management approach within microgrid systems. Moving forward to section 2, we outline
ambitious energy management goals within microgrid systems. The microgrid architecture and the proposed
multi-agent’s system architecture are presented in section 3.1. In section 3.2, we develop an adaptive algorithm
designed to coordinate interactions among microgrid components. Section 4 focuses on practical simulation
and validation, translating theoretical expertise into empirical confirmation of the efficiency and adaptability
of our algorithmic framework in various real microgrid scenarios. Throughout each of these sections, our goal
is to comprehensively present our innovative energy management approach and highlight its transformative
potential for sustainable energy utilization within microgrid systems.

2. ENERGY MANAGEMENT GOALS FOR A SUSTAINABLE MICROGRID: PROPOSED

APPROACH
Within the scope of our approach to microgrid management using a multi-agent system, several
objectives have been established to optimize energy utilization and enhance the overall system performance.
Among these objectives, we focus on four primary goals that are of paramount importance for the sustainable
development of the microgrid. First, our priority is to ensure the environmental sustainability of the microgrid.
The seamless integration of renewable energy sources is a fundamental objective in this regard. By maximizing
the proportion of renewable energy in the energy mix, our aim is to significantly reduce fossil fuel consumption
and greenhouse gas emissions [11], [12]. This approach will contribute to environmental preservation and
facilitate a transition towards cleaner and renewable energy. Furthermore, cost minimization is a key objective
in microgrid management [13]. Through the intelligent use of batteries and supercapacitors, we seek to reduce
electricity procurement costs from the main grid by storing excess energy generated from renewable sources.
This optimization of resource management aims to achieve maximum economic efficiency.
Ensuring the reliability and resilience of the microgrid is also a major priority. By employing batteries
and supercapacitors to ensure stable power supply, we aim to promptly detect production or demand
fluctuations within the system. This approach will maintain network stability and ensure uninterrupted power
supply, even during external disruptions. Lastly, energy efficiency lies at the core of our objectives. Through
optimal management of energy sharing among renewable sources for mutual DC and AC load feeding, coupled
with the use of machine learning algorithms to anticipate energy production and demand, we aim to maximize
the utilization of available resources and minimize energy losses. This approach will enable us to optimize the
overall efficiency of the microgrid and reduce energy wastage.

Int J Pow Elec & Dri Syst, Vol. 15, No. 2, June 2024: 1235-1246
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1237

2.1. Cause-and-effect relationships among primary objectives

These four primary objectives are closely interconnected, with cause-and-effect relationships
highlighting their interdependence. These cause-and-effect relationships underscore the interdependence of
primary objectives, illustrating how targeted actions in one domain can have positive implications in others,
contributing to a balanced and sustainable overall energy management in the microgrid. The diagram in
Figure 1 illustrates these various relationships.

Figure 1. Cause-and-effect relationships among primary objectives

A balanced and integrated approach to energy management will synergistically achieve these goals.
The use of a multi-agent system will enable a coherent and coordinated optimization of these objectives,
ensuring optimal energy management for the microgrid [14], [15].
a) Environmental sustainability <=> energy efficiency:
- Cause: By adopting measures to enhance energy efficiency, the microgrid reduces energy consumption,
contributing to the preservation of natural resources and the reduction of greenhouse gas emissions.
- Effect: Improved environmental sustainability prompts the microgrid to embrace more efficient energy
practices, as heightened awareness of the importance of reducing energy consumption is reinforced.
b) Cost minimization <=> energy efficiency:
- Cause: Through improved energy efficiency, the microgrid reduces energy consumption, leading to
decreased electricity procurement and maintenance costs.
- Effect: Cost minimization motivates the microgrid to invest in energy efficiency solutions, as substantial
long-term financial savings can be realized.
c) Cost minimization <=> network reliability and resilience:
- Cause: Cost minimization drives the microgrid to optimize resource utilization and invest in cutting-
edge technologies to lower operational costs, potentially enhancing network reliability and resilience.
- Effect: Greater network reliability and resilience mitigate costly disruptions and economic losses,
thereby contributing to long-term cost minimization.
d) Energy efficiency <=> network reliability and resilience:
- Cause: More efficient energy utilization reduces demand peaks and network strain, potentially
improving network stability and resilience.
- Effect: Enhanced network reliability and resilience maintain optimal energy performance by reducing
the risk of outages and interruptions that could affect energy efficiency.
e) Network reliability and resilience <=> environmental sustainability:
- Cause: A more reliable and resilient network can better handle fluctuations in renewable energy
production, ensuring optimal utilization of these clean energy sources and reinforcing environmental
sustainability.
- Effect: Environmental sustainability drives the microgrid to invest in technologies and practices that
enhance network reliability and resilience, preserving the environmental benefits of renewable energy
sources.

