Experimental verification of a generating unit excitation neuro-controller

A feedforward neural network (FFNN) based excitation neuro-controller (ENC), which integrates the functions of both the AVR and the PSS, has been implemented and experimentally verified on-line on a physical model of a single-machine infinite-bus power system. The FFNN has a simple structure and needs only the terminal voltage and generator speed to be measured. The ENC output signal depends on the on-line training and a modified error function used in training. The performance of the proposed ENC has been investigated by a number of experimental tests for a variety of operating conditions and disturbances.