9520456 Starrett Current trends in power system regulation and planning are causing complex, unpredictable, nonlinear phenomenon to limit the economic operation of power systems. The analysis and computations required to ensure that the system will operate within its reliability criteria are becoming increasingly difficult and time consuming. The resulting data is often hard to interpret. The goal of this research is to develop computer- based expert systems to: process data, identify patterns, classify system conditions according to expected problems, suggest scheduling or control changes, and control the system. Artificial neural networks (ANN's) are useful for pattern recognition and data classification. Depending on the type of network used, ANN's can be trained to give a desired output under certain conditions, or they can be allowed to recognize patterns in data without pre-defined output specifications. An ANN advisory system is proposed that will monitor system data, process patterns and warn the operator of dangerous system conditions. Fuzzy logic systems formulate output decisions based on sets of rules that apply within ranges of system conditions. Thus they can efficiently deal with uncertain conditions and a range of operating points. The classification results of the neural networks can be used to form a flexible, yet predictable control system using fuzzy logic control. The ANN/fuzzy logic system will be designed to allow adjustments in response to changing power system conditions. ***
