Available online at www.sciencedirect.

com
ScienceDirect

Energy Reports 8 (2022) 163–174

www.elsevier.com/locate/egyr

7th International Conference on Green Energy Technologies, ICGET 2022 July 28–30, 2022,
Frankfurt, Germany

A multi-agent based protection in distribution networks including

distributed generations
E. Abbaspoura , B. Fania ,∗, E. Heydarian-Forushanib , A. Al-Sumaitic
a
Department of Electrical Engineering, Najafabad Branch, Islamic Azad University, Najafabad 85141-43131, Iran
b Department of Electrical and Computer Engineering, Qom University of Technology, Qom 1519-37195, Iran
c Advanced Power and Energy Center, Department of Electrical Engineering and Computer Science, Khalifa University, Abu Dhabi,

United Arab Emirates

Received 18 October 2022; accepted 22 October 2022
Available online xxxx

Abstract
This paper proposes a strategy for protection coordination in distribution networks with DGs penetration. The main objective
of the present work is to employ the potential of Multi-Agent Systems (MASs), which have recently been given considerable
attention in the context of smart grids. A modified scheme is proposed to minimize the number of agents while preserving the
necessities of protection coordination under different contingencies. A new structure is defined for relay agent (RA) to operate
in centralized or decentralized modes. Each RA contains communication and protection units, to detect an appropriate mode
according to the required time to clear the fault and maintain coordination with other agents. IEC-61850 standard including
GOOSE and SMV protocols used to send necessary commands and share data. The proposed protection scheme has 8 stages,
which removes the fault from the network with the help of services and functions defined on the proposed structure of the
relay agent. These steps include gathering network information, sending commands, and updating relay settings. In this regard,
impacts of DGs are analyzed to show how they would change the operation of protective relays and their corresponding
zones. Results of this research ensure the proper performance of protective devices taking into account appropriate settings and
configurations.
© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
Peer-review under responsibility of the scientific committee of the 2022 7th International Conference on Green Energy Technologies.

Keywords: DG penetration; Protection coordination; Multi-agent systems; Distribution systems

1. Introduction
Distributed Generations (DGs) are known as one of the most important strategies in the electricity market, since
they provide us a more reliable operation from the standpoints of network, power quality, environmental issues, etc.
However, the existence of DG sources in distribution systems is somewhat challenging due mainly to change on the
∗ Corresponding author.
E-mail address: [email protected] (B. Fani).

https://doi.org/10.1016/j.egyr.2022.10.394
2352-4847/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.
org/licenses/by/4.0/).
Peer-review under responsibility of the scientific committee of the 2022 7th International Conference on Green Energy Technologies.
E. Abbaspour, B. Fani, E. Heydarian-Forushani et al. Energy Reports 8 (2022) 163–174

level of short circuits and direction of current flowing into the network lines [1,2]. For example, in case of islanding,
distribution network could become isolated from the bulk power network resulting in severe variations on levels of
stable voltages and frequency that may subsequently lead to damage equipment and subsystems. Besides that, the
lack of co-ordination between protective devices, incapability of faults’ detection, malfunction in the performance
of protection units and more issues can be attributed to the operation of DGs in many cases [3,4].
In the literature, there are several schemes to improve the protective functionalities in distribution networks
including DGs. Unfortunately, the majority of them are impractical because of high costs, complexity of operation
and infeasibility to develop in practice. Accordingly, putting forward applicable ideas to address these problems
has been of great value and with the advent of intelligent electronic devices. It is somehow possible to provide
appreciable solutions to cope with above issues.
Table 1 reviews these methods and their relevant challenges. Reducing the impact of DGs based on their operation
scheduling or limiting fault current, changing the network protection system, using adaptive methods as well as
voltage-based protections are some of the solutions to solve the protection system problems. The various benefits
of exploiting DGs have led researchers to find new solutions to reduce obstacles associated with increasing the
penetration of these resources in distribution networks. One of the main obstacles in distribution networks with high
penetration of DGs is disrupting the protection scheme. Researchers have so far proposed several solutions to reduce
the protection problems of DG resources. According to Table 1, only solutions based on the telecommunication
methods can adapt their system to the network conditions and ensure the correct operation of the protection system
at any time.

