Data-Driven Predictive Control Strategy to Improve Robust Performance for Three-Level Inverters With Reduced Common-Mode Voltage
Yaopeng Huang, Tong Liu, Qicai Ren, Alian Chen, Hang Zhang, Wei Wang
Abstract
This article proposed a generalized extended state observer (GESO)-based data-driven predictive control (DDPC) strategy with multi-vector for three-level inverters, which is thoroughly independent of system parameters while obtaining a satisfactory multi-objective control performance considering common mode voltage (CMV) reduction. In the proposed strategy, the predictive dynamic of output current is achieved by DDPC, and only the system input and output knowledge is needed. Meanwhile, due to the complex dynamics and difficulty of identifying by DDPC, the system uncertainties can still degrade the control performance if not solved. In this regard, the GESO is introduced to estimate the unknown uncertainties as a new extended state, which is further compensated in the control law, enabling a significant robust performance enhancement. Then, aiming to reduce the CMV, virtual vectors are constructed into the candidate vectors to replace the small vectors with high CMV magnitude. Moreover, three optimal vectors are selected to synthesize the reference voltage, where the current harmonic and the neutral-point imbalance are simultaneously suppressed. Finally, the validity of the proposed strategy is proved by simulation and experiments on a three-level T-type inverter.