Historically the design of Offshore Wind Turbines (OWT) depends on various influence factors which are set by engineers. Typically, most parameters are determined during the initial design phase and have less consideration on the influences of the entire life-cycle of a structure, leading to the over-exploitation of a single design, like monopiles. In this work, we define the design process as a multi-objective optimization problem and use Artificial Intelligence (AI) to discover multiple optimal solutions, while providing feedback, in the form of feature importance explanations for generated structures. Our approach results in efficient designs, while explanations can improve engineers' understanding of alternative design possibilities.


Offshore wind turbine design is a complex problem governed by numerous influence factors like site-specific boundary conditions, industry standards, costs, etc. We focus on the design of the substructure of an OWT, as the principal differentiator to onshore wind energy and one of the main cost drivers, which accounts for up to one-third of the total cost (Wang, 2018) and offers therefore considerable scope for cost reduction (Perez 2015). Traditionally, the design of such a structure is based mainly on the operational behavior of the structure, the status quo of commonly used designs, and the experiences/preferences of the engineers.

As a result, the substructure space is dominated by a few designs, mostly monopiles or four-legged X-brace jackets (4C Offshore, 2022). Exploration of new designs is ongoing in research and industry fields, such as the twisted jacket form (Chen, 2016), but the progress is slow due to the complexity of constraints considering variable load cases. Furthermore, such conventional approaches rely on industry standards and guidelines, such as standard BSH (BSH, 2007) and DNV-OS-J101 (Veritas, 2014), in which the requirements considering all phases of the life-cycle of a structure (e.g. manufacturing, maintenance, decommissioning, etc.) are still under development. This issue leads again to limited design options.

This content is only available via PDF.
You can access this article if you purchase or spend a download.