2.2. Mechanisms for achieving primary objectives

The sub-objectives of integrating or maximizing renewable energies, collaborative energy sharing for
enhanced efficiency, optimized use of storage, demand and production prediction, and control play a crucial
role in achieving the four primary objectives of microgrid energy management. For these objectives, the use
of specific technological tools proves essential. Firstly, the multi-agent system emerges as a significant asset
for coordinating the diverse entities within the microgrid [16], such as renewable energy sources, storage
systems, and consumers, allowing effective communication and decentralized decision-making.
Simultaneously, renewable energy prediction through advanced models and artificial intelligence algorithms
Multi-objective algorithm for hybrid microgrid energy management based on multi-agent … (Ilham Tyass)
1238  ISSN: 2088-8694

is a key element in anticipating the availability of clean energy sources, thereby facilitating their optimal
integration into the microgrid [17]–[19]. Thus, the microgrid can aim for advanced, resilient, cost-effective,
and environmentally friendly energy management, providing a promising energy future for local communities.
The diagram presented below in Figure 2 illustrates the objectives and means associated with energy
optimization within our microgrid. This visual representation aims to highlight the interconnectedness between
various energy targets. Each objective constitutes a fundamental pillar of our comprehensive approach.

Figure 2. The objectives and means associated with energy optimization

3. THE DEVELOPED ALGORITHM FOR MICROGRID MANAGEMENT USING A MULTI-

AGENT SYSTEM APPROACH
3.1. Modelling the smart grid as a multi-agent system
In this section, we delve into modeling the microgrid as a multi-agent system. This approach considers
the microgrid components as individual agents, each with its own set of behaviors, roles, and interactions. By
doing so, we can effectively capture the complex dynamics and interactions within the microgrid system. The
agents can encompass various entities such as renewable energy sources, loads, energy storage systems, and
even the central grid connection. The microgrid's operations and decision-making processes are distributed
among these agents, allowing for autonomous actions and local optimization strategies. This modeling
approach enables us to achieve a holistic view of the microgrid's behavior as well as taking into account the
interactions between the different components and their responses to various conditions, such as changing
energy availability and demand fluctuations.

3.1.1. The proposed architectural plan for microgrid

The envisaged design for the microgrid entails the integration of renewable energy sources (RE)
alongside a diverse array of AC and DC loads. This configuration is supervised by an innovative modular
system rooted in MAS. The chosen configuration is a hybrid architecture, combining the benefits of both DC
bus and AC bus architectures by employing AC/DC and DC/AC power converters. Renewable energy source
generators are connected to a DC bus, where energy is stored using energy storage devices and subsequently
converted into alternating current (AC) through DC/AC power converters to supply AC loads. This architecture
provides greater flexibility in energy management by leveraging the advantages of direct current energy storage
and alternating current energy injection into the electrical grid [20], [21].The typical hybrid bus configuration
is depicted in Figure 3 [22].

3.1.2. Proposed multi-agent system

Our energy management approach is defined by the collaborative interaction among various
components within the system and the grid to accomplish predetermined goals. In our model, agents function
as entities integrated across the grid infrastructure, encompassing renewable energy generation sources, energy
storage units, and intelligent sensors in residential settings. Our model introduces the deployment of eight
distinct agent types, each delineated alongside their interrelationships as depicted in Figure 4.

Int J Pow Elec & Dri Syst, Vol. 15, No. 2, June 2024: 1235-1246
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1239

