Temporal coherence importance sampling for wind turbine extreme loads estimation

Estimating long return period extreme wind turbine loads is made especially difficult by the large response variability for “the same” environmental conditions. To alleviate this, we have “opened up the black box” of the turbulent wind generation stage of the simulations. Exploiting the notion of “temporal coherence” allows us to manipulate the turbulent inflow to target extreme wind conditions, while at the same time quantifying “how probable these are”. The resulting importance sampling load estimates achieve a significantly lower exceedance probability (i.e., they represent much longer return periods) than estimates using the same number of samples (i.e., the same computational resources) but only a standard Monte Carlo estimate. This paper presents the underlying methodology and some preliminary results. We find that for some loads the method works remarkably well, but for other loads challenges remain.