Table 1. An investigation on the proposed approaches.

Reference Method Description Drawbacks
[5–7] Planning on DG -Limiting DG Current -Affect the reliability of the distribution network
-Optimizing Place & Capacity -Restrictions may cause inconvenient for their owners.
-Disconnecting
[8,9] Fault current -Using fault current limiters -FCL Impedance measurements are difficult to find
limiting -High investment cost to empower it withstanding high
levels of fault current from distribution networks.
[10–14] Changing -Distance Relay -Capital cost is high with double cost of relays, CT, and
Protection System -Differential Relay CB required.
-Directional Relay -Relay setting method is harder than traditional OC relay.
-Pilot Protections -Operative to protect lines but not capable of protecting
-Extra Breakers buses in connection with DGs/loads.
[15–17] Adaptive -Multi Group Settings -Consuming large storage space for computational memory
Protection System -Optimizing Group Setting and is costly.
-Centralized Protection System -Requiring new calculations by changing the network.
-The large amounts of data should be available for an
optimal solution.
-Existing the central unit can reduce reliability.
[18,19] WAMS and -Measuring current and voltage -Of further obligations for real-time transfer of wide-area
WAPC parameters data
[20] PMU and Agents Measurement and -The calculations can be more complex and very costly
communication unit correlation
[21] Hierarchical Gathering and exchanging -Of complicated coordination models for a large system
mode of MAS information from and to the grid with many agents

The presence of various renewable-based DGs as well as information and communication systems make the
structure of the electricity network more complex. Due to different operational functions, the requirements for the
communication system are different. The IEC 61850 protocol can achieve the goal of a single standard through
using the object-oriented modeling technology and the prospect of communication-oriented extensible architecture.
This standard is the foundation for smart substations, where the data can be shared among various equipment of
the smart substations will be under the IEC 61850 standard [22].
The use of intelligent electronic equipment was one of the other solutions that could be considered in the
implementation of protection scheme. With the beginning of using this equipment, another concept so-called agent
164
E. Abbaspour, B. Fani, E. Heydarian-Forushani et al. Energy Reports 8 (2022) 163–174

was introduced. The agent is an intelligent entity that could make decision individually or under human guidance.
With the development of the agent in the network and combining it with phasor measurement units, new methods
have been introduced with the aim of improving the security of protection schemes against cyber-attacks as well as
isolating faulty areas of the network [22].
Along with the development of communication structures and the use of intelligent agents, as well as the ability
of the IEC 61850 standard to send and receive network raw data, a Multi-Agent System (MAS) has been introduced.
These communication structures can be implemented in both peer mode and hierarchical mode depending on the
requirement.
The authors in [21] proposed a hierarchical multi-agent structure in which different network agents collaborate
together to modify the performance of the protection system by rotating network information in three telecommu-
nication layers. An important aspect that should be considered in using the MAS is the structure complexity due
to the increase in network agents. This could be a challenging issue particularly in hierarchical communication
structure. However, the use of peer mode structure could be widespread if a more acceptable protection scheme is
provided in comparison with other methods.
In this paper, a new multi-agent system is defined, which operates in two centralized and decentralized modes.
Relays are the only agents in this system working together to clear the fault at the right time. In comparison with
the hierarchical multi-agent systems operating only in centralized mode, relay agents consist of two communication
and protection units so that with the help of defined services and functions, they can calculate the required operation
time and switch between two centralized and decentralized operational modes if needed. Protection scheme has the
following features:
• Reducing data density on the network due to reducing the number of agents,
• Reducing dependence on the central controller at the time of fault,
• Updating the protection settings of the relays through the local database,
• Reliable performance for high current faults and improving operation time of protections.
The rest of the paper is organized as follow. Section 2 introduces the protection structure of the MAS. The modified
structure is proposed in Section 3. The numerical analyses on the modified IEEE 33-bus test system are carried out
in Section 4. Finally, Section 5 concludes the paper.

