Intelligent Agents in Transmission Network Protection

Transcription

1 SYNOPSIS OF Intelligent Agents in Transmission Network Protection A THESIS to be submitted by Anuradha.Charugalla for the award of the degree of Master of Science (by Research ) Department of Electrical Engineering Indian Institute of Technology Madras Chennai June 2008

2 1. Introduction Electrical Power System has undergone Rapid change in the past decade. Faster, more reliable and better coordinated protection is even more critical under new environment than it has been in the past. New methods are needed to meet this challenge. Traditional protection systems rely upon standalone units that use local measurements as the basis for most decision making. Communication plays very important role in these legacy systems. A new era of decentralized control and efficiency demands has led to an environment demanding efficiency and reliability that pushes these legacy methods to their limits. Distributed Artificial Intelligence (DAI) is a subfield of Artificial intelligence research dedicated to the development of distributed solutions for complex problems regarded as requiring intelligence. Presently DAI has been largely supplanted by the field of Multi-Agent Systems. Distributed Artificial Intelligence (DAI) systems can be defined as cooperative systems where a set of agents which are often heterogeneous, act together to solve a given problem. DAI systems are based on different technologies like, e.g., distributed expert systems, planning systems or blackboard systems [1]. 2. Literature Review In the available literature, Intelligent Agents has been proposed in Power Systems for four areas namely, SCADA [2], State Estimation [3], Restoration [4], and Protection [5-9]. For SCADA application Intelligent Electronic Devices (IED s) were replaced by Intelligent Agents (IA s), in order to reduce the bandwidth of communication and to react with environment. For State Estimation the total network has decomposed into relatively small number of areas and by coordinating and communicating agents of all areas, the time of operation for State Estimation has reduced. For Restoration architecture consists of several Bus agents and one Facilitator agent was proposed. By coordination among the neighbouring bus agents and facilitator agent, reasons for any outage can be find easily and restored the network at a faster rate. For Protection application, Intelligent Agents find solution for Adaptability of relay protection schemes, differential scheme for transmission lines and improving time delay for backup relays. Intelligent agents also find to be suitable for Negotiations in Market games, Scheduling, error detection and correction, Maintenance Scheduling, etc where decision making plays important role [10]. 1

3 3. Motivation for the Study Literature review on power system protection shows that there exists scope for the development of efficient solution technologies to improve the reliability of the protection scheme. Study on DAI (Distributed Artificial Intelligence) reveals that the DAI techniques can be applied for complex and dynamic environments like power systems. In the literature, few attempts have been made to implement DAI techniques in power systems. Yet there is no generalized architecture to implement DAI to power system in large scale. The motivation for the present research is due to the following. Electrical Power systems comprises of a network of complex systems interconnected by strong cause-effect relations and several variables associated with them. The complex interconnected nature of the power system problem can be addressed by means of distributed tasks or processes linked with other similar processes. These processes exchange information during the solution procedure. Coordination and Communication among the distributed tasks or processes (agents) during problem solving forms the basis for representation and modeling of complex networks. These agent-based approaches promise a new methodology to the effective solution of power systems. Transmission network protection is one such example where coordination and communication among processes (agents) plays a vital role in the efficient operation of a power system. Most of the large scale problems are efficiently handled by dividing them into smaller subproblems and solving them independently. Distributed Problem Solving using Decomposition and Coordination approach is the motivation for the solution of large scale problems. DAI and Intelligent Agents provide one such means to develop intelligent agent based architecture for distributed problem solving. The inspiration of above two points forms the basis of motivation for present research work on Intelligent Agents in Transmission Network Protection. 2

4 4. Objective and Scope of the Work The objectives of the present research work are summarized as follows: 1. Proposal of Intelligent Agent (IA) architecture for the Transmission Network Operations. 2. Detailed implementation of Agent Protection of transmission system considering Differential Protection for Transformers, Transmission Lines and Bus bars. 3. Implementation of the Coordination and Communication among the different agents for the proposed protection algorithm. The scope of the present research is as follows: Simulation of Multi-Agents is a wide area of research, encompassing design, representation, analysis, modeling and, simulation. The scope of the proposed work is limited to the following aspects The present study is limited to the feasibility of application and implementation of MAS in Linux Environment for establishing of the communication among agents. The present study is limited to the Development of Algorithm for Protection of Transmission Network using Intelligent Agents. 5. Description of the Research Work Distributed Artificial Intelligence (DAI) finds wide application in Restructured power system operations involving Autonomy and Decision-making. Fig. 1 gives the architecture of multi agent system for complex problems. Agent A Full operation which needs group of agents Communication Fig.1. Distributed Architecture of Multi Agent System 3

