Abstract

Cascaded H-bridge multilevel (CHB-ML) inverters are an attractive alternative for supplying power to ac grids as they have high reliability and offer an acceptable quality of voltage at their output terminals. In order to achieve efficient operation in these CHB-ML inverters, they must work at low switching frequencies. The finite-control set model predictive control (FCS-MPC) scheme is a very intuitive strategy for controlling this type of converter, but traditional FCS-MPC algorithms generally have a steady-state error when operating at low sampling frequencies and/or if there are parameters mismatch in the prediction model, regarding those of the real system. In this article, a grid-connected CHB-ML inverter that uses an improved FCS-MPC scheme is proposed. The proposed strategy eliminates the steady-state error in an MPC operating at low sampling frequencies and maintains correct operation when a change in the control reference occurs. Experimental results from a grid-connected CHB-ML inverter with three units (seven levels) demonstrate the feasibility of the proposal.