Factory acceptance test FAT

What it is (precise)

The definition that actually exists lives in a process-industry standard, not a power one. IEC 62381 defines a factory acceptance test at 3.1.8 as an activity, including inspection and testing, to demonstrate that the automation system, subsystem or component is in accordance with the specification, and adds that it is typically conducted at the vendor's facility. Read the word "typically".

Location is not part of the definition, so a FAT run at an integrator's staging yard or a container assembly line is still a FAT provided it demonstrates conformity to the specification. Clause 1.1 sets the standard's job: requirements and checklists for FAT, FIT, SAT and SIT, carried out to demonstrate that the automation system meets the requirements of the applicable specification.

Edition 3.0 added the factory integration test (FIT) at 3.1.9 — inspection and testing at the vendor's facility to demonstrate that the merging of some or all of the various subsystems and components into one overall automation system functions in accordance with the specification.

Its site counterpart, the site integration test (SIT) at 3.1.20, demonstrates that the merging is completed and that all components work together in accordance with the specification. That pair exists because an equipment-level FAT does not answer the integration question, and on a multi-vendor battery plant the integration question is the one that hurts.

The standard also draws its own boundary, and that boundary is the single most useful line here: Clause 1.2.1 states that engineering and manufacturing activities prior to or after the FAT, FIT, SAT and SIT, such as loop checks and commissioning, are not covered by the document.

IEC 62381 is therefore not a commissioning standard and cannot be cited as one. What it does give you is closure machinery — a punch list, defined at 3.1.16 as a list of all problems found, the corrective action to be taken, party responsible and date completed, then 5.4 FAT punch list, 5.5 documentation of FAT including a FAT report, and 5.6 FAT completion.

Why it matters in a real grid-scale project

FAT is where a large fraction of the contract price changes hands. A 2025 Morgan Lewis survey of US utility-scale storage procurement describes a typical milestone set — execution of the agreement, notice to proceed, delivery of design, delivery of major materials to the factory, factory acceptance testing, delivery to site, and substantial completion or commissioning — and gives indicative percentages: approximately 60% of the overall contract price may be paid by factory acceptance testing, 80% to 90% by delivery, and approximately 10% at commissioning.

Treat those as reported market practice from a law-firm publication, not a standard. The shape holds regardless of the exact split: most of the money is committed long before the commercial operation date.

That is why an equipment-level FAT is a thin gate for a multi-vendor plant. EPRI's ESIC Energy Storage Commissioning Guide records the failure directly — a battery subsystem passed a factory acceptance test and shipped, a power conversion subsystem from a different vendor passed its own factory acceptance test and shipped, and when integrated the system did not operate, extensive field work was required, and each vendor blamed the other.

EPRI's advice is blunt: either buy an integrated system, or plan for significant on-site integrated system acceptance testing. If your scope splits the battery and the power conversion system across two suppliers, budget for the second option.

FAT also sits inside the schedule rather than beside it. EPRI groups the main commissioning-related activities during construction and installation as rough-in inspections, FAT and final field inspection, and notes that completing them is typically associated with reaching a mechanical completion project milestone.

The same guide flags a procurement trap: FAT attendance is coordinated through the contract, many parties may travel to the factory, and any requirement for the FAT should be indicated early in procurement because there are costs involved that need to be exposed. Put the witnessing requirement into the specification at bid stage, or negotiate it later at the vendor's price.

Typical values and standards

For energy storage specifically, EPRI describes the FAT as traditionally centred on single pieces of equipment rather than field-assembled systems, typically witnessed and conducted at the originating factory floor prior to shipment.

Its Storage Wiki puts the same thing more carefully: witnessing a FAT at the vendor's manufacturing facility before shipment is common practice, not a requirement, and the tests are typically quality-control related, checking that the components of the ESS have been built to specification before leaving the factory. For large or modular systems the FAT is sampling-based, resting on statistical results of a random selection of units or modules, and the results are a deliverable even if the owner does not personally witness the testing.

EPRI's list of what a BESS FAT typically covers is worth lifting into a specification: a compliance test for the power conversion system (for UL 1741 compliance) that can be extended to a packaged system (UL 9540); bench or hardware-in-the-loop testing of the various control systems so they integrate and operate together, which is where sensors, metering and alarms, remote control and monitoring, and data systems get verified along with correct register mapping; and verification of all control functions plus system performance at full and partial power, energy ratings and efficiency for packaged systems.