2. Protection scheme based on Multi Agent Systems (MAS)

The proposed algorithm uses the principles of MAS, and its performance is based on connection and data
exchange associated to DGs, protective devices and distribution network configuration using a communication
path. The controlling structure of MAS includes three connecting levels, which is shown in Fig. 1. The first
level of connection refers to distributed agents such as relay–agent, distributed generation–agent, load–agent, and
breaker–agent and so on.
To manage the connection among the agents belonging to the first level, a head is considered corresponding to
each group of agents including Relay–Agent-Group (RAG), Distributed Generation–Agent Group (DGAG), Load–
Agent Group (LAG) and Breaker–Agent-Group (BAG). In other words, these heads create the second level of MAS.
The third level as central processing unit performs three main following duties:
• Gathering all instantaneous data associated to the network via agents existing at the second level.
• Detecting any events in the network by means of gathered data.
• Sharing the information throughout the network on the basis of detected state among all heads and organiz-
ing/monitoring the cooperation among agents to attain a suitable setting and performance.
In an MAS-based scheme, all agents at different levels can fulfill expected functions as well as gathering–exchanging
data, deciding and running commands issued by the central system [21]. For instance, when a fault happens at some
location of the network, the relay agents first send the network data to the central system by means of RAG. The
central system will then accumulate data from different sources as well as other heads and will subsequently issue
necessary information (and commands) to the RAG for clearing fault as well as preserving the suitable time margin
between the main and backup relays as fast as possible. By doing so, the RAG transfers information to downstream
objective relays.
165
E. Abbaspour, B. Fani, E. Heydarian-Forushani et al. Energy Reports 8 (2022) 163–174

Fig. 1. Hierarchical multi-agent structure.

IEC-61850 as a dedicated protocol to distribution systems is employed for making connection among agents and
delivering information throughout the network. This protocol is based on Generic Object-Oriented Substation Events
(GOOSE) considering each network device as an element. In addition, the GOOSE protocol supports a high-speed
peer-to-peer communication, helping MAS-based protection schemes to be more practical. The GOOSE protocol is
used for receiving and sending communication signals in the MAS structure. One of the most important capabilities
of such a technology is to establish a direct connection between two devices so that one can communicate to the
other for data exchange in 4 ms and vice versa [23]. Blocking or informing signals used in the MAS structures
propagate through this protocol. Indeed, agents use this protocol to alert or inform other agents in the network.
Another important protocol in this standard is SMV (Sampled Measured Value) that can be used for transferring
voltage and current data. This protocol helps MAS to share the sensitive parameters of network in microseconds.
As a result, all data gathered or shared by agents is related to SMV structure in IEC-61850 standard.
While using a three-level control along with grouping in the middle level is a solution for maintaining the
coordinated operation among devices when a fault is present, such a situation is problematic when the network size
is large and the number of agents increases. If so, the number of states describing power system stability would be
highly increased. In more detail, it is necessary to define all possible states associated with each agent as a pattern
for being processed so that the system becomes enabled to make decision and to exchange correct information to
each objective agent.
Clearly, there would a possibility of malfunction or disturbance when sending and receiving information among
agents in the presence of concurrent changes at the same time. The need for a physical communication infrastructure
making suitable connections among different levels with an appropriate bandwidth for delivering volumes data is
a big pitfall of this scheme. Besides, the time required for communicating throughout the network can complicate
the matter under discussion.

3. Proposed MAS structure

To solve the previously discussed problems in Section 2, the amount of data should be reduced. To achieve
this goal, the control levels of MAS must be modified so that no disturbance should be occurred in exchanging
necessary data and performing protective devices setting. In fact, by reconstructing the MAS and omitting the need
for decision making at the central process unit, this is not further required for employing the second level agents
(heads). Meanwhile, the number of agents at the first level can be reduced to decrease the size of information. To
design such a construction, we should first find which agents at the first level can perform the functions of other
agents. The most efficient and sensitive devices of the network should be selected for this purpose.
Although power circuit breakers may have proper performance for determining network configuration as
well as the connectivity of loads and sources, they cannot determine the statuses of voltage, current and
frequency. Therefore, power circuit breakers are not appropriate candidates for improving the protective system
and co-ordination.
Loads as zones connected to the distribution network can observe the current flowing into their own branch
(s). If there is no load in a section of distribution network, data associated with such a node is unavailable. More
166
E. Abbaspour, B. Fani, E. Heydarian-Forushani et al. Energy Reports 8 (2022) 163–174

