Zhenyao Sun / The HongKong Polytechnique University
Junkai Wen / The HongKong Polytechnique University
Mingyuan Jiang / The Hong Kong Polytechnique University
Shuangxia Niu / The Hong Kong Polytechnique University
Multistep prediction is widely employed to improve the performance of model predictive current control (MPCC). However, its computational complexity grows exponentially with the prediction horizon, severely limiting its feasibility for real-time applications. Furthermore, the control performance of multistep MPCC is highly sensitive to model accuracy and parameter variations. To address these challenges, we propose an event-triggered model-free predictive current control (ET-MFPCC) strategy. The ET-MFPCC framework introduces a boundary for the current tracking error, ensuring that the voltage vector is updated only when the error trajectory intersects the boundary. The current trajectory is then extrapolated to estimate the duration time for each candidate voltage vector. A cost function based on the instantaneous average switching frequency is formulated to determine the optimal voltage vector. To mitigate mis-switching caused by system noise and parameter uncertainties, an ultra-local model combined with an extended state observer is employed to estimate the current and compensate for disturbances. Comparative simulation results validate the effectiveness and superiority of the proposed ET-MFPCC method over conventional MPCC approaches.