Machine learning model for wind direction and speed prediction

The capacity to estimate wind direction and speed is essential for both the generation of renewable energy and the forecasting of weather. Near the ground, the performance of the mechanistic models that are the foundation of conventional forecasting is quite low. We will explore a different data-driven strategy that is based on supervised learning. We train supervised learning algorithms utilising the previous history of wind data. We use data from individual locations and horizons to conduct a systematic comparison of a number of algorithms, during which we change the input/output variables, the amount of memory, and whether or not the model is linear or nonlinear. According to our findings, the ideal design as well as the performance of the system varies depending on the region. Our technique achieves an improvement in performance of 0.3 m/s on average when it is applied to datasets that are accessible to the public. © 2024 Inderscience Enterprises Ltd.