precisely, if DG is connected to or disconnected from a network’s node, agents related to the loads cannot detect
any event at an objective branch due to lack of observe-ability. The same can be told regarding any change in the
current of a node under network variations. Hence, loads are not appropriate candidates for consideration as agents
of control layer.
The selection of DGs as the agents for coordinating protective devices can produce accurate information about
connection status and amount of power injection to the network. Thus, at the first glance, DGs might be suitable
options for the use as main agents in the modified MAS. However, DGs can only detect and dispatch the information
about their own statues and the branches they are connected to. Therefore, DGs fail to provide sufficient information
required for the proper performance of protection systems.
Obviously, any change made by agents (in the modified MAS) will be propagated to the agent heads and Central
Processing Unit (CPU). As a result, it will lead to change the voltage, current, amount and direction of active and
reactive power injections either directly or indirectly. Due to the fact that the amount of passing current through
a protective device is a criterion for operation of protection system, relays can be selected as the main agents of
modified MAS so that all necessary information regarding other agents can be gathered by relays and therefore, there
is no need for data processing in CPU. Fig. 2 depicts the structure of modified RA structure in the MAS scheme.

Fig. 2. The modified structure of relay agent.

4. Proposed protection scheme

Fig. 2 illustrates the needed circle for a relay agent for clearing the fault. Given that each unit has its own relevant
duties, eight stages are introduced to isolate the network fault. In the first stage, the network information is shared
until an agent can imagine the network map. Then, it is necessary to select or obtain an appropriate configuration
file during a process. To issue certain commands, before entering the protection unit, network agents have to receive
final commands and information from other agents. In the protection unit, the current passing through a relay along
with its operation zone are recognized. According to this information, the operation time is calculated and the trip
command is issued in case of a need. In the following, we will examine the details of both communication and
protection levels.

4.1. Communication unit

In brief, two main objectives are expected from the communication unit:
• Sharing distribution system information in the shortest possible time;
• Managing operations and agents’ settings.
These objectives are more discussed in the following.
167
E. Abbaspour, B. Fani, E. Heydarian-Forushani et al. Energy Reports 8 (2022) 163–174

4.1.1. Sharing data

Each agent shares various data on the network through this module through four services including Current
Data, Health Data, Breaker Condition, and Zone Information. The current passing through each relay, which is
an important criterion for network status detection, is shared by the agents. Due to the fact that the number of
agents has been reduced in the proposed scheme, the correct operation of each agent is very important for network
protection. For this purpose, the health status of each agent is also shared in the network. The breaker status of each
agent is another important piece of information that helps neighboring agents determine the status of the network.
Due to the fact that each agent needs to have information about the zones connected to other agents, agents also
advertise their protected zones on the network.

4.1.2. Name table

Each agent in the network is notified of the presence or absence of neighbor agents. In this situation, each agent
always identifies its upstream and downstream agents. This helps keep other agents informed when a new agent is
added to the network.

4.1.3. Zone data

By receiving information of neighbor agents, each agent recognizes its own protection zone. By receiving the
information and completing the network communication map, the agents are able to identify the main and backup
protections of each zone when a fault does occur.

4.1.4. Database
The information regarding the received settings of each agent that has been done in different conditions is stored
in this section. This database can work completely independently if it could load the appropriate settings locally on
the relay.

4.1.5. Controller
This section contains four services to manage the operation of agents in different situations. Self-control is
responsible for issuing the command of breaker. Neighbor control is responsible for managing the performance of
upstream and downstream agents. This feature could create an interrupt in the operation of relays or force them
to act. The alert signal forces neighboring agents to check the database and suggests settings to them. If the agent
needs to receive protection settings from the central processing unit, it activates the setting signal service and sends
the desired signal over the network.

4.2. Protection unit

Relays settings and performances must be adapted to any changes occurring in the network as a result of adding
or removing DGs as well as utilization variations. Therefore, the proposed scheme should coordinate relays in an
efficient manner. To achieve this goal, all necessary data including measured quantities should be circulated among
agents. The protection unit of each agent is expected to gather the necessary data and the communication unit can
disseminate them. Accordingly, the coordination (among relays) is preserved with the help of communication sector.
Two following duties are assigned to the protection unit of each agent:

• Analyzing the input current.

