Gray-box modeling of electric drive systems using neural networks

This paper presents the use of gray-box modeling to model electric drives. In gray-box modeling the system model is partitioned into a known and an unknown part. The known part of the model is derived from physical principles while the unknown part is modeled using a black-box model. In the case of electrical machines the electric part of the system is well understood from the corresponding governing physical laws, while the mechanical part of the system could be too complex or unknown. The application of this approach is investigated on a DC drive system. We present the use of neural networks as the black-box model for an unknown static nonlinearity. We study the issues of network architecture and of algorithms for parameter estimation