Robust model predictive control of electrical drives considering measurement noises

Abstract

In the realm of model predictive current control (MPC) for electrical drives, the multiple vector MPC offers significant advantages with respect to its high current tracking accuracy and reduced current ripples, making it a promising alternative to conventional single-vector MPC. To compensate for the system disturbances due to model mismatches, extended-state observers (ESOs) are often integrated. However, the observer bandwidth relating to the disturbance rejection capability is usually compromised, since high bandwidth will inevitably amplify high-frequency measurement noises. To address this coupling issue between high frequency noises in steady state current and the disturbance rejection capability in transient scenarios, a novel triple-vector MPC is proposed in this paper. Specifically, the MPC incorporating an active damping method is proposed to improve the robustness against parameter mismatch and disturbances. Moreover, a dedicated adaptive current predictor is proposed to estimate current with one step delay compensation, where an additional coefficient is designed to suppress measurement noises independently. To further enhance the flexibility of the current predictor, the adaptive mechanism based on the operation status is introduced to adjust the introduced coefficient. Comprehensive experiments have been conducted to demonstrate the superiority in both high robustness and noise suppression of the proposed controller.