Wind Farm Wake Steering Control via Deep Reinforcement Learning

Offshore wind farms frequently encounter challenges such as slow wake recovery and the regular arrangement of wind turbines, resulting in significant wake effects and substantial power generation losses. Wake steering control, which adjusts the yaw angles of upstream turbines to shift the wake center away from downstream units, represents an effective approach to enhancing the economic performance of wind farms. However, existing methods exhibit sensitivity to initial conditions and cannot address environmental uncertainties. This paper introduces a wake steering control model for wind farms based on deep reinforcement learning (DRL). The model achieves online optimization by combining offline pre-training with online fine-tuning and conducts online learning through continuous interaction with the environment. Experimental results demonstrate that this approach effectively handles uncertainties in wind farms and holds significant prospects for engineering applications.