This study proposes a novel approach that employs the kernel method as a regression model to demonstrate the dependency of magnet flux linkage on the applied current, which is suitable for magnet temperature estimation. This model can estimate the flux linkage with a mean relative error of less than 2% in comparison with that obtained using finite element analysis. The magnet temperature is estimated by comparing the magnet flux linkage under loading conditions with the values obtained from the regression models built under fixed temperatures. The accuracy of the results obtained using the magnet temperature estimation method is approximately the same as that of the results obtained using the look-up table, suggesting that the proposed approach is suitable for non-linear motor property modeling.
- Machine learning
- Temperature estimation
- Variable Leakage Flux IPM (VLF-IPM)