Determination of output current THD of multilevel inverter by ANN

Abstract

In recent years, the use of multi-level inverters in power electronics applications has been increasing. The most important advantage of the multi-level inverter circuits compared to the classical two-level topology is that the output voltage is closer to the sinusoidal waveform, operating at a lower switching frequency and high efficiency. Also, low Total Harmonic Distortion (THD) current waveforms occur at the output of multi-level inverter circuits. In this study, a three-phase five-level cascaded inverter is designed by simulating it in MATLAB-Simulink. It is aimed to predict the THDi value of the load current of the inverter. By changing the parameters such as switching frequency, modulation index and load value in the designed multi-level inverter circuit, a dataset is produced for the THDi value of the load current by Fast Fourier Transform (FFT) analysis with MATLAB-Simulink. The generated dataset is used in the training and testing phase of the Artificial Neural Network (ANN) system, which is an artificial intelligence technique. Thus, the THDi value of the load current is estimated. In this way, estimating the parameter determines the inverter circuit with the least THDi before the experimental implementation. At the final estimation phase, the average percent error (APE) value of the current value for the test is obtained as 0.5177 % and 0.5718 %, respectively.