Non-stationary Extreme Value Models with Periodic Change for the Extreme Design Wind Speed

Sheng Dong,Yanchun Li,Mengyuan Wang,Shanshan Tao
DOI: https://doi.org/10.1016/j.oceaneng.2024.118823
IF: 5
2024-01-01
Ocean Engineering
Abstract:The accurate calculation of extreme design wind speeds for offshore wind turbines is crucial. Under changing environmental conditions, extreme wind speeds also exhibit non-stationary variations such as trends, jumps, and cycles. Generalised additive models for location, scale, and shape (GAMLSS) are probabilistic models commonly used for handling non-stationary extreme value sequences. They often treat the location or scale parameters of common probability distributions as functions of time to describe the trends or variability in extreme sequences. In this study, we selected the annual maximum wind speed series during tropical cyclones on Waglan Island in Hong Kong, which does not exhibit a trend or jump component but shows clear periodicity. Our main work focuses on constructing a probabilistic model using GAMLSS that can handle periodic non-stationary variations and accurately estimate design wind speeds under non-stationary conditions using methods such as the expected number of exceedances, design life level, equal reliability, and average design life level. In addition, we compared these calculations with those obtained using time series decomposition and synthesis methods. The findings demonstrate that when a project's service life is characterised by varying cycle intervals, the design wind speeds obtained using both methods exhibit similarities.
What problem does this paper attempt to address?