• Choosing an appropriate performance mode corresponding to network utilization.

The protection unit includes five functions:

4.2.1. Agent name

Each protection equipment in the network is known by a unique name to others. This name is a string by which
the agents are distinguished from each other.
168
E. Abbaspour, B. Fani, E. Heydarian-Forushani et al. Energy Reports 8 (2022) 163–174

4.2.2. Input current

Obviously, the base on which a protective system works is the amount of passing current through it. Thus,
an accurate measurement on observed current is of essential importance. In this regard, each agent is expected
to analyze the amount of current via current transformer (CT) and to determine any changes occurred in the
network parameters. Once the changes are detected and analyzed, agents can adapt themselves and also share useful
information among other agents to become more resilient against new variations. Changes are also considered as
events to be furthered used by operators for detecting disturbances in the network. The related changes are all saved
as new events to be employed for further settings. An analysis on the currents feeding agents can reveal the pattern
of changes in the network parameters. By measuring the objective parameters, an adaptive mode of performance
can be reached to meet the needs of the network under these changes.

4.2.3. Operation zone

Each agent must operate accurately in its own protection zone. So, it is essential that this zone be identified.
This function shows how a relay can recognize its protection zone and not operate when a fault does occur outside
that zone.

4.2.4. Operation time

The operating time of a relay is specified in two ways. In the first case, the relay calculates the operating time
according to its settings and issues the interruption command. In this case, the settings are either read from the
database or loaded on the relay through the central unit. In the second case, a functional signal sent by neighbor
agents, and the agent acts accordingly. In this case, the agent may experience a delay in its performance so that
it has to wait 300 ms for the network protection operation. Also, it may experience an immediate performance so
that it will operate in less than 10 ms.

4.2.5. Active setting

The protection settings for the relay agent operation are specified in this section. These settings include the
characteristic curve and the I pickup current. Needless to say, this section changes if the relay settings need to be
updated.

4.3. MAS based protection scheme

The suggested protection scheme intends to solve the time problem in the hierarchical MAS structure according
to Fig. 3. A time margin is considered for the operation of the major and backup protections when the fault does
occur. This time margin depends on the considered standard and could be different in each network.

Fig. 3. The proposed protection scheme.

169
E. Abbaspour, B. Fani, E. Heydarian-Forushani et al. Energy Reports 8 (2022) 163–174

4.3.1. Decentralized method

Fig. 4 shows how this method works. At first, the upstream and downstream agents are identified in the network.
Each agent sends its own and neighborhood protection zones to the network so that other agents could form their
default network map. In the next step, the passing current through the agent is shared and the closest agent to the
fault location is determined according to it. Note that the other backup and main agents are recognized in accordance
with the network map. Afterward, the main agent searches for the appropriate protection setting in its database to
load on itself. If this setting does not exist, the main agent has the responsibility of locking the other agents to
prevent their malfunction. Then the self-control section issues the breaker trip command. If the fault until then
does exist in the network, the main agent sends the trip command for the backup protection through the neighbor
control with the aim of clearing the fault in an appropriate time. However, if there is a suitable setup group, the
main agent will send the alert signal to the backup agent so that this agent adjust itself in accordance with the main
agent. The core agent then would be calculating its operating time and sending the breaker the order of tripping.
If the backup agent has a proper configuration, it will wait for operation of the main agent. If the main agent is in
healthy condition and its breaker operates properly, the fault will be eliminated. Otherwise, the backup agent will
act 0.3 s after the operation of the main agent. If there is no proper setting on the backup relay, the backup agent
will completely remove the fault with almost 0.3 s interruption.

4.3.2. Centralized method

If there is enough time to circulate information in the MAS structure, the agents will wait to receive the required
commands from the central processing unit. Fig. 5 shows how this mode operates. According to Fig. 5, the agents
first send a setting signal to the central unit requesting a change in their settings. The central unit detects the network
topology based on the received information. Then, the central unit refers to its search table to select the appropriate
settings for coordination of the main and backup protections. If these settings exist, the settings are sent to the main
and backup protections, and the fault will be cleared. If no settings are available, it is necessary to calculate the
new processing information and new settings. At this stage, if the processing time exceeds the considered threshold
for operation of the protections, the central unit sends the trip command to the main relays.

