Content uploaded by Mansour Selseleh Jonban
Author content
All content in this area was uploaded by Mansour Selseleh Jonban on May 26, 2023
Content may be subject to copyright.
Journal of Building Engineering 74 (2023) 106866
Available online 22 May 2023
2352-7102/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
Contents lists available at ScienceDirect
Journal of Building Engineering
journal homepage: www.elsevier.com/locate/jobe
Full length article
Multi-agent-based energy management of multiple grid-connected
green buildings
Seyedeh Samaneh Ghazimirsaeid a, Mansour Selseleh Jonban b,∗,
Manthila Wijesooriya Mudiyanselage c, Mousa Marzband a,d,
Jose Luis Romeral Martinez b, Abdullah Abusorrah d,e
aNet Zero Industry Innovation Centre, Campus Masterplan, Teesside University, Middlesbrough TS1 3BA, United Kingdom
bMCIA Center, Electronic Engineering Department, Universitat Politecnica de Catalunya, Terrassa, Spain
cNorthumbria University, Electrical Power and Control Systems Research Group, Ellison Place NE1 8ST, Newcastle upon Tyne, United Kingdom
dRenewable Energy and Power Systems Research Group, Center of Research Excellence in Renewable Energy and Power Systems, King
Abdulaziz University, Jeddah 21589, Saudi Arabia
eDepartment of Electrical and Computer Engineering, Faculty of Engineering, K.A. CARE Energy Research and Innovation Center, King
Abdulaziz University, Jeddah 21589, Saudi Arabia
ARTICLE INFO
Keywords:
Multi-agent-systems
Home microgrid
Energy management system
Demand side management
ABSTRACT
Integration of distributed energy resources (DER) in electrical microgrids introduces residential
green buildings (RGB) with a promising decrement in fossil fuel consumption. This novel
concept compromises numerous challenges such as controlling DERs and consumers in a home
microgrid (H-MG), based on historical data and market clearing price (MCP), which requires
multi-objective analysis and an energy management system (EMS). Initially, tremendous issues
emerged when integrating these systems with RGBs. Therefore, multi-agent systems (MAS) are
capable of utilizing parallel computing as a control method, where each RGB represents as an
agent with independent decision-making capability, while productively cooperating with other
agents. In this paper, an effective EMS has been presented with MAS (EMS-MAS) for DER in a
neighborhood grid, accompanied by several RGBs. The RGB includes controllable and uncontrol-
lable devices by residents and building management systems (BMS), where controllable devices
are HVAC and home appliances (e.g: dishwashers, washing machines), and uncontrollable
devices are lighting systems, flexible heating and cooling demands, respectively, along with
several electrical loads (e.g: lighting system, A/C, refrigerator, etc.) and thermal loads (e.g
HVAC systems, water heater, ovens, etc.) with retailers who sell and buy electricity to/from
residents. Finally, the results confirm that the proposed model has significantly enhanced the
overall energy efficiency and the profit of individual RGBs, and optimally managed the devices
in RGBs while encouraging demand response (DR) load programs, retailers and MCP reduction.
1. Introduction
High penetration of distributed energy resources (DERs) necessitates the proper management and control of energy. In particular,
two control strategies have been described such as generation side and demand side management [1–3]. Demand side management
is immensely appropriate for residential green buildings (RGB) due to the financial incentives [4–7]. However, the most challenging
∗Corresponding author.
E-mail address: mansour.selseleh.jonban@upc.edu (M.S. Jonban).
https://doi.org/10.1016/j.jobe.2023.106866
Received 29 May 2022; Received in revised form 15 May 2023; Accepted 16 May 2023
Journal of Building Engineering 74 (2023) 106866
2
S.S. Ghazimirsaeid et al.
Nomenclature
Acronyms
RGB residential green building
AEL aggregated electrical load
ATL aggregated thermal load
CHP combined heat and power
DR demand response
DW dish washer
DER distributed energy resources
DSO distributed system operator
EES electrical energy storage
ESP electrical solar panel
EV electrical vehicle
GB gas boiler
HHW heat and hot water
H-MG home Microgrid
MCP market clearing price
MO-TE market operator based on transactive energy
NG natural gas
NRL non-responsive load
PV photovoltaic
REF refrigerator
RET retailer
RLD responsive load demand
SBP system buy price
SSP system sell price
SOC state-of-charge
TD thermal dump
TES thermal energy storage
TSP thermal solar panel
TE transactive energy
Indices
e / h / t / i electricity/ heat/ time steps/ number of H-MG
j∈ {CHP,TSP}thermal DERs
k∈ {ESP,CHP}electrical DERs
m∈ {DW,EV,REF,AEL}electrical consumers
l∈ {HHW,ATL,TD}thermal consumers
Constant values
𝑆𝑂𝐶 𝑥,𝑆𝑂𝐶 𝑥,𝑃𝑥,𝑒∕ℎ,𝑃𝑥,𝑒∕ℎminimum/ maximum SOC/ power of X during charging and discharging mode, x∈
{ES+,ES-,EV+,EV-,TES+,TES-}
𝐸Tot,𝑥 total value of X capacity
𝑇𝑦,𝑇𝑦minimum/ maximum value of 𝑦temperature, 𝑦∈ {REF,HHW}
𝑃𝑗
𝑒∕ℎ,𝑃𝑗
𝑒∕ℎminimum/ maximum electrical thermal power j
T𝑦
INI, TRED , TINC initially temperature/ the amount of temperature reduction each time the REF compressor is turned
on/ the amount of temperature increase each time HHW is turned on
𝜁𝑗
𝑒∕ℎelectrical and thermal efficiencies j
𝑇HHW,𝑇HHW minimum/ maximum values of temperature in HHW
𝐸𝑥,𝐸𝑥minimum/ maximum values of energy in 𝑥
Journal of Building Engineering 74 (2023) 106866
3
S.S. Ghazimirsaeid et al.
𝜋𝑧,𝜋𝑧minimum/ maximum values of price bids by z, z∈ {𝑗, 𝑘, 𝑚, 𝑙}
𝜋NG
𝑡natural gas price
𝜆MCP
𝑡MCP prediction value during each time interval t (£/kWh)
Decision variables
𝑋Ret
𝑡,𝑋ES
𝑡,𝑋TES
𝑡,𝑋DR
𝑡binary variable of retailer, electrical energy storage, thermal energy storage, demand
response
𝑃𝑚,𝑒
𝑡,𝑃𝑙,ℎ
𝑡consumed electrical/ thermal power by m/ l at time t
𝑃𝑗,𝑒
𝑡,𝑃𝑘,ℎ
𝑡electrical/ thermal power generated by j/ k at time t
𝜋𝑧,𝑒
𝑡,𝜋𝑧,ℎ
𝑡electrical/ thermal price bids by z at time t
𝑃Ret+,i
𝑡,𝑃Ret-,i
𝑡,𝑒 electric power sold/ bought by H-MG i to/from the retailer
𝜆MCP,𝑠
𝑡Market clearing price by using the S optimization method (£/kWh)
S=1: particle swarm optimization (PSO)
S=2: harmony search (HS)
S=3: differential evolution (DE) algorithm
S=4: bat algorithm (BAT)
objectives could be identified as, defining the optimum value of DERs, loads shifting, supply and demand balancing, and enhancing
the total revenue. Similarly, an effective energy management system strongly depends on the RES structure presented in the
neighborhood grid, DER and local energy storage (ES), and communication between H-MGs and retailers, in order to monitor
and investigate the behavior of the system and effective responses to different scenarios. The multi-agent system (MAS) with the
properties such as self-learning, asynchronous and parallel computing, scalability, re-development, and re-usability, could influence
the performance of RGBs [8,9]. In MAS, each part of the system is controlled by an autonomous individual smart unit that follows
its own defined goals. These units consist of highly interdependent operations and efficient mutual communication to optimize the
decisions and the ultimate results.
Although, enormous contributions were widely presented on electrical energy management strategies with DER and ES for RGBs,
the studies with the peer-connected RGBs (multiple RGBs in the neighborhood systems) to minimize the fossil energy consumption,
have not been considered in the literature. However, in the proposed model, the RGB with the neighborhood of other RGBs has
formed a set of H-MGs, where each RGB is independently managed and controlled by a smart agent, while cooperating with others
to share power and enhance the revenue. In this concern, the chief downsides (DS) in the previous studies could be outlined as
•DS1: Non-existence of a method to exchange energy through the neighborhood RGB resources and, supply the consumer’s load
demand from generated energy in RGBs [10–14].
•DS2: Absent to design a model with the participation of consumers in demand response (DR) program and present a solution
to calculate market clearing price (MCP) [15–22]. However, in the proposed model, MCP value was calculated by Nash
equilibrium point, which is the optimum capacity, with the contribution of all the players in the market.
•DS3: Absence of a solution, which is based on optimization algorithms to implement and evaluate the optimum clearance
value of the market process while introducing a pay-off for all the market players [23–26].
•DS4: Not presents a solution to determine the strategy and behavior of residential customers as prosumers, and absent to
introduce a strategy for the consumers to participate in the market [27–29].
•DS5: Non-existence of an algorithm to reach the collective profit of all the players and dealing with the behavior of the players
with different objectives in the optimization process [30,31].