Figure 3. The typical configuration of the hybrid bus microgrid

Figure 4. The proposed multi-agent’s system

A key feature of our system lies in its capacity to store excess energy generated by renewable sources.
Storage agents, whether batteries or supercapacitors, intervene based on the fluctuation of energy production
and demand. The decision to use supercapacitors or batteries depends on the dynamics of load fluctuations and
energy production fluctuations. Supercapacitors are favored for rapid and frequent changes, both for load
supply and excess energy storage, while batteries are better for more stable or gradually changing conditions
[23]. The algorithm assesses these conditions in real-time to make informed choices between supercapacitors
and batteries, ensuring efficient load supply and energy storage in the microgrid.
Furthermore, our system ensures consistency and coordination among different energy sources and
load types. For instance, surplus solar energy is utilized to power DC loads, while excess wind energy caters
to AC loads. If necessary, this surplus energy can be redistributed to power other load types through
bidirectional converters.
In the multi-agent system, various agents play distinct roles in managing the microgrid effectively.
The supervisor agent oversees the entire system, gathering data from other agents, conducting analyses, and
making global decisions to optimize renewable energy utilization while balancing supply and demand [24].
Wind and photovoltaic energy prediction agents utilize models to forecast future energy production, aiding in
precise energy management planning. [25]. AC and DC demand agents monitor the energy needs of their
respective loads, adjusting consumption based on predictions [26]. The battery agent manages energy storage,
determining when to store or release energy. The supercapacitor agent intervenes when energy fluctuations
exceed a set threshold, rapidly supplying energy as needed. Finally, the network agent handles connections to
the external grid, serving as a backup when other sources are insufficient or storage systems are depleted.
Power and information flow within the microgrid operate in tandem to ensure efficient energy
management and distribution [27]. The power flow initiates from renewable energy sources such as wind and
solar, which supply energy to AC and DC loads according to their specific demands. Any surplus energy can
Multi-objective algorithm for hybrid microgrid energy management based on multi-agent … (Ilham Tyass)
1240  ISSN: 2088-8694

be stored in batteries or supercapacitors for later use. If local energy demands exceed available supply, the grid
can be utilized as a supplementary source. Simultaneously, the information flow facilitates coordination among
agents. Prediction agents furnish forecasts to demand and storage agents, enabling proactive energy planning.
Demand agents relay current and projected energy requirements to storage agents and the supervisor for
optimization. Storage agents, informed by this data and predefined thresholds, make decisions regarding energy
storage and release. Finally, the supervisor integrates information from all agents to make global decisions
aimed at maximizing renewable energy utilization and meeting energy demands effectively.
The interactions between agents within the microgrid system are vital for efficient energy management
and optimization. Initially, prediction agents generate forecasted data on energy production, considering factors
like weather conditions and historical trends. This predictive information is shared with demand and storage
agents, enabling informed decision-making based on anticipated energy availability. Storage agents, in
particular, rely on these predictions along with real-time energy demands to determine optimal storage strategies.
They assess forecasts and current demands continuously, deciding whether to store excess energy during peak
production periods or release stored energy to meet sudden spikes in demand. As the central coordinator, the
supervisor agent plays a crucial role in optimizing microgrid operations. By analyzing predictions, energy
storage decisions, and real-time demand, it formulates strategies to maximize renewable energy utilization. One
unique aspect of the system is its dynamic priority allocation between supercapacitors and batteries.
Supercapacitors take precedence during significant energy fluctuations, offering rapid response times to address
sudden energy needs efficiently. While the system prioritizes renewable energy utilization and storage solutions,
there are instances where local resources may be insufficient. In such cases, the grid connection to the external
power network is activated as a last resort to ensure continuous energy supply.

3.2. Algorithm development

In this section, we delve into the creation of the algorithm. Our objective is to design an intelligent
and adaptive algorithm that orchestrates the interactions among agents within the microgrid. To achieve this,
we formulate a set of rules, guidelines, and decision matrices that guide the behavior of each agent. These rules
are carefully crafted to align with the objectives of maximizing renewable energy, efficient energy storage, and
responsiveness to demand. The algorithm takes into account real-time data from prediction agents, decisions
from the energy storage agent, and the overarching strategy defined by the supervisor.
The algorithm addresses three key aspects of energy management. Firstly, it focuses on optimizing
the utilization of renewable energy sources and ensuring efficient distribution to DC and AC loads based on
input parameters such as demands and energy production. Secondly, it efficiently manages excess energy from
photovoltaic and wind sources by storing it in batteries or supercapacitors, aiming to enhance overall system
performance. Lastly, the algorithm addresses remaining net demand by evaluating fluctuations in DC and AC
loads. If there is a net demand for DC loads, and fluctuations exceed a threshold, the algorithm draws from
stored capacitor energy to meet the demand, adjusting the capacitor's energy level accordingly. This algorithm
has been implemented using the Python programming language, with a specific library dedicated to simulating
multi-agent systems and microgrids called Mesa which stand for multi-agent simulation environment. This
library provides a set of tools and functionalities to efficiently create, simulate, and evaluate interactions and
agent behavior within the microgrid.

