Solutions

TurbineX and the studies behind it.

Parametrica builds its software on one modelling chain: atmosphere and ocean, structural and aeroelastic response, surrogates and uncertainty. TurbineX runs wind turbine loads and fatigue in the browser. The studies further down use the same chain, and each one adds validated cases to the software.

Software

TurbineX

Live · turbinex.app

TurbineX

Cloud platform for wind turbine aeroelastic simulation and IEC 61400 load analysis, built on OpenFAST: design load cases, fatigue, wake effects and lifetime extension, with certification-style PDF reports. It is not a certification tool.

Open TurbineX →

Side-by-side maps of mean wind speed over three planned offshore wind farms off the US East Coast: the mesoscale simulation on the left shows a diffuse cluster wake, the large-eddy simulation on the right resolves individual turbine wakes and their interaction.
Mesoscale (a) against large-eddy simulation (b) over three planned offshore wind farms off the US East Coast. The mesoscale model captures the cluster wake but underestimates the internal wake losses between closely spaced turbines. From Sanchez-Gomez, Deskos, Optis, Lundquist, Sinner, Xia and Musial (2026), Wind Energy Science 11, 2009–2036, CC BY 4.0.
Technical due diligence

Wind resource & wake in complex conditions

The independent physics check on the two assumptions that move a project's energy yield most: what the wind actually does over the site, and how much of it the turbines take from each other. Where the terrain is steep, the array is large, or a neighbour sits upwind, the engineering models behind a standard assessment are being run outside the conditions they were tuned in, and the error lands directly in P50 and P90.

  • Wake and blockage in large arrays Cluster wakes, long-range wakes between neighbouring farms and array blockage resolved with turbulence-resolving simulation rather than a superposition rule, across the stability regimes where engineering wake models diverge most.
  • Flow over complex geometry Steep and forested terrain onshore; cliffs, islands, coastal transitions and platform structures offshore. Separation, speed-up and recirculation that linear flow models cannot represent by construction.
  • Independent review of an energy yield assessment A second opinion on someone else's wake losses, uncertainty budget and site suitability, written so a lender's independent engineer can act on it.

Independent review is usually the right place to start, and often the sharper product: a lender’s engineer wants a second pair of eyes on the assessment in front of them, not a third assessment. Where a full energy yield assessment is the right scope, we run it within the IEC 61400-15 framework and state plainly what is modelled, what is measured, and what the uncertainty rests on.

Turbine, controller, platform and wake solved together rather than in sequence. Rendered from a coupled simulation run on TurbineX.
Coupled multiphysics

Coupled aeroelastic loads & floating dynamics

Aero-elastic simulation for onshore and offshore turbines; aero-hydro-servo-elastic coupling for floating platforms; site-specific design verification across regimes.

  • Coupled aero-hydro-servo-elastic simulation Turbine, controller, platform, hydrodynamic and mooring response resolved together. Fatigue and extreme-load envelopes built from first principles.
  • Motion-aware wake interaction For floating arrays where platform pitch and surge change the wake shape; standard wake models systematically miss this.
  • Site-specific design verification Load case analysis with measured site conditions, atmospheric stability effects and reference-case calibration.
Left: cumulative fatigue damage for four turbine components rising at different rates and crossing the damage limit at different years, with the pitch bearing reaching the limit before the twenty-five year design life. Right: the resulting distribution of annual availability, with the median and the tenth percentile marked.
Damage accumulates at a different rate in every component, so the one that binds the asset is rarely the one a fleet curve predicts. Carrying that through to an availability distribution, rather than a single availability number, is what puts a defensible band around OPEX. Schematic, not a result.
Reliability engineering

Asset reliability, availability & lifetime

A component-level view of what installed assets actually do over their lifetime: reliability modelled as an engineering question, turbine by turbine, with the same aeroelastic physics that drives certification.

