Multiagent Resource Planning System for Utility Provision

Sergei Kozhevnikov

, Petr Skobelev

and Miroslav Svítek

Czech Institute of Informatics, Robotics and Cybernetics Jugoslávských partyzánů 1580/3, Prague 6, Czech Republic

Samara State Technical University, Molodogvardeyskaya street 244, 443100, Samara, Russian Federation

Czech Technical University in Prague, Faculty of Transportation Sciences, Konviktska 20, Prague 1, Czech Republic

Keywords: Multiagent Planning, Smart Grid, Utility Provision, Real-time, Ontology.

Abstract: The intensive economic and social development of smart cities faces the constraints of stable utility provision.

This paper presents a multiagent client-centric smart grid management system for integrated gas, heat and

electricity networks. The system is based on the new approach of agent’s negotiation implementing the

strategies of straight and reverse recursion planning. It can be used as part of the Smart grid and Micro grid

concepts to reduce the price for consumers and decrease the negative impact of peak loads for the suppliers.

This approach corresponds to the fundamental principles of modern complexity theory, which uses the

fundamental principles of self-organization and evolution inherent in the natural world.

1 INTRODUCTION

The intensive economic and social development of

smart cities faces the constraints of stable utility

provision. The constantly increasing demand for

resources reveals several limitations related to the

high cost of production, ecology, the level of energy

losses in the network, and outdated organization

methods. One more limitation is the infrastructure of

a residential area as a conglomerate of complex

technical systems embedded in dense urban or

industrial development landscapes. Every city has its

characteristics and is essentially unique. Some are

characterized by cogeneration of electricity and heat,

while others use combined district heating schemes

with private boilers.

These factors show great potential for adapting

the smart grid (SG) methodology for electricity, heat,

and gas supply systems.

This paper presents a multiagent client-centric

smart grid management system for integrated gas,

heat and electricity networks.

The article's scientific novelty lies behind the

agents' newly presented logic, the negotiation

protocol, where users are part of the decision-making

process, and the software approach of agents and

https://orcid.org/0000-0002-2606-282X

https://orcid.org/0000-0003-2199-9557

https://orcid.org/0000-0003-3971-6850

ontologies combination. The practical value is the

implementation of the complex solution based on the

SG concept that can optimize supply-demand model

of all three types of utilities (gas, heat, electricity) in

one software taking into account of specific

infrastructure constraints. Experimental results have

shown that system implementation can reduce the

price for the end-user.

2 STATE OF THE ART

The term "Smart grid" is still quite vague and for this

research the smart grid is defined through basic

features, among them (Report Australian energy

operator, 2020):

 optimal resources load to increase efficiency,

reduce losses and devices failure;

 active role of consumers in the network;

 the ability to resist new, unplanned events;

 synchronous operation of all network elements.

Multiagent Systems (MAS) now are SG enabling

technology because they can possess all mentioned

features, they can give a distributed software

framework and bring intelligence to the grid (Brena,

Handlin, Angulo, 2015). Agents are widely used as

Kozhevnikov, S., Skobelev, P. and Svítek, M.

Multiagent Resource Planning System for Utility Provision.

DOI: 10.5220/0010990600003116

In Proceedings of the 14th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2022) - Volume 1, pages 433-441

ISBN: 978-989-758-547-0; ISSN: 2184-433X

433

the instrument of negotiating on the low sensor level.

In (Nguyen and Flueck, 2015), the idea of an agent-

based distributed power flow solver for unbalanced

radial distribution systems based on MAS is

presented. Since the 2000's researchers have tried to

use intelligent agents because of three main key

features: autonomy, local view, and decentralization.

However, most methods use agents to model lines,

switching devices, voltage regulators, transformers,

distributed energy storage systems, and batteries (for

example to solve the backward/forward sweep

technique problem to solve the power flow

iteratively) (Boudaoud, Labiod, Boutaba, Guessoum,

2000).