5 One of the first steps in the Restructuring process [11] of the power industry has been the separation of the Transmission activities from the electricity generation activities. The Transmission system having significant economics of scale consequently had a tendency to become a monopoly. The Transmission Companies (TRANSCO s) are those entities, which own and operate transmission wires. Their prime responsibility is to transport the electricity from generators to customers, and making available the transmission wires to all entities in the system [12]. The TRANSCO s can be classified as National or Regional depending on the Operating level of Voltage. Normal software s are acceptable where the environment is controlled and static and the user has to do a small number of tasks. With the environments becoming more complex and with data streaming in from hundreds of different sources at the same time, agents can keep track of and store data, monitor unusual activity and finally take action all without the need of human intervention. The Agents can have sub- agents which can divide tasks among themselves in order to have parallel operation as shown in Fig. 2. Master Agent AGENT 1 AGENT 2 SUB AGENT 1 SUB AGENT n SUB AGENT 1 SUB AGENT n Fig. 2. Agent Hierarchy 5.1 Transmission System Operations and Architecture The various operations related to TRANSCO [13] are broadly classified as, 1. Contingency Analysis (CA) 2. Monitoring and Data Acquisition (SCADA). 3. Network Topology Processing (NTP) 4. State Estimation (SE) 5. Congestion Management (CM) 6. Maintenance Scheduling (MS) 7. Transmission Expansion and Planning (TEP) 8. Load Flow Analysis (LFA) 9. Transmission Pricing (TPR) 4

6 10. Estimation of Data for Reliability Analysis (ESDRA) 11. Risk Assessment (RA) 12. Protection (PT) All these different tasks are grouped into five categories basing on the similarity of the work to form the architecture for transmission system. Five main agents of proposed architecture are 1. Measurement Agent (MA) [CA, SCADA, NTP, SE]. 2. Operation Agent (OA) [CM, MS, TEP] 3. Pricing Agent (PA) [LFA, TRP ] 4. Economic Agent (EA) [ESDRA, RA] 5. Protection Agent (PTA) SCADA PTA NTP State Estimation Relay Agent Contingency Analysis Load Flow Analysis Congestion Management Fault Identification MA Domain Common Identification Maintenance Scheduling Back-up agent Determination of Data N-1 Contingency Analysis Expansion and Planning Risk Assessment PA Cost Allocation OASIS ISO OA EA Fig. 3 Detailed Communication of the Proposed Architecture for TRANSCO Fig. 3 shows the detailed communication aspects of the proposed architecture for Transmission system. Communication between agents is shown herein Fig. 4. 5

7 ISO, GENCO PA OA MA EA PTA Fig. 4 Communication of Combined Agents 5.2 Present Protection System and Need for Intelligent Agents Protection plays very important role in the reliable operation of the integrated complex network. With the recent developments in advanced technology, digital and numerical protections are widely being used by the utility. SCADA forms the back bone for realizing the advances in power system protection. The task of protection of power network consists of an integration of various logical steps which can be practically realized by means of coordinating devices. This can be viewed as the distributed task performed by various agents involving cooperation and coordination amongst them. In the present scenario more protection problems are caused by relay maloperation and inappropriate circuit breaker operation. The causes of relay maloperation are [14]: 1. Inappropriate design or application of relay. 2. Inappropriate relay settings. 3. Human error. 4. Component maloperation. Intelligent Agents finds solution for above reasons as they can keep track of and store data, monitor unusual activity and finally take action all without the need of human intervention. 5.3 Proposed Protection Algorithm In the proposed Algorithm, each of the equipment, which is to be protected, is having one relay agent (RA) working on behalf of it. Apart from relay agent, each of the equipment is having one fault identifier and isolator (F I/I A) agent and one back-up agent (BUA). RA gets input data from corresponding measurement agents (MA) and based on the scheme of protection for that particular 6

