Data-Driven Predictive Control with Inherent Update Method for Two-Level Voltage Source Inverters

This paper proposes an inherent and effective method for updating the database which is critical for the model-free predictive controller. In contrast to existing update methods which require complex analytical expressions, a unified analytical update model is proposed as a sum of the actual current and input control gradients. The effectiveness of the proposed controller is validated in a two-level voltage source inverter connected to the grid and an RLE load. The simulation results show that DDPC effectively cancels the stagnant mode and compared with MPC, DDPC provides better current performance at normal conditions and has higher robustness under parameter mismatches.