MAS is widely used in Smart Homes (Li,

Logenthiran, Woo, 2015) to optimize the energy

consumption on a local house/flat level. Agents help

to plan the optimal solution of energy consumption,

but they do not provide a real-time response to all

unpredictable events. They are used to achieve high

comfort level, energy efficiency, and energy price

through negotiation between devices.

In (Omarov, Altayeva, 2018) the methodology of

MAS is used in intelligent control systems that covers

all the monitored zones of a building and, if

necessary, provides the greatest possible overall

comfort in the building while reducing the required

electric power.

A more comprehensive simulation approach that

accounts for the MAS-related protocols as described in

the FIPA specification is presented in (Le, Bui, Ngo,

2018). It shows promising results for system evaluation

under various settings and design trade-offs.

Agents communication is performed to transfer

neighbouring information between subsystems. It is

shown the possibility of parallel work in different

parts of the distributed network (Shum, 2106).

Agent approach is fundamentally based on

negotiation to find the optimal (or close to the optimal

solution). For this purpose, Combinatorial Auctions

can also be used. The proposed combinatorial

auctions algorithms showed an advantage over a more

rigid scheduling algorithm (Brena, Handlin, Angulo,

2015). This approach can predict and plan but not

manage the system in real-time.

One of the main advantages of MAS is the ability

to work in decentralized systems for electricity

provision (Svítek, Skobelev, Kozhevnikov 2020). It

Is Used Extensively in Research Projects (Morte,

2016) To Develop Distributed Control Systems

Comprised of a Network of Communicating Units.

The Task To Be Solved Is The Issue of Complexity

That Scales Up Exponentially, Limiting The

Development of Smart Grid Technologies.

Decentralization of The Network With The Help

of Multiagent Systems for Electric Vehicles

Infrastructure Is Described in (Jordán, Palanca, Del

Val, Julian, Botti, 2018). Agents Collect, Evaluate

and Manage Data from Elements To Create an

Optimal Cooperation Algorithm.

in (Loni, Parand, 2017) The Game Theory for The

Smart Grid Topic Is Implemented. The Game Theory

Models The Behaviour of Independent and Rational

Agents To Maximize The Profit. Authors Survey

Several Game Theory-based Applications, Incredibly

Cooperative Game Theory To Solve Relevant

Problems in Micro Grids.

a Tremendous Analysis of The MAS Application

in SG Is Done in (Mahela Et Al., 2020).

Comprehensive Overview of Multi-Agent Systems

for Controlling Smart Grids. CSEE Journal of Power

and Energy Systems.). Completed The Review of

General Concepts of Smart Grids and MAS,

Technologies and Standards, Intelligent Agents in SG

and Commercial Projects The Authors See The

Future of MAS in Coordinated Control Replacing

SCADA Systems.

based on The Topic Domain Review, We Can

Define The Trend for MAS Coordinated Control of

The SG and Lack Decision-Making Solutions That

Can Be Applied for Limited Tasks With Severe

Constraints. on The Other Hand, It Is Clear That The

Multiagent Approach Can Be Successfully Used for

Distributed Problem Solving in Decentralized SG

Architectures.

in Our Project, We Extend The SG Concept from

Electricity To All Types of Resources (Electricity,

Gas, Heat). for Citizens It Is a Part of Utility

Provision of The Smart City Concept in General

(Přibyl, Horák, 2015).

The State-of-The-Art Review Highlights The

Main Advantages of MAS. Multiagent Resource

Planning System for Utility Provision Can Plan and

Optimize Utility Provision as a Multi-Criteria Task,

Moving from Determining The Optimum Under

given Constraints (Reliability, Quality,

Environmental Friendliness) To Finding The Optimal

Level Ratio of These Properties.

3 MULTIAGENT SYSTEM FOR

UTILITY PROVISION

3.1 System Functions

Base on the state of the art and market analysis the

consumer-friendly MAS with the following functions

was developed. Main system functions:

ICAART 2022 - 14th International Conference on Agents and Artiﬁcial Intelligence

434

 development of a model and plan of resource

production and consumption processes taking

into account the current situation at the network

objects;

 real time optimization of production/

consumption plan of resources;

 simulation of production/consumption process

for three types of resources in any predefined

scale;

 coordination of decision-making process with

end-users through two-way interaction;

 conducting simultaneous analysis of multiple

planning options with appropriate allocation of

resources for optimization purposes.

