Wong, Sow Kum (1996) An adaptive agent architecture for the design of complex decision support system with applications to power system protection. PhD thesis, Victoria University of Technology.

Abstract

In the early 1990s, there was a notable lack of intelligent systems in the area of power system protection. Moreover, there was no legacy software that would design a coordinated protection scheme for a power system, although there was considerable interest in such applications shown by people from the industry. However, to build a good, robust and efficient application system, the architecture must be designed properly and with care. It is the architecture that eventually builds up the core of an operational system. Therefore, serious thought and consideration must be given to the study of the architecture. Unfortunately, literature review showed that not many research projects were undertaken in this area. This prompted investigation into the analysis and design of a generic agent architecture for the development of an intelligent system which could be applied to any application domain. The approach used here to design the agent architecture is an adaptive and incremental approach. One of the main attractive features of the agent architecture is the dynamic knowledge component of the system. The knowledge base can grow and modify itself automatically so that it can adapt itself to the ever changing environment and demands. The design of the agent architecture is developed and built using a layered architecture approach. Layering is a very powerful technique that enables effective representation and integration of various techniques and paradigms into one single architecture. Communication is performed via message passing between the adjacent layers. One of the advantanges of this adaptive approach is the development of intelligent agents which imitate the behaviour of human experts during problem solving. The agents attempt to apply their past experiences or skills when resolving a problem. Over the time, as they go through more problem solving tasks, their experiences, skills and knowledge also grow. The technique used to selectively retains the experience, copes with the memory requirements for knowledge expansion, while the organisation of the memory layer allows the agent to confidently add the reliability factor to the proposed solution. There are basically three main types of agents developed for the implementation of an intelligent multiagent system in this thesis. These are the Interface agent, Coordinator agent and Design agent. The thesis closes with a final conclusion regarding the work accomplished in the thesis. A number of recommendations regarding possible enhancements to the system for future investigations are also presented.

Search Google Scholar

Repository staff login