Model verification

TurbineX runs OpenFAST under the hood. To give that statement weight we benchmark the platform against an external, peer- reviewed code-comparison study and publish the protocol, reference data, and acceptance criteria. The current benchmark is the IEA Wind 22 MW Reference Turbine of Collier et al. (2024).

Verification manual (PDF)

Download the full manual:

Download verification manual — IEA 22 MW (PDF)

13 pages · v1.0 · benchmark against Collier et al. 2024, J. Phys.: Conf. Ser. 2767 052042.

What the manual covers

Four verification tests (VT1–VT4) span the modelling layers used by the platform's certification pipeline:

• VT1 — Mass and inertia. Static check from the ElastoDyn summary. Blade mass, 1st and 2nd mass moment, RNA mass, tower + monopile mass. Acceptance: ±2 % of cross-code mean.
• VT2 — Isolated blade natural frequencies. First eight blade modes via free-decay. Acceptance: ±1 % on mode 1, ±4 % on modes 2–3, ±5 % on torsion and higher.
• VT3 — Steady-state operation 3–25 m/s. Closed-loop runs at 1 m/s steps, rigid tower, zero tilt, no gravity. 15 output channels compared. Acceptance: TurbineX inside the four-code [min, max] band (Bladed, HAWC2, OpenFAST, QBlade) padded ±5 % for digitisation, on ≥20 of 23 wind speeds.
• VT4 — Turbulent DLC 1.1 statistics and DELs. Class B NTM, seven wind speeds bracketing rated, six seeds each, ten-minute realisations. Mean / std / DEL on blade root moments. Acceptance: mean inside cross-code band ±5 %, DEL inside cross-code band ±10 % (the wider pad absorbs the reduced seed count vs the 18-realisation reference matrix).

Why benchmark, not certify?

The manual deliberately frames acceptance as “inside the cross-code spread” rather than “matches code X exactly.” Bladed, HAWC2, OpenFAST and QBlade disagree with each other by single-digit percents on most quantities (and by more on blade root M_z, peak loads above rated, and damping ratios); penalising TurbineX for sitting anywhere inside that band would be artificial. The band itself is the published cross-code uncertainty for current-generation very-large turbines.

How TurbineX uses this

• The four VT cases are tracked in-repo under model_verification/IEA22MW/ with driver scripts, per-case OpenFAST inputs, and digitised reference data.
• Zeph reads the reference values directly from the same data files when you ask “what should X be on the 22 MW? ” — see Zeph guide.
• Certifier-grade reports for studies on the IEA 22 MW can include a Model verification annex pulled from this benchmark — see Reports.

Other turbines

The 22 MW manual is the current public benchmark. The IEA 15 MW (Rinker et al. 2020) and NREL 5 MW (OC3/OC4/OC5) are on the roadmap; reference data files are already digitised in-repo under model_verification/data/ but the run-and-compare loop is not yet automated.

Reference

W. Collier, D. Ors, T. Barlas, F. Zahle, P. Bortolotti, D. Marten, C. S. L. Jensen, E. Branlard, D. Zalkind, K. Lønbæk, “Aeroelastic code comparison using the IEA 22 MW reference turbine,” J. Phys.: Conf. Ser. 2767, 052042, 2024. doi:10.1088/1742-6596/2767/5/052042.