Predictive control approach for multicellular converters

The classic methods for controlling power converters based on average model allows a good control of the transitory state. However, the steady state (waveform, subharmonic, etc) is not always completely controlled. This article shows how to obtain an optimal periodic cycle from an average reference in currents and voltages. The optimal cycle is then used as a steady state reference for a closed loop predictive control. Moreover, the real time implementation is ensured by a neural network. Simulations and experimental results for a four-level three-cell converter verify the performance of the method.