Characterization of Multimodality in Wind Farm Layout Optimization

The phenomena of multiple optima in wind farm layout optimization (WFLO) problems is investigated. The choice of optimization algorithm and cost of solving WFLO problems is driven by the degree of local optimality in the design space; however little work has attempted to characterize this. Here, an engineering wake model is utilized with a multi‐start gradient‐based optimization approach to determine multimodality for circular wind farm design problems. It is found that the number of local minima increases with more turbines and farm size. However, all local minima found are characterized by either 1) all turbines spread around the perimeter, 2) one turbine in the middle with the remainder spread on the perimeter, or 3) some turbines in the middle with the rest on the perimeter. The first type of optima has superior power production. For the difference layouts there is approximately a 2%–3% difference between the best and worst optima in most cases. Design space maps show that wake interaction creates regions of good and poor performance in the design space that generate the local minima observed.