This paper proposes a strategy to obtain the optimum energy management by MAS theory for the H-MGs in the neighborhood
structure. Each H-MG is defined as an agent with several properties such as automation, adaptability, and simultaneous interaction
with other agents. Moreover, the proposed model has facilitated several types of DERs to increase the profit for all the players by
optimization algorithms. On the consumer’s side, the residential customers represent MCP as prosumers and consumer during DR
program and the H-MGs earn an additional profit by selling the surplus energy to the neighbors and retailers. Nevertheless, the
energy exchange between RGBs has influenced the consumer accessibility to energy generation resources, which results to escalate
competition between players in the electricity market. In addition, DR strategy has contributed to reduce the peak load and gain the
overall profit, by shifting load demand from peak hours to non-peak hours. For example, the MCP has increased during peak hours,
whereas the MCP has reduced in the non-peak hours, to optimize the load demand in the system. Several literature were investigated
implementing the EMS for H-MG with the contribution of MAS method [7,9,32,33]. Overall, the main research contributions (RC)
of this study could be highlighted as follows:
•RC1: A navel framework, based on MAS is presented to create a smart structure between H-MGs to exchange energy and
supply load through the neighborhood RGB resources (Eliminates DS1 and DS2).
Journal of Building Engineering 74 (2023) 106866
4
S.S. Ghazimirsaeid et al.
•RC1: In the presented framework, for controlling and managing power in the neighborhood grids, DR strategy is considered
with the consumers and prosumers involvement for an effective MCP that increases overall profit by shifting demand from
on-peak price to off-peak price (Eliminates DS3).
•RC1: Introduces a comprehensive and smart algorithm to exchange power through the resources among H-MGs, and the
possibility of consumers load supply by reliable access to the DERs of other H-MGs (Addresses DS4 and DS5).
2. A review of MAS
MAS is an intelligent autonomous system with computerized entities called agents. An agent is located in an environment which
could share autonomous actions to solve conflict issues. These agents are able to interact with its background and also with each
other to accomplish the delegated objectives. On the other words, MAS is a framework for optimizing the activities of agents by
combining artificial intelligence and mathematical tools for decision-making [34]. The agents are divided into three categories such
as passive agent, active agent, and cognitive agent [12,13,35,36]. The passive agent or reactive agent is called the agent without
the target, which has low communication capacity and respond to a commanded action. The active agent can be described as agent
with goals, while the complex calculations are controlled by the cognitive agent which is capable of mutual communication.
Agent environment is defined as an external section to entities which communicates, assess, and investigates the system. The
three main categories of agent environments are virtual, discrete, and continuous. Virtual environments are dealing with artificial
intelligent (AI) and, discrete agent environment is introduced when there are a finite number of actions. Similarly, if the numbers
of agent performances are unlimited, it is to be explained as a continuous agent environment. The agent environments are further
grouped according to the determinism, accessibility, episodic/non-episodic, and static/dynamic properties. Determinism refers to
the accuracy of the action while accessibility defines the possibility to gather complete details. Further, when the future actions are
independent from the present actions, it is described as an episodic environment, whereas in non-episodic background, the future
agent’s performance is influenced by the current actions. Static and dynamic environments are determined by the nature of changes
in the operating conditions.
The advancement of MAS has facilitated numerous advantages in the contemporary world. In particular, the agents in MAS are
autonomous, self-aware, and independent which solve complicated situations by dividing them into smaller problems with simplified
computations to increase the processing power [37]. Moreover, these agents are capable of local views instead of global views, to
maintain the simplicity of the network. Decentralization is one of the key benefits of the MAS where the controlling of the system is
distributed to enhance the system’s reliability and efficiency. For example, the failure of an individual agent does not result in the
overall system failure and entities could simply add or remove from the system without affecting other elements [24,38]. Therefore,
these benefits make it ideal for energy management in green buildings.
The applications of the MAS have diversified into enormous fields. Specifically, MAS plays a prime role in AI and graphical
applications, transportation, logistics, census, power systems, and smart grids, to balance the dynamic load and enhance the
self-healing in the system. The properties of smart agents in H-MGs could be listed as follows [25,28,39,40]:
•Agents are autonomous and self-decision makers.
•They are able to perceive changes in the environment.
•Capable of changing the decisions and consequently modifying the actions according to the existing environmental conditions.
•Agents are able to react in their environment. In other words, they change their environment through their behaviors.
•Agents could communicate with each other in order to enhance the efficiency in a distributed decision system.
•In non-supervisory control, agents consist of a specific level of autonomy to carry out a set of necessary measures, without
permission from a central controller. In fact, this autonomy dependents on defined goals for each agent.
•In a large system, the agents share information together to enhance community benefits.
•Agents are accompanied by specific behaviors for fulfilling targeted goals by resources, skills, and services.
•Similar agents contain different behaviors, which are related to their decisions.
3. The structure of EMS based on MAS (EMS-MAS)
The proposed model is an EMS based on MAS. In the structure, several agents have been introduced to interact with each other
and gain a common goal. For example, the generation sources, consumers and prosumers presented in a H-MG act as individual
agents. Generation agents are responsible to monitor and control power levels and defining the optimum generating point, along
with balancing the load between the generation side and the demand side [41]. ES and DR agents are responsible for monitoring
SOC condition and controlling the corresponding loads (prosumers) [42,43]. The DR agents are also responsible for load shedding
during peak hours. Further, the ES agent’s main goal is to ensure that the energy storage system is used optimally to balance
the supply and demand of power in the H-MG. The behaviors of the ES agent include monitoring the SOC of the energy storage
system, monitoring the demand and supply of power in the H-MG, and controlling the loads to ensure that the energy storage
system is used optimally. The ES agent also communicates with other agents in the H-MG to share information about the available
power and demand and cooperates with them to achieve the objectives of the MCP. Ultimately, the ES agent plays a crucial role
in achieving energy autonomy for the H-MG by efficiently managing the energy storage system and coordinating with other agents
to ensure that the system operates in a sustainable and self-sufficient manner. When a DR event occurs, the ES and DR agents are
responsible for monitoring the SOC condition of ES systems and controlling corresponding loads (prosumers). They may shed or
Journal of Building Engineering 74 (2023) 106866
5
S.S. Ghazimirsaeid et al.
Fig. 1. Information exchange and relation among agents in the MAS structure.
reduce loads in response to the DR event, in order to balance the load and reduce overall power consumption. The shedding of
loads is managed through a process called load curtailment, which involves reducing the power consumption of non-critical loads
or temporarily switching them off. The shedding of loads is shared among the Home-Microgrids (H-MGs) through a coordinated
effort among the ES and DR agents, NSOs, CNSO, and retailers. The ES and DR agents monitor the SOC condition of ES systems and
control corresponding loads, while the NSOs communicate information about power generation and consumption to the CNSO agent.
The CNSO agent uses this information to evaluate the optimum price bids and define the amount of electrical/thermal energy to be
supplied by each NSO, in order to balance the load and ensure that the demand for power is met. The retailers declare the purchasing
and selling power to the CNSO agent, who uses this information to make decisions about the ideal biddings in the load distribution
system. Ultimately, the shedding of loads is shared among the H-MGs in a way that balances the load and minimizes overall power
consumption during the DR event. This coordinated effort among multiple agents helps to ensure the effective implementation of
DR and the efficient use of energy resources in the EMS system. As shown in Fig. 1, agents of the MAS in the proposed structure
are mutually cooperating to achieve the objectives in the market.
According to the figure, the topology consists of four levels such as retail level, central neighborhood system operator (CNSO)
level, neighborhood system operator (NSO) level, and H-MG level. At the H-MG level, generators, consumers and prosumers
communicate with NSO operators as cognitive agents to receive information such as generated and consumed power. Therefore,
the decision-making information is accessed by NSO. Further, the NSOs, CNSO and retailers perform as reactive agents to share
information on surplus power and inadequate power in the H-MG. Thereafter, CNSO agent selects the specific H-MGs with excess
power and insufficient power based on the information from the NSOs. The retailer agents distribute the details about power
availability and the demand, among buyers and sellers. Ultimately, CNSO agent selects the most appropriate buyers and sellers
for heat and electricity according to the offers received from NSOs and retailers. Therefore, it is confirmed that CNSO can work
with other agents, such as energy producers, distributors, and consumers, to establish a competitive and fair market for energy
trading. To measure and evaluate the optimum price bids, the CSNO can:
•Define the market rules: The central system operator can establish rules and regulations for the market, such as the types of
energy products traded, the bidding process, and the clearing and settlement procedures.
•Receive and process bids: The central system operator can receive bids from energy producers and distributors and process
them according to the market rules. The central system operator can use a sophisticated software system to handle the bidding
process and ensure that all bids are evaluated fairly.
•Monitor the market: The central system operator can monitor the market closely to detect any market manipulation or abuse.
The central system operator can use advanced market monitoring tools to analyze market data and detect any unusual trading
behavior.
Journal of Building Engineering 74 (2023) 106866
6
S.S. Ghazimirsaeid et al.
Fig. 2. Method of exchanging information in the proposed MAS.
•Collaborate with other agents: The central system operator can collaborate with energy producers, distributors, and
consumers to ensure that the market is efficient and effective. The central system operator can work with these agents to
set up efficient trading mechanisms, such as auctions or spot markets, that help to ensure fair pricing.
•Settle trades: The central system operator can settle trades according to the market rules. The central system operator can
ensure that all trades are settled promptly and accurately and that all parties receive their payments or energy products as
agreed.
In addition, the CNSO agent also plays a crucial role in promoting energy autonomy by encouraging the use of renewable
energy sources and minimizing dependence on non-renewable resources. By continuously monitoring and analyzing data on energy
availability and demand, the CNSO agent can help to identify opportunities for the integration of renewable energy sources into the
power grid, which can ultimately lead to a more sustainable and self-sufficient energy system. Overall, the CNSO agent is a critical
component of an effective energy management system, helping to ensure a reliable and efficient supply of energy while promoting
energy autonomy and sustainability. The steps of transferring information between agents are listed (shown in Fig. 2) as follows:
•NSO declares the generating power and the consuming power to communicate with HMGs and consumers.
