Input-output model identification of small and large induction motors using multilayer feedforward neural networks

A multilayer feedforward neural network is used to realize a moving-average model of an induction motor. This input-output representation generalizes earlier, widely-used static load models to include dynamic loads. The new model can easily be integrated in existing power flow and stability programs. The validity of the proposed methodology is demonstrated with the successful identification of small and large induction motors described by first-order and third-order differential models, respectively. The neural network-based models can accurately predict the machine real and reactive power consumptions under simultaneous or separate voltage and frequency disturbances.