Integrating intelligent systems into a cooperating community for electricity distribution management

摘要

Systems in which semi-autonomous problem-solving agents communicate and cooperate with one another represent an exciting vision of future computing environments. However, if this vision is ever going to result in commercially viable systems, then consideration must be given to the large software base that exists within many organisations. Success requires the ability to incorporate pre-existing systems alongside purpose-built agents in a cooperating community. This requirement is vital because the former represent a substantial resource investment that companies cannot afford to consign to the scrap heap. We report on our experiences of constructing cooperating communities that contain elements that were pre-existing and some that were developed specifically for incorporation into an integrated environment. The general purpose framework of ARCHON (ARchitecture for Cooperative Heterogeneous ON-line systems) provides the underlying technology to facilitate cooperative problem solving, and the exemplar domain is the real world problem of electricity distribution management. The actual application being developed is called CIDIM (Cooperating Intelligent systems for DIstribution system Management). An evolving methodology for designing and developing a mixed system such as this is outlined, based on our experiences in CIDIM and several other real-world industrial applications. It specifies a hybrid top-down and bottom-up approach to integration, identifies the important characteristics that shape multi-agent problem analysis, and outlines key factors that impinge upon the design of the community. this methodology is then used to motivate the design decisions for the CIDIM application. Finally the process of instantiating the individual agents is discussed, some helpful guidelines on testing and evaluating future applications are given, and the implementation of one of CIDIM's cooperative scenarios is described in depth.