EPRI's indicative schedule for a roughly 1 MW-class BESS allows about two days, with the utility representative, system integrator, site controller vendor, PCS vendor and battery vendor all potentially involved.

The component standards a project FAT actually witnesses are the routine tests. IEC 60076-1:2011 Clause 11 splits transformer testing into routine, type and special; for dry-type units, IEC 60076-11:2018 Clause 14.2 lists winding resistance, voltage ratio and phase displacement, short-circuit impedance and load loss, no-load loss and current, applied voltage, induced voltage withstand, and partial discharge.

Type tests such as lightning impulse and temperature rise, and special tests such as fire behaviour and seismic, are not normally repeated per unit at a project FAT — that is practice, not a clause you can cite. For switchgear, IEC 62271-1:2017+AMD1:2021 Clause 8 gives the routine list: dielectric test on the main circuit, tests on auxiliary and control circuits, resistance of the main circuit, tightness, and design and visual checks.

How it shows up in specs, studies and contracts

On a real BESS specification, FAT appears per subsystem rather than as one system-level event. Entergy Louisiana's model BESS scope book requires the seller to develop and submit a factory test plan for the power conversion system to be approved by the owner, and handles battery module factory test data separately.

It then ties the paperwork to project gates: OEM FAT and shop test reports for the listed equipment are due prior to initial energization, while commissioning turnover documentation and SCADA FAT results are due at substantial completion. Notice what is being gated — the document, not the test. That is the pattern to look for whenever you read a submittal table.

Standard-form contracts often do not use the words at all. The World Bank Standard Procurement Document for Plant (Design, Supply and Installation), used for Bank-financed plant procurement, contains neither FAT nor SAT anywhere in its 265 pages. What it has instead is GCC Sub-Clause 23.1: the contractor shall at its own expense carry out, at the place of manufacture and/or on the site, all such tests and inspections of the plant and any part of the facilities as are specified in the contract.

FIDIC's Yellow and Silver Books are generally described as using a third vocabulary again — Tests on Completion, Taking-Over Certificate, Tests after Completion — though FIDIC's contract text is copyrighted and no authorised copy could be checked for this entry. Whichever form you are on, the contract's defined terms govern and FAT is only your project's shorthand.

Witnessing there is a right, not a precondition. GCC 23.2 to 23.4 entitle the employer and project manager to attend at the employer's own cost, require reasonable advance notice, oblige the contractor to provide a certified report of the results, and let the contractor proceed if they do not attend. Hold points and witness points — the mechanism engineers actually argue about — have no standards-body definition that could be located as of mid-2026.

The working industry definitions: a hold point is a mandatory verification point beyond which work cannot proceed without approval by the designated authority; a witness point is one that authority may attend. The substance appears in contract documents even where the labels do not: GCC 23.10 forbids covering up any part of the facilities or foundations on site before the required test or inspection.

Scoping the matrix and handing off to site

Scope is the lever that decides whether a FAT is worth attending, so write it as a matrix rather than a paragraph: one row per deliverable subsystem, with columns for the venue, the document that row is tested against, whether attendance is a hold point or a witness point, the deliverable that closes the row, and the project gate that deliverable is tied to. The rows follow the commercial packages rather than the physics.

DC block enclosures and their battery management arrive under one purchase order, the AC block power conversion equipment under another; high-voltage scope such as the main power transformer, the auxiliary transformer and the MV switchgear is frequently owner-supplied, with its own witness list and its own factory; and most balance of plant — foundations, cable, the grounding grid — never sees a factory test at all, because there is no factory article to test.

Where one BESS integrator holds the whole scope, the internal boundaries in that matrix are theirs to close. Where the scope is split, every boundary is a place where both sides can pass and the interface between them stays untested — which is exactly the failure EPRI's lessons-learned case describes.

The deliverable column is where FAT paperwork stops being a formality. IEC 62381 supplies the closure machinery but not the report contents that make a result usable two years later, so the specification has to name them: the serial numbers of the units actually tested and the firmware revision each one carried, the calibration status of the instruments that produced the readings, the version of the point list or register map the control checks ran against, and the ambient and load conditions in force during the run.

