A Wind Speed Forecasting Model Using Bi-Directional LSTM

Abstract

Accurate wind speed forecasting is essential for optimizing wind energy generation, ensuring grid reliability, and supporting meteorological planning. This study presents a wind speed forecasting model based on a Bi-Directional Long Short-Term Memory (Bi-LSTM) network to capture complex temporal dependencies in time series data. The Bi-LSTM model was trained on historical wind speed data and tested across various forecasting schemes with several evaluation metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Coefficient of Determination (R²). The result in the 24-hour forecast shows that the model achieved an MSE of 0.0454, RMSE of 0.2131, MAE of 0.1700, MAPE of 0.0443, and R² of 0.7103. On the 48-hour forecast, the model recorded an MSE of 0.0326, RMSE of 0.1807, MAE of 0.1353, MAPE of 0.0378, and R² of 0.8564. The 72-hour forecast achieved an MSE of 0.0290, RMSE of 0.1702, MAE of 0.1302, MAPE of 0.0336, and R² of 0.9429. The model demonstrated strong generalization on the full testing set with an MSE of 0.1432, RMSE of 0.1959, MAE of 0.1423, MAPE of 0.0708, and R² of 0.9810. These findings highlight the Bi-LSTM model's effectiveness and robustness for wind speed forecasting, offering a valuable tool for wind energy management and operational planning.