A novel approach to generate bias-corrected regional wind infeed timeseries based on reanalysis data

Abstract

Weather data, particularly from reanalysis models, are often applied in simulations of infeed patterns for renewable energy systems. The reanalysis datasets provide spatially differentiated weather timeseries for historical years. However, their exactness in wind power applications deserves detailed scrutiny. Notably, the physical model abstracts from boundary layer friction. Abstaining from physical flow models, scientific scholars proposed ex-post bias correction methods to better depict local wind speeds. Yet, such bias correction often is performed on national aggregated figures, as public data is scarce. In this work, a dataset of approx. 23,000 wind turbines for Germany is used to assess deviations between simulated and measured energy infeed for four different years. In line with other studies, we identify in detail that in flat terrain, simulations based on reanalysis data overestimate measured results. In topographically complex regions, a minor overestimation and occasionally an underestimation can be observed. Multilinear regression at turbine level shows that these deviations can be explained by local factors. Reanalysis data in combination with bias-correction based on local factors from 2016 enhance energy output simulations on turbine level on average by 71% for 2020, 93% for 2021 and 97% for 2022.