•Information from the generators, consumers and prosumers are received from NSO.
•NSO distributes optimum load on power generated by the existing DERs in each H-MG and reports the amount of electrical/
thermal power availability and the demand to the CNSO. CNSO is responsible for global communication in local NSOs to
accomplish load demand in the H-MGs connected to each other. Further, the retailers declare the purchasing and selling
power to CNSO.
•CNSO cooperates with other agents to measure and evaluate the optimum price bids.
•CNSO defines the ultimate decisions such as the amount of electrical/thermal energy to be supplied by each NSO and calculates
the ideal biddings in the load distribution system, along with informing the final results to other agents for market clearing.
The implementation process of an algorithm with DERs in MO-TE (Market Operator based on Transactive Energy) to reduce the
electricity price while increasing the profit in power generation is presented in Fig. 3. This MO-TE structure consists of three main
units such as TOAT (Taguchi Orthogonal Array Test) unit, TE (Transactive Energy) unit, and MCP (Market Clearing Price) units.
According to Fig. 3, generated solar power, MCP, the load demand, SBP (System Buying Price) and SSP (System Selling Price) are
Journal of Building Engineering 74 (2023) 106866
7
S.S. Ghazimirsaeid et al.
Fig. 3. The process of implementing the proposed algorithm structure.
defined as intermittent parameters, where the TOAT controls the number of test to reduce the unnecessary testing load by providing
reliable statistical data. In addition, the MCP unit is responsible for measuring the MCP during each time period.
3.1. Communication protocol
In order to exchange data between agents, Controller Area Network-open (CANopen) protocol can be used as a communication
protocol [44]. The CAN bus is a fast field bus control system is used for decentralization, intelligence and network control that
can transfer data between different CAN stations [45,46]. The CANopen protocol is a suitable selection for exchanging information
between agents due to its expandability, cost-effectiveness in setting up, reliability and fast operation.
3.2. Taguchi orthogonal array test (TOAT) unit
Precise information forecasting is a vital issue in augmenting the accuracy of market structure modeling and planning in a
smart system. Nonetheless, implementing this matter is unattainable due to the stochastic nature of effective factors in renewable
energy generation, demand quantities, and energy prices. Diverse approaches have been presented to apply the uncertainty elements,
including Monte Carlo simulation, information-gap decision theory, and structures based on fuzzy modeling, which all differ in terms
of computational volume, time, and the number of application scenarios. One of the swiftest manners to consider uncertainty is
Taguchi’s orthogonal array testing (TOAT), which diminishes the computation volume and performance time by scrutinizing fewer
scenarios. Indeed, orthogonal arrays are employed to analyze the impact of factors on the response of average and variation in
Taguchi’s designs to achieve a balanced plan that the factor’s levels have the same weight. Therefore, each factor’s independent
Journal of Building Engineering 74 (2023) 106866
8
S.S. Ghazimirsaeid et al.
assessment is feasible since a distinct factor does not influence the estimation of another one. TOAT, with the least number of
scenarios, leads to selecting the optimal test scenarios in the worst case to protect the uncertainty sources in all situations. Besides,
it provides the possibility of discrete factors investigation exposing appropriate statistic data in unsure operation space. Hence, by
substantial scenario reduction, a robust solution can be accessible. Accordingly, in dynamic optimization problems which apply
several repetitions to determine the optimal point and even in designing cases with high sensitivity, utilizing this method saves time
effectively. In fact, TOAT’s purpose is to create a process that has less volatility and chaos to address instability with little or no
control and execute an assured procedure. An orthogonal array is a matrix that is represented as 𝐿𝐻(𝐵𝐹). H and F represent the
number of rows and columns of the matrix, respectively, and B indicates the number of levels of the matrix elements (uncertainty
variables). Also, F expresses the maximum number of uncertainty factors examined by the array. This method consists of three
primary steps as follows:
•Step 1: Opting for the orthogonal matrix according to the number of uncertainty’s factors in the problem.
•Step 2: Creating n initial test scenarios by using distribution functions.
•Step 3: Computing the possibility of occurrence of the generated scenario via employing probability distributions.
3.3. Transactive energy (TE) unit
TE unit has designed by several methods such as particle swarm optimization (PSO), harmony search (HS), differential evolution
(DE) and bat algorithm (BAT) methods (see algorithm 1 in Appendix A). Each approach has used electrical and thermal parts for
providing initial values to the variables. As observed in algorithm 1 (in Appendix A), when there is a power surplus occurred in the
system, the CHP obtain the priority and supplies the required load, and further demand is accomplished by discharging ES. However,
when the supplying part of electrical demand is not existing, the absent load demand is calculated and shifted to another time period,
where the MCP is comparatively lower. Moreover, the further energy surplus is obtained by purchasing from other retailers. On the
other hand, when the excess power is generated in the H-MG with the DR conditions at the initial stage of DR load requirement, the
ES is turned into charging mode. Thereafter, the further electrical power produced by each H-MG in the retail market is considered
as the disposal power. When there is a thermal power shortage (see algorithm 2 in Appendix A), the H-MG is engaged to provide the
thermal service prior to discharging the TES (if TES has the possibility of discharging), otherwise thermal power is purchased from
other H-MGs. In contrast, if excess thermal power is occurred, first the TES is turned to charging mode and in case of continued
excess generation, the amount of surplus power is supplied to achieve thermal power demand in other neighborhoods H-MG. Overall,
the utilization of TE units with the proposed control algorithm helps to improve the energy management system’s efficiency and
reliability by integrating different energy sources and storage systems, enabling optimal utilization of resources while minimizing
operational costs and reducing environmental impacts.
3.4. MCP unit
Market clearing price (MCP) is generally defined as the highest accepted offer. In the electricity market, the value of generated
and consumed power for each generation and consumption resource, and the proposed price is revealed to the market operator. In
particular, the generated power is arranged in ascending order while the value of the consumed power is sorted in descending order.
In this pace, the generators and consumers along with retailers, declare the highest offer price to purchase and sell the power. The
ultimate MCP value is defined by this unit for the objective functions of each market player. In addition, MCP creates a relationship
between power generation and power consumption. Further explanation regarding this unit has presented by authors in [47]. In
addition, by involving consumers and prosumers in the MCP, the framework can help to ensure that the demand response strategies
are implemented effectively and efficiently, leading to a more reliable and sustainable energy system, where the DR strategies in
the presented framework can help to improve the efficiency, stability, and profitability of the power grid.
4. Power grid under study
The general schematic of the system under study is shown in Fig. 4. The system has inumber of H-MGs where the electrical and
thermal DERs and as well, consumers are installed. Each H-MG consists of electrical and thermal stores, and a set of generation
resources including GB, TSP, ESP, CHP along with consumers containing NRL and RLD. In addition, this model consists of
controllable and uncontrollable devices, where controllable devices are the devices which can be controlled by the residents and the
building management system such as HVAC system and home appliances (e.g: dishwashers, pool pumps, dryers, washing machines).
Uncontrollable devices are the devices that cannot be controlled by the residents or the building management system, e.g: lighting
system, and flexible heating and cooling demands. Electrical loads are devices operated by electricity, e.g:- lighting, A/C, refrigerators
and computers, while thermal load are produce or consume heat such as HVAC systems, water heaters, and ovens.
5. Mathematical implementation of the problem
In this section, problem is mathematically formulated using key components in market structure based on transactive energy.
This framework is easily expandable for other electricity distribution systems with high level of consumer participation.
Journal of Building Engineering 74 (2023) 106866
9
S.S. Ghazimirsaeid et al.
Fig. 4. Schematic of neighborhood system with several H-MGs (continues and black line show the electrical part, gray dash shows the thermal part and the
dash-point is related to gas branch).
5.1. Objective functions of the participants in MO-TE
The defined objective functions on maximizing the generators and retailers profit, while minimizing the consumers costs are as
follows:
max ∑
∀𝑡∑
∀𝑖∑
∀𝑗∑
∀𝑘
(R𝑘,𝑒
𝑡,𝑖 +R𝐸𝑆 −,𝑒
𝑡,𝑖 +R𝑗,ℎ
𝑡,𝑖 +R𝑇 𝐸𝑆 −,ℎ
𝑡,𝑖
−C𝑗,ℎ
𝑡,𝑖 −C𝑇 𝐸𝑆 +,ℎ
𝑡,𝑖 −C𝐸𝑆 +,𝑒
𝑡,𝑖 −C𝑘,𝑒
𝑡,𝑖 ) × 𝛥𝑡
(1)
max ∑
∀𝑡∑
∀𝑖
(RRet-,𝑒
𝑡,𝑖 −CRet-,ℎ
𝑡,𝑖 ) × 𝛥𝑡 (2)
min ∑
∀𝑡∑
∀𝑖∑
∀𝑙∑
∀𝑚
(C𝑝,ℎ
𝑡,𝑖 −C𝑚,𝑒
𝑡,𝑖 ) × 𝛥𝑡 (3)
where, R𝑘,𝑒
𝑡,𝑖 and R𝑗,ℎ
𝑡,𝑖 are accordingly the electrical and thermal revenue from the DERs k and j in the H-MG i. The RES-,𝑒
𝑡,𝑖 and RTES-,ℎ
𝑡,𝑖
are respectively the revenue from ES and TES electrical and thermal discharge related to H-MG i at time t. Also, RRet-,𝑒
𝑡,𝑖 and RRet+,𝑒
𝑡,𝑖
are respectively the revenue and cost values of selling and buying electrical power from/to retailer for H-MG i. Alongside, C𝑝,ℎ
𝑡,𝑖 and
C𝑚,𝑒
𝑡,𝑖 are respectively electricity costs related to consumers p and m at H-MG i.
