Cold aisle of a high density computing hall

Model verification for large computational loads

Verification your transmission planner will accept

Nyquist delivers measured verification for large computational load facilities, so the dynamic model filed on your behalf reflects the facility as built rather than the specification it was designed from.

The platform Who we build for

1,500

MW disconnected in 82 seconds, July 2024

3 in 4

Large load models reported inadequate

2027

Registration of computational loads begins

50

MW threshold for registration

Who we build for

Built first for the operators and developers who carry the obligation

Primary

Data center operators and developers

A facility of 50 MW or more at transmission voltage becomes a registered entity, owes its transmission planner a dynamic model, and must verify that model on a schedule. The obligation sits with you, and so does the cost of getting it wrong.

See what we deliver

50 MW

Registration threshold

100 kV

Connection voltage

2027

Registration begins

Yours

Whose obligation it is

And for the counterparties who receive the evidence

Utilities and transmission

You receive the model and carry the study risk it creates.

View

System operators

Your planning cases rest on a load class nobody has characterised.

View

Transmission substation transformer bank at dusk

Where the obligation lands

The connection point your facility is studied at, and the system your model is asked to represent.

Measured response of many facilities to one regional disturbance

Fleet oscillogram

Response of many computational load facilities to a single regional disturbance.

Measure

At the fence and behind it

Measured response, captured at the facility during real system disturbances.

Model

Fitted, not asserted

Dynamic models derived from that response and delivered in the form the planning study consumes.

Reference

Across the fleet

Comparative context drawn from facilities of similar design, so a site can be positioned before it energises.

The measurement

A disturbance is an experiment the grid runs once.

The same disturbance reaches every facility on the path, and each one answers it differently. Measured consistently, those answers become comparable and a class of facility can be characterised. Measured in isolation, they never do.

Once

How often a disturbance happens

Every site

Answers it differently

What we install

Installed at the facility existing metering position. No change to the compute environment and no interruption to operations.

Where it lands

Evidence your planner can put straight into a study.

Platform

What you receive

An engagement produces four things you can hand to a counterparty. How we produce them is our work. What you receive is below.

Delivered as an engagement, not a software subscription. Nothing is installed on your network and nothing needs a security review of our code.

Measured response

01 The model

A dynamic model of your facility

Delivered in the formats your transmission planner's studies consume, and built to represent the facility as it was actually constructed and commissioned.

02 The verification

Documentation that the model holds up

The package that accompanies the model and demonstrates it represents the facility, prepared in the form your planner and your regional entity expect to receive.

03 The record

The evidence behind both, kept

Held on your behalf under your own compliance obligation, and produced whenever it is asked for, including years after the work is done.

04 The context

Where your facility sits

Comparative context against facilities of similar design, so a site can be positioned against its class rather than assessed in isolation. Context no single owner can produce alone.

Engagements

Start with a scoping assessment

A short, paid engagement covering your registration position, the obligation dates that apply to each site, and where measured evidence changes a study outcome. It produces a written scope before anyone commits to a programme.

Request a scoping assessment

What we need to begin

01

Interconnection voltage and connected load at each site.

02

In service or planned energisation date.

03

The transmission planner or utility your models are filed with.

Nothing site identifying leaves the engagement without written instruction.

Industries . Data centers

A registered class, with a verification obligation attached

A facility hosting 50 MW of computational load or more, connected at 100 kV or above, becomes a registered entity and owes its transmission planner a dynamic model of its own behaviour on a defined schedule. Verifying that model is the facility owner's obligation, not the utility's. For sites still in design, the least disruptive point to resolve it is inside the commissioning programme.

CLO 001

Modelling data and verification

Modelling data provided to your transmission planner, refreshed on a schedule, with verification required after entering service and again after changes that alter dynamic response.

CLO 002

Operational data and communications

Documented data specifications, compliance with operating instructions, and communication capability with your operator.

CLO 003

Protection and disturbance monitoring

Protection coordination, and disturbance recording capability at the facility supply, to defined accuracy and timing.

What an unverified model costs

Equipment built to cover an assumption nobody can check

Study cycles repeated because the submitted data does not survive review

Non firm service, or a curtailment condition that was never priced

Energisation delay, which at a modern campus is the largest of these by an order of magnitude

What we deliver

A matched pair of dynamic models derived from measured response

Verification documentation in the form your planner expects

Supporting records held on your behalf and produced whenever they are requested

Positioning of the facility against comparable sites

Compliance schedule

Dates run from registration and from your in service date, not from a single industry deadline.

