Multi Agent Systems and Their Applications-new Good

of 11
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  Full Terms & Conditions of access and use can be found at  Journal of International Council on Electrical Engineering ISSN: (Print) 2234-8972 (Online) Journal homepage: Multi-agent systems and their applications  Jing Xie & Chen-Ching Liu To cite this article:  Jing Xie & Chen-Ching Liu (2017) Multi-agent systems and their applications, Journal of International Council on Electrical Engineering, 7:1, 188-197, DOI:10.1080/22348972.2017.1348890 To link to this article: © 2017 The Author(s). Published by InformaUK Limited, trading as Taylor & FrancisGroupPublished online: 14 Jul 2017.Submit your article to this journal Article views: 6975View Crossmark data  JOURNAL OF INTERNATIONAL COUNCIL ON ELECTRICAL ENGINEERING, 2017VOL. 7, NO. 1, 188󲀓197 Multi-agent systems and their applications Jing Xie a  and Chen-Ching Liu a,b§ a School of Electrical Engineering and Computer Science, Washington State University, Pullman, WA, USA; b School of Mechanical and Materials Engineering, University College Dublin, Dublin, Ireland ABSTRACT  The number of distributed energy components and devices continues to increase globally. As a result, distributed control schemes are desirable for managing and utilizing these devices, together with the large amount of data. In recent years, agent-based technology becomes a powerful tool for engineering applications. As a computational paradigm, multi-agent systems (MASs) provide a good solution for distributed control. In this paper, MASs and applications are discussed. A state-of-the-art literature survey is conducted on the system architecture, consensus algorithm, and multi-agent platform, framework, and simulator. In addition, a distributed under-frequency load shedding scheme is proposed using the MAS. Simulation results for a case study are presented. The future of MASs is discussed in the conclusion. 1. Introduction 1.1. Motivation of distributed control  he energy system is evolving rapidly. Indeed, rom smart buildings to smart grids, digital technologies are producing numerous data streams that oer in-depth inormation about the system. For example, the num- ber o distributed intelligent electronic devices (IEDs) and distributed energy resources (DERs) continue to increase globally. he total number o installed pha-sor measurement units (PMUs) has increased to more than 2000 in North America [1]. In addition, many more remote control switches have been installed in distribution systems to enhance the resilience o dis- tribution eeders. Furthermore, development o renew- able energy contributes to the increase in DERs. By 2017, the total capacity o utility-scale solar panels operating in the U.S. is expected to be over 20 GW [2]. Decentralized and distributed control schemes are needed or managing and utilizing these widely distributed devices.wo reasons behind the increasing number o devices are the growing investment driven by policies and lower hardware prices. Moore’s law has been valid or a long period o time. Hardware is becoming cheaper and bet-ter in perormance. Tis provides the oundation or distributed control. Considering the increasing scale o the number o devices in a system, distributed methods are needed to shif the computational burden to local controllers.At the same time, integrating renewable energy gen-eration, energy storage, and electric vehicles brings new challenges [3,4]. A critical problem comes rom the decen- tralized ownership o energy system components. For example, in Pullman’s distribution system (Washington, U.S.A.), a 1 MW flow battery is installed by Avista and a 72 kW solar panel is established on the campus o Washington State University (WSU). While the electric energy trading and regulatory environment is evolving, decentralized ownership o energy system components becomes a problem or centralized control methods with participation rom a large number o consumers. Another difficulty arises rom the random nature o these compo-nents. For instance, the energy demand, arrival time, and departure time o electric vehicles (EVs) are random [5,6]. Centralized methods are complex and impractical or modeling and managing these stochastically behaved EVs. 1.2. Why multi-agent systems?  Modeling and computation tasks are becoming much more complex as the size continues to increase. As a result, it is laborious and difficult to handle using centralized methods. Although motivations to apply multi-agent KEYWORDS Multi-agent system (MAS); intelligent agent; multi-agent platform; distributed control; system architecture; consensus algorithm; smart grid; under-frequency load shedding ARTICLE HISTORY Received 14 January 2017 Accepted 26 June 2017 © 2017 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (, which permits unrestricted use, distribution, and reproduction in any medium, provided the srcinal work is properly cited. CONTACT Jing Xie § Dr. Chen-Ching Liu is a Visiting Professor at the School of Mechanical and Materials Engineering, University College Dublin, Dublin, Ireland   OPEN ACCESS  JOURNAL OF INTERNATIONAL COUNCIL ON ELECTRICAL ENGINEERING 189 research on the system architecture, consensus algorithm, and multi-agent platorm, ramework, and simulator. A MAS-based UFLS scheme is discussed in Section 4 with a study case. Te uture o MASs is discussed in the conclusion. 