Selective Harmonic Elimination of PUC-5 MLI Using Machine Learning

In this paper, an artificial neural network technique is introduced for the application of Selective Harmonic Elimination (SHE) for a five-level packed U cell inverter. SHE is a low-frequency modulation approach for multilevel converter control and harmonic elimination. The proposed ANN-SHE method involves computing optimal switching angles through a system of nonlinear equations to reduce total harmonic distortion (THD) based on Fourier series analysis. The proposed method is based on a multilayer perceptron algorithm, which is a type of ANN that is well-suited for solving nonlinear problems. The MLP algorithm was trained on a dataset of switching angles and modulation indices, and it was able to learn the relationship between these two variables. The change in switching angles for all possible values of the modulation index could then be estimated using the trained ANN. The paper elaborates the proposed ANN's programming procedures utilizing the MATLAB/Simulink environment. The implemented prototype of inverter tested in real time with various modulation indexes using a DSP embedded board i.e., TMS320F28379D. The simulation results and experimental data are compared and presented, demonstrating the close match between the two. The ANN model offers swift and accurate estimation of switching angles for each of the modulation index, making it an efficient alternative to traditional SHE methods.