HARMONIC SOURCES IDENTIFICATION ON THE POWER GRID WITH SOLAR POWER PENETRATION USING MACHINE LEARNING BASED ON FEATURE EXTRACTION

Abstract

This paper aims to identify and analyze the sources of harmonic distortion in power grids with solar power penetration. The increasing integration of solar power into the power grids presents challenges in maintaining power quality and ensuring system stability. Among these challenges, harmonic distortion caused by the nonlinear operation of photovoltaic (PV) systems, particularly inverters, has become a critical issue. A comprehensive methodology is proposed, using single-point measurement methods combined with machine learning based on feature extraction to accurately locate and quantify harmonic sources at PCC (point of common coupling). The features of the 3-phase current and 3-phase voltage waveform at PCC are saved in image and numerical data by MATLAB/Simulink data generation model. Applying a machine learning (ML) model as a Random Forest Classification (RFC) model will extract the features based on numerical data of waveforms. With current and voltage input at PCC under the single-point method, the RFC model will identify what these waveforms are the same as the waveform of the datasets. From that, classify the characteristics of current and voltage waveform at PCC with the direction, phase angle, amplitude, and order of the harmonics, as well as the voltage amplitude and phase angle of the power grid and solar power source. The results help identify the responsibilities of the involved parties and enhance the effectiveness of harmonic quality management.