Quantifying and clustering the wake-induced perturbations within a wind farm for load analysis

Abstract

Abstract The main objective of the HIPERWIND project is to reduce the Levelized Cost Of Energy by controlling the uncertainties in a chain of numerical models simulating an offshore wind farm. This paper studies the wake-induced perturbations of the ambient wind conditions inside a wind farm. A theoretical floating offshore wind farm off the coast of South Brittany in France is considered as a test case, and a numerical model simulating the wake of this farm is exploited. This model gives an analytical prediction of wake deficit and added turbulence, also considering the influence of the floaters’ position due to mean wind forces. An uncertainty propagation on the wake model is performed, considering the multivariate joint distribution of ambient wind conditions as input. This results, for each turbine, in a wake-modified probabilistic distribution on the wind speed and turbulence intensity averaged over the rotor. A new approach for clustering these wake-modified distributions is proposed in this paper to regroup the wind turbines exposed to similar environmental conditions, leading to a similar structural response. This analysis is essential to ensure the practical feasibility of the reliability analysis of an overall project with the computational cost of aero-servo-hydro-elastic simulations of offshore wind turbines. Several methods classically used for clustering are compared. Finally, four clusters only are sufficient to represent the whole set of turbines in the wind farm.