Here we proposed a condition evaluation method for offshore wind turbines by statistical inference using strain data. To determine the unknown statistical parameters of the data and its confidence level, we first assume the data obey the normal distribution or Weibull distribution, and the unknown parameters are solved by maximum likelihood estimation. Then statistical inference of the distribution with unknown parameters is performed using the Chi-square tests. We find that the monitoring data obey the Weibull distribution at a significance level of 0.05. A method based on statistics for qualitative evaluation of the operation condition of offshore wind structures is proposed at last.


In recent years, the renewable energy industry has developed rapidly, and offshore wind power has made an irreplaceable contribution to the use of renewable energy worldwide. However, the high cost of offshore wind structure design and construction, coupled with the extremely harsh marine environment, results in the high cost of offshore wind power development, which is the biggest defect of the offshore wind power industry. The objective of this paper is to reduce the operation and maintenance cost of offshore wind power through condition monitoring information mining.

In China, the service life of offshore wind turbines is 25 years. During its entire life, the offshore wind turbine condition monitoring system usually generates huge amounts of data, which may be structured or unstructured and may include or lack time information. There are different methods used worldwide to deal with these data. Such as Bayesian methods (Li and Shi, 2012), statistical pattern recognition paradigm (Martinez et al., 2016), component reliability estimations (Scheu et al., 2017), probabilistic surrogate modeling (Singh et al., 2022), etc.

In the field of civil engineering (bridge structure), many scholars have already done some work in structural condition evaluation based on monitoring data. Abdullah et al. (Abdullah et al., 2021) used the Gumbel distribution model combined with strain data to evaluate the reliability of fatigue life of automotive leaf springs under variable amplitude road load strain data; Lei et al. (Lei et al., 2022) used empirical mode decomposition (EMD) algorithm and rain flow counting method for signal preprocessing and statistical analysis of on-site monitoring data, and proposed a method based on strain monitoring data to evaluate the fatigue remaining life of rigid suspension brackets in highway arch bridges; Gulgec et al. (Gulgec et al., 2020) proposed a method based on deep learning to estimate strain response using acceleration data to evaluate the fatigue damage of horizontal bending beams under different loads. Professor Li Zhaoxia (Li et al., 2003) proposed a reliability evaluation method for fatigue life of bridge sections by using strain data on large-span steel bridges and obtaining typical time histories through statistical analysis. Pasquier et al. (Pasquier et al., 2014) proposed a prediction method using data features to reduce the uncertainty of fatigue life assessment, and verified its reliability through experiments and bridge measurements. Ye et al. (Ye et al., 2012) proposed a method for evaluating the fatigue life of steel bridges using long-term dynamic monitoring. By comparing the daily stress spectra of Qingma Bridge on different dates, it was found that under normal traffic and wind conditions, the daily stress spectra of different dates were similar. Therefore, a universal standard daily stress spectrum considering both traffic (highways and railways) and typhoon effects was obtained by proportionally combining the standard traffic stress spectrum and the standard typhoon stress spectrum. The typical dynamic strain response of the structure was determined and the fatigue life of key welding details on the bridge was evaluated. Deng et al. (Deng et al., 2015) evaluated the impact of different strain components on fatigue damage using strain monitoring data from 2006 to 2009 on the Runyang Bridge, and found that the effect of temperature on the stress range spectrum can be ignored. However, a large number of low-stress cycles caused by random interference can lead to incorrect equivalent stress ranges and cycles, resulting in inaccurate fatigue life results.

This content is only available via PDF.
You can access this article if you purchase or spend a download.