IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 22, NO. 1, JANUARY 2007 433 Implementing Multiagent Systems Technology for Power Distribution Network Control and Protection Management Ioannis S. Baxevanos and Dimitris P. Labridis, Senior Member, IEEE Abstract—Power distribution network protection and restora- tion are significant issues in terms of quality and reliability of the supplied electrical energy. In this paper, an extended research in possible high-end protection and control methods is presented. The authors intend to show the efficiency of combining modern IT tech- niques with the equipment provided by distribution automation evolving technology. As a result, a flexible and versatile multia- gent system (MAS) is proposed. The effected MAS is fault isolation and power restoration oriented, as shown in the case studies. The structure of the proposed system (software and hardware aspects) is introduced. Both microscopic and macroscopic procedures are revised. Power line communications technology is implemented. Results acquired from computer simulations are evaluated. Con- clusions concerning the perspectives of implementing the proposed system are finally presented. Index Terms—Artificial intelligence, cooperative systems, power distribution protection, underground power distribution lines. I. INTRODUCTION D ISTRIBUTION system protection and restoration are important applications of the artificial intelligence tech- nology. Researchers have already associated available knowl- edge and experience of domain experts with computational systems dedicated to centralized control and protection. As a result, several rule-based expert systems have been proposed including logical rules for switching actions for fault isola- tion and load transfer between adjacent feeders [1], heuristic rules (compiled from the dispatcher’s experience) for locating distribution system faults [2] and to some extent, presenting user-friendly graphical user interfaces (GUIs) as practical restoration aid systems, to assist the decision making of super- visory-control and data-acquisition (SCADA) operators [3]. Expert systems have also been applied to the general concept of power system restoration as shown in [4]–[6]. Possessing an auxiliary role, expert systems usually require information concerning the topology of the network and the nature of the network’s components, as well as remote metering of elec- trical values. Therefore dedicated communication channels are postulated, including fiber optics or copper-wire telecommu- nications and recently, wireless technologies such as Global System for Mobile Communications (GSM) and terrestrial Manuscript received June 6, 2005; revised December 9, 2005. This work was supported by the Operational Program for Education and Initial Vocational Training (O.P. “Education”)-Heracletus. Paper no. TPWRD-00334-2005. The authors are with the Department of Electrical and Computer Engineering, Power Systems Laboratory, Aristotle University of Thessaloniki, Thessaloniki GR 54124, Greece (e-mail: baxi@auth.gr). Digital Object Identifier 10.1109/TPWRD.2006.877085 trunked radio (TETRA) [7]. However, issues regarding time and computational resources, costs and physical aspects of the problems restrain the operability of the above systems. Efforts toward distributing activities and intelligence fall into the scope of distributed AI (DAI) and multiagent systems (MAS). Implementing the agent technology and especially MAS for infrastructures, such as the power grid, seems to be quite promising [8], [9]. In the case of distribution system fault management, the decomposition of the problem is quite effec- tive [10]. Approaching the complex problem of distribution grid management in a distributed way converges with significant areas of applications for MAS, namely distributed solving of problems and solving distributed problems [11]. The utility industries will benefit from the implementation of multiagent technology, because there will be no need for exclu- sive central control and protection of the distribution grid. Su- pervision of procedures that will take place locally, as close as possible to the problematic sectors, will only be needed. Several heterogeneous components with possible incompatible behav- iors will combine actions under the guidance of software agents that will cooperate in order to accomplish their joint goals. The objectives of the MAS approach to distribution network management are still the same with the ones of centralized sys- tems: to minimize the duration of outages as well as the cus- tomer interruptions, in favor of the overall distribution system reliability. Customer-based distribution reliability indices, es- pecially customer average interruption duration index (CAIDI) and system average interruption duration index (SAIDI) [12], will be potentially improved. However, DAI and MAS tech- nology, if implemented, may actually result in self-managing automated networks [13]. The MAS project presented in this paper intends to reinforce self-managing capabilities provided to the network, with adap- tive behaviors which are necessary for applications in many dif- ferent network topologies. II. PRESENT WORK APPLICATION FIELD This particular research and the effected project were trig- gered off by the imperative need for really rapid fault local- ization, isolation and restoration procedures. A brief layout of a distribution network that falls into the scope of this work is presented in Fig. 1 [14]. Open-loop structures as well as radial medium-voltage (MV) feeders are considered. The distribution network consists of both underground (res- idential, commercial, and industrial customers) and overhead (industrial and agricultural customers) MV lines. Statistical data indicate that underground lines usually sustain permanent 0885-8977/$20.00 © 2006 IEEE