Abstract The drift of the neutral-point voltage (NP-V) at the DC-side capacitor of the T-type three-level grid-connected inverter directly affects the grid-connected performance. The model predictive control (MPC) accuracy is directly affected by the time variation and non-modeling of grid-connected inverter system parameters. In order to overcome the defect of NP-V imbalance in traditional MPC when the modulation is relatively high, a virtual neutral vector and a virtual compensation vector are constructed to participate in the control vector set of MPC without affecting NP-V in this paper. The single objective cost function of current tracking is established, and the small vector is optimized according to the DC bus capacitance state without setting the weight factor to further control the NP-V balance. An extended state observer (ESO) is proposed to observe the non-modeling part of the grid-connected inverter system and add the observation results to the grid-connected reference voltage. The experimental results show that the optimal MPC strategy eliminates NP-V low-frequency oscillations, and its total harmonic distortion (THD) is lower than that of the traditional MPC strategy.
CHEN Long,ZHOU Yangzhong. An improved model predictive control based on virtual space vector for T-type three-level grid-connected inverters[J]. Electrical Engineering, 2022, 23(11): 37-43.
2022, Vol. 23 
