Two Artificial Intelligence methodologies are validated to improve wind speed predictions for wind farms

Last year, wind energy accounted for 23.2% of all energy injected into the Spanish electricity system, according to data published by Red Eléctrica in its latest 2024 report. Although wind power leads national energy production, its dependence on weather conditions and its inherently intermittent nature present challenges. Therefore, fine-tuning wind speed prediction data for these infrastructures is a key task to optimize the management and performance of wind turbines.

This is precisely what the AYRNA group at the University of Córdoba (UCO) has proposed, using Artificial Intelligence to help fill the sails of wind power, as it were. The team has confirmed two methodologies trained on over 13 years of data, capable of predicting extreme speeds with greater accuracy than traditional methods, using variables such as wind components at different altitudes, pressures, and air temperatures.

Both systems are based on artificial neural networks, inspired by the human brain, and ordinal classification systems, which categorize wind speeds from lowest to highest intensity, rather than predicting specific speeds.

As explained by researcher Antonio Gómez, with the Department of Computer Science and Artificial Intelligence at the UCO, both methodologies have been trained to forecast four different wind speed ranges—low, moderate, high, and extreme—with time horizons of 1, 4, and 8 hours. Each of these categories is associated with not only a specific wind speed range, but also an estimated range of wind energy production.

While the first model performs similarly across all four wind classes, the second excels with more severe events, notes David Guijo, another author of the study. In fact, for gusts exceeding 20 meters per second, which fall into the extreme wind category, the system outperforms traditional methods and can predict speeds with over 94% accuracy. This is particularly valuable for anticipating extreme wind events, allowing turbines to be shut down to prevent damage or collapse.

"Energy companies must periodically estimate the energy they will put on the grid, which underscores the need to refine forecasts for optimal predictions," emphasizes researcher Pedro Antonio Gutiérrez. He notes that while both systems can be extrapolated to different wind farms with relative ease, the models were trained on a specific farm featuring particular conditions. Therefore, applying them to other settings would require retraining and validation.

This work, conducted in collaboration with researchers in the Department of Signal Theory and Communications at the University of Alcalá, is part of the national NEXO research project. The project aims to develop Artificial Intelligence models for applications to renewable energy, various meteorological events, and the field of medicine.