Reliability ensemble averaging reduces surface wind speed projection uncertainties in the 21st century over China

Accurate prediction of future surface wind speed(SWS)changes is the basis of scientific planning for wind turbines.Most studies have projected SWS changes in the 21st century over China on the basis of the multi-model ensemble(MME)of the 6th Coupled Model Intercomparison Project(CMIP6).However,the simulation capability for SWS varies greatly in CMIP6 multi-models,so the MME results still have large uncertainties.In this study,we used the reliability ensemble averaging(REA)method to assign each model different weights according to their performances in simulating historical SWS changes and project the SWS under different shared socioeconomic pathways(SSPs)in 2015-2099.The results indicate that REA considerably improves the SWS simulation capacity of CMIP6,eliminating the overestimation of SWS by the MME and increasing the simulation capacity of spatial distribution.The spatial correlations with observations increased from 0.56 for the MME to 0.85 for REA.Generally,REA could eliminate the overestimation of the SWS by 33%in 2015-2099.Except for southeastern China,the SWS generally decreases over China in the near term(2020-2049)and later term(2070-2099),particularly under high-emission scenarios.The SWS reduction projected by REA is twice as high as that by the MME in the near term,reaching-4%to-3%.REA predicts a larger area of increased SWS in the later term,which expands from southeastern China to eastern China.This study helps to reduce the projected SWS uncertainties.

aging structures in the presence of incomplete deterioration information

The performance of an aging structure is commonly evaluated under the framework of reliability analysis, where the uncertainties associated with the structural resistance and loads should be taken into account. In practical engineering, the probability distribution of resistance deterioration is often inaccessible due to the limits of available data, although the statistical parameters such as mean value and standard deviation can be obtained by fitting or empirical judgments. As a result, an error of structural reliability may be introduced when an arbitrary probabilistic distribution is assumed for the resistance deterioration. With this regard, in this paper, the amount of reliability error posed by different choices of deterioration distribution is investigated, and a novel approach is proposed to evaluate the averaged structural reliability under incomplete deterioration information, which does not rely on the probabilistic weight of the candidate deterioration models. The reliability for an illustrative structure is computed parametrically for varying probabilistic models of deterioration and different resistance conditions, through which the reliability associated with different deterioration models is compared, and the application of the proposed method is illustrated.