  • Component-level damage modelling Site-specific fatigue and damage progression on blade root, drivetrain, bearings, tower, foundation and, for floating, mooring.
  • Availability & downtime distributions Physics-based failure timing translated into availability distributions and OPEX-uncertainty bands a technical reviewer can defend.
  • Lifetime extension & repowering Load reassessment and remaining-life estimation for ageing fleets, built to survive scrutiny from owners, certifiers and lenders.
Four-panel coupled mesoscale-to-microscale simulation of Hurricane Laura: a 4.5 km mesoscale grid, a 1.5 km vortex-following nest, and turbulence-resolving large-eddy simulation domains at 33 m and 11 m grid spacing.
Coupled mesoscale–microscale simulation of Hurricane Laura (2020). Panels (a) and (b) are the mesoscale domains at 4.5 km and 1.5 km grid spacing; panels (c) and (d) are turbulence-resolving large-eddy simulations at 33.3 m and 11.1 m. Reproduced from Deskos et al. (2026), Wind Energy Science, CC BY 4.0; coupling methodology in Sanchez-Gomez et al. (2025).
De-risking extreme weather

Extreme-weather resilience

Tropical cyclones, derechos and severe storms break most of the assumptions in the standards. We model what a specific storm does to a specific machine, running the mesoscale-to-microscale chain down into turbulence-resolving large-eddy simulation and carrying it through to component-level structural response, then turning that into design, operating and survival decisions.

  • Survivability & stop-strategy studies What a storm does to this machine at this site, and which operating decisions change the outcome.
  • Forensic event reconstruction Physics-based post-event analysis of failures, like our published derecho reconstruction, built for root-cause clarity.
  • Mesoscale-to-LES hazard modelling Tropical and extratropical hazard resolved to the site, grounded in our peer-reviewed research programme.

See the derecho reconstruction →  ·  Read the tropical-cyclone research brief →

Two conceptual curves. Left, a vulnerability function: damage stays at zero below an onset wind speed, rises through a steep band and saturates at full damage. Right, a loss exceedance curve: annual probability of exceedance decays with loss, and the return-period loss is read off at 0.4 percent annual probability.
The two functions behind a per-risk number. A vulnerability curve maps hazard intensity onto damage for a specific machine; convolving it with the site hazard gives the loss exceedance curve an underwriter reads a return-period loss from. Schematic, not a result.
Asset & risk intelligence

Site-specific risk & performance studies

Physics-based hazard, exposure and response for a named asset, lease area or portfolio: the analysis a developer can stake a competitive bid on, an operator can plan around, and a lender's independent engineer will accept. When risk transfer needs a number, the same study gives brokers and their markets evidence built for scrutiny.

  • Damage and loss for a named asset Site-specific extreme-weather hazard combined with the asset's fragility: likely damage and loss for the actual machine at the actual site.
  • Downtime & availability exposure Physics-based failure and repair timelines as downtime distributions, and what mitigation changes them.
  • Decision-ready reports Margins, mitigation and monitoring evidence, detailed enough to challenge and transparent enough to defend.
Left: wind-speed RMSE for five datasets, all falling within 0.97 metres per second of one another. Right: the same five datasets scored as fit, borderline or not fit against four decisions, siting, financing, O and M windows and trading, where they are clearly not interchangeable.
The same five datasets, two ways of looking at them. On the usual accuracy statistic they sit within 0.97 m/s of each other and it tells you nothing about whether any of them is good enough. Scored at the asset against independent in-situ measurement, in the units each decision is actually made in, they are not interchangeable at all. Demonstration ingest with declared error models: not a measurement of, or a claim about, any named commercial or Copernicus product.
Data analytics

Which data is fit for this decision?

Every number rests on a dataset, and the usual accuracy statistics do not tell you whether that dataset is good enough for the decision in front of you. A global error statistic on wind speed says nothing about whether a product will misprice a capacity factor, miss a ramp, or fail to flag a low-generation week at your site. We score data in the units of the decision instead.

