Neural Network Based Selective Harmonic Elimination with Improving the THD in Low Modulation Indexes

Multilevel converters are on the state-of-the-art in power conversion systems due their improved voltage and current waveforms. Selective Harmonic Elimination Pulse Width Modulation (SHEPWM) switching strategy is commonly applied for the elimination of low order harmonics in the multilevel converter with stepped waveform. In this paper, a new approach of this switching algorithm is utilized to a seven-level cascaded converter to produce the required fundamental voltage and in the same time cancel out specified order harmonics. In this paper to solve SHE nonlinear transcendental, resultant theory has been used. Result angels obtain from resultant theory are trained to neural network, since the neural network is trained, gives the best angles for the entire modulation index. In the seven-level converter, the switching angles can be chosen to produce the desired fundamental output while making the fifth and seventh harmonics identically zero. But the main drawback is that for some ranges of the modulation index (m) there are not any solutions in resultant theory, to overcome this problem and improving the THD at output voltage in lower modulation indexes a DC−DC buck converter has been used to have adjustable dc source in input of converter to coordination between modulation index and output voltage. The simulation results have been carried out using SIMULINK/MATLAB present the effectiveness of the SHEPWM strategy for the proposed converter. In addition, the experimental results of proposed topology prototype have been presented to validate its practicability.