Weight factor selection for model-based predictive control of a four-level flying-capacitor inverter

Flying-capacitor (FC) converters need, besides an output current control, a control of the FC voltages to be able to produce appropriate output voltage levels. The output currents and capacitor voltages can be actively controlled by model-based predictive control (MBPC). This MBPC has several control parameters such as weight factors in the cost function. The selection of these weight factors and other control parameters of MBPC is not trivial and needs an objective criterion. A method using mean square error (MSE) values of the controlled variables is proposed. The feasibility of the MBPC and the MSE analysis method for multilevel inverters is demonstrated by experiments. The good correspondence between simulations and experiments clearly shows that the results from a design analysis of MBPC in simulation is applicable on the practical setup. Two models in the prediction step of the MBPC are discussed in this study: using a simplified model, where the star-point voltage is neglected, and using an unsimplified model. The simplification, although reducing the computational requirements, results in suboptimal control quality and more importantly a less robust control. The experiments validate the conclusion that using the unsimplified model extends the range of good weight factors and improves the control quality of the current and capacitor voltages.