8 equipment, performs simulations. It then sends the data to FI/IA to confirm the fault and to send the trip signal to corresponding circuit breakers. Back-up agent works for back-up protection. Corresponding CT RELAY AGENT (3) FAULT I/ I AGENT (4) Corresponding PT MEASUREMENT AGENTS (1) CORRESPONDING C.B s (2) BACK-UP Agent (5) Fig. 5 Block diagram of Proposed Algorithm Fig. 5 provides the block diagram of the proposed algorithm. The MA consisting the corresponding CT and PT of the device, sends the data to RA to perform simulation. The RA after performing the simulation sends the information to corresponding F I/I A. This Fault Identifier agent decides the corresponding fault based on the information it gets from RA, and sends information to both CBA and BUA. CBA trips corresponding C.B s. this is shown in fig. 5. C.B Equipment C.B LAYERS CT PT CT PT AGENTS LAYER-1 MA 1 MA 2 MA LAYER-2 CBA 1 Data Data CBA 2 CBA LAYER-3 Corresponding Relay Agent Simulation Data RA LAYER-4 Trip Corresponding Fault I/ I Agent Trip F I/I A LAYER-5 Report Corresponding Back-up Agent Report BUA Fig. 6 Details of working of Intelligent Agents for Protection 7

9 Fig. 6 shows the details of the working of Intelligent Agents for Differential Protection of Power Systems. The Proposed Architecture has five layers; each layer describes the operation of particular agent in that layer. Layer_1 Measurement Agents (MA): This Measurement Agents gets current data from Power System and sends that data to Relay Agent which is in Layer 3. Layer_2 Circuit Breaker Agents (CBA): CBA s get input from the Fault Isolator Agent and send the trip signal to corresponding Circuit Breakers in Power Systems. Layer_3 Relay Agent (RA): RA performs the simulation work like digitalizing the signal which it got from MA. It sends simulation data to Fault Identifier and Isolator Agent (FI/IA). Layer_4 Fault Identifier and Isolator Agent (FI/IA): This Agent decides whether the fault is there or not basing on the simulation data it obtained from RA. If it identifies the fault, it will send the trip signal to corresponding Circuit Breaker Agents and Back-up Agent. Layer_5 Back-Up Agent (BUA): This provides Back-Up protection when ever there is a fault in Circuit Breaker. This Agent gets information from F I/IA and CBA. If it does not get report from CBA, it sends the trip signal to another CBA which is near to it. In this way it is provides Back-Up Protection. 6. Implementation and Results Important implementation aspects of the proposed work is described in this section 6.1 Implementation Issues Working of the MAS for the transmission network protection is implemented on Linux workstations connected via 100 Mbps Ethernet. The simulation program is coded in C. Inter-processor communication was done through MPI (Message Passing Interface) [15]. If the system becomes more complex, we can use two or three systems using LAM/MPI (Local Area Multi-computer / Message Passing Interface) in which, each system had its own memory from which it accessed data (distributed memory). The tests were done on 2 bus, 4 bus and 9 bus systems. Fig. 7 illustrates the communication between processes (Agents) distributed on three computers using MPI. 8

10 Fig. 7 Communication with MPI The structure of the program implementation is given in Fig. 8. The program is started with creating a class of data which contains input information. The program is started with main function and after declaring the variables, the MPI status should be checked and the size of the program and complexity of the program can be given by specifying rank and size. Measurement processes read input from the data file and sends it to the relay processes to perform the simulation. The communication among the processes can be given using MPI. Finally the program execution time can be calculated. Fig. 8 Pseudo Code of the Proposed Algorithm 9

11 6.2 Detailed Implementation for 2-bus system Fig. 9 gives the implementation aspects of the proposed algorithm. Network Equipment 1 Simulation Environment simulink/simpowersystems Data Acquisition simulink/simpowersystems tool Equipment n Send trip signal if there is any fault Send data to program in C++/MPI Fig. 9 Implementation Aspects of MAS For any network considered, both side data of the all equipment of network can be collected from SIMULINK/Simpowersystems tool. Any power system network can be drawn in this tool to get waveforms and sampled data. This sampled data is given input to proposed algorithm implemented in Linux environment using C++/MPI. Basing on the data it identifies the fault and sends the trip signal. Fig. 10 (a) shows the two bus system considered for the simulation of MAS. It consists of 2 transformers, 1 transmission line, 2 bus-bars and 6 circuit breakers. Each equipment has nine agents. Hence total agents for 5 equipments are 5 9 =45 Agents + 6 CBA s= 51 Agents. Fig. 11 is the two-bus system drawn in SIMULINK/Simpowersystems, Fig. 11 provides both side current waveforms before and after the fault on transmission line G G TF Transmission line 1 TF 2 BB 1 BB 2 Fig. 10(a) Two Bus System Model 10

