Neural plug-in motor coil thermal modeling

With the latest advancement in technologies, many motor controls and protections now rely on the use of online motor data fed through embedded motor models (including thermal models) stored in microprocessors where the critical decisions are made. The accuracy of these embedded motor models has a direct effect on the performance and reliability of the motors. In this paper, the authors propose to use an artificial neural network plug-in modeling concept to significantly increase the accuracy of the lumped-parameter motor thermal modeling approach. The neural plug-in approach could preserve the fast calculation response of the lumped-parameter model, preserve the physical meaning of thermal parameters, and increase the overall accuracy of the thermal model. This paper discusses and compares the motor thermal dynamics and its estimates by the three different models, lumped-parameter model, conventional neural network model, and neural plug-in model. The preliminary modeling results clearly indicate that the neural plug-in modeling approach is superior than either the lumped-parameter approach or the neural network approach