Solar PV Power Prediction System Based on Machine Learning Approach

Abstract

Electricity power is an essential need for the development of technology and industries. It is an essential component of modern human life. Using fossil and other petroleum components for electricity generation harms the environment. Many countries and industries have started using renewable sources like solar systems for electrical power generation. However, the main inconvenience of these systems is that they are unpredictable. The purpose of this work is to develop a machine learning-based method to estimate the generated power of PV solar systems based on environmental data such as sun irradiation, wind speed, and others. Before implementing Machine Learning techniques, the built system goes through a feature selection stage to identify the most influential parts. This step improves system performance by removing unnecessary data. Only three ML approaches were used and compared: CNN (Convolutional Neural Network), SVR (Support Vector Regression), and (RF) Random Forest. The SVR outperforms the competition.