Posters

Abstract

The accurate estimation of performance degradation in wind turbines is crucial for predicting a wind turbine lifetime energy yield. While the topic received increasing attention over the recent years, a literature review revealed that many studies did not account for important influencing factors like availability, external wake, and changes in the wind farm environment. Considering that literature generally concludes in degradation loss factors significantly beyond the usual industry consensus, there is a need to establish further evidence that the observed trends are actually related to degradation rather than external factors. Additionally, existing large-scale studies are geographically concentrated (most data-rich countries being Germany, the UK, the US and Sweden) and sometimes show policy-influenced patterns (e.g. higher degradation after the end of the support scheme). This study seeks to close both gaps by adding data for data-poor regions (France and Belgium mainly), while excluding losses due to unavailability, and excluding external factors (wake environment, control strategy, tree growth, construction activities), while trying to identify explanatory factors such as the manufacturer, gearbox type, and the site wind resource.   The dataset consists of 12,800 months of data from 215 turbines located in 37 wind farms spread over Belgium, France and Ireland. The samples have varying durations, from 2 to 10 years. The data was filtered to isolate degradation solely due to degradation, excluding changes due to the neighboring environment. The long-term extrapolated 100% availability-corrected production of each turbine is reconstructed on a monthly basis and normalized to the average long-term production evaluated from the first year, yielding a performance index whose evolution is analyzed.    Datasets shorter than 40 months show less significant trends and are responsible for a higher spread on the resulting degradation figures and are discarded.   A fleet-wide analysis reveals a statistically significant average yearly degradation of 0.36% with a highly significant p-value of , which is higher than what is often considered in pre-construction estimates. At a wind turbine level, a statistically significant degradation is observed for more than half of the turbines. Despite correcting for the said factors, the degradation estimate is aligned with other academic results which mostly range from 0.12% to 0.9% per year.   A sensitivity analysis was performed on parameters such as manufacturer, terrain type, drivetrain technology, or long-term capacity factor as an assumed proxy of turbine wear. The resulting reduction on sample size reduces the significance of the analysis. Yet, pairwise comparisons revealed significant differences between some turbine manufacturers and between drive-train types, which opens the door to further classification of losses.   Extending the analysis with more data is however required to draw more significant conclusions on the driving factors to long-term performance degradation, and its distribution through time.