5.2. Technical and economic constraints
This section enlists the various technical and economic constraints.
5.2.1. Total electrical and thermal equilibrium
Eqs. (4) and (5) define that the total power generated by electrical and thermal generators during each time interval is equal
to the total demand of electrical and thermal consumers.
∑
∀𝑖∑
∀𝑘
(𝑃𝑘,𝑒
𝑡,𝑖 +𝑃ES-,𝑒
𝑡,𝑖 + (1 − 𝑋Ret
𝑡)⋅𝑃Ret-,e
𝑡,𝑖 )
=∑
∀𝑖∑
∀𝑚
(𝑃𝑚,𝑒
𝑡,𝑖 +𝑃ES+,𝑒
𝑡,𝑖 +𝑋Ret
𝑡
⋅𝑃Ret+,e
𝑡,𝑖 )
(4)
∑
∀𝑖∑
∀𝑗
(𝑃𝑗,ℎ
𝑡,𝑖 +𝑃TES-,ℎ
𝑡,𝑖 ) = ∑
∀𝑖∑
∀𝑙
(𝑃𝑝,ℎ
𝑡,𝑖 +𝑃TES+,ℎ
𝑡,𝑖 )(5)
5.2.2. Retailer constraints
Eq. (6) describes the cost of buying electrical power from retailer to H-MG, whereas Eq. (7) represents the offer price range of
purchasing the power for H-MG decided by the retailer.
CRet-,𝑒
𝑡,𝑖 =𝜋Ret-,𝑒
𝑡,𝑖 ×𝑃Ret-,e
𝑡,𝑖 (6)
0≤𝜋Ret-,𝑒
𝑡,𝑖 ≤𝜆SBP
𝑡(7)
RRet+,𝑒
𝑡,𝑖 =𝜋Ret+
𝑡,𝑖 ×𝑃Ret+,𝑒
𝑡,𝑖 (8)
Journal of Building Engineering 74 (2023) 106866
10
S.S. Ghazimirsaeid et al.
0≤𝜋Ret+
𝑡,𝑖 ≤𝜆SSP
𝑡,𝑒 (9)
The revenue of selling electrical power from H-MG i to the retailer is determined by Eq. (8), while Eq. (9) shows the price bid
of H-MG i to the retailer.
𝑃Ret+,𝑒
𝑡,𝑖 ≤𝑋Ret
𝑡×𝑃Ret (10)
𝑃Ret-,𝑒
𝑡,𝑖 ≤(1 − 𝑋Ret
𝑡) × 𝑃Ret (11)
𝑃Ret ≤(𝑃ESP,𝑒
𝑡,𝑖 +𝑃CHP,𝑒
𝑡,𝑖 +𝑃ES-,𝑒
𝑡,𝑖 )(12)
Lastly, Eqs. (10) and (11) present the exchanged power constraints between H-MG and retailer.
5.2.3. H-MG i constraints
ES and TES constraints in H-MG i
CES+,𝑒
𝑡,𝑖 =𝜋ES+,𝑒
𝑡,𝑖 ×𝑃ES+,𝑒
𝑡,𝑖 (13)
0≤𝜋ES+,𝑒
𝑡,𝑖 ≤𝜆MCP,e
𝑡(14)
RES-,𝑒
𝑡,𝑖 =𝜋ES-
𝑡,𝑖 ×𝑃ES-,𝑒
𝑡,𝑖 (15)
0≤𝜋ES-,𝑒
𝑡≤𝜆MCP,𝑒
𝑡(16)
where, CES+,𝑒
𝑡,𝑖 ,RES-,𝑒
𝑡,𝑖 ,𝜋ES+,𝑒
𝑡,𝑖 and 𝜋ES-,𝑒
𝑡,𝑖 respectively show the cost, revenue, and price bid resulting from buying/ selling electrical
power by ES in H-MG i. Eqs. (17) and (19) represent the maximum and minimum charging/ discharging of ES in H-MG i.
𝐸ES,𝑖 ≤𝐸ES,𝑒
𝑡,𝑖 ≤𝐸ES
𝑖(17)
𝑃ES-,𝑒
𝑡,𝑖 ≤𝑃ES-
𝑖×𝑋ES
𝑡,𝑖 , 𝑃 ES-,𝑒
𝑡,𝑖 ≥0(18)
𝑃ES+,𝑒
𝑡,𝑖 ≤𝑃ES+
𝑖×𝑋ES
𝑡,𝑖 , 𝑃 ES+,𝑒
𝑡,𝑖 ≥0(19)
𝑃ES-,𝑒
𝑡,𝑖 ×𝛥𝑡 ≤(𝐸ES
(𝑡−1),𝑖 −𝐸ES
𝑖)(20)
𝑃ES+,𝑒
𝑡,𝑖 ×𝛥𝑡 ≤(𝐸ES
𝑖−𝐸ES
(𝑡−1),𝑖)(21)
𝐸ES,𝑖
𝑡,𝑒 =𝐸ES,𝑖
𝑡−1,𝑒 + (𝑃ES+,𝑖
𝑡−1 −𝑃ES-,𝑖
𝑡−1 ) × 𝛥𝑡 (22)
The maximum charging/ discharging limits determine by Eqs. (20) and (21), where the energy existing in Eq. (22) states energy
equilibrium in ES.
CTES+,𝑖
𝑡,ℎ =𝜋TES+,𝑖
𝑡,ℎ ×𝑃TES+,𝑖
𝑡,ℎ (23)
0≤𝜋TES+,𝑖
𝑡,𝑒 ≤max(𝜋HHW,𝑖
𝑡,ℎ , 𝜋TD,𝑖
𝑡,ℎ )(24)
Eq. (23) explain the cost resulting from buying thermal power by TES in the charging mode, where price bid interval is given
in Eq. (24).
RTES-,𝑖
𝑡,ℎ =𝜋TES-,𝑖
𝑡,ℎ ×𝑃TES-,𝑖
𝑡,ℎ (25)
0≤𝜋TES-,𝑖
𝑡,ℎ ≤min(max(𝜋CHP,𝑖
𝑡,ℎ , 𝜋GB,𝑖
𝑡,ℎ ), 𝜋TSP,𝑖
𝑡,ℎ )(26)
where, RTES-,𝑖
𝑡,ℎ in Eq. (25) is defined the revenue of selling thermal power generated by TES in the discharging mode. The 𝜋TES-,𝑖
𝑡,ℎ in
Eq. (26) is represented the price bid range for selling thermal power.
𝐸TES,𝑖 ≤𝐸TES,𝑖
𝑡,ℎ ≤𝐸TES,𝑖 (27)
𝑃TES-,𝑖
𝑡,ℎ ≤𝑃TES-,𝑖 ×𝑋TES,𝑖
𝑡, 𝑃 TES-,𝑖
𝑡,ℎ ≥0(28)
𝑃TES+,𝑖
𝑡,ℎ ≤𝑃TES+,𝑖, 𝑃 TES+,𝑖
𝑡,ℎ ≥0(29)
The TES maximum and minimum charging/ discharging limitations are given in Eqs. (27) to (29).
𝑃TES-,𝑖
𝑡,ℎ ×𝛥𝑡 ≤(𝐸TES,𝑖
𝑡−1 −𝐸TES,𝑖)(30)
Journal of Building Engineering 74 (2023) 106866
11
S.S. Ghazimirsaeid et al.
𝑃TES+,𝑖
𝑡,ℎ ×𝛥𝑡 ≤(𝐸TES,𝑖 −𝐸TES,𝑖
𝑡−1 )(31)
𝐸TES,𝑖
𝑡,ℎ =𝐸TES,𝑖
𝑡−1,ℎ + (𝑃TES+,𝑖
𝑡−1,ℎ −𝑃TES-,𝑖
𝑡−1,ℎ ) × 𝛥𝑡 (32)
Eqs. (30) and (31) derive the discharge/ charge maximum limit for the energy in TES, whereas Eq. (32) defines the energy
equilibrium in TES.
EV constraints in H-MG i
if 𝑋EV,𝑖
𝑡= 1 ⟹𝑃EV+,𝑖 ≤𝑃EV+,𝑖
𝑡,𝑒 ≤𝑃EV+,𝑖 (33)
SOCEV,𝑖
𝑡≤SOCEV,𝑖 (34)
SOCEV,𝑖
𝑡=SOCEV,𝑖
𝑡−1 −
𝑃EV+,𝑖
𝑡,𝑒 ×𝛥𝑡
𝐸EV,𝑖
Tot
(35)
SOCEV,𝑖
𝑡=SOCEV,𝑖
⟹𝑋EV,𝑖
𝑡, 𝑃 EV+,𝑖
𝑡,𝑒 (36)
Eq. (34) describes that SOCEV,𝑖
𝑡of automobile battery during each time interval must be less than its maximum value, wherein,
Eq. (35) is the automobile battery power equilibrium constraint.
CEV+,𝑖
𝑡,𝑒 =𝜋EV+,𝑖
𝑡,𝑒 ×𝑃EV+,𝑖
𝑡,𝑒 (37)
0≤𝜋EV+,𝑖
𝑡,𝑒 ≤𝜆MCP
𝑡,𝑒 (38)
The cost of purchasing electrical power by EV is presented in Eq. (37), and (38) shows the scope of the offer of EV price for
buying power.
ESP constraints in H-MG i
Eq. (39) presents the limitations of ESP generated power.