  • Skill in the currency of the decision Capacity-factor error, ramp and weather-window skill, low-generation detection and probabilistic forecast skill, scored at the actual asset location rather than averaged across a continent.
  • Does higher resolution buy a better decision? Reanalysis, satellite-derived, blended and downscaled products differ in resolution, measurement type and provenance. We test whether the costlier product changes the answer, and it frequently does not.
  • Combining sources, measured rather than asserted Atmospheric, marine and radiation products carry different errors in different regimes. We quantify what combining them actually adds, cross-validated against independent in-situ observation.
  • Tiered accuracy by decision Siting, financing, O&M planning and trading do not tolerate the same error. We say which accuracy tier each decision needs, so you stop paying for precision that changes nothing.

Validation runs against independent in-situ measurement, met masts, buoys and ground radiation stations, rather than scoring one model against another.

Fifteen-minute-ahead probabilistic forecasts of active and reactive power, showing the median and the 10 to 90 per cent band against actuals over four held-out winter days.
Fifteen-minute-ahead quantile forecasts, median and 10–90% band, against actuals across four fully held-out days. The band covers 80.0% of outcomes against 80% nominal, which is the property that decides whether an operator can act on it. Method, metrics and figures →
Grid-edge analytics

Forecasting & grid-edge intelligence

Probabilistic forecasting of generation and load with calibrated uncertainty bands, compressed to run on the edge device next to the meter. Built for grid operators, aggregators and asset operators whose decisions live in the upper percentiles, not the average.

  • Calibrated probabilistic forecasts Quantile forecasts of active and reactive power whose bands mean what they say: 80% nominal, 80% measured on held-out data.
  • Edge-native deployment Containerised services with standard interfaces (REST, MQTT), model bundles under a megabyte and sub-millisecond inference on low-cost hardware.
  • Decision KPIs, not accuracy scores Missed events against alarm budgets, congestion risk at the feeder, value quantified on the operator's own terms.
Studies

Studies use the same chain as the products.

Where a question is not yet covered by TurbineX, we answer it as a study on the same models. Each study ends in a document for an underwriting call, a placement, a bid or a close, and starts with a scoping call and a fixed quote.

If the question turns out not to need us, we will say so on the first call.

  • 01

    Asset risk assessment

    For one named asset: likely damage under extreme conditions, failure modes, repair time and downtime, written for an underwriting or lending decision.

  • 02

    Fixed-scope study

    One site, one asset, one hazard, answered at a fixed time and a fixed price. Typically four to six weeks and one written deliverable.

  • 03

    Technical due diligence

    A second opinion on submissions, energy yield assessments and certification packages. Days, not months. Available as a standing retainer where the questions keep coming.

  • 04

    Forensic loss engineering

    Blade, gearbox, bearing and foundation failures: what happened, why, and what it means for cover.

  • 05

    Collaborative R&D

    A work package inside a research consortium, or a joint proposal written from scratch. Send the call identifier and the deadline and you will have an answer on fit within a day.

  • 06

    Subcontract or white-label

    For consultancies, certifiers and model vendors: we build the hazard, response or calibration science behind your deliverable, you keep the client and the brand. Confidential and conflict-checked before anything starts.

Why the numbers hold

Built to survive the challenge.

Every number we hand over eventually meets a sceptic: a design reviewer, a lender's independent engineer, a certifier, a control room. Three properties decide whether it holds in that room, and we build for all three.

  • Traceable to the physics: hazard and response resolved from first principles on open, auditable codebases, not fitted to a short loss history and not hidden in a black box.
  • Tested across the uncertainty: the answer stated with the range around it, and the sensitivity to the assumptions that drive it.
  • Documented end to end: every assumption written down, every result a third party can re-run.

Which number are you having to defend?

A study, a work package in a proposal, a second opinion, or specialist capacity behind your own deliverable. Tell us the question and the date you need it answered by.

Talk to us