Temperature-Aware Model Predictive Current Control for PMSM Drives Under Thermal Parameter Drift

Abstract

Model predictive current control (MPCC) provides fast torque dynamics for permanent magnet synchronous motor (PMSM) drives, but its accuracy rests on machine parameters that drift as winding and magnet temperatures rise during regular operation. Existing remedies-disturbance observers and model-free predictors-act at the voltage level and therefore cannot correct the reference-level error introduced when the torque command is converted into a current reference through an outdated flux-linkage value. This paper presents TA-MPCC, a temperature-aware MPCC that estimates winding and magnet temperatures online with a two-node lumped-parameter thermal network, propagates them to the resistance and flux-linkage values of both the prediction model and the reference conversion through a feed-forward adaptation law, and delegates fast residual errors to an extended state observer. Because the dominant thermally induced error is corrected where it enters, the compensation remains effective where voltage-level schemes saturate. In a simulated $\mathbf{3 0}$-min heat run, the median torque deviation of TA-MPCC is $\mathbf{0 . 0 2 3 ~ N} \cdot \mathbf{m}$, 72\% below the strongest baseline, and it stays below $0.07 \mathrm{~N} \cdot \mathrm{~m}$ up to a magnet temperature of $105^{\circ} \mathrm{C}$, a gain confined to the thermal axis: under $\pm \mathbf{4 0 \%}$ inductance mismatch THD still rises from $8.0 \%$ to $9.5-10.3 \%$, an increase the thermal adaptation law does not address, while nominal current THD is preserved (8.1\% versus 8.0\%).