An Improved Model Predictive Controller for 27-Level Asymmetric Cascaded Inverter Applicable in High-Power PV Grid-Connected Systems

Abstract

The asymmetric cascaded H-bridge multilevel inverter (MLI) topology is more suitable for high-power/high-voltage applications compared with other topologies if the highest voltage H-bridge (the main cell) is modulated with the grid frequency (50/60 Hz). With specific modulation indexes, the power that passes through the main cell is more than 80% of the total power. This improves the efficiency and limits leakage current to the grid. However, achieving such a task is challenging using linear controllers and it was not proposed before in the literature using nonlinear controllers. This article presents a novel finite control set model predictive control (FCS-MPC) algorithm suitable for high-voltage/high-power single-stage central inverter connecting photovoltaic (PV) system to the grid. In addition to tracking the reference, multiple constraints are proposed; the first constraint is to modulate the main cell with the grid frequency, and the second constraint is to achieve nonregenerative operation in the lower voltage (the auxiliary) H-bridge cells to ensure the unidirectional power flow for the prerectifier. It should be noted that such constraints using weight factors added to the cost function in conventional FCS-MPC are not applicable to the implementation because the application requires separate independent control signals to individual H-bridge cells. Therefore, the proposed FCS-MPC controller implements the constraints using individual algorithms added to the respective H-bridge cell control signals. The proposed algorithm results in better mymargin THDs in the output current waveform compared with previous schemes. Furthermore, a duty-cycle optimization algorithm is implemented in addition to the conventional algorithms to achieve reduced errors at a relatively lower sampling frequency. The controller is verified and implemented using 2kW hardware prototype for grid-connected operation.