Solar panel energy production forecasting by machine learning methods and contribution of lifespan to sustainability

Abstract

The struggle to protect the atmosphere and the environment is increasing rapidly around the world. More work is needed to make energy production from renewable energy sources sustainable. The integration of energy with machine learning provides numerous advantages. In this study, the solar energy system, which is one of the main renewable energy sources, is considered. Support Vector Machine (SVM), K-nearest neighbor, Random Forest, Artificial Neural networks, Naive Bayes, Logistic Regression, Decision Tree, Gradient Boosting, Adaptive Boosting, and Stochastic Gradient Descent are used to forecast energy production. Forecast experiments are conducted in a region with high solar radiation and high temperature. Thus, there is an opportunity to examine overheated solar panels as well. A small-scale but adequate weather station is installed right next to the solar panel. Inputs such as temperature, pressure, humidity, and solar radiation obtained from the atmosphere with sensors are used. Obtained data are processed utilizing an Arduino microcontroller, data are recorded with C# software, and machine learning training is performed using Python programming. According to the results, the best performance is provided by SVM. This study provides guidance on whether solar energy systems investments are appropriate in the relevant region.