Unlocking Precision in Wake Modelling

The Critical Crossroads in Wind Energy Assessment

Wind energy, especially in large offshore and high-density onshore farms, is no longer just about optimising performance; it's about mastering wake physics.

Wake losses—where downstream turbines operate in the reduced wind speed wake of upstream units—can be on the order of 10% to 35% of a project's total energy yield in high-density farms.¹ With wind farm development demanding an increasingly fast turnaround, an unexpected systematic performance loss can eliminate the expected profit from an entire project pipeline.

This critical reality forces us to ask a first-principles question:

Are the core tools we use to model multi-billion-dollar assets fit for purpose?

The Problem with the Empirical Legacy

Many engineering tools still rely on analytical single-turbine wake models, often based on approaches from the 1980s, which treat each turbine's wake independently. These models simplify the complex interaction of multiple wakes by using approximate superposition models—a mathematical construct that combines individual wake predictions.²

While this legacy approach has proven useful for decades and remains computationally fast, it faces significant limitations today. This is especially true in complex modern wind farms characterised by closer spacing, large turbine diameters, and non-neutral atmospheric conditions (like stability effects).

To cope with these factors, analytical models require an increasing number of empirical corrections and tuning parameters that are highly site-specific. Research shows that model predictions are particularly sensitive to errors in these tuned parameters. For instance, the choice of superposition method can result in relative differences on the order of 100% for key parameters like the wake expansion coefficient.³

The inherent challenge is that while well-calibrated empirical models can sometimes achieve mean absolute errors under 15% in overall array efficiency predictions, this accuracy is often achieved through error compensation at the individual turbine level.

In other words, errors in wake-deficit modelling for upstream turbines are masked by compensating errors downstream, making the model unreliable for novel designs or layouts. High-fidelity comparisons between RANS and LES solvers reveal significant discrepancies, highlighting that modelling choices strongly influence power-deficit results, with different models showing divergent predictions, especially for turbines deeper in the array.⁴

The Rapid Evolution of Industry Requirements

Importantly, the proportion of the wind industry relying exclusively on empirical analytical models is decreasing rapidly, especially in offshore projects.

For bankable energy predictions, particularly for farms with significant wake losses, it is increasingly advisable to use high-fidelity or mid-fidelity CFD assessments to manage the associated uncertainties.⁵ However, legacy tools are still being used in many "bank-grade" project workflows, albeit as part of ensemble wake modelling, creating "interpretation" challenges.

First Principles and the Mid-Fidelity Breakthrough

A robust approach is to return to physical first principles. For modern wind farm design, this transition has led to the emergence of mid-fidelity field models—solvers that simulate the entire waked wind field directly, rather than superimposing individual wake models. Physics-based mid-fidelity models provide a crucial balance between computational speed and physical accuracy. From a literature review I compiled;

• Parabolic 3D RANS (three-dimensional Reynolds-Averaged Navier–Stokes) solvers
• The Dynamic Wake Meandering (DWM) model*
• LiDAR-informed Pseudo-2D RANS models

Many of these are now available as open-source tools or commercial offerings.⁶

Crucially, this class of tool acts as a field model. It simulates the entire waked wind field, which inherently eliminates the need for the empirical wake superposition model used in legacy approaches². These physics-based approaches incorporate critical physics that analytical models often might miss, including:⁷

• Eddy viscosity turbulence closure parameterised by local shear ⁸
• The temporal movement (meandering) of the wake and wake-added turbulence ⁹
• Stability corrections for ambient shear and turbulence decay ¹⁰

Field-based models like the Pseudo-2D RANS approach have demonstrated significant improvements in predictive accuracy, reflecting their enhanced physical fidelity.⁸

The Reality of RANS and LES Trade-Offs

RANS solvers, while more accurate than purely analytical approaches, still depend on turbulence closure models that introduce uncertainty.

When compared to high-fidelity LES (Large Eddy Simulations), different RANS models (standard k-ε, realisable k-ε, SST k-ω) produce notably different predictions of turbulence intensity and downstream power output. The realisable k-ε model generally shows better agreement with LES, though notable differences remain.¹¹

Mastering the New Toolset

The evolution of wake modelling—from empirical correlations to physics-based mid-fidelity solvers and high-fidelity LES—requires that professionals continuously up skill.

Understanding the nuances between these modelling fidelities, the impact of atmospheric stability (particularly offshore), and the limitations of wake summation methods is no longer optional.¹²

To stay competitive and ensure project bankability, professionals must master both the foundational physics and the capabilities of these new tools.

Invest in mastery: Explore advanced educational opportunities to fully grasp wake evolution, superposition methods, and turbulence physics.

This mastery includes the critical evaluation of new tools. For example, understanding that even 'LES' may be run with a coarse-grid solution, trading individual wake resolution for operational speed. High-fidelity LES requires fine mesh resolution and adequate time-stepping; coarse-grid solutions may sacrifice critical near-wake physics.¹³

This article is a contribution to the ongoing discussion on wake modelling. The physics are complex and the tools are evolving rapidly, and readers are encouraged to consult primary literature for the most current information.

Understanding both the strengths and limitations of available wake models is crucial—each approach has its own use case. For professionals seeking to deepen their expertise, specialized educational opportunities, like theWake Masterclass, as well as engaging with recognised experts like Dr Wolfgang Schlez, founder of ProPlanEn, who provides industry 'capacity building' based on over 30 years of experience—can offer valuable insights.

* The Dynamic Wake Meandering model is understood to be a well-established mid-fidelity approach in wind farm modelling that accounts for time-dependent wake behaviour.

** Please note that  ProPlanEn is a solution partner of OWnD, within the CPS and API segment.