Cooperating Intelligent Systems for Electricity Distribution

1 INTRODUCTION Artificial Intelligence (AI) techniques, and expert systems technology in particular, has often been used to tackle the most difficult automation problems. This is especially true in industrial applications where conventional software and teams of operators were unable to cope with the demands of rapidly changing, complex environments. After more than a decade of exploitation, hundreds of expert systems have been deployed. However as this technology has proliferated and individual systems have increased in size and complexity, some well documented problems have begun to emerge. These include the fact that systems do not easily scale up, that they are brittle and that consistency is difficult to achieve. Various techniques for circumventing these problems have been advocated. These include building an extremely large base of common-sense knowledge (Guha and Lenat 1990) and allowing the sharing and reuse of knowledge (Neches et al. 1991). However the approach pursued in this work was to build systems of smaller, more manageable components which can communicate and cooperate. In such multi-agent systems: knowledge, resources, control and authority are distributed amongst community members who then work together, in a coordinated and coherent manner, to solve problems. The ARCHON project has developed a general-purpose framework which enables multiple problem solvers, some of which may be pre-existing, to be interconnected so that they can cooperate whilst solving problems (Jennings 1991, Wittig 1992). This paper concentrates on the project's CIDIM (Cooperating Intelligent Systems for DIstribution System Management) application which is in the domain of electricity distribution management.