Machine Learning Based Sensorless Current Prediction of a Multilevel Inverter

Abstract

Renewable energy, transportation, and the power industry have all shown a great deal of interest in multilevel inverters. The inverter's power rating can be increased without necessitating higher ratings on individual devices by increasing the number of voltage levels it has. This reduction in harmonic distortion increases as the number of voltage levels increases and becomes more significant. However, in practice, there are variations that are inflicted on the output waveform and these distort the waveform as well as the poor power quality. This paper shows that it is possible to enhance the functionality of multilevel inverters by using machine learning algorithms. Because of these algorithms one could have accurate and effectual ways of enhancing the efficiency of the inverter. In this paper, machine learning algorithms which includes Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and MultiLinear Regression (MLR) were applied to the MATLAB data set of a three-phase, three-level diode clamped multilevel inverter with RL Load. In contrast, multilevel inverters are much more advanced power conversion instruments than the conventional inverters and these are capable of providing more flexibility in controlling the process of conversion of DC into AC. Multilevel inverters facilitate generation of an AC waveform which can be a sine wave or steady wave and the standard inverters allow only for a square wave or pulse wave. This is achieved through the use of many DC voltage levels available in the device, which minimize distortion or what could be referred to as harmonics and improve the quality of the power delivered.