Use of Artificial Intelligence and Sizing and Simulation Software in Photovoltaic Plants

Abstract

This academic study investigates the application of sizing and simulation software, such as PV*SOL and PVSYST, to analyze actual data collected from solar plants in Petrolina (PE), Messias (AL), and Piranhas (AL), comparing it with meteorological data from nearby stations. The study's objective is to assess the accuracy and effectiveness of these tools in implementing, testing, and monitoring solar plants. Essential factors in photovoltaic project design include meteorological data, site shading, module orientation, geographic location, temperature-induced losses, electrical components, equipment, and climate change considerations. The analysis covers January to December 2023, using hourly data from reliable meteorological inputs. These software tools aid in system sizing by incorporating multiple factors and estimating energy output, which is crucial to closely matching predicted energy production with actual performance. The quality of meteorological databases and mathematical models impacts software performance, necessitating efforts to filter, qualify, and catalog data sources. Production results indicated an annual output of 3,796 MWh for the Petrolina plant and 1,027 MWh for the Messias II plant, with measured data showing a 5% to 12% variation from estimated figures. Furthermore, the study incorporates Neural Designer, a machine learning-based neural network software, to conduct additional comparative analyses. The findings provide insights into site selection, equipment, plant characteristics, operational practices, and alignment of energy production with software predictions, offering recommendations for improvements and identifying potential locations for future solar farms.