Hybrid Time-Series Forecasting with VMD and LogTrans for Wind Energy Applications

Accurate Wind Speed Forecasting (WSF) is important for managing the wind energy, improving the turbine performance, and maintaining the power grid stability. This study proposes a new hybrid model that combines Variational Mode Decomposition (VMD) with Log-Sparse Transformer (LogTrans), a Transformer-based Deep Learning (DL) model that uses log-sparse attention to better capture the patterns in time-series data. VMD is used to break down the original Wind Speed (WS) signal into several simpler components, making it easier for the forecasting model to learn meaningful patterns. These components are then processed by the LogTrans model to make accurate predictions. The model is tested using WS data collected from Garden City, and its performance is evaluated at 5-min, 10-min, and 30-min forecasting intervals. The results show that the proposed VMD-LogTrans model gives more accurate and reliable predictions than benchmark DL models and other hybrid approaches. The proposed model offers an efficient solution for short-term WSF and can help improve the integration of wind energy into the power systems.