All of that is cheap to record on the day and effectively impossible to reconstruct afterwards.

Two entries settle later arguments on their own. Firmware moves between the factory and the site far more often than hardware does, so a control function verified at FAT may be running different code by energisation unless the report pins the revision.

And a register map that shifts by one point between the factory bench and the site turns a verified telemetry path into a mislabelled one, with the control hierarchy above it reading the wrong quantity in perfect good faith. Open punch-list items should carry forward into the site punch list by name: FAT completion closes the test, not the defect.

Then say plainly which rows a factory cannot fill. A works test runs on a shop supply into whatever load the facility can provide, so everything downstream of the equipment terminals is absent — no MV collection system, no point of interconnection, no utility source impedance, no live path to a control room. Grounding is the clearest case.

Routine tests for a dry-type unit under IEC 60076-11 Clause 14.2 include voltage ratio and phase displacement, so the works test does confirm the transformer matches its own nameplate. But the nameplate quotient is a rated-voltage ratio — 50 for a 34.5 kV / 690 V unit — and it equals the per-winding turns ratio only when both windings share a connection type, so a Dyn unit reads 50 x sqrt(3) = 86.6 and a YNd unit 50 / sqrt(3) = 28.9.

Whether the vector group suits the site's grounding scheme is a design question no routine test asks, and whether the resulting source is effectively grounded — X0/X1 ≤ 3 and R0/X1 ≤ 1 with both ratios positive, the classical IEEE C62.92-family criterion — is a property of the installed zero-sequence network, computed per location rather than certified at a works. Protection relays divide the same way: settings can be injection-tested against a settings file at the factory, coordination against the real upstream source cannot.

Metering and timing close the list. A factory check on a revenue meter is largely a review of calibration certificates against the declared class — 0.2S/0.5S under IEC 62053-22, or the separately defined 0.2/0.5 under ANSI C12.20, which are different scales and should never be quoted as one — while installed accuracy also depends on the instrument transformers feeding it and their burden as wired.

Anything that depends on the plant's clocks has no factory equivalent either: time synchronization across devices, and the real round trip from a remote terminal unit out to the system operator or back in as an automatic generation control setpoint, exist only once the communications path does.

Those rows belong to the site acceptance test, the site integration test, and the capability testing that follows — the P-Q capability test among them. Assign them there in the matrix, and let the FAT column state that it does not reach them, so nobody discovers the gap by looking for a result that was never scheduled.

Common pitfalls

The most expensive mistake is assuming the acronym means the same thing to everyone. The DOE Energy Storage Handbook, Chapter 21, does not use FAT for the factory test at all.

It calls that the factory witness test (FWT), which each subsystem must pass before shipping and which the handbook calls the first real step of the commissioning process, and it reserves FAT for the on-site Functional Acceptance Test, which takes place at the completed installation connected to the utility grid. Two authoritative energy-storage sources, one acronym, opposite locations. Define the acronym in the contract's definitions clause and stop assuming.

Second, passing a FAT does not move risk. The World Bank plant GCC says so at 23.9: neither the execution of a test or inspection, nor attendance by the employer or project manager, nor the issue of any test certificate releases the contractor from any other responsibilities under the contract.

Combine that with EPRI's sampling point — for large or modular systems the FAT rests on statistical results of a random selection of units or modules — and the honest reading is that a passed FAT tells you a sample of the equipment matched the specification on the factory floor, and nothing beyond that. If the specification does not state the sample plan, neither does the certificate.

Third, watch the scope of what you are citing. EPRI warns that the capabilities of the testing facility must be understood before the tests are defined, giving the example of testing an inverter with undersized load banks, which prevents the full charge and discharge capability from being tested while still allowing some function and protection-based testing — so a FAT certificate can be entirely genuine and still prove nothing about rated power.

And do not build a commissioning procedure out of IEC 62381, because Clause 1.2.1 puts loop checks and commissioning outside its scope. IEC 62933-2-1:2017 will not fill the gap either: it is scoped to unit parameters and testing methods, with performance test methods also in IEC TS 62933-2-2, and neither sets an acceptance-test framework.