12 Fig. 10(b) Matlab / Simulink Diagram for 2-bus system G G MA Layer CBA Layer RA Layer FI/IA Layer BUA Layer MA =[0~29]; RA = [30~34]; F I/I A= [35~38]; BUA = [40~44]; CBA= [45~50]]; TRF1 = [0~5, 30, 35, 40, 45, 46] ; BUS1 = [6~11, 31, 36, 41, 46, 47]; TL1 = [12~17, 32, 37, 42, 47, 48] ; BUS2 = [18~23, 33, 38, 43, 48, 49]; TRF2 = [24~29, 34, 38, 44, 49, 50] Fig. 11 Detailed Process of communication for two-bus system Fig 11 shows detailed communication among the different processes for two-bus system. As discussed above, two-bus system needs 51 processes (agents) to implement the proposed algorithm. 11

13 In the system each agent is connected to a set of agents (not all remaining agents). So we need to provide the communication with that set of agents, in order to achieve the reliability of the system. Fig. 11 gives the detailed communication aspects among the different processes of two-bus system. Here processes 0-29 are measurement agents which are in measurement layer of the proposed architecture. Processes are relay agents for the corresponding five equipments, are in relay layer. Processes are fault identifier agents for the corresponding five equipments, are in fault identification and isolation layer. Processes are back-up agents for the corresponding five equipments, are in back-up layer. Processes are circuit breaker agents of the system, are in circuit breaker layer. All these processes do not need to have communication with all others. We need all agents working for particular equipment should be well coordinated and FI/IA of particular equipment should be coordinated with FI/IA of its neighboring equipments. Data File M A T L A B S I M U L I N K Process 12~14 10,000 Samples* 3 phases Process 15~17 10,000 Samples* 3 phases Relay Agent of Trans. line (32) Performs the differential logic with inputs on both sides Multi Agent Protection System Fault Identifier and Isolator agent (37) Identifies fault and send trip to C.B s Back-up Agent (42) C.B (47) C.B (48) Off-line On-line Fig.12 Detailed Implementation for a Fault on Transmission Line Fig. 12 gives the detailed implementation for a fault on transmission line. The current signals of both sides of transmission line are generated from Simulink/ SimPowerSystem and the sampled data is given as a input data file to multi-agent protection system. The relay process of transmission line (32) performs differential logic and send the out put to FI/I agent (37). This agent identifies the fault and send trip to corresponding C.B s (47,48) and Back-up agent. These C.B s send report to backup agent hence the faulted transmission line can be isolated. 12

14 6.3 Case Studies: The program has also been extended to two case studies of i) four-bus and ii) nine-bus systems respectively. The CPU time is approximately 18 microseconds for execution of a two-bus system and 25 microseconds for nine-bus system respectively. Even though the number of equipments is tripled, the computation time is almost same. This is the advantage of distributing the task. 7. Contributions of the Study Important contributions of the research work are the following 1. Proposal of a Basic Architecture of Multi-Agent Systems for transmission network operations. 2. Development of an Algorithm incorporating intelligent agents and their implementation for transmission network protection. 3. Implementation of Communication and Coordination among the various agents in LAM/ MPI. 8. Conclusions and Results Important conclusions arising out of the research work are given as: 1. Multi Agent based Architecture for Operation of a TRANSCO in a Restructured Power system has been proposed. This is achieved by broadly dividing Agents into five types which may belong to either Cognitive or Reactive type of Agents. 2. The communication and coordination between agents has been illustrated by means of parallel operation among Sub-Agents. Feasibility of application and Implementation of an Intelligent Agent System (IAS) for Power System Differential Protection using C++ under LINUX environment using MPI as communication media is performed. 3. The proposed scheme improves Reliability, Speed, Selectivity, and Sensitivity which are the desirable characteristics of a relay. The working of proposed method is found to be modular, scalable and adaptable to various types of protection and coordination among the devices. 9. Scope of Future Work Multi agent system architecture can be proposed for GENCO & DISCO operations. Agent architecture can be implemented for all the Agents apart from Protection. 13

