Posters

Abstract

The increasing density of offshore wind farms amplifies mesoscale wake interactions, reducing regional energy yield and complicating spatial planning. While engineering wake models are computationally efficient, they fail to capture large-scale atmospheric dynamics, whereas numerical weather prediction (NWP) models resolve these effects but at prohibitive computational cost. To address this gap, we propose a weather-driven surrogate model that predicts wind speed deficit fields caused by wake interactions directly from atmospheric conditions, enabling rapid hourly assessments using long-term “no-wind-farm” datasets. The method combines proper orthogonal decomposition (POD) and convolutional neural networks (CNN). Paired NWP simulations with and without wind farms define wake snapshots as wind speed deficits (ΔU = U_farm − U_nofarm). POD compresses these deficit fields into dominant spatial modes and temporal coefficients, reducing dimensionality while preserving wake structure. A CNN maps meteorological predictors—such as wind speed, stability, and turbulence metrics—to POD coefficients, allowing reconstruction of full deficit fields without rerunning NWP simulations. A case study over the Dogger Bank cluster (300°–330° wind sector) demonstrates that the surrogate model accurately reproduces the spatial footprint and evolution of wakes, achieving strong temporal correlations near wind farms. While the model tends to underestimate deficit magnitudes, it reliably captures wake extent and structure, even under varying atmospheric conditions. Performance declines in noisy scenarios, motivating future improvements through spatial weighting and deficit-based filtering. This framework offers a computationally efficient alternative for evaluating mesoscale wake impacts across multiple scenarios, supporting long-term planning and operational strategies. By leveraging atmospheric reference datasets (e.g., ERA5, NEWA), it enables wake-aware assessments without costly NWP reruns, accelerating the integration of wake effects into offshore wind development workflows.