Parameter Optimization of floating foundations for Offshore Wind Turbines Based on Machine Learning

With the increasing global demand for renewable energy, offshore wind power generation has attracted great attention, especially the development of floating offshore wind turbines .Offshore floating wind turbine foundations involve multiple key dimensional parameters that strongly interact and influence the system's extreme motion response. Traditional optimization methods struggle to handle these complex couplings, necessitating AI-driven approaches like neural networks for intelligent parameter optimization. However, limited dataset availability in this emerging field poses a challenge. This paper proposes a small-sample-based optimization method for wind turbine foundation design. The HexaSemi-submersible FOWT was optimized by tuning four key parameters :column spacing (L), platform draft (T), column diameter (D), and caisson height (h). OpenFAST simulated motion responses, followed by a BP neural network modeling parameter-displacement relationships. Coupled with genetic algorithms, this approach reduced platform displacement by 16.03% versus the original design configuration.随着全球可再生能源需求持续攀升，海上风力发电已获得广泛关注，其中漂浮式海上风力涡轮机（Floating Offshore Wind Turbine, FOWT）的发展更受瞩目。海上漂浮式风电基础包含多个关键尺寸参数，这些参数间存在强烈的交互耦合作用，会对系统的极端运动响应产生显著影响。传统优化方法难以应对这类复杂的耦合关系，因此亟需借助神经网络等人工智能驱动的手段开展智能参数优化。但该新兴领域面临数据集规模有限的挑战。本文提出一种面向风电基础设计的小样本优化方法。以六立柱半潜式漂浮式风电基础（HexaSemi-submersible FOWT）为研究对象，通过调整四项关键参数：立柱间距（L）、平台吃水深度（T）、立柱直径（D）以及沉箱高度（h）进行优化设计。首先通过OpenFAST软件模拟运动响应，随后构建BP神经网络（Backpropagation Neural Network）以建模参数与平台位移间的关联关系；结合遗传算法后，该优化方案相较初始设计构型，将平台位移降低了16.03%。