Data-driven Models for Wind Speed Forecasting in Malacca State

Abstract

Wind energy is a type of renewable energy that has received much attention in the electricity market. This study focuses on forecasting wind speed at Malacca state’s Station. This paper proposes using a Group Method of Data Handling (GMDH) model with a hyperbolic tangent transfer function in forecasting wind speed data in Malacca state. The performance of this model is compared with three other models: Artificial neural network (ANN) hyperbolic tangent, ANN log sigmoid and GMDH model. In addition, this study investigates the impact of using different transfer functions in data-driven models for forecasting wind speed. Usually, the application of the GMDH model only implements the second-order polynomial as a partial description. However, in this study, we implement the hyperbolic tangent transfer  function in the GMDH model. The forecasting model’s accuracy is evaluated using mean  absolute percentage error (MAPE), root mean square error (RMSE) and mean absolute error  (MAE). The results show that the GMDH model with hyperbolic tangent is highly accurate with more negligible computational power than other models. The GMDH-hyperbolic tangent  model managed to improve the forecasting performance of the conventional GMDH model by 11.21% in mean absolute percentage error and outperforms other ANN models.