Flux Linkage Tracking-Based Permanent Magnet Temperature Hybrid Modeling and Estimation for PMSMs with Data-Driven-Based Core Loss Compensation

For permanent magnet synchronous machine (PMSM) drive, accurate magnet temperature is critical. The popular model-based magnet temperature estimation can be affected by core loss effect especially in the high-speed conditions. This article proposes a novel hybrid approach for accurate magnet temperature modeling and estimation, in which the estimation model is established by tracking the flux linkage variation, while the data-driven-based model is proposed to compensate the core loss effect. Specifically, the flux linkages in the rotating frame are projected into a new frame to derive the estimation model establishing the relationship between flux linkage variation and magnet temperature, in which the inverter distortion effect is canceled to improve the model accuracy. Based on this estimation model, the core loss effect is modeled, which indicates that the core loss influence is highly nonlinear and dependent on operating conditions. Hence, a radial basis function-based network is employed to model and compensate the core loss effect, and the network training is derived from the proposed model. The proposed hybrid approach can effectively improve the estimation performance especially at the high-speed conditions. Extensive experiments and comparisons are conducted on a laboratory interior PMSM drive to evaluate the proposed approach under various operating conditions.