An on-line self-learning power system stabilizer using a neural network method

Based on the extensive theoretical analysis of a self-learning algorithm, a novel on-line neural network self-learning algorithm is proposed. This algorithm aims to learn the inverse dynamics of a controlled system. Samples can be easily obtained by the measurements. A reference model or a given orbit is used to generate ideal system responses. A scheme for on-line real-time implementation of such a controller is given. The proposed algorithm has been used to design a self-learning power system stabilizer. Simulation results show that the proposed self-learning neural network based PSS is very effective in damping out the lower frequency oscillations.