Research on optimization design of wind turbine steel tower foundation based on machine learning

Zhichen Liu, Linggang Wei, Jialing Wang, Bin Hu, Yinglong Song, Zuhua Zhang,
Research on optimization design of wind turbine steel tower foundation based on machine learning,
Structures,
Volume 80,
2025,
109840,
ISSN 2352-0124,
https://doi.org/10.1016/j.istruc.2025.109840.
(https://www.sciencedirect.com/science/article/pii/S2352012425016558)
Abstract: Wind power generation is a crucial approach to addressing global warming. The optimization of wind turbine construction costs is essential, as current foundation design calculations are relatively inefficient, prone to errors or overestimation, and require rapid estimation of engineering quantities during the bidding process. Currently, there has been no research utilizing machine learning for the design of wind turbine foundations. This study identified 9 main factors affecting foundation engineering quantities and 4 key output variables representing these quantities. A high-quality dataset with over 200 data entries was compiled. Nine machine learning models were employed, with hyperparameters optimized using random search and Bayesian optimization. Input variables were analyzed via Pearson coefficients and importance ranking, while model performance was evaluated using statistical metrics, goodness-of-fit, and analytic hierarchy process. Results showed that GBDT and CatBoost performed best (e.g., R > 0.95), with rotor diameter, horizontal bending moment, and seismic intensity being dominant factors-consistent with prior research. The study enabled comprehensive foundation quantity computation, along with cost and carbon emission estimates. These findings provide valuable insights for wind turbine foundation design and optimization, supporting the wider adoption of wind power generation.
Keywords: Wind turbine foundation; Machine learning; Engineering quantities; Importance analysis; Analytic hierarchy process