Compensating the Measurement Error in Model-free Predictive Control of Induction Motor via Kalman Filter-based Ultra-local Model

In model predictive control, ensuring the accuracy and robustness of the prediction model is crucial. A Kalman filter is a self-correction method commonly used as an observer for state estimation in uncertain applications. Model-free predictive control utilizes an ultra-local model for prediction purposes. Precise measurements and feedback gains are required for accuracy. This study proposes a new ultra-local prediction model based on the Kalman filter, replacing the extended state observer with the proposed model for disturbance observation. The Kalman filter-based prediction model is applied to the model-free predictive control of the induction motor. The method is validated with experimental results, comparing it to the extended state observer-based prediction model, using a 4kW induction motor setup.