In current wind turbine design standard, the long-term extreme responses of operational and parked wind turbines with various mean recurrence intervals (MRIs) are estimated from probability distribution of short-term 10-min extreme response under the assumption that the short-term extremes are statistically independent. This study examines the adequacy of this critical assumption through a Monte Carlo simulation procedure. The 10-min mean wind speed series are simulated based on translation process theory with prescribed Weibull distribution and power spectrum. The extreme response series are then generated using the distribution of extreme response under various mean wind speeds. The distribution of annual extreme response is determined from simulated samples and compared to that from distribution of short-term extreme. The results illustrate that the short-term extreme responses can be considered to be independent, while the mean wind speeds exhibit certain level of correlation. This study also presents improved methods for estimating long-term extreme response of parked turbines by using more accurate modeling of distribution tail of mean wind speed. A mixed distribution is suggested, which combines bulk distribution estimated from moderate wind speed data and tail distribution estimated by fitting the excesses above a given threshold with generalized Pareto and three parameters Weibull distributions. A new method of directly using annual maximum wind speed distribution is also proposed that takes into account the independent number of wind speed process in terms of extremal index. The results reveal the importance of better modeling of wind speed distribution tail in the estimation of extreme response of parked turbines.
- Extremal index
- Extreme value distribution
- Generalized Pareto distribution
- Peaks-over-threshold method
- Wind turbines