3.2 Types and Characteristics of

System Objects

The main objects in the system are: suppliers

(produce one or several types of resources),

consumers (consume one or several types of

resources), network elements (transmit the resource

from supplier to consumer).

3.2.1 Suppliers

Focusing on statistical and forecast demand data

suppliers make plans for resource production.

Produce resources of one or several types:

 Gas. The supplier represents one of the state

gas supply system branches - a gas distribution

station. In the model context, the gas supplier's

production capacity is equal to its throughput

capacity.

 Heat. Can be presented as city level suppliers

(thermal power station) acting through the

public heat supply network, or local suppliers.

The thermal power station can work as

cogeneration plant and can operate in two

modes: thermal - heat load is the priority,

electric - electric load is the priority.

 Electricity. Electricity can be supplied by

cogeneration plants, hydroelectric power

plants, or local generators.

3.2.2 Consumers

Consume resources requested from suppliers, and

delivered through networks corresponding to the

three types of resources. The sum of all requests

forms a general graph of resource consumption,

affecting the volume of the resource production.

Consumers can be divided to industrial (factories,

plants, etc.), offices, and household users. The main

difference between them in volumes of resource

consumption, range of planning horizon, and

discreteness of consumption plans.

The supplier of resources can also be a consumer.

3.2.3 Storages

Can be both consumer and supplier. Storages can

accumulate a certain amount of resources - gas or

electricity. Storages are characterized by the capacity

and the ability to return the accumulated resource.

3.2.4 Transformers

Resource transformers can change the types of the

transmitted resource (for example, change gas to

electricity).

3.3 Types and Characteristics of

Network Channels

There are three types of channels provided by the

resource-supply network - gas, electric and heat.

3.3.1 Gas Supply Channel

Presented in the system as trunk and city pipelines.

The trunk pipeline delivers the gas to distribution

stations where the pressure is lowered to the required

level.

3.3.2 Heat Supply Channel

Delivers heat, in this project (as usually in real cities),

this channel is not reconfigurable part of the

infrastructure network.

3.3.3 Electricity Supply Channel

The structure (way of distribution) of the network can

be changed to optimize the electrical model of the

network.

3.4 System Ontology

Integrated resource supply network pursues the goal

of transporting gas, electric, and heat resources from

producers to consumers. Visually such a network can

be represented as a connected graph (Figure 1). The

paths in this case can be called channels and divided

into three types (gas, electricity, heat channels).

Objects are divided into four types - supplier,

consumer, storage, transformer. For every element we

can also define subspecies, based on the relationship

of these objects to one or another type of resources.

Multiagent Resource Planning System for Utility Provision

435

In the developed system, the description of the

main elements was done in the network ontology

editor. It allows creating the predefined ontology

classes of objects, to assemble the network, start

planning, and simulate the process of resource

distribution in real-time.

Figure 1: Ontology of a problem domain.

Network ontology editor allows the creation of

the predefined and new classes of main and auxiliary

objects (substations, solar panels) to assemble the

network. New object classes can be included in the

network by extending the ontology and adding classes

of appropriate agents.

3.5 Description of Planning Algorithms

3.5.1 MAS Approach

In this project MAS implies matching of every object

with its own software agent (SA). SA description

includes the characteristics of the simulated object

that are stored and can be changed in the ontologically

descriptive knowledge base (network ontology

editor). Dynamically generated solutions are

achieved in the process of communication of agents

that is carried out by transmitting the predefined types

of messages.

The optimization problem of the model is solved

based on the market approach. It can be described as:

 expression of all optimization parameters is

developed through universal cost

characteristics (internal virtual currency);

 the rules of agents' behaviour aimed at

increasing the local profit of every agent by

reducing the costs in the virtual currency,

which leads to dynamic improvement of the

network parameters;

 automatically finding the optimal dynamic

equilibrium of the multiagent system by

exchanging virtual currency;

 applying internal agent penalty and bonus

functions to control the dynamics of the

network automatically;

 use of scalar convolution on parameters of

optimization of the network agents' state;

 use a component-by-component variation of

optimization parameters of the multiagent

system in order to align the indicators.

