A joint optimization framework for power and fatigue life based on cooperative wake steering of wind farm

Reducing the cost of wind energy relies on maximizing power production and extending the service life of wind farms by decreasing fatigue load. This paper offers new insights into multi-objective optimization by integrating an accurate and efficient ANN-based power and lifetime prediction framework into a novel double-stage control scheme, extending the service life while maintaining wind farm power production to meet energy demand. By combining the ANN yawed wake model with Bayesian machine learning, the novel double-layer machine learning framework strikes a balance between accuracy and efficiency for power and lifetime optimization in their respective stages. The power tradeoff benefits of the proposed scheme are assessed in a simplified 5-turbine row, alongside a parametric analysis of inflow conditions. Results demonstrate that the proposed framework can effectively extend wind farm lifetimes with minimal power tradeoff, achieving nearly an 8 % lifetime extension at the expense of less than 2 % power reduction. Incorporating the load effect into multi-objective yaw control enhances the positive impact of turbine yaw in the back row compared to considering power alone. Aside from the decline in the initial and near-critical stages, the optimal lifetime extension fluctuates slightly with the power tradeoff. The initial stable point can serve as a reference for determining the optimal allowable power reduction coefficient in multi-objective cooperative control. This double-stage scheme demonstrates significant practical potential for general wind farm control optimization, effectively balancing the core objectives of maximizing power generation and extending turbine lifetime. Meanwhile, the ANN-based wind farm predictions bridge the gap in rapid and accurate evaluation of wind turbine power and lifetime during the control phases.