3.2.1. Sub-algorithm 1: Renewable energy allocation and cooperation algorithm for load supply
This sub-algorithm focuses on optimizing renewable energy utilization and cooperation in supplying
loads. Given input parameters such as DC and AC demands, photovoltaic and wind energy productions, it
calculates power distribution to DC and AC loads. The algorithm considers net demands and prioritizes
renewable sources to ensure efficient energy allocation:
- Initialize the power variables for different sources and loads:
PDC, PAC, PSPV, PSW, PSB, PSS, PSG all set to 0.
Loop through the data series:
For i in range (length of DDC):
- Calculate net demand for DC and AC loads:
Net_DDC = max (0, DDC[i] - PPV[i])
Net_DAC = max (0, DAC[i] - Pw[i])
- Supply DC loads using photovoltaic production:
PSPV = min (DDC[i], PPV[i])
PDC = PSPV
- Supply AC loads using wind production:
PSW = min (DAC[i], Pw[i])
PAC= PSW

Int J Pow Elec & Dri Syst, Vol. 15, No. 2, June 2024: 1235-1246
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1241

- Utilize wind production if photovoltaic production is insufficient for DC loads:

if Net_DDC > 0:
PSW += min (Net_DDC, Pw[i] - PAC)
PDC += min (Net_DDC, Pw[i] - PAC)
Net_DDC = max (0, Net_DDC - min (Net_DDC, Pw[i] - PAC))
- Utilize photovoltaic production if wind production is insufficient for AC loads:
if Net_DAC > 0:
PSPV += min (Net_DAC, PPV[i] - PDC)
PAC+= min (Net_DAC, PPV[i] - PDC)
Net_DAC = max (0, Net_DAC - min (Net_DAC, PPV[i] - PDC))

3.2.2. Sub-algorithm 2: Energy excess management for battery and supercapacitor

This sub-algorithm focuses on efficiently managing energy excess from photovoltaic (PV) and wind
sources by storing it in batteries or supercapacitors without exceeding their maximum capacity. The main goal
is to optimize the utilization of surplus energy to enhance system performance. In the subsequent section, we
will delve into a specific scenario within the algorithm, focusing on the case where the excess energy stems
solely from photovoltaic production. This scenario is captured by the condition "If PPV[i] - DDC[i] > 0 and Pw[i]
- DAC[i] < 0". While we will elaborate on the particulars of this case, it is important to note that the underlying
reasoning remains consistent across all three scenarios, regardless of whether the excess energy originates from
wind energy, photovoltaic energy, or a combination of both:
- Calculate the surplus energy from photovoltaic (PV) and wind sources:
Pexcess = PPV[i] + Pw[i] - DDC[i] - DAC[i]
- Check if there is surplus energy to manage:
if Pexcess > 0:
- Determine the energy storage mechanism based on production and demand conditions:
If PPV[i]- DDC[i] >0 and Pw[i]- DAC[i] <0:
- Calculate the fluctuation in PV production:
fluctuationPV = abs (PPV[i] - PPV[i-1])
if fluctuationPV >= rapid fluctuation threshold:
- Store the surplus energy in supercapacitors up to their capacity
SOC+= min (Pexcess , CAPmax - SOC)
Pexcess -= min (Pexcess , CAPmax - SOC)
if Pexcess >0:
- Store the rest in battery up to their capacity
BOC += min (Pexcess , BATmax - BOC)
else:
- Store the surplus energy in battery up to their capacity
BOC += min (Pexcess , BATmax - BOC)

3.2.3. Sub-algorithm 3: Enhancing load supply efficiency through optimized storage selection and grid
support
This sub algorithm is responsible for addressing the remaining net demand by either utilizing the
energy stored in the battery/supercapacitor or by drawing from the grid. The process begins by calculating the
fluctuations in demand for DC and AC loads compared to their previous values. If there is still a net demand
for DC loads, the algorithm evaluates the demand's fluctuations. If the fluctuations exceed a threshold, the
algorithm aims to first draw from the capacitor's stored energy to meet the demand. If the net demand is within
the capacitor is remaining capacity (above a certain minimum threshold), the DC load is supplied directly from
the capacitor's energy, and the capacitor's energy level is adjusted accordingly.
If the remaining net demand is greater than the capacitor's capacity, the algorithm ensures the load is
supplied using the battery's stored energy. If the net demand still exceeds the battery's remaining capacity, the
grid is tapped to fulfill the remaining demand. Similar principles apply for the case of AC loads. The same
logic is followed for handling cases where the demand fluctuations exceed a predefined threshold The code
effectively manages load supply using the available energy sources, considering different storage options and
resorting to the grid when necessary.
In the subsequent sections of this sub algorithm, we will delve into the scenario where the net direct
current (DC) demand is positive (net DC demand > 0), and the fluctuations in this demand are deemed
significant. We will elaborate on how this positive net DC demand is catered to, prioritizing the utilization of
energy stored in the supercapacitor, followed by the battery, before resorting to grid energy if necessary. It is
important to note that the reasoning and approach presented in this section is also applied to other similar cases,
such as scenarios where the net DC demand is positive but with insignificant fluctuations, as well as situations
Multi-objective algorithm for hybrid microgrid energy management based on multi-agent … (Ilham Tyass)
1242  ISSN: 2088-8694

