Multi-objective selective harmonic mitigation for cascaded H-bridge multilevel inverters connected to photovoltaic systems using hierarchical multi-output support vector regression

A key issue in designing an effective multilevel inverter is to generate high quality output waveform with low-order harmonics to meet grid codes. Selective harmonic mitigation (SHM) fulfills the grid code requirement and eliminates or reduces bulky and costly tuned filters. SHM becomes complex when the dc links of a cascaded H-bridge (CHB) inverter are varied over time (when PVs are used as dc links). This paper presents a new multi-objective selective harmonic mitigation formula for a CHB inverter connected to a photovoltaic system. This multi-objective formulation considers the supply voltage requirements of standard EN 50160 to choose optimal switching angles. To predict optimal switching angles for all possible input intervals, dc link voltages, and modulation index variations, a learning algorithm called the hierarchical multi-output support vector regression is proposed. Simulation results are presented that validate the advantages of the proposed structure.