𝑃ESP,𝑖 ≤𝑃ESP,𝑖
𝑡,𝑒 ≤ESP, 𝑖 (39)
In addition, Eq. (40) is determined the revenue resulting from ESP generated electrical power, and the corresponding price bid
interval is defined in Eq. (41).
RESP,𝑖
𝑡,𝑒 =𝜋ESP,𝑖
𝑡,𝑒 ×𝑃ESP,𝑖
𝑡,𝑒 (40)
0≤𝜋ESP,𝑖
𝑡,𝑒 ×𝜆MCP,𝑖
𝑡,𝑒 (41)
TSP constraints in H-MG i
Eq. (42) derived the generated thermal power income by TSP. Eq. (43) is shown the price bid interval for selling power by TSP.
RTSP,𝑖
𝑡,ℎ =𝜋TSP,𝑖
𝑡,ℎ ×𝑃TSP,𝑖
𝑡,ℎ (42)
0≤𝜋TSP,𝑖
𝑡,ℎ ≤max(𝜋TES-,𝑖
𝑡,𝑒 , 𝜋CHP,𝑖
𝑡,ℎ , 𝜋GB,𝑖
𝑡,ℎ ,)(43)
CHP constraints in H-MG i
Eq. (44) expresses the CHP power generation limits, whereas in Eqs. (45) and (46), FUCHP,𝑖
𝑡,𝜁CHP,𝑖
𝑒1and 𝜁ℎCHP,𝑖 are represented
fuel, electrical and thermal efficiency of CHP, respectively.
𝑃CHP,𝑖 ≤𝑃CHP,𝑖
𝑡,𝑒 ≤𝑃CHP,𝑖 (44)
𝑃CHP,𝑖
𝑡,𝑒 =FUCHP,𝑖
𝑡×𝜁CHP,𝑖
𝑒1+𝜁CHP,𝑖
𝑒2(45)
𝑃CHP,𝑖
𝑡,𝑒 =𝜁CHP,𝑖
𝑒1×
𝑃CHP,𝑖
𝑡,ℎ
𝜁ℎCHP,𝑖 +𝜁CHP,𝑖
𝑒2(46)
CCHP,𝑖
𝑡=𝜋NG
𝑡×FUCHP,𝑖
𝑡(47)
CCHP,𝑖
𝑡≤𝜋CHP,𝑖
𝑡≤2 × CCHP,𝑖
𝑡(48)
RCHP,𝑖
𝑡,𝑒 =𝜋CHP,𝑖
𝑡,𝑒 ×𝑃CHP,𝑖
𝑡,𝑒 (49)
RCHP,𝑖
𝑡,ℎ =𝜋CHP,𝑖
𝑡,ℎ ×𝑃CHP,𝑖
𝑡,ℎ (50)
Eq. (47) is described the cost of power generation by CHP, and the respective price bid is shown in Eq. (48). Further, Eqs. (49)
and (50) state the revenue of selling electrical and thermal powers generated by CHP.
Journal of Building Engineering 74 (2023) 106866
12
S.S. Ghazimirsaeid et al.
GB constraints in H-MG i
The limits of the power generated by GB is given in Eq. (51), while Eq. (52) expresses the GB cost from generating thermal power.
Similarly, Eq. (53) provides the amount of fuel consumed by GB and Eq. (54) is derived the price bid range for selling power.
0≤𝑃GB,𝑖
𝑡,ℎ ≤𝑃GB,𝑖
𝑡,ℎ (51)
CGB,𝑖
𝑡,ℎ =𝜋NG
𝑡,ℎ ×FUGB,𝑖
𝑡(52)
FUGB,𝑖
𝑡=𝑃GB,𝑖
𝑡
𝜁GB
ℎ
(53)
CGB,𝑖
𝑡,ℎ ≤𝜋GB,𝑖
𝑡,ℎ ≤2 × CGB,𝑖
𝑡,ℎ (54)
The revenue of selling thermal power by GB is shown in Eq. (55).
RGB,𝑖
𝑡,ℎ =𝜋GB,𝑖
𝑡,ℎ ×𝑃GB,𝑖
𝑡,ℎ (55)
5.2.4. Consumers constraints in H-MG i
DR constraints in H-MG i
Eq. (56) is determined that the value of shiftable power should be less than or equal to the sum of total consumed power minus
the total generated power. Furthermore, Eq. (58) is represented that the DR limit between two consecutive intervals must not exceed
a certain limit.
𝑃DR-,𝑖
𝑡≤(𝑃TCP,𝑖
𝑡−𝑃TGP,𝑖
𝑡)⋅𝑋DR-,𝑖
𝑡(56)
𝑃DR+,𝑖
𝑡≤(𝑃TGP,𝑖
𝑡−𝑃TCP,𝑖
𝑡)⋅(1 − 𝑋DR-,𝑖
𝑡)(57)
𝑃DR+,𝑖
𝑡≤𝑘𝜖×𝑃NRL,𝑖
𝑡× (1 − 𝑋DR-,𝑖
𝑡)(58)
−𝑘𝑡≤(𝑃DR+,𝑖
𝑡−𝑃DR+,𝑖
𝑡−1 )≤𝑘𝑡(59)
ATL and AEL constraints in H-MG i
Eqs. (60) and (61) are the costs of buying electric and thermal power by AEL and ATL, wherein Eqs. (62) and (63) explain the
price bid interval for buying power by AEL and ATL.
CAEL,𝑖
𝑡,𝑒 =𝜋AEL,𝑖
𝑡,𝑒 ×𝑃AEL,𝑖
𝑡,𝑒 (60)
CATL,𝑖
𝑡,𝑒 =𝜋ATL,𝑖
𝑡,𝑒 ×𝑃ATL,𝑖
𝑡,𝑒 (61)
𝜆MCP
𝑡,𝑒 ≤𝜋AEL,𝑖
𝑡,𝑒 ≤2 × 𝜆MCP
𝑡,𝑒 (62)
max(𝜋TES-,𝑖
𝑡,ℎ , 𝜋CHP,𝑖
𝑡,ℎ , 𝜋GB,𝑖
𝑡,ℎ , 𝜋TSP,𝑖
𝑡,ℎ )≤𝜋ATL,𝑖
𝑡,ℎ ≤
2 × max(𝜋TES-,𝑖
𝑡,ℎ , 𝜋CHP,𝑖
𝑡,ℎ , 𝜋GB,𝑖
𝑡,ℎ , 𝜋TSP,𝑖
𝑡,ℎ )(63)
TD constraints in H-MG i
Eq. (72) presents the cost of buying thermal power by TD, where Eq. (73) states the offer price range.
CTD,𝑖
𝑡,ℎ =𝜋TD,𝑖
𝑡,ℎ ×𝑃TD,𝑖
𝑡,ℎ (64)
0≤𝜋TD,𝑖
𝑡,ℎ ≤min(𝜋TES-,𝑖
𝑡,ℎ , 𝜋CHP,𝑖
𝑡,ℎ , 𝜋GB,𝑖
𝑡,ℎ , 𝜋TSP,𝑖
𝑡,ℎ ,)(65)
REF constraints in H-MG i
The expenses of buying power by FER is derived in Eq. (69), while Eq. (70) is defined the offer price interval when purchasing
energy.
{if 𝑇REF,𝑖 ≤𝑇RET
𝑡≤𝑇REF,𝑖 𝑋REF,𝑖
𝑡= 0
Otherwise 𝑋REF,𝑖
𝑡= 1 (66)
𝑋REF,𝑖
𝑡= 1 ⟹𝑃REF,𝑖
𝑡,𝑒 =𝑃REF,𝑖, 𝑇 REF,𝑖
𝑡=𝑇REF,𝑖
𝑡−1 −𝑇RED,𝑖 (67)
𝑋REF,𝑖
𝑡= 0 ⟹𝑃REF,𝑖
𝑡,𝑒 = 0, 𝑇 REF,𝑖
𝑡=𝑇REF,𝑖
𝑡−1 +𝑇RED,𝑖 (68)
CREF,𝑖
𝑡,𝑒 =𝜋REF,𝑖
𝑡,𝑒 ×𝑃REF,𝑖
𝑡,𝑒 (69)
0≤𝜋REF,𝑖
𝑡,𝑒 ≤𝜆MCP
𝑡,𝑒 (70)
DW constraints in H-MG i
Journal of Building Engineering 74 (2023) 106866
13
S.S. Ghazimirsaeid et al.
