Predictive Machine Learning for Advancing PV Energy Integration in Electrical Grids

Abstract

This work explores the application of a machine learning technique to predict both voltage and power variations in electrical grids. The Support Vector Regression (SVR) was selected and its performance is assessed on small-scale (IEEE 4-bus) and large-scale (IEEE 123-bus) networks, focusing on the influence of the epsilon parameter (0.1 & 0.2). Using normalized datasets and Radial Basis Function (RBF) kernel, the model's accuracy in predicting powers and phase voltages for different buses is evaluated. The findings show that SVR with epsilon=O.l delivers higher accuracy, making it a reliable tool for the prediction of PV energy integration impact into electrical grids.