Industries . Utilities and transmission

You carry the study risk of a model you did not produce

Interconnection studies for large computational loads rest on parameters supplied by the customer and derived from equipment specifications. Where those parameters cannot be relied on, the conservative assumption becomes the design basis, and the cost of that assumption is carried by your system.

Response density across a regional footprint

Reported by the industry, to the regulator

3 in 4 of reported large load capacity has no adequate dynamic model on file

1 in 3 planners hold large loads whose transient behaviour is unknown to them

Most entities have no established commissioning process for large loads

Engagements

Independent measurement of a large computational load in your queue

Technical review of a submitted model before it becomes a study basis

Commissioning measurement delivered as a contracted service

Comparative context on how a proposed facility is likely to behave

Industries . System operators and markets

Planning cases built on a load class nobody has characterised

Contingency sizing, ride through assumptions and queue triage all rest on a representation of large computational load that has not been checked against a real facility.

Fleet oscillogram across many facilities

Measured response, disturbance sequence

Fleet characterisation

How many distinct behavioural classes exist across the large computational load in your footprint, and which queued projects belong to each. Aggregate answers, without site disclosure.

Validated reference models

Generic models for your planning cases with parameters derived from measured facilities, so the assumption in the case has a provenance you can point at.

Event reconstruction

Post event analysis that identifies which element acted, rather than only how much load was lost.

Canadian jurisdictions. Reliability standards reach Canadian entities through provincial adoption, and the timing differs by province. For IESO, AESO and Canadian transmission owners the physics and the fleet are identical, the compliance calendar is not, and engagements are scoped accordingly.

Footprint scale

Behavioural classes across a footprint, rather than a single site studied in isolation.

Company . About us

Measurement is the missing layer in grid planning

Every other class of asset connected to the transmission system is characterised by observation. Generators are tested. Inverter based resources are validated against recorded disturbances. The largest and fastest growing category of load on the network is characterised by paperwork.

Nyquist exists to close that. We measure large computational load facilities, derive their dynamics from what they actually do, and hold the resulting record so it can be used more than once.

How we work

Evidence over assertion

Every deliverable is traceable to a recording. Where a parameter cannot be measured, we say so in writing rather than estimating quietly.

Your data stays yours

Records are held as your agent under your own compliance obligation. Nothing site identifying reaches a shared reference without written instruction.

Vendor neutral

We sell no equipment and represent no manufacturer. The value of an independent measurement depends on it being independent.

Company . Careers

Build the measurement layer

Small team, field work, and problems that have not been solved because nobody has had the data. Roles are based in the United States with travel to facilities during commissioning.

Power Systems Modelling Engineer Engineering Dynamic modelling and system identification. You have produced models a transmission planner accepted.

Field Measurement Engineer Field operations Field installation, commissioning test attendance and site coordination.

Data Platform Engineer Engineering Time series ingestion and alignment at fleet scale, in the formats the industry exchanges.

Applications to careers at nyquist.energy

Insight

One fault. 6 voltage collapses. 1,500 MW in 82 seconds.

On 10 July 2024 a lightning arrester failed on a 230 kV line in the Eastern Interconnection. Automatic reclosing produced 6 successive faults in 82 seconds, each cleared correctly in 42 to 66 milliseconds. Coincident with them, approximately 1,500 MW of data center load disconnected. None of it was disconnected by utility equipment.

Why a specification cannot predict it

The disconnection was decided by supply transfer logic and an embedded rule counting disturbances within a time window. Neither appears on a specification sheet, and neither is visible from outside the fence.

Why it recurs

A second recorded instance involved facilities with no supply backup at all, shutting down directly on voltage dip. Different mechanism, same outcome. The behaviour is a property of the installed plant, not of the class.

Contact

Request a briefing

Engagements begin with a scoping assessment covering your registration position, the obligation dates that apply to each site, and where measured evidence changes a study outcome. Tell us which of these you are, and we will reply within two business days.

Data centers

Interconnection voltage, connected load and energisation date.

Utilities

The evidence standard your planning function would accept.

System operators

The assumption in your planning case with the weakest provenance.

Enquiries, engagements and careers

info@nyquistenergy.com

Include your interconnection voltage, connected load and energisation date and we will come back with a scoping outline rather than a brochure.

Request a scoping assessment

Reply within 2 business days