Extreme Wind Turbine Response During Operation

Estimation of extreme response values is very important for structural design of wind turbines. Due to the influence of control system and nonlinear structural behavior the extreme response is usually assessed based on simulation of turbulence time series. In this paper the problem of statistical load extrapolation is considered using techniques from structural reliability theory. Different simulation techniques to estimate extreme response characteristics are described and compared, including crude Monte Carlo simulation, Importance Sampling, and splitting methods such as the Russian Roulette and the Double and Clump algorithm. A statistically consistent technique is described for including statistical uncertainty and assessing the extreme 50-year response using simulated time series and conditioned on the model parameters. The peak over threshold method together with the Maximum Likelihood Method provides a tool to obtain consistent estimates incl. the statistical uncertainty. An illustrative example indicates that the statistical uncertainty is important compared to the coefficient of variation of the extreme response when the number of 10 minutes simulations at each mean wind speed is limited to 10.Estimation of extreme response values is very important for structural design of wind turbines. Due to the influence of control system and nonlinear structural behavior the extreme response is usually assessed based on simulation of turbulence time series. In this paper the problem of statistical load extrapolation is considered using techniques from structural reliability theory. Different simulation techniques to estimate extreme response characteristics are described and compared, including crude Monte Carlo simulation, Importance Sampling, and splitting methods such as the Russian Roulette and the Double and Clump algorithm. A statistically consistent technique is described for including statistical uncertainty and assessing the extreme 50-year response using simulated time series and conditioned on the model parameters. The peak over threshold method together with the Maximum Likelihood Method provides a tool to obtain consistent estimates incl. the statistical uncertainty. An illustrative example indicates that the statistical uncertainty is important compared to the coefficient of variation of the extreme response when the number of 10 minutes simulations at each mean wind speed is limited to 10