High-Fidelity Model for Interior Permanent Magnet Synchronous Machines Considering the Magnet Saturation and Spatial Harmonics Based on Deep Neural Network

Interior permanent magnet synchronous machines (IPMSMs) exhibit relatively large spatial harmonics in phase voltage and high nonlinearity in torque production due to the magnet saturation and reluctance torque. In order to accurately simulate and evaluate the control performance of the motor drive system, high-fidelity motor models considering these nonlinear phenomena are always appreciated. In this paper, an IPMSM modelling technique is proposed based on deep neural networks (DNNs). The proposed modelling technique could represent the nonlinear characteristics and spatial harmonics in torque but independent of the complex mathematical models or empirical formulas, thus the proposed method could significantly eliminate the difficulties of nonlinear IPMSM system modelling.