Fast real power contingency ranking using a counterpropagation network

This paper proposes a fast real power contingency ranking approach which is based on a pattern recognition technique using a forward-only counterpropagation neural network (CPN). The power system operating state is described by a set of variables which compose the pattern. The corresponding performance indices of various contingencies can then be recognised by a properly trained counterpropagation network. A feature selection method is also employed for reducing the dimensionality of the input patterns. When compared with a full AC load flow the proposed method is more superior and has good pattern recognition ability.