With the increasing number of wind farms over the last decades and the availability of large datasets associated with them, a growing body of wind energy-related research has shifted towards data-driven and machine learning techniques for modeling, optimization, and control of wind farms.

However, most current data-driven algorithms were developed for relatively simple problems, so the enormous complexity of fluid flows in real wind farms poses unique challenges for data-driven modeling.

Zehtabiyan-Rezaie et al. reviewed a collection of recent studies on data-driven modeling of wind farms and offered a comprehensive overview of available approaches, objectives, methodologies, and utilized data. They outlined outstanding challenges and opportunities.

“Any data-driven models for wind energy applications should ideally be interpretable and have some degree of generalizability,” said author Mahdi Abkar. “The most popular strategy to achieve these goals is to incorporate known physics into models.”

The review revealed most of the studies utilized the purely data-driven approach, which suffers from lack of physical explainability due to its black-box nature. The authors concluded that a shift to physics-guided data-driven models that yield higher accuracy and interpretability in predictions is necessary.

The researchers pointed out the data used in most studies are not publicly available, or the data generation process is not described in a way that helps facilitate further research. They highlighted examples of public datasets from other research fields where data-driven methodologies achieve top performances. They noted that publicly trained models and datasets can advance data-driven research in the growing world of wind energy.

Source: “Data-driven fluid mechanics of wind farms: A review,” by Navid Zehtabiyan-Rezaie, Alexandros Iosifidis, and Mahdi Abkar, Journal of Renewable and Sustainable Energy (2022). The article can be accessed at http://doi.org/10.1063/5.0091980.