Surrogate-assisted power optimization framework for heterogeneous Wind farms under turbine upgrade scenarios

With the progressive retirement and retrofitting of aging wind turbines, heterogeneous wind farms are increasingly emerging worldwide. Owing to land-use constraints and economic considerations, many wind farms adopt in-situ replacement strategies in which outdated turbines are replaced by larger-capacity units at existing locations. This practice, however, poses significant challenges to both power prediction accuracy and large-scale optimization efficiency due to the intensified wake interactions induced by turbine heterogeneity. To address these challenges, this study first introduces a non-uniform three-dimensional (3D) wake modeling framework, which provides a more reliable basis for power evaluation in heterogeneous wind farms. Subsequently, a Random Forest surrogate-assisted adaptive particle swarm optimization (RFSA-APSO) method is developed, in which a surrogate-guided particle selection mechanism and a center perturbation replacement strategy are jointly employed to effectively alleviate the high computational cost and premature convergence commonly encountered in high-dimensional wake optimization problems. The proposed optimization framework is further applied to in-situ turbine replacement scenarios, and comprehensive numerical investigations are conducted on three representative wind farm layouts, including a single-row configuration (1 × 8), a regular grid (4 × 4), and a large-scale staggered layout (7 × 7). The results demonstrate that the proposed framework can effectively mitigate wake-induced power losses under various wind conditions, achieving total power improvements of 5.78%–9.38% compared with the individual optimum operation mode. Further sensitivity analyses indicate that turbine replacement strategies exert a pronounced influence on both the baseline power production and the achievable optimization potential. These findings highlight the importance of physics-informed and site-specific decision-making in the planning and upgrading of heterogeneous wind farms.