A dynamic recurrent neural network for wide area identification of a multimachine power system with a FACTS device

Multilayer perceptron and radial basis function neural networks have been traditionally used for plant identification in power systems applications of neural networks. While being efficient in tracking the plant dynamics in a relatively small system, their performance degrades as the dimensions of the plant to be identified are increased, for example in supervisory level identification of a multimachine power system for wide area control purposes. Recurrent neural networks can deal with such a problem by modeling the system as a set of differential equations and with less order of complexity. Such a recurrent neural network identifier is designed and implemented for supervisory level identification of a multimachine power system with a FACTS device. Simulation results are provided to show that the neuroidentifier can track the system dynamics with sufficient accuracy