Performance Analysis of N-Beats and Regression Learners for Wind Speed Forecasting and Predictions

Abstract

Wind energy has enormous potential to fulfil industrial and other power requirement demands specifically in remote areas. The amount of power generated from the wind turbine depends on several factors, namely wind speed, wind direction, rotor area, the height of the tower, etc. As the wind speed is highly dynamic, it highly affects the power generation capacity of the windmills. Thus, it is highly desired that wind shall be monitored as well as forecasted earlier to prevent any sudden ups or downs in the power generation. This manuscript presents a regression-based methodology to predict the wind speed using XGBoost and AdaBoost regression learners. Their learning capabilities have been compared using mean absolute error. XGBoost is found to have lesser value of MAE at 0.392. Parallelly, N-Beats, the time series forecasting model is trained to forecast the wind speed. This way, the present study showcases the utility of time series forecasting method to accurately predict and forecast the wind speed. © 2024, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.