Energy-Optimal and Thermally-Robust Predictive Flux Control of Industrial Induction Motor Drives

Variable-speed induction motor drives spend most of their service life at partial load, where rated-flux field-oriented control (FOC) is inefficient and where loss-minimizing control (LMC) recovers a large part of the loss. LMC, however, is brittle in two ways that matter in industry: it is tuned isothermally, so as the windings heat the rotor-resistance drift detunes the field orientation and corrupts torque; and it treats the loss-optimal flux as a quasi-static set-point, so an abrupt load rise from a light-load, low-flux condition forces a slow flux rebuild that throttles torque. This paper proposes a thermally-adaptive economic model predictive controller (TA-EMPC) that retains the energy optimum of LMC while removing both weaknesses. A temperature-coupled total-loss model (machine copper and core loss plus inverter conduction and switching loss) is minimized over a finite horizon subject to a torque-delivery constraint; a reduced-order two-node thermal observer updates the loss-defining resistances online without a temperature sensor; and a load-demand-aware flux-reservation term pre-magnetizes the machine ahead of anticipated torque rises. In simulation on a representative 7.5 kW drive, TA-EMPC matched the energy of static LMC to within 0.3% across pump, conveyor, and fast-cycling duty profiles—both saving 1.4–2.3% of cycle energy relative to rated-flux FOC, and up to about 14 efficiency points at very light load—while, unlike LMC, holding the steady torque error below 0.5% when the winding temperature rose to about 90 °C (a resistance increase of roughly 26%) (against an 8% error for the non-adaptive scheme) and reducing the torque undershoot during a light-to-heavy load step from about 23% to near zero. An operation-count analysis indicates that the condensed quadratic-program formulation with move blocking is executable within the 100 µs sampling interval on a production digital signal controller for the chosen control horizon (analytical estimate; on-target measurement is future work). The contribution is thus energy-efficient operation delivered with the torque robustness that loss minimization alone does not provide.