Assessment and optimisation of regional scale wind farm deployment using machine learning.

The impact of inter-farm wakes is a growing issue as offshore wind is scaled up to meet renewable energy needs. High-fidelity simulations which capture such wake effects under potential future build-out scenarios are required to enable regional-scale planning which can mitigate wake impacts. Here, we present a machine learning-based workflow for estimating power losses due to inter-farm wake effects, suited to efficient analysis and optimal planning of future build-out. We apply this tool to the assessment of planned build-out in the North Sea. We estimate that percentage power losses due to inter-farm wakes will more than double compared with their current level, reaching 2.4%, and that increased losses in summer will exacerbate natural seasonal variability in resource. Our tool also facilitates sensitivity analysis and optimisation of wind farm fleets with respect to a variety of design choices. In this work we optimise total fleet power output with respect to small adjustments in future farm locations, finding that wake-induced losses can be reduced by one third via careful spatial planning, corresponding to annual economic gains of £160m compared with current plans.