In this paper, a hybrid neural network based approach is proposed for fast
voltage contingency screening and ranking. The developed hybrid neural netw
ork is a combination of a filter module and ranking modular neural network.
All the selected contingency cases are applied to the filter module, which
is trained to classify them either in critical contingency class or in non
-critical contingency class using a modified BP algorithm. The screened cri
tical contingencies are passed to the ranking modular neural network for th
eir further ranking. The ranking modular neural network reduces a K-class p
roblem to a set of K two-class problems with a separately trained network f
or each of the simpler problems. Total load demand, real and reactive pre-c
ontingency line-flows and terminal voltages in the contingent element, alon
g with a topology number corresponding to the contingent element, are selec
ted as input features for the neural networks. The continuous values of vol
tage performance index are classified into four classes (levels) according
to their severity, and the modular neural network is trained for this multi
-class classification problem. The effectiveness of the proposed method is
demonstrated by applying it for contingency screening and ranking at differ
ent loading conditions for IEEE 30-bus system and a practical 75-bus Indian
system. Once trained, the hybrid neural network gives fast and accurate sc
reening and ranking for unknown patterns and is found to be suitable for on
-line applications at Energy Management Systems. (C) 1999 Elsevier Science
Ltd. All rights reserved.