where the net alternating current (AC) demand is positive. In summary, this particular case offers an intricate
insight into the methodology adopted to address diverse energy demands while adhering to storage and grid
constraints:
- Verify the load requirements
if Net_DDC > 0 :
fluctuationDemandDC = abs(DDC [i] - DDC [i-1])
if fluctuationDemandDC > DFth :
if Net_DDC <= (SOC - LCmin):
- Supply the DC load with the energy stored in the capacitor
PSS = Net_DDC
PDC += PSS
SOC -= PSS
else:
- Supply the DC load with the energy stored in the capacitor up to the minimum threshold
PSS = SOC - LCmin
PDC += PSS
SOC -= PSS
Net_DDC -= PSS
if Net_DDC <= (BOC - LBmin):
- Supply the DC load with the energy stored in the battery
PSB = Net_DDC
PDC += PSB
BOC -= PSB
else:
PSB = BOC - LBmin
PDC += PSB
BOC -= PSB
Utilize grid energy to meet the remaining net demand
PSG = Net_DDC - PSB
PDC += PSG

4. SIMULATIONS AND VALIDATION

In order to comprehensively evaluate the effectiveness of our proposed approach, we conducted
simulations using Python's Multi-Agent Simulation Environment. These simulations involved generating
demand and production profiles through Python scripts. We systematically explored a range of scenarios to
highlight the algorithm's potential impact on optimizing the management of usage priorities among renewable
energy sources, storage devices, and the grid. Our analysis extended to effectively handling the storage of
surplus and deficit energy, dynamically allocating resources between batteries and capacitors based on
fluctuations in demand and production. Through these simulations, we aimed to demonstrate how our algorithm
enhances energy distribution and storage efficiency, facilitating the optimal utilization of renewable sources
while effectively addressing variations and fluctuations in energy demand and supply.
The graph in Figure 5 illustrates the energy sharing between photovoltaic (PV) and wind sources in a
microgrid. The two-colored bars represent the sharing of excess energy between the two sources. It highlights the
dynamics of cooperation between renewable sources. The result presented in the circular graph of Figure 6
illustrates the sharing of excess energy between DC and AC buses. When renewable sources generate more energy
than required to power the associated loads, our algorithm directs the surplus solar energy towards AC loads,
while the 'Wind to DC' segment represents the share of excess wind energy shared with DC loads. This leads to
an optimal utilization of available resources and an overall increase in energy efficiency within the microgrid.
The graph in Figure 7 visually presents the distribution of power sources in a microgrid to meet the
energy demands of both direct current (DC) and alternating current (AC) loads. Each colored bar in the graph
corresponds to a specific power source and illustrates its role in providing energy to the loads. The segment
labeled 'PV power generation' shows the contribution of photovoltaic panels in supplying energy specifically
to DC loads. Similarly, the 'WIND power generation' segment represents the portion of energy from wind
sources that is utilized to power AC loads. Furthermore, the graph includes segments such as 'Power supplied
by battery,' 'Power supplied by supercapacitor,' and 'Power supplied by the grid.' These segments indicate the
contribution of different energy storage and supply mechanisms. The 'Power supplied by battery' segment
reflects the role of the battery in meeting energy demands, the 'Power supplied by supercapacitor' segment

Int J Pow Elec & Dri Syst, Vol. 15, No. 2, June 2024: 1235-1246
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1243

signifies the contribution of the supercapacitor, and the 'Power supplied by the grid' segment represents the
energy sourced from the external grid.

Figure 5. Energy sharing between buses: PV to AC Figure 6. Excess energy sharing between buses
and wind to DC

By analyzing the results across various simulation scenarios and carefully scrutinizing the data
presented in Table 1. In the presence of fluctuations, the algorithm prioritizes the optimized utilization of
energy stored in the supercapacitor after the depletion of renewable resources, with occasional utilization of
the supercapacitor to ensure uninterrupted supply (scenario 4). On the contrary, in simulation scenarios 2 and
3, priority is given to the battery for powering loads, as long as the minimum storage threshold is not reached.
When there is an excess of renewable energy production, our algorithm favors storage in batteries, unless
notable production fluctuations occur. In the latter case, excess energy is stored in the supercapacitor, as
showcased in scenario 8 (refer to Tables 2 and 3). Furthermore, when renewable resources are depleted and
storage devices have reached their minimal thresholds, our algorithm invokes grid power to cater to load
demands, as illustrated by scenario 10. These outcomes vividly demonstrate the flexibility and efficiency of
our algorithm in managing diverse energy conditions. It adeptly adapts to demand and production variations,
strategically harnessing different available storage sources to ensure reliable and optimized energy supply,
while concurrently reducing dependence on the conventional grid.

