Online d-q axis inductance identification for IPMSMs using FEA-driven CNN

The permanent magnet synchronous motor (PMSM) is the most commonly used option for electric vehicles, because it has a straightforward design and a comparatively high power-density. For the sake of healthy monitoring and sophisticated parameter-dependent control theories for PMSMs, determining the parameters of PMSMs is crucial. Precise identification of the inductance is required due to its coupled and nonlinear connection with other electromagnetic properties. In this paper, a convolutional neural network (CNN) model is designed to identify the d-q axis inductances of an interior permanent magnet synchronous motor (IPMSM). The model is trained with datasets obtained by finite element analysis (FEA) methods. Simulation validates that the proposed model performs excellently in terms of online identification, yielding maximum bias values of 2.96 % for the q-axis inductance and 2.11% for the d-axis inductance. The proposed method achieves accurate inductance online identification providing a new solution to handle nonlinear industrial problems.