2. System design – system architecture, agent type, and consensus algorithm System design is critical as it covers many aspects in the development o MASs, e.g. agent models, coordination, data collection, and interaction among agents. In this sec-tion, a literature survey o system architecture, agent type, and consensus algorithm is presented.  2.1. System architecture and agent type Tere are abstract and concrete architectures [8]. Components and the basic engine structure are defined in the abstract architecture, which needs to be as generic as possible. Starting rom an abstract architecture, the con-crete architecture is developed by assigning a type to each component and implementing each macro instruction o the engine.Intelligent agents are classified into several types with respect to their unctionalities and decision-making mechanisms. (1) Purely reactive agents make decisions using only the present inormation without reerring to historical data. Tus, they utilize direct mapping rom sit-uation to action and respond to the environment directly. For example, in [21], reactive agent techniques are uti-lized to build up car-ollowing models based on artificial neural networks. (2) Logic-based agents make decisions through logical deduction. In [22], a method is developed or handling multiple hypotheses and perorming log- ic-based ault diagnosis. Scenarios rom the Italian power system are used or evaluation. (3) Belie-desire-intention (BDI) agents are built using symbolic representations o the intentions, belies, and desires o agents. In [23], the stages o autonomy determination or sofware agents are discussed. Te recognition o potential autonomy is provided utilizing the BDI paradigm. (4) Layered archi-tectures incorporate several sofware layers. Each layer combines agents that deal with different abstract levels o the environment. For example, in [14,15], a SPID sys- tem is designed in the layered architecture or integrating  various types o protection systems and deense schemes in different levels.  2.2. Consensus algorithm Te cooperation o agents is achieved through inormation interaction to reach a consensus. Te perormance and systems (MASs) or researchers rom various disciplines are different, as indicated in [7], the major advantages o using multi-agent technologies include: (1) individuals take into account the application-specific nature and envi- ronment; (2) local interactions between individuals can be modeled and investigated; and (3) difficulties in mod-eling and computation are organized as sublayers and/or components. Tereore, MASs provide a good solution to distributed control as a computational paradigm. In addi-tion, artificial intelligence (AI) techniques can be utilized. In [8], an agent is defined as a computer system that is situated in an environment that is capable o autonomous actions in this environment to meet its design objectives. In [9], a MAS is defined as a system that comprises two or more agents, which cooperate with each other while achieving local goals. 1.3. Applications of MASs MASs have been applied to various problems, including market simulation, monitoring, system diagnosis, and remedial actions [10–12]. In [13], a substation physical security monitoring (SPSM) system is proposed within the ramework o strategic power inrastructure deense (SPID) [14,15]. It monitors rom remote the physical security o power substations. In [16], an approach is developed to prevent interconnected power systems rom catastrophic ailures. It uses a MAS-based deense sys-tem that allows agents to have adaptive decision criteria. In [17], an integrative and flexible method is proposed that uses agent-based modeling or assessment o market designs. Agents are acilitated by Q-learning. Compared with the competitive benchmark, they can exploit market flaws to make higher profits. For remedial actions, a dis-tributed under-requency load shedding (UFLS) scheme is developed in [18] using the MAS. It is improved and being implemented as a demonstration or the RIAPS platorm [19]. Details o the UFLS scheme are presented in Section 4.Progress in AI, hardware, and sensor technologies have been achieved by the MAS community, resulting in agent technologies that are applied successully to real-world industrial problems. A project [20] was supported by U.S. Department o Energy (DOE) to transer the VOLRON M   sofware platorm to ranormative Wave. Details o VOLRON M  developed by Pacific Northwest National Lab (PNNL) are discussed in Section 3.2. In addition, it provides ranormative Wave with technical support to develop products and services that improve the operating efficiency o buildings and resilience o power grids. Tis project indicates that industry is adopting agent technologies. Te remaining o this paper is organized as ollows. A literature survey is presented in Sections 2 and 3, including  190 J. XIE AND C.