Short-term solar power forecasting using different machine learning models

Abstract

Pronosticar la producción de energía solar fotovoltaica se ha vuelto cada vez más relevante durante la última década. En este trabajo se lleva a cabo el entrenamiento y la evaluación de diferentes modelos de Machine Learning para generar predicciones de potencia de corto plazo, basadas en la potencia actual y variables meteorológicas. Se usaron datos de tres ubicaciones geográficas diferentes en el mundo

Photovoltaic power production forecasting has become increasingly relevant over the past decade. In this study, training and evaluation of different Machine Learning models is performed in order to generate short-term predictions based on current power production and weather variables including temperature, relative humidity, wind speed and direction, cloud cover, and direct radiation. Three different datasets were used to train each model, from three different geographic locations in the world