Table 1. Demand fluctuations

Scenarios Fluctuations in DC demand Fluctuations in AC demand Need for storage devices
S1 NaN NaN NO
S2 5.00% 0.00% YES
S3 10.00% 20.00% YES
S4 -15.00% -30.00% YES
S5 -5.00% 15.00% YES
S6 25.00% -15.00% NO
S7 -20.00% 5.00% NO
S8 15.00% 5.00% NO
S9 -15.00% 5.00% NO
S10 10.00% 20.00% YES

Table 2. Production fluctuations

Scenarios PV fluctuation Wind power fluctuation Excess production
S1 NaN NaN YES
S2 5.00% -20.00% NO
S3 10.00% 10.00% NO
S4 -5.00% -25.00% NO
S5 0.00% 10.00% NO
S6 25.00% -10.00% NO
S7 -5.00% 5.00% YES
S8 -5.00% 20.00% YES
S9 -5.00% -15.00% YES
S10 5.00% -5.00% NO

Multi-objective algorithm for hybrid microgrid energy management based on multi-agent … (Ilham Tyass)
1244  ISSN: 2088-8694

Table 3. Evolution of storage level

Scenarios Evolution of battery storage level Evolution of supercapacitor storage level
S1 NaN NaN
S2 -10.00% 0.00%
S3 -15.00% 0.00%
S4 0.00% -5.00%
S5 -5.00% 0.00%
S6 0.00% 0.00%
S7 0.00% 0.00%
S8 0.00% 20.00%
S9 0.00% 0.00%
S10 -5.00% -5.00%

Figure 7. Power sources contributing to load supply component

5. CONCLUSION
In conclusion, our study delves into the context of renewable energies and microgrid systems,
highlighting their potential for local and sustainable energy generation. However, the efficient management of
energy resources presents a pivotal challenge in optimizing their utilization due to the intricate interplay
between renewable energy sources, storage mechanisms, and consumption patterns. Our primary objective has
been the development of a multi-objective optimization algorithm tailored for energy management within
microgrids. This algorithm prioritizes the seamless integration of renewable energy sources and ensures
effective storage management between batteries and supercapacitors. At the core of our approach lies the
utilization of a multi-agent system, facilitating efficient coordination of energy distribution and utilization
within the microgrid. By harnessing the synergy among renewable energy sources, storage units, and variable
loads, the algorithm aims to enhance energy efficiency while maintaining a consistent and balanced power
supply to the connected loads.
Our efforts to comprehensively evaluate the efficacy of our proposed approach involved simulations
conducted within Python's multi-agent simulation environment, covering a range of scenarios. These
simulations aimed to showcase the algorithm's impact on optimizing usage priorities among renewable energy
sources, storage devices, and the grid. Notably, our algorithm dynamically allocates resources between
batteries and capacitors based on demand and production fluctuations, effectively managing the storage of
surplus and deficit energy. The outcomes of our study unveil the algorithm's versatility and efficiency in
managing diverse energy conditions, adeptly adapting to variations in demand and production. By strategically
leveraging various available storage sources, it ensures dependable and optimized energy supply while
diminishing reliance on conventional grid sources. In summary, our work advances the understanding and
application of energy management within microgrids, offering a valuable contribution to the optimization of
renewable energy utilization. Through the integration of innovative algorithms and multi-agent systems, we
pave the way for more resilient and sustainable energy solutions in the microgrid landscape.

Int J Pow Elec & Dri Syst, Vol. 15, No. 2, June 2024: 1235-1246
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1245