Eqs. (73) and (74) express the cost for buying power by DW, and price bid scope for buying power, respectively.
if 𝑋DW,𝑖
𝑡= 1 ⟹𝑃DW,𝑖
𝑡,𝑒 =𝑃DW,𝑖,DTDW,𝑖
𝑡=DTDW,𝑖
𝑡−1 + 1 (71)
if DTDW,𝑖
𝑡=DTDW,𝑖
⟹𝑃DW,𝑖
𝑡,𝑒 = 0, 𝑋DW,𝑖
𝑡(72)
CDW,𝑖
𝑡,𝑒 =𝜋DW,𝑖
𝑡,𝑒 ×𝑃DW,𝑖
𝑡,𝑒 (73)
0≤𝜋DW,𝑖
𝑡,𝑒 ≤𝜆MCP
𝑡,𝑒 (74)
HHW constraints in H-MG i
Finally, the HHW limitations could be evaluated as below:
{if 𝑇HHW,𝑖 ≤𝑇HHW
𝑡≤𝑇HHW,𝑖 𝑋HHW,𝑖
𝑡= 0
Otherwise 𝑋HHW,𝑖
𝑡= 1 (75)
𝑋HHW,𝑖
𝑡= 1 ⟹{𝑃HHW,𝑖
𝑡,𝑒 =𝑃HHW,𝑖
𝑇HHW,𝑖
𝑡=𝑇HHW,𝑖
𝑡−1 +𝑇INC,𝑖 (76)
𝑋HHW,𝑖
𝑡= 0 ⟹{𝑃HHW,𝑖
𝑡,𝑒 = 0
𝑇HHW,𝑖
𝑡=𝑇HHW,𝑖
𝑡−1 −𝑇INC,𝑖 (77)
CHHW,𝑖
𝑡,ℎ =𝜋HHW,𝑖
𝑡,ℎ ×𝑃HHW,𝑖
𝑡,ℎ (78)
0≤𝜋HHW,𝑖
𝑡,ℎ ≤max(𝜋TES-,𝑖
𝑡,ℎ , 𝜋CHP,𝑖
𝑡,ℎ , 𝜋GB,𝑖
𝑡,ℎ , 𝜋TSP,𝑖
𝑡,ℎ ,)(79)
6. Results and discussion
In this section, the simulation results of each four methods is presented. The system under study has three H-MGs named as, A,
Band C, which include different DER and consuming sources. The specifications are described further in Appendix B.
The total power generated by electrical and thermal DERs in each GB, and the total electrical power sold and purchased between
GBs and retailers with four of optimization methods is illustrated in Fig. 5. As it is observed in Fig. 5(a), maximum power generated
by electrical DERs in GB (#A) is obtained by HS method due to the absent of power transferring from GB to retailer. Further, the total
power allocated for DR+ has the least value compared to other methods. The higher amount of total power produced in this H-MG
facilitates to sell energy to other neighborhood H-MG, concurrently increasing the revenue of H-MG (#A). Besides, the conditions
for H-MG (#B) is different compared to other H-MG, as the value of DE method is relatively higher when generating power, due to
purchasing the power from another retailer. From Figs. 5(b) and 5(c), it is observed that GB #B in the PSO optimization method
has a greater interaction with the retailer compared to other methods. Due to the average value of electrical MCP in PSO method is
lower than other methods, GB #B facilitates the numerous consumers with lower MCP buying price from the retailer. Furthermore,
the DR+ consumed power by HS method is 27% of the total consumed (DR+) power by other optimization methods. In particular,
the HS method tempts to purchase more energy from the retailers to produce more RLD loads in the MO-TE model.
According to Fig. 5(a), in GB #C, the value of total power generated by BAT method is the maximum compared to other
optimization methods. Moreover, the power exchange value with retailer for GB #C has the highest limit for BAT, as illustrated from
Figs. 5(b) and 5(c). The reason is that, the value of sum DR- has reached its lowest possible limit with regards to other methods (6%),
while about 26% of total DR+ power shared to BAT method which is a significantly higher value among the other implemented
methods. Since the average value of electrical MCP in the BAT method is lower with respect to HS and DE methods, the consumers in
this GB could receive the power for a lower price. In addition, it is noticeable that the minimum value of electrical MCP is obtained
in the BAT method.
In contrast, the amount of the total power generated by thermal DER for each GB is shown in Fig. 5(d). In GBs #A and #C,
most of thermal power is produced by HS method, while in GB #B the total energy is generated using BAT method. However, the
average value of thermal MCP by HS and BAT methods is lower than other methods due to the occurrence of a smart selection on
further power generation by thermal DER. This is because, the minimum value of thermal MCP is achieved in these methods and
maximum value of thermal MCP is gained in DE and PSO methods which could lead to a significant increase in the thermal power
cost. Therefore, less thermal power is generated by DE and PSO methods while optimizing the profit of GB owner and satisfying
the consumer thermal power requirement.
In Fig. 6, the consumed load profile is presented for each GB. According to Fig. 6(a), the consumption peak value by PSO and
BAT methods in GB #A has been shifted to non-peak intervals. In fact, the average of MCP value in peak intervals is high in
every optimization methods. However, the consumers are managed to pay a less amount of price to GB owner and retailers when
exchanging power, due to the participation of consumers in the DR program. This is when the total value of DR+ in BAT method is
about 28% of total DR+ among the other types. In addition, it is expected that PSO method follows a similar pattern for consuming
resources to increase the DR+ value. Evaluation expresses that about 26% of DR+ generation among implemented methods are
shared by PSO.
As evaluated in GB #C, the total values of DR- in DE and BAT are equal, which is nearly 28% of the total DR- proposed by all
the methods. The minimum value of total DR in HS method represents the lack of shifting demand from one time interval with
Journal of Building Engineering 74 (2023) 106866
14
S.S. Ghazimirsaeid et al.
Fig. 5. The total amount of electrical and thermal powers consumed by each GB using different optimization methods.
Fig. 6. Consumed load demand profile in the GBs.
high price to another with much lower price. This is because, the cost of electricity production by the entire H-MGs contains the
maximum value for all the methods due to a 28% decrement compared to the DE method, which facilities the minimum electricity
production cost. The HS method is utilized to gain the maximum GB #B value of DR+, as a result of DR- value demonstrates a
notable reduction by using this method. Due to high cost of generated power, the proposed algorithm desires to reduce the power
consumption in the GB. Most significantly, although the amount of generation in the BAT method has the highest value after HS,
the total value of DR- has become minimum relative to other methods. Therefore, BAT method has preferably increased the DR+
value. In GB #C, the DR+ and DR- values are maximum in the PSO method with regards to other types. In addition, PSO method
has the minimum value of electricity cost after DE method.
It is concluded that, more value of DR+ is supplied by BAT method and meanwhile reduces the DR- value to minimum value.
As discussed before, the electricity generation cost and the average electrical MCP in the BAT method compared to other methods
is high during 24 h performance of the grid under study. Thereby, supplying the DRs during relevant times reduces the power cost
for consumers.
Figs. 7(a) and 7(b) describe the values of electrical and thermal MCPs obtained from simulation by each optimization methods. In
fact, Fig. 7(a) explains that MCP values in all approaches have been reduced to predicted MCP to occur in 100% of time period. At the
Journal of Building Engineering 74 (2023) 106866
15
S.S. Ghazimirsaeid et al.
Fig. 7. MCP profile in each time interval under the 24 h performance of the system under study using different optimization methods.
beginning, PSO method has performed successfully in reducing the MCP compared to BAT method which the weakest performance
is presented. In the morning, PSO approach shows a most effective performance in MCP reduction, while the worst functionality is
occurred in HS method where the MCP increases for around 83%.
During the latter time intervals, HS performance was the worst performance among all the methods, as opposed to the
functionality at the initial stage. The PSO method has achieved lower MCP value relative to DE (about 34%). Although PSO has the
best operation during this time interval (afternoon), it has the worst performance among all the methods during the afternoon to
evening. However, HS method has gained the best performance during afternoon to sunset where the electrical MCP has reached
its minimum value, which Is nearly 78% of the time compared to PSO method. During the last few hours of the day, HS method
has obtained a greater value for MCP which is about 22% of the time. Overall, the best performance for MO-TE structure over 24 h
period has achieved by HS method, which is approximately 6%, 9% and 62% of times less MCP value respectively in comparison
to PSO, BA and DE. In Fig. 7(b), expresses that the PSO method has a considerable proportion in decreasing the thermal MCP, at
the beginning, and midnight to morning. Therefore, it has obtained the minimal value compared to other optimization methods.
The worst result during this time interval is for BAT, about 77% of the time because it gains a higher thermal MCP value as a
result of utilizing PSO method. In the morning, the best performance is obtained by DE, while the PSO method has notably reduced
the functionality due to the decline in thermal MCP by about 45% of the time. In the time period of 12:00 to 18:00, DE has presented
the best performance compared to others. In contrast, BAT has a worse performance of 70% weaker than DE. Further, during the
final hours of the day, BAT has the best performance relative to others. Considering all the GBs, performance of DE algorithm is 2%
preferable than BAT, and 28% better than HS and PSO. Therefore, DE has the best performance in reducing the MCP value.
7. Conclusions
In this paper, an effective energy management algorithm based on MAS has proposed for several H-MGs connected in a
neighborhood grid. Optimization algorithms were suggested in order to increase the DERs utilization while maximizing the profit
for all the market players. Further, the DR strategy has improved the model efficiency and consumer profitability by shifting the
demand from peak hours (high MCP) to non-peak hours (low MCP). Energy exchange among the GBs has influenced the competition
in the power market while encouraging consumers to access different DERs. Moreover, the presence of active consumers (referred
to as prosumers) has enhanced the total revenue. The main contributions of the proposed structure could be summarized as follows.
∙The model has introduced an effective decision making strategy which is equally distributed among the players (consumers,
prosumers and retailers) with improved participation in the H-MG.
∙The simulation results have shown without shortage in the implemented scenarios that even with fast changes in the strategies
of the players participating in the market, easily take their proper strategy without knowing the pricing date of other players.
This performance has been obtained by the effect of other players on the MCP price.
∙The difficulty in increasing the grid network and complexity of the decision making variables in the EMS based one MAS, has
been eliminated by introducing a smart algorithm in optimization methods.
∙The proposed structure has increased the awareness for all the market players about the participation of DR program with
advanced management.
Journal of Building Engineering 74 (2023) 106866
16
S.S. Ghazimirsaeid et al.
CRediT authorship contribution statement
Seyedeh Samaneh Ghazimirsaeid: Conceptualization, Methodology, Formal analysis, Investigation, Writing – original draft.
