Hierarchical neural network adaptive power system stabilizer

An adaptive power system stabilizer based on feedforward neural networks is proposed in this paper. A hierarchical architecture of neural networks consisting of two subnetworks is used. One subnetwork is used for the identification of the dynamics of the power system under study, and the other one is used as a stabilizer. The weights of the neural network stabilizer are adjusted according to the difference between the output of the neural network identifier and a desired output track. Both the neural network stabilizer and the neural network identifier are trained in succeeding stages by the backpropagation algorithm. An application of this scheme for regulating the speed of a synchronous power generator under fault conditions is described.