REFERENCES
[1] M. S. Sami et al., “Energy management of microgrids for smart cities: A review,” Energies, vol. 14, no. 18, 2021, doi:
10.3390/en14185976.
[2] V. Prema and K. U. Rao, “Predictive models for power management of a hybrid microgrid &#x2014; A review,” in 2014
International Conference on Advances in Energy Conversion Technologies (ICAECT), Jan. 2014, pp. 7–12, doi:
10.1109/ICAECT.2014.6757053.
[3] F. Ye, Y. Qian, and R. Q. Hu, “Self-sustaining wireless neighborhood area network design for smart grid,” in 2014 IEEE Global
Communications Conference, GLOBECOM 2014, 2014, pp. 2715–2720, doi: 10.1109/GLOCOM.2014.7037218.
[4] Z. Boussaada, A. Remaci, O. Curea, O. B. Driss, H. Camblong, and N. M. Bellaaj, “Management Approach for Microgrid Operation
Using Multi Agent System (MAS) Technique Man-agement Approach for Microgrid Operation Using Multi Agent System
Management Approach for Microgrid Operation Using Multi Agent System (MAS) Technique,” 2017, [Online]. Available:
https://hal.science/hal-01631167.
[5] V. H. Bui, A. Hussain, and H. M. Kim, “A multiagent-based hierarchical energy management strategy for multi-microgrids
considering adjustable power and demand response,” IEEE Transactions on Smart Grid, vol. 9, no. 2, pp. 1323–1333, 2018, doi:
10.1109/TSG.2016.2585671.
[6] H. S. V. S. K. Nunna and S. Doolla, “Demand response in smart distribution system with multiple microgrids,” IEEE Transactions
on Smart Grid, vol. 3, no. 4, pp. 1641–1649, 2012, doi: 10.1109/TSG.2012.2208658.
[7] Z. Wang, L. Wang, A. I. Dounis, and R. Yang, “Multi-agent control system with information fusion based comfort model for smart
buildings,” Applied Energy, vol. 99, pp. 247–254, 2012, doi: 10.1016/j.apenergy.2012.05.020.
[8] J. Chen, R. K. Jain, and J. E. Taylor, “Block Configuration Modeling: A novel simulation model to emulate building occupant peer
networks and their impact on building energy consumption,” Applied Energy, vol. 105, pp. 358–368, 2013, doi:
10.1016/j.apenergy.2012.12.036.
[9] B. Zhao, M. Xue, X. Zhang, C. Wang, and J. Zhao, “An MAS based energy management system for a stand-alone microgrid at high
altitude,” Applied Energy, vol. 143, pp. 251–261, 2015, doi: 10.1016/j.apenergy.2015.01.016.
[10] H. S. V. S. Kumar Nunna and S. Doolla, “Multiagent-based distributed-energy-resource management for intelligent microgrids,”
IEEE Transactions on Industrial Electronics, vol. 60, no. 4, pp. 1678–1687, 2013, doi: 10.1109/TIE.2012.2193857.
[11] W. Zhong et al., “IDES: Incentive-driven distributed energy sharing in sustainable microgrids,” 2014 International Green
Computing Conference, IGCC 2014, 2015, doi: 10.1109/IGCC.2014.7039166.
[12] M. Hamidi, A. Raihani, and O. Bouattane, “Sustainable Intelligent Energy Management System for Microgrid Using Multi-Agent
Systems: A Case Study,” Sustainability (Switzerland), vol. 15, no. 16, 2023, doi: 10.3390/su151612546.
[13] V. Lešić, A. Martinčević, and M. Vašak, “Modular energy cost optimization for buildings with integrated microgrid,” Applied
Energy, vol. 197, pp. 14–28, 2017, doi: 10.1016/j.apenergy.2017.03.087.
[14] A. S. Nair et al., “Multi-Agent Systems for Resource Allocation and Scheduling in a Smart Grid,” Technology and Economics of
Smart Grids and Sustainable Energy, vol. 3, no. 1, p. 15, Dec. 2018, doi: 10.1007/s40866-018-0052-y.
[15] S. S. Binyamin and S. Ben Slama, “Multi-Agent Systems for Resource Allocation and Scheduling in a Smart Grid,” Sensors, vol.
22, no. 21, 2022, doi: 10.3390/s22218099.
[16] S. Sen and V. Kumar, “Microgrid control: A comprehensive survey,” Annual Reviews in Control, vol. 45, pp. 118–151, 2018, doi:
10.1016/j.arcontrol.2018.04.012.
[17] I. Tyass, T. Khalili, M. Rafik, B. Abdelouahed, A. Raihani, and K. Mansouri, “Wind Speed Prediction Based on Statistical and
Deep Learning Models,” International Journal of Renewable Energy Development, vol. 12, no. 2, pp. 288–299, 2023, doi:
10.14710/ijred.2023.48672.
[18] Y. P. Faniband and S. M. Shaahid, “Forecasting wind speed using artificial neural networks - A case study of a potential location
of saudi arabia,” E3S Web of Conferences, vol. 173, 2020, doi: 10.1051/e3sconf/202017301004.
[19] A. Tokgoz and G. Unal, “A RNN based time series approach for forecasting turkish electricity load,” in 26th IEEE Signal Processing
and Communications Applications Conference, SIU 2018, 2018, pp. 1–4, doi: 10.1109/SIU.2018.8404313.
[20] A. Idda, M. E. Slimani, S. Bentouba, and Y. Hammaoui, “Différentes Configurations du Système PV pour l’Alimentation Sans
Interruption ( ASI ) : Application au Relais GSM,” Journal of Advanced Research in Science and Technology, pp. 574–582, 2017.
[21] B. Wichert, M. Dymond, W. Lawrance, and T. Friese, “Development of a test facility for photovoltaic-diesel hybrid energy
systems,” Renewable Energy, vol. 22, no. 1, pp. 311–319, 2001, doi: 10.1016/S0960-1481(00)00024-0.
[22] I. Tyass, O. Bouamrane, A. Raihani, K. Mansouri, and T. Khalili, “Hybrid Renewable Energy System Investigation Based on Power
Converters Losses,” Lecture Notes in Electrical Engineering, vol. 745, pp. 615–625, 2022, doi: 10.1007/978-981-33-6893-4_57.
[23] K. Daniel, “Development of Battery/Supercapacitor Hybrid Energy Storage Scheme for Grid Connected Solar PV Source,”
University of Nairobi, 2023.
[24] L. Wang, Z. Wang, and R. Yang, “Intelligent multiagent control system for energy and comfort management in smart and
sustainable buildings,” IEEE Transactions on Smart Grid, vol. 3, no. 2, pp. 605–617, 2012, doi: 10.1109/TSG.2011.2178044.
[25] M. Afrasiabi, M. Mohammadi, M. Rastegar, and A. Kargarian, “Multi-agent microgrid energy management based on deep learning
forecaster,” Energy, vol. 186, 2019, doi: 10.1016/j.energy.2019.115873.
[26] R. Jabeur, Y. Boujoudar, M. Azeroual, A. Aljarbouh, and N. Ouaaline, “Microgrid energy management system for smart home
using multi-Agent system,” International Journal of Electrical and Computer Engineering, vol. 12, no. 2, pp. 1153–1160, 2022,
doi: 10.11591/ijece.v12i2.pp1153-1160.
[27] Z. Jian, Q. Ai, C. Jiang, X. Wang, Z. Zheng, and C. Gu, “The application of Multi Agent System in Microgrid coordination control,”
2009, doi: 10.1109/SUPERGEN.2009.5348277.

