Wind speed forecasting using a hybrid model considering the turbulence of the airflow

Abstract

Listen

Translate article

This work presents the conformation of a hybrid model aimed at forecasting the wind speed one step ahead. The model comprises the Multiplicative Cascade and Persistence. To couple both techniques, a transition point was used, in this case, 6.6 m/s, defined from the spectrum of wind speed near the ground. The Multiplicative Cascade was used when speeds were higher than the transition point, while Persistence was used when lower. Two time series measured in Tamaulipas and Veracruz, Mexico, were used to build the hybrid model. The performance of the individual models concerning the hybrid model was compared, finding that the adjustment of the hybrid model exceeds that of the Multiplicative Cascade and Persistence models for both series. Different error metrics were used in the quantitative analysis, where the hybrid model was the most competitive. For example, in Mean Absolute Percentage Error, the difference in performance between the hybrid model and Persistence is 2.72% points for the 10-min time series. In contrast, this difference is 27.89% points for the hourly time series. Finally, this work shows that for the cases analyzed, differentiating the effects of turbulent from non-turbulent flows and modeling with the proper techniques improves forecasting accuracy.