The proposed multiagent technology based on the

market approach provides universal and dynamic

optimization for software-configurable networks.

3.5.2 Type of Agents

In the developed system, the planning and resource

allocation task is solved by agent's negotiation and

successive approximations from essential, simple,

fast solutions to a more complex and efficient one

according to the available time. Software agents

optimize their target parameters, identify conflicts,

negotiate, make concessions and achieve a solution in

the interests of the whole system.

The main types of agents in the developed system

are:

 Order agent (represents the consumers and

requests for specific resource. Every consumer

can generate various number of orders for

different resources. Every request creates a

separate agent).

 Scene agent (system agent. Responsible for

agent’s management process).

 Resource Agent (represents the suppliers and

amount of resources of various type can be

provided. Every supplier creates one agent for

every type of available resource).

3.5.3 Initial Parameters and General

Planning Description

Every consumer has its own resource consumption

plan, visually represented (Fig. 2) as a two-

dimensional graph with the axes of time and volume

(GСal, m3 and kW), respectively, with discretization

by hours/days. Users set consumption plans with a

monthly horizon based on the average rate typical for

this area.

Users set up, edit, and approve the plans privately.

The system summarizes requests, creating three

aggregated plans for each type of resource in general.

The system then identifies peaks in these

aggregate schedules and starts the negotiation

process. It looks for the consumers who have invested

the most significant contribution to the plan and then

starts to trade to decrease request and ensure the

sufficiency and stability of the whole plan.

ICAART 2022 - 14th International Conference on Agents and Artiﬁcial Intelligence

436

Figure 2: Consumption plan.

After negotiating, the system approves a resource

consumption plan that satisfies all users. There is now

a penalty if consumers use fewer/more resources than

requested and planned as a preventive measure

implemented to oblige consumers to comply with the

established plan.

There are three main phases of planning in the

system: initial flat planning, proactive orders and

resources phases.

3.5.4 The Initial Flat Planning Phase

The initial planning phase is shown on Fig.3.

Figure 3: Flat planning diagram.

The flat planning phases runs as follows:

a) The Order agent receives a scheduling request

message from the Scene agent.

b) The Order agent determines the list of potentially

suitable resource suppliers for planning this

order.

c) A request to the Scene is generated to obtain a

list of resource suppliers capable of performing

the order. Capability is based on the type of

energy resource, network channels, and links

between suppliers and consumers.

d) A list of resource suppliers suitable for the order

is returned.

e) The Order agent subscribes through the message

service to receive information from the Resource

agents about the price needed to perform the

request.

f) Through the agent repository, the Order agent

receives the list of Resource agents (received in

step C) and forms a list of potential further

request messages for the price of placement of

the request.

g) Through the message service, the Order agent

sends to every Supply agent a request message

for the price of order placement for a certain time

interval. The message specifies the size of the

request, the time interval from t1 (now) to t2

(planning horizon). On this step only open and

available slots are taken into account.

h) The Order agent gets responses from the

Resource agents to the request-message about the

price and time. The result is a set of all possible

placements on suitable resources. The answer

message can provide several variants of one

order placement because it can fit several

variants in the planning interval due to the

different prices at all possible t1-t2 time

intervals.

i) The Order agent sorts the possible variants and

keeps those where the price is less than the initial

amount of virtual money left.

j) On every received interval a set of possible

variants of placement is determined to calculate

the value of the target function. Two values sort

the obtained list: decreasing of the target function

and the price of resource placement.

k) The list of sorted options for order placement is

used by the Order agent. It sends the planning

request message and tries to schedule the order

on the resource - first in the list. The message

contains the order ID, the desired time, the price

in virtual currency that the order can pay for the

placement.

