Photovoltaic power predictor module based on historical production and weather conditions data

Abstract

In recent years the demand for electrical energy has increased significantly. Usually, the electrical grid covers this demand. However, this fuel energy is known for its significant carbon footprint. For that reason, different mechanisms to bring cleaner energies have been explored, like hydraulic, wind, thermal, and one of the most popular solar energy. Although solar energy is abundant and environmentally friendly, the photovoltaic energy that comes from the sun, solar production is subject to different external perturbations, such as environmental conditions. Therefore it has been necessary to develop other methods based on statistics, machine learning, or deep learning to make solar forecasting and predict production and weather conditions. Specifically, this work proposes an evaluation of three different deep learning models to predict irradiance, temperature, and production of a photovoltaic system located in the city of Cartagena, Colombia. Those are irradiance and temperature using the historical data on production and weather conditions. This data has been registered on a web platform for seven months, from January 1, 2022, until June 28, 2022.