Adaptive Fossa Search with Tiered Surrogates for Wind Blade Aerodynamics

Wind turbine blade design requires expensive Computational Fluid Dynamics (CFD) simulations to evaluate aerodynamic performance, making conventional optimization prohibitively costly. Surrogate-assisted metaheuristics offer a promising alternative, yet most approaches rely on single-fidelity approximations that struggle to balance accuracy and computational efficiency. This study proposes a Multi-Fidelity Surrogate-Assisted Fossa Optimization Algorithm (MF-SAFOA) for wind blade aerodynamic design. The method integrates three fidelity levels: XFOIL for rapid 2D airfoil analysis (low-fidelity), QBlade for Blade Element Momentum evaluation (medium-fidelity), and OpenFOAM for 3D RANS simulation (high-fidelity). An adaptive Kriging surrogate model is constructed from low- and medium-fidelity samples, while an Expected Improvement infill criterion selectively triggers high-fidelity evaluations only in promising regions. The Fossa Optimization Algorithm explores the design space using opposition-based learning and a memory mechanism, with repair operators enforcing structural and geometric constraints. Validation on the NREL 5-MW reference turbine demonstrates that MF-SAFOA achieves 97.3% of the Annual Energy Production (AEP) obtained by pure high-fidelity optimization while reducing CFD calls by 78.4%. Compared to single-fidelity surrogate methods and standard FOA, the proposed approach converges 2.1× faster with superior solution quality. Ablation studies confirm that the multi-fidelity strategy outperforms both low-fidelity-only and random high-fidelity infill approaches. The MF-SAFOA framework provides an efficient, license-free pathway for high-dimensional wind blade optimization, making expensive CFD-based design accessible for academic and industrial applications.