Multiagent Resource Planning System for Utility Provision

437

l) Upon the receiving of a message of successful

order placement, an amount equal to the

difference of the initial amount of money and the

cost of scheduling for the selected resource is

credited to the virtual currency account.

m) When the order is successfully scheduled, the

Resource agent credits the amount received from

the system to his account.

n) In case of the order is not scheduled the next

possible resource from the list receives a request

for order placement. The planning can be

unsuccessful because of the increase in the

resource rate at the requested interval, or there is

another order on this time interval, or the starting

time of the order left.

o) In case of the order does not find the appropriate

resource (from the list) and allocation place, it

stays unplanned.

3.5.5 Phase of Proactive Analysis and

Redistribution of Orders

The system has the option of proactive agent’s

behaviour and order analysis (agents proactively look

for better options to place the order after the

completed first round of planning). In this case, the

negotiation process runs as follows:

a) The Scene agent selects all the orders that are

planned but not yet executed at the moment.

b) For each selected order, the current value of its

target function is calculated.

c) The list of selected orders is sorted by the

ascending value of the target function.

d) The scene agent sends request messages for the

moving price for a certain number of the most

dissatisfied with its placement orders (at the

beginning of the list).

e) The Order agent that receives the message

requests a Scene agent the list of resource

suppliers capable of performing the required list

of orders and sorts it according to their

desirability.

f) For each resource from the received list

(including the current resource supplier where

the order is currently planned), a sample of

possible placement points is made, taking into

account the order size, performance, minimum

time step (ξ) and planning horizon (T). In Fig. 4,

sample points are indicated with round dots.

Figure 4: Sample point for order placement.

g) For all selected sample points, the Resource

agent calculates and sends the Scene agent the

price for placing the order.

h) In case this time interval is already occupied with

scheduled orders, the system selects the orders

that interfere.

i) The Scene agent sends the request to the

preselected orders for the possibility to move for

the specified time left or right (Fig. 5).

Figure 5: Scheduled orders (green) interfere with the new

order (red).

j) The preselected Order agents receive the shift

request and start the process of negotiation with

the Resource agent.

k) The preselected Order agents send the request to

the corresponding Resource agent to find out the

price of placing the order on another time

interval.

l) If moving to the new time interval influences

other scheduled orders, their agents also request

to move. This request contains the pre-calculated

value and direction of the shift. Recursive

descent continues until the next order to be

shifted can move to the new position unhindered,

or up to the counter limiting the recursion depth

(set in the settings).

m) In the reverse recursion, the affected agents send

response messages containing information about

all required moves and its total cost (Fig. 6).

n) As a result, each placement point has its cost. The

cost is calculated as the price for the resource and

the price for moving the affected orders.

Figure 6: Straight and reverse recursion.

o) From the whole set of possible placement points,

the system excludes those with the price

exceeding the order's available virtual money.

Next, the system excludes allocation options that

do not improve the target function's value if

approved.

p) The remaining set of allocation points is sorted in

descending order of their efficiency, expressed in

ICAART 2022 - 14th International Conference on Agents and Artiﬁcial Intelligence

438

improving the value of the system's target

function. In case of equality, the points with less

value are preferred.

q) In the response message (for the Scene agent),

the Request agent specifies the full cost and the

array of required shifts necessary to move the

request to the sorted set's first (best) point.

r) Changes are approved and performed if the total

cost specified in the message is less than the sum

of the virtual money of the Order agent and the

virtual money returned by resources for

unscheduling the order.

s) Prior to reallocation, the system checks the

relevance of the schedule and it is subsequently

blocked. If changes are detected in the schedule,

the reallocation is not performed.

t) As the next step the system tries to remove all

orders of the schedule affected by the

reallocation. If unsuccessful, the reallocation is

not performed, and the schedule returns to its

original state.

Elements removed on the previous step are moved

to new positions.

3.5.6 Phase of Proactive Resource Analysis

and Load Redistribution

The system has the option of proactive resource

analysis (agents proactively look for better resource

load options after the completed first round of

planning). In this case the negotiation process runs as

follow:

a) The order whose de-scheduling leads to the

highest efficiency of the system is selected.

b) The order is removed from the schedule (in this

case the schedule of proactive resource is

blocked from editing by other processes):

 The virtual money paid for placement in the

schedule is returned.