15 10. References 1. Gerhard Weisis, Multiagent Systems, MIT press, Cambridge, R. Biernatzki et al, Agent Technology used in power systems, Proceedings of the 7th AFRICON Conference in Africa, Volume 2, pp , Sept Mikel M. Nordman, matti Lehtonen, Distributed Agent based State Estimation for Electrical Distribution networks, IEEE trans on Power Sys., vol-20, May T.Nagat, H.Sasaki, A Multi-Agent Approach to Power System Restoration, IEEE Trans. on Power Syst., Vol-7, pp , Yacushi Tonitia et al, A Cooperative Protection System with an Agent Model, IEEE Transactions on Power Delivery, Vol. 13, No 4, pp , Oc.t Yang Ming Yo, Approach to Agent Based Adaptive Protection System, Proceedings of IEEE TENCON,Vol-3, pp-28-31, Oct Ming-yu Yang Yang-li Zhu, Study on Adaptive Distance Protection Using Multi Agent Technology 7 th Intl., Power Engineering Conference, pp Vol. 2, Dec D.V.Coury, An Agent Based Current Differential Relay for Use with a Utility Intranet, IEEE Transactions on Power Delivery, Vol.17, No.1, pp , Nov R.Giovanini et al., Improving Local & Backup Protection using Wide area Agent, 8 th IEE Intl. Conf. on Developments in Power System Protection, Vol. 2, pp , April C. Rehtanz, Autonomous Systems and Intelligent Agents in Power System Control and Operation, Springer Verlag, L.L.Lai, Power System Restructuring and Deregulation, John Wiley, Kankar Bhattacharya, Math H.J.Bollen, Jaap E.Daalder, Operation of Restructured Power Systems, Kluwer Academic publishers, A.J.Wood, B.F.Wollenberg, Power Generation Operation & Control, 2 nd edition Wiley, New York, Qun Qiu, Risk Assessment of Power System Catastrophic Failures and Hidden Failure Monitoring and Control System, Dissertation submitted to Virginia Polytechnic Institute, Peter S.Pacheco, Parallel Programming with MPI, Morgan Kaufmann publishers, Inc,

16 11. Proposed Contents of the Thesis Table of Contents Acknowledgements Abstract List of Tables List of Figures Chapter-1 Introduction 1. Introduction to Power Systems 2. Motivation for the Study 3. Literature Survey 4. Scope and Objectives of the Present Work 5. Organization of the Thesis Chapter 2 Distributed Artificial Intelligence (DAI) and Intelligent Agents (IA) 1. Introduction to AI and DAI 2. Intelligent Agents in DAI 3. Summary Chapter 3 Intelligent Agent in Power Systems 1. Literature Survey on Intelligent Agents for Power systems 2. Summary Chapter 4 Intelligent Agent Architecture for Transmission System Network 1. Intelligent Agents for Transmission Network 2. Classification of Agents 3. Description of Agents 4. Summary Chapter 54 Transmission Network Protection Using Intelligent Agents 1. Literature for Intelligent Agents in Protection 2. Need for Intelligent Agents in Protection Scheme 3. Proposed Architecture for Protection 4. Summary Chapter-6 Implementation Details 1. Implementation Scheme 2. Simulink/SimPowerSystems 3. Linux/C++ 4. Message Passing Interface (MPI) 5. Summary Chapter-7 Discussion and Results 1. Case Study 1 Two Bus System 2. Case Study 2 Four Bus System 3. Case Study 3 Nine Bus System 4. Summary 15

17 Chapter-8 Conclusions and Future Work 1. Conclusion 2. Scope for Future Work Appendix Publications References 12. List of Papers Published Based on this Thesis 1. Anuradha.Ch, K.S. Swarup, Power System Protection using Multi Agent Systems with Message Passing Interface, International Conference on Power System Analysis Control & Optimization (PSACO), Andhra University, Waltair, pp , March Anuradha.Ch, K.S. Swarup, Transmission Line Differential Protection Scheme Incorporating Intelligent Agents, Communicated to IEEE Transactions on Power Delivery, April Anuradha.Ch, Balakrishna.P, K.S.Swarup, Multi-Agent Based TRANSCO Operation in a Restructured Environment, Communicated to IEEE Transactions on Power Systems, April Anuradha.Ch, Balakrishna.P, K.S.Swarup, Multi-Agent Systems for Transmission Operation in a Restructured Power System, Communicated to INDICON, IIT Kanpur, May,