Hybrid model to improve wind energy prediction considering data granularity

Abstract

The growth of wind energy generation as a renewable source in the transition to sustainable energy poses significant challenges in ensuring reliable production forecasting due to the intermittent nature of wind resources. The implementation of advanced forecasting models has become a priority to optimize its integration into the power grid and ensure the stability of the energy supply. This study focuses on improving wind energy predictions through the use of advanced machine learning techniques. The methodology includes a detailed analysis of different forecasting time horizons, sampling rates, and exogenous variable configurations, comparing traditional models such as SARIMAX, XGBoost, and LSTM neural networks with hybrid approaches. Furthermore, performance metrics such as R2, MAE, RMSE and MAPE are evaluated to assess the accuracy and reliability of the proposed models. The results demonstrate that the selection of the optimal model depends on the forecasting horizon, data granularity, and available resources, maximizing precision and efficiency for each scenario.