Previsão da eficiência dos módulos fotovoltaicos utilizando técnicas de aprendizado de máquina

Abstract

The advancement of solar energy as a renewable source of electricity has sparked interest in understanding and predicting the performance of photovoltaic (PV) modules under different environmental conditions. This work aims to predict the efficiency of PV modules using Machine Learning (ML) techniques. Predicting the efficiency of modules is essential for effective planning and monitoring of solar energy systems, as well as a key component in decision-making related to PV generation. Initially, the principles and ML techniques used in module efficiency prediction are presented. Among the explored ML algorithms are Artificial Neural Networks (ANN), Support Vector Machine (SVM), and Linear Regression (LR). The selection of the most suitable algorithm among the explored ones is based on criteria such as performance, interpretability, and scalability. The construction of prediction models is carried out using pre-prepared datasets, which are crucial to ensure the quality and integrity of the data used in model training and validation. The models are trained based on historical data that includes information about environmental variables such as solar radiation, temperature, wind speed, among others, and the corresponding module efficiency. The evaluation, crucial for determining their effectiveness and generalization capability, is based on metrics such as Root Mean Squared Error (RMSE) equal to 1.8148, 2.6412, and 3.2831; Mean Absolute Error (MAE) of 1.0972, 2.2471, and 3.1025; Mean Squared Error (MSE) of 6.9841, 8.5032, and 13.4529; and Coefficient of Determination (R2) with results of 0.8912, 0.8601, and 0.8107. Among the developed ML models, the Artificial Neural Network (ANN) model performs the best, surpassing the performance of traditional models in the literature for predicting PV module efficiency.