Wind farm efficiency is constrained by wake interactions, where downstream turbines operate in the disturbed air of upwind machines, reducing power output and increasing structural loads. Wake-mixing control techniques address this challenge by dynamically actuating upstream turbines to excite and accelerate wake recovery, demonstrating significant power production gains. However, optimizing these strategies has proven difficult because evaluating both power benefits and fatigue implications requires computationally expensive high-fidelity simulations that accurately capture turbulent flow dynamics and structural response.
Researchers have now developed a practical framework using Gaussian process regression and large-eddy simulations (LES) to overcome this limitation. The approach identifies optimal control parameters and quantifies corresponding fatigue loading through two key advances. First, using LES data, the team optimized frequency and amplitude settings for pitch-based wake excitation, finding maximum power gains of 7.5% at a Strouhal number of 0.25 and pitch amplitudes near 4° in a two-turbine configuration. These results provide actionable guidance for control system design.
Second, the researchers developed a surrogate model capable of predicting fatigue loads across various wind farm configurations and control settings. Rather than relying on conventional engineering wake models, which cannot reproduce periodic excitation effects, the surrogate model integrates rotor-plane inflow characteristics from LES with aeroelastic simulations. The Gaussian process regression approach efficiently captures the relationship between wake overlap, turbine spacing, control parameters, and damage equivalent loads—the critical metric for turbine lifetime assessment.
This dual capability transforms wind farm design by enabling simultaneous evaluation of power gains and load penalties. Operators can now make informed trade-offs, selecting control strategies that maximize revenue while maintaining acceptable structural integrity. The model reduces reliance on expensive full-simulation analysis, accelerating the development and deployment of advanced wind farm control methods. As wake-mixing techniques continue maturing, this framework provides essential decision-support for realizing their economic potential without compromising turbine reliability.
