Direct model predictive control of five-level dual flying capacitor active neutral point clamped converters with modified sphere decoding

The five-level dual-flying-capacitor active-neutral-point-clamped converter (5L-DFC-ANPC) is an interesting topology, particularly for high-power photovoltaic, wind power, automotive applications etc. For such topology, direct model predictive control (DMPC), which combines the objective realizations and modulation stages into one computational process, is an attractive method. However, the computational burden to solve the so-called 'mixed-integer nonlinear optimization problem' of DMPC for this 5L-DFC-ANPC topology is very high, because of the great amount of the available switch combinations, particularly for longer prediction horizons. This work applies the modified sphere decoding method for PMSM drive systems with the 5L-DFC-ANPC topology, within the DMPC concept. The computational burden reduction is validated. The current harmonic distortions are seen evidently lower at two step predictions.