Improving wind power modelling in the UK through spatially resolved bias correction of reanalysis winds

Reanalysis datasets are widely used in wind energy modelling and power system analysis, particularly in regions where long-term observational records are unavailable. However, their application at wind-farm scale remains challenging, as reanalysis wind fields often exhibit systematic biases linked to simplified physical representations, observational uncertainty, and coarse spatial resolution. In particular, limited spatial resolution restricts the ability of reanalysis data to represent local variability that is critical for accurate wind power simulation.

To address this challenge, we investigate a spatially differentiated bias correction strategy that differs from conventional nationally uniform adjustment schemes. The method adopts a cluster-based representation of wind farm locations, allowing bias correction factors to vary across groups of spatially coherent sites. This framework is applied to a large fleet of UK wind farms to assess its performance under realistic modelling conditions.

Using multi-year operational data from UK wind farms, we evaluate monthly wind power simulations driven by corrected and uncorrected reanalysis winds. The spatially differentiated correction yields reductions in simulation error exceeding 30% relative to baseline ERA5-driven results, demonstrating clear improvements over uniform correction approaches. To assess robustness across reanalysis products, the same methodology is applied to MERRA-2, where comparable performance gains are observed.

Beyond aggregate error metrics, the analysis reveals pronounced regional structure in reanalysis wind speed errors across the UK. Underestimation is most evident in areas of complex terrain, including the Scottish Highlands and mountainous regions of Wales, whereas wind farms located on flat inland plains and offshore sites exhibit relatively minor and more spatially consistent biases. These spatial patterns highlight the importance of accounting for geographic variability when correcting reanalysis wind speeds and demonstrate the value of spatially resolved correction strategies for wind energy applications.