Multi-objective algorithm for hybrid microgrid energy management based on multi-agent … (Ilham Tyass)
1246  ISSN: 2088-8694

BIOGRAPHIES OF AUTHORS

Ilham Tyass is a Ph.D. student at the Electrical Engineering and Intelligent

Systems (EEIS) Laboratory, ENSET Institute in Mohammedia, Hassan II University of
Casablanca, Morocco. Her research focuses on optimization and numerical modeling of a
hybrid wind-photovoltaic system with energy storage. She can be contacted at email:
tyass1@yahoo.fr.

Abdelouahad Bellat is a professor and holds a doctorate in mechanical

engineering and artificial intelligence, which he obtained in 2021. Abdelouahad holds a
doctorate from Hassan II Casablanca University, ENSET Institute. His research mainly
focuses on the design and optimization of wind farms using artificial intelligence algorithms.
He can be contacted at email: bellatabdelouahad@gmail.com.

Abdelhadi Raihani is now a teacher of electronics engineering and researcher

at Hassan II University of Casablanca ENSET Institute Mohammedia Morocco. He received
the B.S. degree in electronics in 1987 and the M.S. degree in applied electronics in 1991 from
the ENSET Institute. He had his DEA diploma in information processing from the Ben M’sik
University of Casablanca in 1994. He received the Ph.D. in parallel architectures application
and image processing from the Ain Chok University of Casablanca in 1998.His research is
focused on medical image processing and electrical systems related to wind and solar energy.
He has published more than 30 publications in various national, international conference
proceedings and journals. He can be contacted at email: abraihani@yahoo.fr.

Khalifa Mansouri is now a teacher of computer science and researcher at the

University Hassan II Casablanca, ENSET Institute. His research is focused on Real Time
Systems, Information Systems, e-Learning Systems, Industrial Systems (modeling,
optimization, and numerical computing). Diploma ENSET Mohammedia in 1991, CEA in
1992 and Ph.D. (Calculation and optimization of structures) in 1994 to Mohammed V
University in Rabat, HDR in 2010 and Ph.D. (computer science) in 2016 to Hassan II
University in Casablanca. He can be contacted at email: khalifa.mansouri@enset-
media.ac.ma.

Int J Pow Elec & Dri Syst, Vol. 15, No. 2, June 2024: 1235-1246