A GA-Neural Network Harmonic Minimization Method for Multilevel Inverters with Unequal DC Sources for Different Modulation Index Values

The paper presents a real-time algorithm to compute the switching angles to control a three-phase multilevel cascaded inverter in order to minimize the THD to improve the EMC of the system. In particular, the proposed method uses an Artificial Neural network, trained by the GA algorithm, to identify the optimal switching angles corresponding to several values of DC sources voltage levels and the modulation index. Moreover, the algorithm searches the best sequence of switching angles to fire the different DC sources of the inverter to work in the condition of minimum Total Harmonic Distortion (THD). The method is validated by simulation results.