Fig. 4. The decentralized protection scheme.

5. Simulation
The modified IEEE 33-bus test system has been selected as a typical distribution network as given in Fig. 6. The
simulation of protection system is performed in the context of ETAP software. The modified network consists of
two 5 MW synchronous generators. The network topology could be changed through two installed breakers, i.e. P1
and P2. The relay agents RA1 to RA8 have been considered to protect the network, while the time margin of the
main and backup protections is set to 300 to 800 ms. The coordination margin of these protections can be seen for
the fault current in their protection zones in Fig. 7(a) and (b). In Fig. 7(a), RA1 and RA2 are coordinated with each
other for a fault current of 1024 A. Also, on this line, RA2 is the backup of RA3 for the fault current of 810 A.
170
E. Abbaspour, B. Fani, E. Heydarian-Forushani et al. Energy Reports 8 (2022) 163–174

Fig. 5. The centralized protection scheme.

Fig. 6. The modified IEEE 33-bus test system.

Also, RA3 and RA4 are coordinated with each other with a time margin of 616 ms. Fig. 7(b) shows the operation
time of feeder’s protections. The RA5 and RA6 operate for tripping fault currents of 1800 and 2140 A. The RA7
is the backup of RA8 for the fault current of 910 A.

5.1. Scenario No. 1: Fault F1 in the presence of DG1

According to the fault zone, the closest protection is the RA3 that has the responsibility to attain a cleared fault
in an appropriate time. In this case, two other agents including RA1 and RA2 also sense the fault current. Due to
the presence of DG1 and hence, the level increment of the short circuit current, the network protection coordination
is disturbed and the performance priority of the agents is changed. In this situation, the coordination margin among
RA3 and RA2 for the fault current 1110 A will be less than 300 ms. Due to the fact that there is not enough time
to update the protection settings, the centralized operation mode is selected, and RA3 is recognized as the main
agent. In this case, the protection zone of other agents is determined. Table 2 shows the database status for RA2
and RA3. Due to the fact that there are no suitable settings for this amount of current, it is necessary to send the
lock signal to other agents. In this case, the RA2 waits for the acting RA3 and operates in 495 ms if RA3 does
not operate properly due to its malfunction or malfunction of its breaker. Fig. 8(a) shows the performance status of
these two agents.
171
E. Abbaspour, B. Fani, E. Heydarian-Forushani et al. Energy Reports 8 (2022) 163–174

Fig. 7. (a) Operation status of RA1 to RA4; (b) Operation status of RA5 to RA8.

Fig. 8. (a) Operation of RA2 and RA3 in Scenario No. 1; (b) Operation of RA1 and RA2 in Scenario No. 2.

5.2. Scenario No. 2: Fault F2 in the presence of DG2 while P1 and P2 connected

In this case, RA7 and RA8 must act according to the fault location. Depending on the network status, if RA8
does not act, RA3 should operate, and on the other hand, if RA7 does not act, RA1 and RA2 should act. The RA8
and RA3 agents have an appropriate time margin and send the setting signal to receive new settings. If there is no
information for the coordination of these two agents, it is necessary for the central unit to calculate the new settings.
If there are some delays in the calculation time, the central unit sends the trip command to RA8. Otherwise, new
settings are sent for both agents and the main and backup agents are specified. Table 3 shows the required settings
needed to coordinate these two agents for 560 A in the central unit settings table. On the other hand, RA7 with
3150 A is another main agent for fault clearing. In this situation, the coordination margin of the main agent with
RA1 and RA2 agents is disturbed by passing 2500 A, 1100 A and there is no time to update the protection settings.
In this case, there are appropriate settings on the RA7 database to be coordinated with RA1 according to Table 4,
and as a result, an alert signal is sent to RA1 to change the settings. Due to the lack of proper protection settings
for RA2, this agent is locked. According to Fig. 8(b), if RA7 agent does not act, the RA1 and RA2 agents act at
the desired times, respectively.
172
E. Abbaspour, B. Fani, E. Heydarian-Forushani et al. Energy Reports 8 (2022) 163–174

Table 2. The database of RA2 and RA3 agents.

