Harmonic Indices Prediction of LED Lamps Based on Voltage-Dependent Optimized Machine Learning Techniques

Besides the driver circuit structure, the distortion of the applied voltages plays the main role in the field of Power Quality (PQ) emissions associated with power electronic devices, specifically Light Emitting Diode (LED) lamps. Thus, a total of 17 distinct voltage waveforms generated by programmable power supply are applied on samples of 8 LED lamps that have the same circuit topologies and the same wattage. These lamps are collected from different manufacturers within the European Union (EU) markets. This paper aims to analyze the current distortion behavior exhibited by each lamp during various applied voltages, with the ultimate objective of developing Machine Learning (ML) models capable of predicting specific harmonic indices. To assess the effectiveness of the optimized proposed models, namely Deep Neural Network (DNN) and Support Vector Machine (SVM), three distinct scenarios are created using varying data set split ratios.