 The data about the old placement on the

resource's schedule is deleted.

 The Order agent send the new request for

planning.

 If the process is unsuccessful at any stage - the

schedule and orders are returned to the initial

state (before the reallocation attempt).

Process of proactive analysis and redistribution of

orders and resources is shown on Fig.7.

Figure 7: Proactive agents diagram.

3.5.7 Penalty Function

The penalty function (Fig. 8) is used to vary through

the special admin. user interface the agent’s objective

function. It makes possible to increase or decrease the

level of agent’s mobility.

Depending on the value of the current KPI, the

components of the agent's objective function have

different characteristics with a maximum at the

optimum value.

On the interval (0, 0.2), when resource proactivity

is activated, step 6 is executed (Fig. 7), on the interval

(0.2, 0.6) step 8 is executed, and on the interval (0.6,

1) step 7 is executed.

Figure 8: The penalty function.

The penalty function can be adjusted to the

specific task or certain logic of the model. In the

developed system and performed test cases the

penalty function is connected with the maximum load

of the suppliers. As the KPI equal the load, the

Multiagent Resource Planning System for Utility Provision

439

minimum penalty equal to the optimal modes of

equipment.

3.6 Developing Scenarios and

Performing Tests

3.6.1 Initial Problem Statement for Test

Scenario Development

For the test scenario the small village had been

chosen. The network and objects configuration of the

village as a set of suppliers and consumers of

resources is presented in Figure 9 (five suppliers of

three different types of resources).

Figure 9: Configuration of the test network.

The parameters of the suppliers are presented in

Table 1. Every supplier can change and decrease the

amount of produced resources.

Table 1: Parameters of the suppliers.

Supplie

Resource type Max. production

Hydroelectric

ower

lant

Electricity 150/225 KwH

Gas supplie

Gas 60000 m3/Month

Thermal

ower

lant

Electricity/Heat 170 KwH

Thermal

ower plant

Electricity/Heat

170 KwH /1800

m3/month

Wind power

station

Electricity 30 KwH

It is assumed that every object has access to the

dispatching service to change the rate of consumption

of resources. All objects have sensors to count the

current level of consumption of resources.

The simulation starts after formalizing the

network of consumers and suppliers of resources,

defining their relationships and parameters. During

the modelling process it is possible to enter additional

events (new additional demand) that influence and

refine consumption forecasts. Other types of events

(new consumer or supplier, issues with the network,

revise of the price) may also change the general

forecasts and the prices and plan of the delivery can

be dynamically recalculated.

In this context, the developed system can show

how intelligent planning and the forecast availability

can reduce the production costs for suppliers and

reduce the price for the users.

3.6.2 Performing Tests

To prove the concept, the test scenarios were

conducted. In the modelled scene a village with 15

private houses with consumption rates is slightly

deviating from the average normative values. The

degree and vector of deviation were set by the system

for each object individually in an automatic random

way, taking into account the given constraints.

Additionally, to the basic scenario, a gas-electric

power generator can be used by private homes. The

factory, with predefined schedule of consumption,

with a given slight fluctuation. The factory has its

own source of energy as a set of solar panels.

Fluctuation in the volume of electricity consumption

by the factory depends on the solar panels capacity.

Thermal power plant loaded to 30% of resource

capacity The load on the turbines of the hydroelectric

power plant is 80%.

All presented consumers are connected to all

types of resources. All suppliers are consumers of all

resources except those that they produce themselves.

Figure 9 shows a simplified diagram of the network

topology designed for the experiment.

Having the initial data, the system builds the

consumption plan of all three resources defining the

peak hours of electricity. As the next step the system

in the proactive phase rebuild a plan reducing the

peak hour’s consumption.

After the basic scenario, new external conditions

of increased demand for electricity from the factory

were implemented. This case leads to the additional

production of electricity by the thermal power plant,

what leads to the additional production of heat for the

lower price for the user in public heat supply network.