Agent name Local database
Fault current (A) Fault current (A) Fault current (A) Fault current (A)
IRA3 = 500 IRA3 = 600 IRA3 = 1600 IRA3 = 2000
IRA2 = 400 IRA2 = 500 IRA2 = 1600 IRA2 = 1800
RA2 TMS:0.12 TMS:0.07 TMS:0.65 TMS:0.69
Ip:200 Ip:320 Ip:200 Ip:200
RA3 TMS:0.075 TMS:0.075 TMS:0.05 TMS:0.05
Ip:190 Ip:284 Ip:190 Ip:150

Table 3. The search table of central processing unit for RA3 and RA8.
Agent name Central controller
Fault current (A) Fault current (A) Fault current (A) Fault current (A)
IRA8 = 560 IRA8 = 1000 IRA8 = 1200 IRA8 = 1800
IRA3 = 560 IRA3 = 1000 IRA3 = 1000 ,IRA3 = 1600
RA3 TMS:0.52 TMS:0.2 TMS:0.43 TMS:1.4
Ip:100 Ip:320 Ip:200 Ip:150
RA8 TMS:0.05 TMS:0.07 TMS:0.06 TMS:0.05
Ip:80 Ip:284 Ip:200 Ip:100

Table 4. The database of RA1 and RA7 agents.

Agent name Local database
Fault current (A) Fault current (A) Fault current (A) Fault current (A)
IRA7 = 2500 IRA7 = 1000 IRA7 = 3150 IRA7 = 1800
IRA1 = 2000 IRA1 = 600 IRA1 = 2500 IRA1 = 1200
RA1 TMS:0.24 TMS:0.08 TMS:0.24 TMS:0.08
Ip:600 Ip:300 Ip:750 Ip:600
RA7 TMS:0.07 TMS:0.1 TMS:0.05 TMS:0.05
Ip:500 Ip:200 Ip:800 Ip:400

6. Conclusion

The MAS structure is one of the well-known methods for solving the upcoming protection challenges as a result
of emerging DG resources in the distribution networks. However, the employment of hierarchical MAS cannot
always ensure protection coordination at the right time. On this basis, this paper proposed a modified MAS structure
so that in addition to solve the hierarchical structure drawbacks, it was able to solve the protection coordination in
both the centralized and decentralized manners by considering the relay agents and changing their structure. Relays
are the only agents of this plan that can share their data and make independent decisions for their operation by
changing the structure and defining specific services and functions. These characteristics provide the possibility of
sending and receiving control signals as well as sharing network current information by GOOSE and SMV protocols,
respectively. Finally, based on the time required to prevent relays from malfunctioning, the proposed solution can
remove the fault in centralized mode with the help of the communication and protection units, and if there is not
enough time, it can be clear in decentralized mode. The definition of new functions and services on relay agents
can provide conditions to detect changes in the network. Detecting current direction on feeders independently of
other additional equipment and also providing a new relationship among agents to detect backup and main agents
based on fault location feeding from different points of network can be considered as future work with the aim of
completing this paper. Finally, the simulation results and numerical analysis have showed the effectiveness of the
proposed method.
173
E. Abbaspour, B. Fani, E. Heydarian-Forushani et al. Energy Reports 8 (2022) 163–174

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
The authors are unable or have chosen not to specify which data has been used.

