● Flagship capability

GridWindow

One engine that follows a large grid connection through its whole life: studied before it is made, monitored while the site runs, and reconstructed after any event in a record anyone can verify.

OwnershipGridWindow is developed and owned by Cloudsail Technologies. SEHRO proposes to commercialise it; the IP arrangement has not yet been completed.

Request a walk-through Company deck

01 Inside GridWindow

One connection, three moments.

Genuine screenshots of the application running on its synthetic Marlow Basin network.

GridWindow overview of the fictional Marlow Basin network with the connection assessment for a proposed 100 MW data centre: not N-1 secure, with the main issues and calculated levers listed. — **Before**The connection assessment for a proposed 100 MW data centre (Option A, evening peak). It is within the stated limits with every circuit in service, and not N-1 secure: losing any of six listed circuits breaches a limit. Calculated levers, such as a 20 MW battery, are shown with their effect.
Verdict
Not N-1 secure
Lowest voltage
0.975 pu
Most loaded
LN-MAR-KOO 78%

GridWindow 3D window around the data centre with circuit LN-MAR-RID out of service; the data-centre bus is marked in breach of its voltage limit. — **Outage**The data-centre window with circuit LN-MAR-RID out of service. The calculated case shows one breach in the window: the Halcyon data centre 132 kV bus falls below its 0.95 pu voltage limit.
Outage
LN-MAR-RID
Data-centre bus
0.9086 pu
Limit
0.95 pu

GridWindow 3D digital twin of the window around the data centre, with buses and circuits labelled with calculated voltages and loadings. — **3D twin**The digital twin of window D3 at 100 MW with no outage selected. Nothing inside the window breaches the stated limits. Terrain and layout are illustrative; every electrical figure comes from the engine.
Halcyon 132 kV
0.979 pu
Saltpan 33 kV
0.997 pu
CN-A-TAN-HAL
51%

GridWindow monitoring view replaying a synthetic operating day at 10:45, with data-source health, the data-centre connection and assumptions from the BEFORE study marked as holding or deviating. — **During**The synthetic operating day replayed at 10:45, tick 58 of 145. Data-centre demand is 36.2 MW, metered and corroborated by network SCADA. Two assumptions from the BEFORE study are flagged as deviations; a coincident voltage sag is labelled inferred, not a proven cause.
Demand
36.2 MW
Deviations
A-DC-1, A-NET-1
Sources
All usable

GridWindow investigation view reconstructing the synthetic day, listing each record with when it happened and when it was known, beside a signed evidence pack summary. — **After**The day reconstructed from the signed evidence pack. Each record shows when it happened and when it was known: the Kooringa North transformer TX-KOO-1 opened at 10:40:00 and was known at 10:45:00.
Assessments
145 logged
Evidence pack
Signed
Learned later
98 records

Captured by GridWindow's automated browser check on 27 October 2026 (code commit b7f1700). Full frames, resized for the web only. Synthetic demonstration: fictional network, terrain and data. Select an image to open it at full size.

02 How it works

Engineering, shown.

Five technical foundations, each with its current evidence.

T.01

AC power-flow modelling

Full Newton-Raphson AC power flow on a bus-branch network model: voltages, angles, and line and transformer loading. Every reported solve is re-solved by an independent solver. A case that does not converge, or where the solvers disagree, is reported as inconclusive and shows no figures.Validated, narrow scope · steady-state AC connection assessment on a published benchmark network, within the declared operating range

T.02

Staged connection capacity

A connection is modelled as it grows, for example 20 MW now and 100 MW later, across operating snapshots such as evening peak and midday. Hosting capacity is found by bisection on full AC solves, and every figure is classified: conclusive, lower bound, at search ceiling, pre-existing or inconclusive. Levers such as a battery, flexible demand or a new circuit are solved as their own cases.Prototype · synthetic network

T.03

Constraints and outages

Thermal and voltage limits are stated up front. Each stage is solved intact and with each circuit on a declared outage list removed, and the first binding constraint is named by element. If any listed outage is unresolved, the N-1 headline reads inconclusive, never secure.Prototype · declared outage list

T.04

Before, during, after

BEFORE records the assumptions a connection was screened on. DURING tests them against measured values only and scores each data source. AFTER reconstructs events with event time and knowledge time kept apart, so no assessment uses data it could not yet have had.Prototype · zero hindsight leakage across 145 time steps

T.05

Reproducible evidence

Every run records code version, package versions and input hashes. Results ship in a signed evidence pack with a standalone, fail-closed verifier, and a reproduce tool re-executes every logged assessment. 213 automated tests, with mutation testing and code review by a separate AI system; professional engineering review is next.Prototype · demonstration signing key; a production trust root is next

03 Validation

Where it stands.

Validated against a published benchmark network of the Australian grid, and exercised end to end on a synthetic regional network.

Capability	Stage	Evidence
AC connection assessment	Validated	Matches the benchmark's published reference solutions within frozen tolerances, with every solve re-checked by an independent solver. Detailed results on request. Proof of concept (TRL 3) for this function only.
Independent solver cross-check	Validated	Every reported solve confirmed by a second solver before a figure is shown.
Before, during and after workflow	Prototype	End to end on a synthetic regional network. Zero hindsight leakage across 145 replayed time steps.
Signed evidence pack and verifier	Prototype	Tamper and forgery cases detected in automated tests.
3D operator interface	Prototype	Every electrical figure on screen comes from engine output. Automated browser check 39 of 39.
Real network model and meter data	Next	Access to one real network planning model under a future confidentiality agreement; NEM12 meter data.
Independent engineering review	Next	Power-systems engineer review of method and results before a customer pilot.

04 Next

See it run.

A private technical walk-through on a benchmark network and a synthetic regional network.

Contact Technical direction