Model‐Free Predictive Control of Induction Motor Using a Correntropy Criterion–Based Unscented Kalman Filter
In conventional model predictive control, three‐level inverter‐fed induction motor systems are susceptible to parameter mismatch, leading to degraded control performance. To enhance parametric robustness against nonlinear dynamics and impulsive noise, this paper proposes a model‐free predictive torque control using a correntropy criterion–based unscented Kalman filter (CCUKF). First, an ultralocal model is employed to consolidate system uncertainties into a lumped disturbance. Second, the sigma‐point sampling method of the unscented Kalman filter accurately captures nonlinear statistical characteristics, avoiding the linearization errors inherent in the extended Kalman filter and improving state estimation accuracy. Furthermore, the correntropy criterion is introduced to optimize the Kalman gain, robustly suppressing non‐Gaussian noise and outliers caused by electromagnetic interference. Experimental results demonstrate improvements in both dynamic response and steady‐state performance, along with effective suppression of torque fluctuations, showing superior performance compared with conventional methods while reducing dependence on motor parameters.