References
[1] Holguin JP, Rodriguez DC, Ramos G. Reverse power flow (rpf) detection and impact on protection coordination of distribution systems.
IEEE Trans Ind Appl 2020;56(3):2393–401.
[2] Shen S, et al. An adaptive protection scheme for distribution systems with DGs based on optimized thevenin equivalent parameters
estimation. IEEE Trans Power Deliv 2017;32(1):411–9.
[3] Yan B, Luh PB, Warner G, Zhang P. Operation and design optimization of microgrids with renewables. IEEE Trans Autom Sci Eng
2017;14(2):573–85.
[4] Mahat P, Chen Z, Bak-Jensen B, Bak CL. A Simple adaptive overcurrent protection of distribution systems with distributed generation.
IEEE Trans Smart Grid 2011;2(3):428–37.
[5] Akram U, Khalid M, Shafiq S. Optimal sizing of a wind/solar/battery hybrid grid-connected microgrid system. IET Renew Power
Gener 2018;12(1):72–80.
[6] Hamidi ME, Chabanloo RM. Optimal allocation of distributed generation with optimal sizing of fault current limiter to reduce the
impact on distribution networks using NSGA-II. IEEE Syst J 2019;13(2):1714–24.
[7] Varma RK, Rahman SA, Atodaria V, Mohan S, Vanderheide T. Technique for fast detection of short circuit current in PV distributed
generator. IEEE Power Energy Technol Syst J 2016;3(4):155–65.
[8] Lim S-H, Park M-K. Analysis on protection coordination of OCRs using voltage components for the application of SFCL in a power
distribution system with DG. IEEE Trans Appl Supercond 2021;31(5):1–6, Art no. 5602106.
[9] Naderi SB, Negnevitsky M, Jalilian A, Hagh M, Muttaqi K. Optimum resistive type fault current limiter: An efficient solution to
achieve maximum fault ride-through capability of fixed speed wind turbines during symmetrical and asymmetrical grid faults. IEEE
Trans Ind Appl 2017;53(1):538–48.
[10] El-Arroudi K, Joós G. Performance of interconnection protection based on distance relaying for wind power distributed generation.
IEEE Trans on Power Deliv 2018;33(2):620–9.
[11] Aghdam TS, Kazemi Karegar H, Zeineldin HH. Variable tripping time differential protection for microgrids considering DG stability.
IEEE Trans Smart Grid 2019;10(3):2407–15.
[12] Khurshaid T, Wadood A, Gholami Farkoush S, Yu J, Kim C, Rhee S. An improved optimal solution for the directional overcurrent
relays coordination using hybridized whale optimization algorithm in complex power systems. IEEE Access 2019;7:90418–35.
[13] Zhang Z, Xu B, Crossley P, Li L. Positive-sequence-faultcomponent-based blocking pilot protection for closed-loop distribution network
with underground cable. Int J Electr Power Energy Syst 2018;94:57–66.
[14] Funmilayo HB, Butler-Purry KL. An approach to mitigate the impact of distributed generation on the Overcurrent Protection scheme
for radial feeders. In: 2009 IEEE/PES power systems conference and exposition. 2009, p. 1–11.
[15] Yazdaninejadi A, Golshannavaz S, Nazarpour D, Teimourzadeh S, Aminifar F. Dual-setting directional overcurrent relays for protecting
automated distribution networks. IEEE Trans Ind Inf 2019;15(2):730–40.
[16] Sharaf HM, Zeineldin HH, El-Saadany E. Protection coordination for microgrids with grid-connected and islanded capabilities using
communication assisted dual setting directional overcurrent relays. IEEE Trans Smart Grid 2018;9(1):143–51.
[17] Ustun TS, Ozansoy C, Ustun A. Fault current coefficient and time delay assignment for microgrid protection system with central
protection unit. IEEE Trans Power Syst 2013;28(2):598–606.
[18] Eissa MM. Challenges and novel solution for wide-area protection due to renewable sources integration into smart grid: an extensive
review. IET Ren Pow Gen 2018;12(16):1843–53.
[19] Luo L, Tai N, Yang Guangliang. Wide-area protection research in the smart grid. Energy Procedia 2012;16(part C):1601–6.
[20] Habib HF, Youssef T, Cintuglu MH, Mohammed OA. Multi-agent-based technique for fault location, isolation, and service restoration.
IEEE Trans Ind Appl 2017;53(3):1841–51.
[21] Liu Z, Su C, Høidalen HK, Chen Z. A multiagent system-based protection and control scheme for distribution system with
distributed-generation integration. IEEE Trans Power Deliv 2017;32(1):536–45.
[22] Cintuglu MH, Ma T, Mohammed OA. Protection of autonomous microgrids using agent-based distributed communication. IEEE Trans
Power Deliv 2017;32(1):351–60.
[23] León H, Montez C, Stemmer M, Vasques F. Simulation models for IEC 61850 communication in electrical substations using GOOSE
and SMV time-critical messages. In: IEEE world conf. factory communication systems. WFCS, 2016.

174