Economic model predictive control (MPC) of induction motor drives selects the flux from a loss model that usually omits saturation, additional load loss, the core-loss increment under pulse-width modulation, and the junction-temperature dependence of switching loss. A thermally adaptive economic MPC is extended with learned components, and each effect is isolated. The exact DC-link power balance shows that losses supplied through the air gap are unobservable without a torque sensor, whereas the additional load loss is observable; its estimate is passed to the stator node of the thermal observer. The magnetizing curve is learned from the reactive-power balance by a set-membership estimator contracted by monotonicity and concavity, and the resulting flux band tightens the current, reserve, and voltage constraints of the economic layer. Fifteen controller variants were simulated on 2.2, 7.5, and 37 kW drives on pump and conveyor cycles, including 600 Monte-Carlo samples. In nominal runs, the winding-temperature error fell from 5.5–10.2 K to 1.1–1.6 K and the torque error from 0.7–1.6% to 0.2–0.4% of rated torque. Across the samples, the stator voltage exceeded its planning limit in one case, against 13–36 per 100 without learning, and conveyor-cycle energy fell by 0.26–0.42 percentage points; pump-cycle differences were not resolved.

Economic model predictive control (MPC) of induction motor drives minimizes a loss model that usually omits saturation, additional load loss, the core-loss increment under pulse-width modulation, and the junction-temperature dependence of switching energy. A published thermally adaptive economic MPC is extended in three ways. First, it is proven that the DC-link power residual is structurally blind to losses supplied through the air gap when torque is reconstructed from measured currents; therefore, only stator-terminal and converter losses can be learned without a torque sensor, and a physics-informed residual model is identified and fed to the thermal observer. Second, the magnetizing curve is learned from the reactive-power balance by a set-membership estimator contracted by monotonicity and concavity. Third, the guaranteed flux band tightens the current, torque-reserve, and voltage constraints of the economic target problem, whose feasible set grows monotonically during learning. In simulations of 2.2, 7.5, and 37 kW drives with pump and conveyor cycles and a structurally mismatched plant, the winding-temperature estimation error decreased from 5.5–10.7 K to 1.3–1.6 K, torque error from 0.74–1.45% to 0.24–0.40% of rated torque, cycle energy came within 0.06–0.23% of an oracle, and voltage-headroom violations were eliminated at an energy cost of 0.03–0.18 percentage points.