Abstract
Sørlige Nordsjø II (SNII) is a planned large-scale offshore wind farm in the Southern Norwegian North Sea near the Danish border, with phased development targeting a total installed capacity of up to 3 GW. For projects of this scale, site assessments are typically performed using computationally efficient analytical models. However, these models—often calibrated using data from smaller, sub-gigawatt wind farms—lack the fidelity to represent the complex flow physics in large offshore parks. In particular, they fail to account for the two-way interaction between extensive wind farms and the thermally stratified atmospheric boundary layer, which significantly influences wake behavior and overall energy yield. This study investigates the boundary layer dynamics within the SNII wind farm, focusing on wake behavior and its coupling with the overlying free atmosphere under stable, neutral, and unstable stratification. We conduct a detailed energy/momentum budget analysis to identify dominant energy sources and sinks that govern wind farm performance. A high-resolution large-eddy simulation (LES) is performed for a representative subdomain of the SNII site, consisting of 120 floating wind turbines, each rated at 15 MW. The LES results are benchmarked against three analytical wake models—TurboPark, Niayifar and Porté-Agel, and Bastankhah and Porté-Agel—all of which use Gaussian wake formulations, with turbulence effects explicitly incorporated only in the Niayifar and Porté-Agel and TurboPark models. The comparison reveals power prediction discrepancies ranging from less than 1% to over 56% between the LES and the analytical models, with the smallest differences under neutral conditions, moderate discrepancies in the unstable case (up to 31%), and the largest deviations under stable stratification. Energy budget diagnostics suggest that these differences stem from the wake models' omission of critical physical mechanisms, including geostrophic forcing, pressure-gradient-driven energy input, and turbulent energy dissipation. Finally, downscaling the wind farm configuration in this study from 120 turbines (1.8 GW) to a 1.5-GW layout with 100 turbines—consistent with the planned Phase I capacity—results in only a modest reduction in total energy yield. Across all atmospheric stability regimes and wake models, the reduced layout maintains comparable or slightly improved capacity factors due to diminished wake losses, particularly under neutral and stable conditions where wake interactions are strongest. For example, in the TurboPark model, the capacity factor under unstable conditions increases from 61.4% to 62.1%, while under stable conditions it increases from 72.9% to 73.3% after downsizing. This study suggests that incorporating key physical processes into site design analyses and wake models, combined with strategic layout optimization, can improve predictive accuracy, enhance per-turbine efficiency, and reduce wake losses under varying environmental conditions, providing a viable pathway to optimize early-phase deployment of next-generation offshore wind farms like SNII with minimal loss of total power output.