Offshore turbine interaction modelling using CFD.ML

As offshore wind projects scale into multi-GW clusters, engineering wake and blockage models face growing accuracy challenges — with differences of up to 5% on annual energy yield at the largest project scales. This whitepaper presents CFD.ML, DNV's new turbine interaction model for offshore energy production assessment.

Request a copy

First name

Last name

Company name

Job title

Country

By completing this form, I confirm that I have read the privacy statement and understood and accept the terms of use.

I would like to receive informational emails with related content in the future from DNV, for example but not limited to invitations to webinars, seminars, newsletters, or access to research that DNV thinks is relevant to me. I can unsubscribe or change my email preferences at any time using the links in the footer of the emails I receive from DNV.

CFD.ML is a machine-learning model trained on DNV's high-fidelity WRF-to-CFD simulations. It predicts wake and blockage losses together as a unified turbine interaction effect, using site-specific atmospheric conditions including boundary layer height, turbulence intensity, shear, and temperature profile. In a blind test across approximately 30 offshore wind farms, CFD.ML v2 demonstrated the lowest bias and lowest uncertainty of all tested models relative to the high-fidelity CFD reference.

The paper covers the motivation for moving beyond patched engineering models, the key physical drivers of turbine interaction effects, how CFD.ML was developed and validated, and how turbine interaction losses are accounted for in DNV's Energy Production Assessment methodology. CFD.ML becomes DNV's standard model for offshore turbine interaction loss assessment in all Energy Yield Assessments from 1 June 2026.

What you will learn

The accuracy gap: Why commonly used engineering wake and blockage models risk overpredicting AEP at the scales of tomorrow's offshore wind farms, and how DNV's benchmarking quantifies this risk
How CFD.ML works: How a Graph Neural Network trained on high-fidelity CFD predicts wake and blockage losses together as a unified turbine interaction effect, using site-specific atmospheric conditions
Validation results: How CFD.ML v2 demonstrated the lowest bias and lowest uncertainty of all tested models in a blind test across 30 offshore wind farms
Multi-fidelity workflow: How to apply CFD.ML iteratively through project development and reserve full CFD where it adds most value — reducing uncertainty progressively across the project lifecycle
EPA integration: How DNV accounts for turbine interaction losses using CFD.ML within its bankable Energy Production Assessment methodology

Who should read this

Offshore wind developers preparing energy yield assessments for large-scale projects
Wind energy consultants delivering energy production assessments under timeline and budget constraints
Investors and lenders requiring confidence that offshore energy forecasts will hold across project lifecycles
Technical specialists evaluating turbine interaction modelling approaches for future-scale clusters

Download the whitepaper to understand how CFD.ML addresses the growing accuracy gap in offshore turbine interaction modelling. CFD.ML becomes DNV's standard model for offshore turbine interaction loss assessment in all Energy Yield Assessments from 1 June 2026.

Whitepaper

Offshore turbine interaction modelling using CFD.ML

About

The prediction of wake and blockage losses together as a unified turbine interaction effect, using site-specific atmospheric conditions including boundary layer height, turbulence intensity, shear, and temperature profile.