󰀭C. LIU unctionalities highly rely on the communication layer, especially the connection topology and associated proto-cols. Te consensus and interactive consistency problems are two important aspects o the Byzantine agreement problem [24]. Te terms o the above three problems are used interchangeably in the literature, although they have different ormal definitions. In this study, the consensus problem is addressed. Te well-known and widely used Paxos and average-consensus algorithms are described in this subsection. In addition, the ault tolerance is discussed.  2.2.1. Paxos algorithm Te Paxos algorithm was srcinally proposed by L. Lamport in 1988 [25]. In 2001, it was published in [26] with a user-riendly presentation. As one o the simplest distributed algorithms, the Paxos algorithm is designed to implement a ault-tolerant distributed system. Te ‘synod’ consensus algorithm [25] serves as the heart o the Paxos algorithm. Each Paxos agent can be a proposer, accep-tor, or learner. I most nodes are available, this algorithm guarantees that agents will converge to one value. It has been used by many sofware products, e.g. (1) the Bigable which is adopted by many products o Google; and (2) the search engine, Bing, o Microsof.  2.2.2. Average-consensus algorithm Te average-consensus algorithm has been applied to many fields or the consensus and cooperation o net-worked MASs. Its convergence and convergent speed are reported in [27] and [28]. As an adaptive distributed algo- rithm, it requires only communication among neighbor-ing agents. Tereore, the communication burden is low. All available agents will reach an agreement, which is equal to the average o their initial values. Te simple mathe-matic model is summarized. A graph, = (  ,  ) , repre-sents the network o n  agents. Te set o agents is denoted by V   = { 1,2, … , n } . A nonnegative n  × n  adjacency matrix B  = [ b ij ]  specifies the interconnection topology o network N  . I an active communication link exists rom agent i  to agent  j , b ij  is a positive value. Otherwise, b ij  = 0. (1) ̇  y  i ( t  ) =  n j = 1 b ij   y   j ( t  ) −  y  i ( t  )  (2) Y  ( t  ) =  col   y  1 ( t  ) ,  y  2 ( t  ) , … ,  y  n ( t  )  (3) Δ= diag  n j = 1 b 1  j , … ,  n j = 1 b ij , … ,  n j = 1 b nj  (4) L  = − B +Δ (5) ̇ Y  ( t  ) = (−Δ+ B ) Y  ( t  ) = − LY  ( t  ) where Δ and L  are the degree and Laplacian matrices, respectively.  2.2.3. Fault tolerance Fault tolerance is ocused on both continuous availability and the elimination o recovery time. As an important perormance metric o ault tolerance, downtime reers to periods o time in which a system is not operational [29]. In order to minimize the downtime, specialized sofware routines are needed to detect ailures o hardware (e.g. sensors, actuators, storage devices, and communication channels) and switch to backup devices automatically. Tis type o sel-checking logic should be provided by the operating system (OS) and governed by the sofware platorm designed or distribution applications. In addi-tion, the capability to remove, disconnect, and repair the problematic devices without disruption to the computer system is critical. Furthermore, important data sources should be checked and calibrated periodically, e.g. PMUs. With respect to the requirements on cyber security, firewalls and access control should be provided by the OS to prevent unauthorized users and access. Encryption is needed to secure the communication among nodes. Te basic mechanisms o encryption should be provided by the OS. Te security level can be enhanced urther with digital certificates i it is supported by the platorm. Once the cyber intrusions are detected and traced promptly and accurately by the OS and platorm, mitigation actions are applied to block unauthorized access. In most applica- tions, it is assumed that non-Byzantine ault is considered. Tereore, no cheating agent(s) is (are) considered in the multi-agent based applications. Both ault tolerance pol-icies and mechanisms become simpler.In general, the OS and platorm should support the recovery o nodes due to sofware or other deects (e.g. power ailure). Te recovery includes two parts: (1) auto- matically restarting the node; and (2) recovering the node to a well-known state. It is expected that multiple states will be recorded and available or recovery. As a result, applications will be able to achieve ault tolerance and decide which state should be selected or recovery. In addition, sel-checking should be perormed immediately once a node restarts, including hardware, sofware, and interaces. I unctionalities o the restored node cannot be guaranteed, the node may be required by applications to remain silent instead o participating in the coordina-tion. Tereore, sel-checking is critical and its reliability matters.Fault tolerance becomes more complicated i commu- nication channels are unreliable. Te outage o communi-cation lines and noisy measurements should be taken into account. Tereore, it is harder to reach a consensus i the underlying communication channels are unreliable. For example, the balanced digraph may become unbalanced
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!