To address the insufficient accuracy of conventional permanent magnet synchronous motor (PMSM) models caused by neglecting magnetic saturation nonlinearity and periodic parameter disturbances, a nonlinear motor system model integrating a Physics-Informed Neural Network (PINN) is developed. By exploiting the differential relationships among incremental inductance, flux linkage, and magnetic energy, the voltage and torque equations considering rotor position variation are derived, and analytical expressions for the derivatives of incremental inductances are obtained. To reduce the computational burden of PINN in system-level simulations, linear and nonlinear approximation strategies based on incremental inductances and their derivatives are proposed, which significantly reduce the frequency of PINN calls while maintaining model accuracy. CPU/GPU collaborative computation and cross-frequency-domain scheduling are further implemented to improve simulation efficiency. Considering the influence of the test bench mechanical dynamics, an electromechanical–magnetic coupled simulation model is established. The accuracy of the proposed nonlinear motor model is validated through two-phase short-circuit tests as well as simulations and test bench experiments under sinusoidal and non-sinusoidal excitations. The results demonstrate that the proposed model accurately captures the nonlinear electromagnetic characteristics of PMSMs while significantly improving system simulation efficiency.
Li et al. (Wed,) studied this question.