Mansour Selseleh Jonban: Formal analysis, Investigation, Writing – original draft, Review & editing. Manthila Wijesooriya
Mudiyanselage: Writing – original draft, Writing – review & editing. Mousa Marzband: Validation, Formal analysis, Writing
– review & editing, Supervision, Funding acquisition. Jose Luis Romeral Martinez: Visualization, Writing – review & editing.
Abdullah Abusorrah: Visualization, Writing – review & editing, Funding acquisition.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared
to influence the work reported in this paper.
Data availability
No data was used for the research described in the article.
Acknowledgments
This research work was funded by Institutional Fund Projects under grant no. (IFPIP: 1205-135-1443). Therefore, the authors
gratefully acknowledge technical and financial support from Ministry of Education and Deanship of Scientific Research (DSR), King
Abdulaziz University (KAU), Jeddah, Saudi Arabia.
Appendix A
See Algorithms 1 and 2.
Algorithm 1 EMS-MAS algorithm (Electrical section)
Require: ⊳For all BAT method, PSO method, DE method, HS method
⊳Hourly prediction data of MG, SOC𝑥, SOC𝑥,𝑃𝑥
𝑒∕ℎ,𝑃𝑥
𝑒∕ℎ, E𝑥
Tot,𝑇𝑦,𝑇𝑦,𝑃𝑗
𝑒∕ℎ,𝑃𝑗
𝑒∕ℎ, T𝑦
INI, TRED , TINC,𝜁𝑗
𝑒∕ℎ,𝑇HHW,𝑇HHW ,𝐸𝑥,𝐸𝑥,𝐸𝑥,𝐸𝑥,𝜋𝑧,𝜋𝑧,
𝜋NG
𝑡,
𝜆MCP
𝑡
⊳x∈ {ES+,ES-,EV+,EV-,TES+,TES-}
⊳y∈ {REF,HHW}
⊳j∈ {CHP,GB,TSP}
⊳z∈ {𝑗, 𝑘, 𝑚, 𝑙}
while t≤24 do
⊳Electrical part
𝑃req
𝑡,𝑒 = (𝑃ESP
𝑡−𝑃𝑛
𝑡,𝑒)(80)
if 𝑃req
𝑡,𝑒 ≥0then ⊳Excess Generation Section
𝑃Ex
𝑡,𝑒 = (𝑃ESP
𝑡−𝑃req
𝑡,𝑒 )(81)
if 𝑃Ex1
𝑡,𝑒 ≥𝑃DR-
𝑡,𝑒 then
𝑃DR+
𝑡,𝑒 = Responding to 𝑃DR-
𝑡,𝑒 completely (Eqs. (56)–(59))
𝑃Ex2
𝑡,𝑒 = (𝑃Ex1
𝑡,𝑒 −𝑃DR+
𝑡,𝑒 )(82)
if SOC𝑡≤SOC ≤SOC𝑡then Fully charged mode
𝑃Ex3
𝑡,𝑒 = (𝑃Ex2
𝑡,𝑒 −𝑃ES
𝑡,𝑒 )(83)
if 𝑃Ex3
𝑡,𝑒 >0then
Dedicating excess power to grid or EWH
else
Exit the loop
end if
else
Go to grid and EWH. Then, exit the loop
end if
Journal of Building Engineering 74 (2023) 106866
17
S.S. Ghazimirsaeid et al.
else
𝑃DR+
𝑡,𝑒 =𝑃Ex1
𝑡,𝑒 and exit the loop (Eqs. (56)–(59))
end if ⊳Shortage Power Section
end if
if 𝑃req
𝑡,𝑒 ≥𝑃CHP
𝑡,𝑒 then ⊳Eqs. (44)–(46)
CHP enters to the circuit
if 𝑃CHP
𝑡,𝑒 ≥𝑃req
𝑡,𝑒 then Go to excess generation section
else
if SOC𝑡> SOC𝑡then ⊳Eqs. (34)–(36)
Discharging Mode
if 𝑃req
𝑡,𝑒 >0then
if 𝑃EV
𝑡,𝑒 ,𝑃DW
𝑡,𝑒 >0then ⊳Eqs. (71)–(72) and (33)
𝑃EV
𝑡,𝑒 +𝑃DW
𝑡,𝑒 = Buy power from Grid and exit the for loop
else𝑃DR-
𝑡,𝑒 =𝑃req
𝑡,𝑒 and exit the for loop
end if𝑃DR-
𝑡,𝑒 =𝑃req
𝑡,𝑒 and exit the for loop
else
𝑃EV
𝑡,𝑒 ,𝑃DW
𝑡,𝑒 > 0
𝑃EV
𝑡,𝑒 +𝑃DW
𝑡,𝑒 = Buy power from Grid and exit the for loop
end if
end if
end if
else
𝑃DR-
𝑡,𝑒 =𝑃req
𝑡,𝑒 and exit the for loop
end if
return Return determine the optimum capacity and profit of the all players.
Algorithm 2 EMS-MAS algorithm (Thermal section)
while t≤24 do
𝑃𝑟𝑒𝑞
𝑡,ℎ = (𝑃𝐸𝑇 𝑃
𝑡,ℎ +𝑃𝐶𝐻 𝑃
𝑡,ℎ −𝑃𝑛
𝑡,ℎ)(84)
if 𝑃𝑟𝑒𝑞
𝑡,ℎ ≥0then ⊳Excess thermal Section
if 𝑆𝑂𝐶 𝑡,ℎ ≤𝑆 𝑂𝐶𝑡,ℎ ≤𝑆𝑂𝐶 𝑡,ℎ then ⊳Fully charged mode
if 𝑃𝐸𝑥1
𝑡,ℎ = (𝑃𝑟𝑒𝑞
𝑡,ℎ −𝑃𝐸𝑆
𝑡,ℎ )>0
Dedicating excess power to thermal dump ⊳Eq. (64)–(65)
else
Exit the loop then
end if
else
Dedicating excess power to thermal dump and exit the for loop
end if
else 𝑃𝑟𝑒𝑞
𝑡,ℎ <0⊳Shortage thermal Section
𝑃𝑟𝑒𝑞
𝑡,ℎ ≥𝑃𝐺𝐵 ⊳Eq. (51)–(53)
Gas boiler enters to the circuit
if 𝑃𝐺𝐵
𝑡> 𝑃 𝑟𝑒𝑞
𝑡,ℎ then
Go to the excess thermal section
else 𝑆𝑂𝐶 𝑡,ℎ > 𝑆 𝑂𝐶ℎ⊳Eq. (27)–(29)
Discharging mode
if 𝑃𝑟𝑒𝑞
𝑡,ℎ >0then
Buy Power from Virtual Source and exit the for loop
else Exit the for loop
end if
Buy power from virtual source and exit the for loop
end if
end if
end while
return Return determine the optimum capacity and profit of the all players.
Journal of Building Engineering 74 (2023) 106866
18
S.S. Ghazimirsaeid et al.
Appendix B
Name of DER Variable Value Name of DER Variable Value
GB
𝜁GB
ℎ85%
HHW
𝑃HHW
𝑡,𝑒 0.5
𝑃GB
ℎ12 𝑇HHW
INI ,𝑇HHW 18
𝑃GB
ℎ3.6 TINC 6
CHP
𝜁CHP
𝑒2−94.6916
ES
𝑃ES+
𝑡,𝑒 30
𝜁CHP
𝑒10.358511 𝑃ES+
𝑡,𝑒 0.34
𝑃CHP
𝑒8𝑃ES-
𝑡,𝑒 30
𝑃CHP
𝑒2𝑃ES-
𝑡,𝑒 0.34
SOCES 0
SOCES 100%
DW 𝑃DW 0.42 Natural gas 𝜋NG
𝑡0.012
EV
𝑃EV+
𝑡,𝑒 3.2
HHW
𝑃HHW
𝑡,𝑒 0.5
𝑃EV+
𝑡,𝑒 0𝑇HHW
INI ,𝑇HHW 18
SOCEV 0 TINC 6
SOCEV 100% 𝑇HHW 36
REF
𝑃REF
𝑡,𝑒 0.12
TES
𝑃TES+ 14.4
𝑇REF 9𝑃TES+ 0
𝑇REF 3𝑃TES- 14.4
𝑇REF
INI 27 𝑃TES- 0
𝑇INI 6
DR 𝑘𝜖5
𝑘𝑡5
References
[1] M.H. Jahangir, S. Eslamnezhad, S.A. Mousavi, M. Askari, Multi-year sensitivity evaluation to supply prime and deferrable loads for hospital application
using hybrid renewable energy systems, J. Build. Eng. 40 (2021) 102733.
[2] E. Rosales-Asensio, D. Icaza, N. González-Cobos, D. Borge-Diez, Peak load reduction and resilience benefits through optimized dispatch, heating and cooling
strategies in buildings with critical microgrids, J. Build. Eng. 68 (2023) 106096.
[3] B. Liu, D. Rodriguez, Renewable energy systems optimization by a new multi-objective optimization technique: A residential building, J. Build. Eng. 35
(2021) 102094.
[4] R. Tang, S. Wang, H. Li, Game theory based interactive demand side management responding to dynamic pricing in price-based demand response of smart
grids, Appl. Energy 250 (2019) 118–130.
[5] R.K. Chauhan, K. Chauhan, Building automation system for grid-connected home to optimize energy consumption and electricity bill, J. Build. Eng. 21
(2019) 409–420.
[6] S.E. Ahmadi, S.M. Kazemi-Razi, M. Marzband, A. Ikpehai, A. Abusorrah, Multi-objective stochastic techno-economic-environmental optimization of
distribution networks with G2V and V2G systems, Electr. Power Syst. Res. 218 (2023) 109195.
