Temperature-enhanced Steinmetz modeling and multifactor interaction analysis of high-frequency magnetic core loss

To address the limited prediction accuracy of Steinmetz Equation (SE) under multi-temperature operating conditions, this study proposes an improved model that explicitly incorporates temperature effects, namely Steinmetz model (SET). On the basis of preserving the power-law structure and physical interpretability of SE, SET introduces an exponential temperature correction term to describe the nonlinear response of core loss to temperature variations, and PSO is employed to achieve global optimization of model parameters. Experimental results based on the MagNet dataset show that, under identical optimization conditions, SET achieves improvements of 58.72% in MAE and 26.83% in MAPE, with the coefficient of determination increasing to 0.9928, and exhibits enhanced robustness under high-loss and extreme operating conditions. Furthermore, by combining statistical analysis with interpretable machine learning methods, the dominant effects of excitation waveform and core material, as well as their significant interaction effects on core loss, are systematically revealed.