Methods for extrapolating extreme loads to a 50 year probability of exceedance, which display robustness to the presence of outliers in simulated loads data set, are described. Case studies of isolated high extreme out-of-plane loads are discussed to emphasize their underlying physical reasons. Stochastic identification of numerical artifacts in simulated loads is demonstrated using the method of principal component analysis. The extrapolation methodology is made robust to outliers through a weighted loads approach, whereby the eigenvalues of the correlation matrix obtained using the loads with its dependencies is utilized to estimate a probability for the largest extreme load to occur at a specific mean wind speed. This inherently weights extreme loads that occur frequently within mean wind speed bins higher than isolated occurrences of extreme loads. Primarily, the results for the blade root out-of-plane loads are presented here as those extrapolated loads have shown wide variability in literature, but the method can be generalized to any other component load. The convergence of the 1 year extrapolated extreme blade root out-of-plane load with the number of turbulent wind samples used in the loads simulation is demonstrated and compared with published results. Further effects of varying wind inflow angles and shear exponent is brought out. Parametric fitting techniques that consider all extreme loads including ‘outliers’ are proposed, and the physical reasons that result in isolated high extreme loads are highlighted, including the effect of the wind turbine controls system. Copyright © 2011 John Wiley & Sons, Ltd.
- Extreme loads
- Cumulative distribution functions
- Parametric fitting
- Wind turbulence
- Principal component analysis and correlation matrix