Neural network based saturation model for round rotor synchronous generator

This paper presents an artificial neural network (ANN) based technique to m odel saturation for a round rotor synchronous generator. The effects of exc itation level, rotor angle, and real power generation on generator saturati on are included in the modeling process. To illustrate the technique, small excitation disturbance tests are conducted on a 7.5 kVA, 240V, 60 Hz, roun d rotor synchronous generator at various levels of excitation and loading. The small excitation disturbance responses are processed by a recursive max imum likelihood algorithm to yield estimates of mutual inductances L-ad and L-aq at each operating condition. By developing a suitable training patter n, variables representative of generator operating condition are mapped to mutual inductances L-ad and L-aq. The developed models are validated with m easurements not used in the training process and with large disturbance res ponses.