A container passes twenty-five checks before it is allowed onto a truck, and the ones that fail are almost never the ones being tested. They are the ones nobody wrote down. The walk below is the order these checks actually happen in, from the incoming inspection bench to the seal on the door.
What the record has to show before anything is powered
- Model, serial number and configuration against the purchase order
- Cell batch identity against the traceability record, not against a brand label
- BMS, EMS and converter firmware revisions, recorded as version numbers
- Approved single-line diagram set beside the as-built wiring
- Insulation resistance at the specified test voltage, with ambient conditions recorded
- Dielectric withstand test on the power circuit, witnessed
- Earthing and bonding continuity at every enclosure entry
- DC polarity, string continuity and torque marks on all terminals
If the firmware revision is not written down, the certificate behind the unit cannot be tied to the unit, and nothing found later can be traced back to the test that should have caught it.
The functional block: battery, BMS and converter
- Cell voltage spread across the pack, reported as a distribution rather than a single maximum
- Temperature sensor map proven by heating one sensor and watching the correct channel respond
- Balancing demonstrated under load, not merely enabled in a menu
- State-of-charge estimate checked against measured energy over a partial cycle
- Protection thresholds proven by injection: over-voltage, under-voltage, over-current, over-temperature
- Contactor close and pre-charge sequence, timed
- Converter start, stop, and the change from grid-following to grid-forming
- Round-trip efficiency measured at agreed load points, with the measurement boundary stated
- Protection relay settings compared against the coordination study
- Communication point map exercised end to end, every register read and written
Two of these carry more weight than the rest. Efficiency is meaningless without the boundary it was measured between, and a threshold proven by simulation is a setting, not a test.
Thermal, safety, and the interlocks that must not be bypassed
- Cooling setpoints, alarm thresholds, and the response to a simulated high cell temperature
- Gas and smoke detection triggered with test gas, with the interlock path proven through to the fire panel
- Suppression signal path verified without discharging the agent
- Liquid cooling pressure and leak test, where the architecture is liquid-cooled
If a temperature channel is mapped one position off, the derating logic protects the wrong cell for the life of the asset, and no site test will find it later.
The last three checks happen with the doors closed
- Weighbridge reading against the documented net weight
- Shock and tilt indicators fitted, photographed and sealed
- Documentation set complete: test reports, UN 38.3 test summary, packing list and seal numbers
The weighbridge check has one reference, the 21,000 kg net weight on Ruibit's datasheet for the 20GP unit, and a few hundred kilograms of difference is normal before the discrepancy becomes a conversation at the port.
What the deviations looked like
Across recent orders the findings were mundane and repetitive. A connector reseated after a second insulation attempt. Two terminals without torque marks. A temperature channel mapped one position off. A firmware version running two revisions ahead of the one named on the test certificate. None of these stopped a shipment, and each was cheaper to settle at the works than at the destination.
What this walk did not show, and what to ask for
What it did not show: a capacity test at the buyer's own duty cycle, an interlock proven under a real fault rather than a simulated one, and packing tested against a real transport profile. Those belong to site acceptance, not to the factory.
FAT scope is not comparable between suppliers unless the same four items appear in both documents.
Source: the four items are the ones that must appear in the FAT scope document for two suppliers to be compared on the same terms.
Ask which checks a third party witnessed and which are self-declared; the answer that all tests are witnessed deserves a follow-up. Ask at which load points efficiency was measured. Ask whether the thresholds were injected; an answer that the BMS is factory-set means the setting was verified, not the protection.
Three of these cannot be repeated once the container leaves the works. The insulation and dielectric tests need the test bay. The weighbridge reading and the seal numbers exist only at that moment. The firmware revision is tied to the certificate as it stands on that day. A and the cover what happens after the container arrives, while the and the published specifications for the and container units are the documents the checks above are measured against.
FAQs
1. What is a BESS factory acceptance test?
The set of checks a container passes at the manufacturing works before it is released for shipment. It covers identification and configuration, insulation and dielectric testing, the functional behaviour of battery, BMS and converter, thermal and safety interlocks, and the weighing and documentation that travel with the unit.
2. How many checks does a container FAT include?
About twenty-five in the sequence used here, grouped into identification and configuration, functional performance, thermal and safety, and the closing three that happen after the doors are sealed.
3. Should a buyer witness the factory acceptance test?
Not every check, but the ones that cannot be repeated. Insulation and dielectric testing need the test bay, the weighbridge reading exists only on that day, and the firmware revision is tied to the certificate as it stands. Those are worth attending or having witnessed by a third party.
4. What does it mean to prove a protection threshold by injection?
Injecting a signal that exceeds the threshold and observing the system act on it. A threshold that has only been configured is a setting rather than a demonstrated response, and the difference matters when a warranty is argued.
5. What cannot be checked after a container ships?
A capacity test at the site's own duty cycle, an interlock proven under a real fault rather than a simulated one, and packing tested against a real transport profile. Those belong to site acceptance instead.
