Application of a new fuzzy neural network to transient stability prediction

This paper presents a new fuzzy neural network (FNN) for prediction of transient stability. This FNN implements fuzzification and defuzzification processes in the neural network architecture. Degree of freedoms or connection weights of the FNN are tuned by a kind of steepest descent learning algorithm. The proposed FNN is used for the prediction of the transient stability status of the power system. To learn the FNN, a feature selection technique based on the transient swings is presented. The whole method is tested on a few well-known test systems. Moreover, the proposed FNN is compared with the standard multi-layer perceptron (MLP) neural network. Obtained results, discussed comprehensively, confirm the validity of the developed approach.