The results of the planning in Figure 10 show the

changes in the price of the utilities for users before

and after implementation of the system.

It is shown, that adaptive rescheduling in the

network of objects connected in one smart grid and

multiagent resource planning system for utility

provision in cooperation with solar panels installation

can reduce the price of the electricity more than 2

times.

ICAART 2022 - 14th International Conference on Agents and Artiﬁcial Intelligence

440

Figure 10: Results of the experiment.

4 CONCLUSIONS

Multiagent resource planning system for utility

provision is designed to: improve the efficiency of

utility model by applying real-time planning. It helps

to optimize the production-consumption model, make

short-term and long-term planning for optimal

operation modes and effectively balance the market

in local micro grids and more complex smart grids.

The multiagent system implements a variety of

methods for the distributed solution of complex

problems.

The solution is sought based on the model of

interaction of software agents, which negotiate in

achieving the goal and seek consensus, for which

conflicts are sought and resolved by mutual

concessions, to achieve the most rational solution to

the common problem.

The system optimizes the processes of production

and consumption of three types of resources in one

model and present the schema for gas, heat, and

electricity supply systems integrity. It shows an

individual approach for every order and resource

through taking into account the various strategies and

criteria, preferences and constraints of all players.

ACKNOWLEDGEMENTS

This work was supported by the European Regional

Development Fund under the project AI&Reasoning

(reg. no. CZ.02.1.01/0.0/0.0/15_003/0000466).

REFERENCES

Report Australian energy operator (2020). Electricity

Statement of Opportunities – https://aemo.com.au/

R. F. Brena, C. W. Handlin, P. Angulo (2015). A smart grid

electricity market with multiagents, smart appliances

and combinatorial auctions. IEEE First International

Smart Cities Conference (ISC2).

C. P. Nguyen and A. J. Flueck (2015). A Novel Agent-

Based Distributed Power Flow Solver for Smart Grids.

IEEE Transactions on Smart Grid, vol. 6.

K. Boudaoud, H. Labiod, R. Boutaba, Z. Guessoum (2000).

Network security management with intelligent agents.

NOMS IEEE/IFIP Network Operations and

Management Symposium.

W. Li, T. Logenthiran, W. L. Woo (2015). Intelligent multi-

agent system for smart home energy management.

IEEE Innovative Smart Grid Technologies - Asia

(ISGT ASIA).

B. Omarov, A. Altayeva (2018). Towards Intelligent IoT

Smart City platform Based on OneM2M Guideline:

Smart Grid Case Study. IEEE International Conference

on Big Data and Smart Computing (BigComp).

Le D.P., Bui D.M., Ngo C.C. (2018). FLISR Approach for

Smart Distribution Networks Using E-Terra Software.

A Case Study. Energies.

C. Shum (2106). Modelling and simulating

communications of Multiagent Systems in Smart Grid.

IEEE International Conference on Smart Grid

Communications (SmartGridComm).

Svítek M., Skobelev P., Kozhevnikov S. (2020) Smart City

5.0 as an Urban Ecosystem of Smart Services. Service

Oriented, Holonic and Multi-agent Manufacturing

Systems for Industry of the Future. SOHOMA 2019.

Studies in Computational Intelligence, vol 853.

Springer, Cham

M. Morte (2016). E-mobility and multiagent systems in

smart grid. 17th International Scientific Conference on

Electric Power Engineering (EPE).

Jordán, J., Palanca, J., Del Val, E., Julian, V., Botti, V.

(2018). A Multi-Agent System for the Dynamic

Emplacement of Electric Vehicle Charging Stations.

A. Loni, F. Parand (2017). A survey of game theory

approach in smart grid with emphasis on cooperative

games. IEEE International Conference on Smart Grid

and Smart Cities (ICSGSC).

O. P. Mahela et al. (2020). Comprehensive overview of

multi-agent systems for controlling smart grids. CSEE

Journal of Power and Energy Systems.

O. Přibyl, T. Horák (2015). Individual Perception of Smart

City Strategies. Smart Cities Symposium Prague

(SCSP).

Multiagent Resource Planning System for Utility Provision

441