Previsão de geração de energia elétrica térmica convencional: um estudo de caso usando regressão linear e redes neurais artificiais

Abstract

This study aims to analyze the performance of two time series forecasting models in estimating conventional thermal electricity generation. The models analyzed, provided by the WEKA software, include Linear Regression (LR) and Artificial Neural Networks (ANNs). The analysis was performed using a dataset on thermal electricity generation from January 1996 to June 2023, provided by the Institute of Applied Economic Research (IPEA). We used the Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE) metrics to measure the performance of the models. These metrics allowed a detailed comparison of the performance of each approach. Our results indicated that the MLP Neural Network obtained the best performance, standing out as the most efficient approach for this dataset.