[7] R. Khalili, A. Khaledi, M. Marzband, A.F. Nematollahi, B. Vahidi, P. Siano, Robust multi-objective optimization for the Iranian electricity market considering
green hydrogen and analyzing the performance of different demand response programs, Appl. Energy 334 (2023) 120737.
[8] J.M. Vidal, Fundamentals of multiagent systems with NetLogo, 2010, Examples.–2010, March 1.
[9] M. Nasir, A. Rezaee Jordehi, M. Tostado-Véliz, S.A. Mansouri, E.R. Sanseverino, M. Marzband, Two-stage stochastic-based scheduling of multi-energy
microgrids with electric and hydrogen vehicles charging stations, considering transactions through pool market and bilateral contracts, Int. J. Hydrogen
Energy (2023).
[10] P. Balakumar, T. Vinopraba, K. Chandrasekaran, Real time implementation of demand side management scheme for IoT enabled PV integrated smart
residential building, J. Build. Eng. 52 (2022) 104485.
[11] M.S. Jonban, L. Romeral, A. Akbarimajd, Z. Ali, S.S. Ghazimirsaeid, M. Marzband, G. Putrus, Autonomous energy management system with self-healing
capabilities for green buildings (microgrids), J. Build. Eng. 34 (2021) 101604.
[12] F. Ahmad, A. Iqbal, I. Asharf, M. Marzband, I. Khan, Placement and capacity of EV charging stations by considering uncertainties with energy management
strategies, IEEE Trans. Ind. Appl. (2023) 1–10.
[13] F.H. Aghdam, M.W. Mudiyanselage, B. Mohammadi-Ivatloo, M. Marzband, Optimal scheduling of multi-energy type virtual energy storage system in
reconfigurable distribution networks for congestion management, Appl. Energy 333 (2023) 120569.
[14] J. Ebrahimi, M. Abedini, A two-stage framework for demand-side management and energy savings of various buildings in multi smart grid using robust
optimization algorithms, J. Build. Eng. 53 (2022) 104486.
Journal of Building Engineering 74 (2023) 106866
19
S.S. Ghazimirsaeid et al.
[15] H. Saber, M. Ehsan, M. Moeini-Aghtaie, M. Fotuhi-Firuzabad, M. Lehtonen, Network-constrained transactive coordination for plug-in electric vehicles
participation in real-time retail electricity markets, IEEE Trans. Sustain. Energy 12 (2) (2020) 1439–1448.
[16] C. Puttamadappa, B. Parameshachari, Demand side management of small scale loads in a smart grid using glow-worm swarm optimization technique,
Microprocess. Microsyst. 71 (2019) 102886.
[17] S. Areekkara, R. Kumar, R.C. Bansal, An intelligent multi agent based approach for autonomous energy management in a microgrid, Electr. Power Compon.
Syst. 49 (1–2) (2021) 18–31.
[18] S.A. Mansouri, A. Rezaee Jordehi, M. Marzband, M. Tostado-Véliz, F. Jurado, J.A. Aguado, An IoT-enabled hierarchical decentralized framework for
multi-energy microgrids market management in the presence of smart prosumers using a deep learning-based forecaster, Appl. Energy 333 (2023) 120560.
[19] A.S. Daramola, S.E. Ahmadi, M. Marzband, A. Ikpehai, A cost-effective and ecological stochastic optimization for integration of distributed energy resources
in energy networks considering vehicle-to-grid and combined heat and power technologies, J. Energy Storage 57 (2023) 106203.
[20] R.K. Chauhan, K. Chauhan, A.Q. Badar, Optimization of electrical energy waste in house using smart appliances management system-A case study, J.
Build. Eng. 46 (2022) 103595.
[21] A. Dargahi, K. Sanjani, M. Nazari-Heris, B. Mohammadi-Ivatloo, S. Tohidi, M. Marzband, Scheduling of air conditioning and thermal energy storage systems
considering demand response programs, Sustainability 12 (18) (2020) 7311.
[22] S. Khemakhem, M. Rekik, L. Krichen, A collaborative energy management among plug-in electric vehicle, smart homes and neighbors’ interaction for
residential power load profile smoothing, J. Build. Eng. 27 (2020) 100976.
[23] R. Ghorani, M. Fotuhi-Firuzabad, M. Moeini-Aghtaie, Optimal bidding strategy of transactive agents in local energy markets, IEEE Trans. Smart Grid 10
(5) (2018) 5152–5162.
[24] K. Saberi-Beglar, K. Zare, H. Seyedi, M. Marzband, S. Nojavan, Risk-embedded scheduling of a CCHP integrated with electric vehicle parking lot in a
residential energy hub considering flexible thermal and electrical loads, Appl. Energy 329 (2023) 120265.
[25] S.E. Ahmadi, M. Marzband, A. Ikpehai, A. Abusorrah, Optimal stochastic scheduling of plug-in electric vehicles as mobile energy storage systems for
resilience enhancement of multi-agent multi-energy networked microgrids, J. Energy Storage 55 (2022) 105566.
[26] E. Namalomba, H. Feihu, H. Shi, Agent based simulation of centralized electricity transaction market using bi-level and Q-learning algorithm approach,
Int. J. Electr. Power Energy Syst. 134 (2022) 107415.
[27] H. Algarvio, F. Lopes, Agent-based retail competition and portfolio optimization in liberalized electricity markets: A study involving real-world consumers,
Int. J. Electr. Power Energy Syst. 137 (2022) 107687.
[28] F. Ahmad, I. Ashraf, A. Iqbal, M. Marzband, I. Khan, A novel AI approach for optimal deployment of EV fast charging station and reliability analysis with
solar based DGs in distribution network, Energy Rep. 8 (2022) 11646–11660.
[29] K. Imran, J. Zhang, A. Pal, A. Khattak, K. Ullah, S.M. Baig, Bilateral negotiations for electricity market by adaptive agent-tracking strategy, Electr. Power
Syst. Res. 186 (2020) 106390.
[30] Y. Liu, Y. Wang, Y. Li, H.B. Gooi, H. Xin, Multi-agent based optimal scheduling and trading for multi-microgrids integrated with urban transportation
networks, IEEE Trans. Power Syst. 36 (3) (2020) 2197–2210.
[31] X. Wang, X. Mao, H. Khodaei, A multi-objective home energy management system based on internet of things and optimization algorithms, J. Build. Eng.
33 (2021) 101603.
[32] M.B. Sanjareh, M.H. Nazari, G.B. Gharehpetian, S.H. Hosseinian, A novel approach for sizing thermal and electrical energy storage systems for energy
management of islanded residential microgrid, Energy Build. 238 (2021) 110850.
[33] C. Tsay, A. Kumar, J. Flores-Cerrillo, M. Baldea, Optimal demand response scheduling of an industrial air separation unit using data-driven dynamic
models, Comput. Chem. Eng. 126 (2019) 22–34.
[34] P.A. Schrodt, Artificial intelligence and international relations: An overview, Artif. Intell. Int. Politics (2019) 9–31.
[35] M. Dastani, L. Van Der Torre, A classification of cognitive agents, in: Proceedings of the Twenty-Fourth Annual Conference of the Cognitive Science Society,
Routledge, 2019, pp. 256–261.
[36] S. Nolfi, Power and the limits of reactive agents, Neurocomputing 42 (1–4) (2002) 119–145.
[37] K. Tazi, F.M. Abbou, F. Abdi, Multi-agent system for microgrids: design, optimization and performance, Artif. Intell. Rev. 53 (2) (2020) 1233–1292.
[38] M. Selseleh Jonban, A. Akbarimajd, J. Javidan, Intelligent fault tolerant energy management system with layered architecture for a photovoltaic power
plant, J. Sol. Energy Eng. 137 (1) (2015).
[39] B.M. Radhakrishnan, D. Srinivasan, A multi-agent based distributed energy management scheme for smart grid applications, Energy 103 (2016) 192–204.
[40] A. Kantamneni, L.E. Brown, G. Parker, W.W. Weaver, Survey of multi-agent systems for microgrid control, Eng. Appl. Artif. Intell. 45 (2015) 192–203.
[41] K.P. Kumar, B. Saravanan, Day ahead scheduling of generation and storage in a microgrid considering demand side management, J. Energy Storage 21
(2019) 78–86.
[42] J. Khazaei, D.H. Nguyen, Multi-agent consensus design for heterogeneous energy storage devices with droop control in smart grids, IEEE Trans. Smart
Grid 10 (2) (2017) 1395–1404.
[43] J.M. Raya-Armenta, N. Bazmohammadi, J.G. Avina-Cervantes, D. Saez, J.C. Vasquez, J.M. Guerrero, Energy management system optimization in islanded
microgrids: An overview and future trends, Renew. Sustain. Energy Rev. 149 (2021) 111327.
[44] H. Boterenbrood, CANopen High-Level Protocol for CAN-Bus, Vol. 20, Nikhef, Amsterdam, 2000.
[45] M. Esro, A.A. Basari, S. Kumar, M. Sadhiqin, Z. Syariff, Controller area network (can) application in security system, World Acad. Sci. Eng. Technol. 35
(2009).
[46] Y. Zhu, F. Zhuo, L. Xiong, Communication platform for energy management system in a master-slave control structure microgrid, in: Proceedings of the
7th International Power Electronics and Motion Control Conference, Vol. 1, IEEE, 2012, pp. 141–145.
[47] M. Marzband, H. Alavi, S.S. Ghazimirsaeid, H. Uppal, T. Fernando, Optimal energy management system based on stochastic approach for a home microgrid
with integrated responsive load demand and energy storage, Sustainable Cities Soc. 28 (2017) 256–264.
