The overall design of large wind farms concerns the selection of turbines, siting of turbine layout and many other engineering decisions. Its optimization holds a great potential for reducing the investment costs and increasing the profit.

Traditionally, wind farm layouts are optimised with physics-based flow models and search & evolutionary algorithms such as Random Search and Genetic Algorithm, which are also a type of AI techniques for solving optimization problems. Many studies have applied machine learning (ML), a subset of AI techniques, to build surrogate models for wind farm flows.

This project will investigate the potential of combining the strengths of ML based surrogate modelling with AI enabled search & evolutionary algorithms, by proposing a framework/methodology to better integrate AI into the workflow of design optimization of wind farms to achieve faster and better results.

The process of building/refining the surrogate model will be integrated with the searching/optimization process guided by the search & evolutionary algorithms to save computational costs and improve optimization results. The research will consider both onshore and offshore applications, with realistic modelling of wind farm costs included.

The fellow will be trained in the application of ML techniques for surrogate modelling of wind farm flows and AI enabled search & evolutionary algorithms for solving complex engineering optimization problems.

A new framework/methodology to integrate AI into the workflow of design optimization of wind farms, which can be generalised to other engineering design problems, will be developed, that can better reduce investment costs and improve profits for future wind farms.

Two secondment periods (3 months). Industrial at EDF UK (Suguang Dou, M21-M23) to improve the cost model of wind farms. Academic at TU-DELFT (Prof. Simon Watson, M30-32) to combine the wind resource assessment techniques developed by DC2 and apply the optimization framework to multiple wind farm design.