Voltage contingency ranking using fuzzified multilayer perceptron

A fuzzified multi-layer perceptron (FMLP) trained by back-propagation algor ithm is proposed for on line voltage contingency analysis and ranking. The input vector consists of fuzzy membership values of loads to different ling uistic categories, while the output vector is defined in terms of fuzzy mem bership values of a voltage performance index in different severity classes . Fuzzifying the loads into linguistic categories using non-linear membersh ip functions enables efficient modeling of uncertainty associated with load s. Angular distance based clustering has been used to determine significant inputs to the fuzzified neural network. Due to the incorporation of fuzzy logic, the method is capable of handling even those contingencies that belo ng to more than one class. The effectiveness of the method has been shown o n IEEE 30-bus test system and 75-bus Indian system and it is found to class ify and rank the contingencies quite accurately for unknown load patterns. (C) 2001 Elsevier Science B.V. All rights reserved.