Classifying faults in high voltage power systems: A rough-fuzzy neural computational approach

This paper introduces an approach to classifying faults in high voltage power system with a combination of rough sets and fuzzy sets in a neural computing framework. Typical error signals important for fault detection in power systems are considered. Features of these error signals derived earlier using Fast Fourier Transform analysis, amplitude estimation and waveform type identification, provide inputs to a neural network used in classifying faults. A form of rough neuron with memory is introduced in this paper. A brief overview of a rough-fuzzy neural computational method is given. The learning performance of a rough-fuzzy and pure fuzzy neural network are compared.