Posters
Come meet the poster presenters to ask them questions and discuss their work
We would like to invite you to come and see the posters at our upcoming conference. The posters will showcase a diverse range of research topics, and will give delegates an opportunity to engage with the authors and learn more about their work. Whether you are a seasoned researcher or simply curious about the latest developments in your field, we believe that the posters will offer something of interest to everyone. So please join us at the conference and take advantage of this opportunity to learn and engage with your peers in industry and the academic community.
PO422: Reducing bias in virtual wind datasets using ensemble correction and machine learning
Jon López de Maturana Echevarria, Head of Wind & Site, Statkraft
Abstract
Mesoscale models and virtual wind datasets are essential tools in early-stage wind resource assessments, particularly when onsite measurements are not yet available. However, they are also subject to large uncertainties: systematic deviations from real measurements of ±30–40% in annual energy production (AEP) are common, typically linked to wind speed biases exceeding 1 m/s. These errors vary by model, country, and site characteristics such as terrain complexity, altitude, roughness, wind directionality, and proximity to existing wind farms or water bodies. We present the Risk Reducer Tool v2.1 (RRTv2.1), a framework that quantifies and corrects systematic biases in mesoscale-derived wind data. The approach was validated against 80 sites across Europe (Norway, Spain, UK, Ireland, Italy), each with reliable onsite measurements. Five industry-standard mesoscale and virtual wind datasets were evaluated, focusing on both energy yield (AEP) and wind speed distribution (Weibull parameters). Our analysis confirms that virtual models tend to systematically overpredict energy levels (mean bias error between +4.3% and +16.3%), with strong regional and model-dependent patterns. By combining machine learning (Random Forest as best performer among other ML algorithms like Ridge Regression, MLP Neural Network, and Support Vector Regressor -SVR-) with an ensemble methodology, RRTv2.1 effectively corrects model bias. After correction, all models showed near-zero bias on average. The ensemble consistently delivered the most robust results: lowest bias, balanced uncertainty and confidence bounds (mean, standard deviation, min, max). While some scatter remains, systematic overestimation risk is largely removed. This enables wind developers to make more risk-aware decisions during early project phases, potentially reducing financial exposure and improving portfolio prioritization.
No recording available for this poster.
