Solar Radiation Forecasting in Photovoltaic Systems Utilizing Machine Learning Techniques

Abstract

This research responds to the growing demand for renewable energy, focusing specifically on photovoltaic systems that harness solar energy as a viable and sustainable solution. The methodology implemented included the analysis of hourly solar radiation data collected during the period 2017-2023. These data were fundamental to make predictions and validate the algorithm used. The purpose of these predictions was to optimize the sizing of a photovoltaic system appropriate for an urban area. For this purpose, a decision tree algorithm was employed within the machine learning technique, using Python software due to its accessibility. The results were stored in an .xlsx file, which simplified the system sizing process. In addition, standard deviation calculations were incorporated to estimate the solar radiation over the next three months, thus allowing an accurate and adequate calculation of the required PV system. In conclusion, the designed PV system was efficiently sized based on the predictive analysis provided by the algorithm. With a peak power of 1.26 kWp and a well-adapted storage configuration, this system is equipped to meet the daily energy demands of 123.5 kWh.