Selective Harmonic Elimination in Cascade Multilevel Inverter with Variable DC Sources Using Artificial Neural Networks

Selective Harmonic Elimination Pulse Width Modulation (SHEPWM) switching strategy is applied for the elimination of low order harmonics in the cascade multilevel inverter with stepped waveform. In this paper, practical implementation of SHEPWM strategy is improved. The case of variable DC sources is investigated and artificial neural network is utilized to perform the calculations. Previous methods have stored the switching angles in a look-up table. The table is one-dimensional and the angles are only function of modulation index. When the case of variable DC sources is considered, it is necessary to make a multi-dimensional look-up table. This look-up table contains a large amount of data without acceptable resolution and also needs multivariate interpolation. The complexity of this method results in harder and expensive implementation of it in practical systems. It is shown that this method can be embedded in the microprocessor systems for real time calculation of the switching angles. In addition, the cost of this hardware is very low. Finally, the results of a prototype system employing the proposed method are presented Copyright (C) 2011 Praise Worthy Prize S.r.l. - All rights reserved.