Data-driven wind farm flow control and challenges towards field implementation: A review

Abstract

Data-driven wind farm flow control (WFFC) is an innovative approach that leverages the collected data and advanced analytics to enhance the performance of wind turbines within wind farms. Its significance lies in its ability to adapt to changing wind and turbine conditions and improve operations, boosting energy yield, extending turbine/component lifetime, and potentially reducing socio-environmental impact and costs, thus supporting the viability and sustainability of wind energy as a renewable power source. This review explores the dynamic field of data-driven WFFC and its challenges towards practical implementation. Building on top of traditional wind farm modelling and model-based control, it details the virtues and limitations of these methods while introducing the concept of data-informed or data-driven flow models that harness data to augment predictive accuracy and control strategies. The analysis then covers the methodologies for power and load surrogates, elucidating the pivotal role of surrogate modelling in enhancing WFFC, and showcasing its value in decision-making and energy optimisation. Furthermore, the growing field of reinforcement learning (RL) is highlighted, showcasing its adaptive potential to revolutionise wind farm control through learning from past interactions. The investigation concludes by identifying key challenges impeding the practical deployment of data-driven WFFC, including data quality concerns, cybersecurity risks, and limitations of the current algorithms. In summary, this comprehensive review presents the ongoing development of data-driven WFFC, emphasising the synergy between traditional methods, surrogate modelling, RL, and the critical challenges to be addressed for successful integration of these methodologies in real-world wind farm operations.