A generator can arrive with a clean factory test report and still be unsafe to hand over. Commissioning is the site work that turns a delivered machine into a standby plant: proving the installation boundary, the dead checks, the first start, the load acceptance and the transfer sequence, then closing every defect before the keys change hands. This checklist follows that order and names the evidence to keep at each step.
Commissioning is not a repeat of the factory test
The factory acceptance test proves the set that is about to ship. The site acceptance test proves the set and the installation together. Commissioning is the process that makes the second test meaningful: the documents are checked, the dead tests are recorded, the first start is observed, the load is applied in a controlled sequence, and every reading is compared with a reference before the site is accepted.
The distinction matters because most handover failures are not engine failures. A set that ran perfectly on a load bank can still fail on site because the exhaust is fighting a long route, the radiator is recirculating hot air, the fuel line is drawing air, the neutral link is wrong, the ATS logic is inverted, or the protection settings were copied from another project. None of those faults belongs to the factory test report, and none of them is found by listening to the engine for a few minutes.
For the factory side, the sequence and the test sheet are covered in factory test and site acceptance. This page starts at the site boundary and ends at the handover pack. It is written as a checklist, not a schedule: the duration of each step, the hold time at each load point and the acceptance tolerance are determined by the equipment manufacturer’s data and the project specification.
| Stage | What it proves | Evidence to keep | Stop condition |
|---|---|---|---|
| Document and boundary review | That the installed system matches the approved design and the interfaces are owned | Approved single-line diagram, layout, protection schedule, ATS logic, load list | Any unresolved design change or unowned interface |
| Dead checks | That the installation is safe to energise and mechanically ready to crank | Insulation, continuity, phase sequence, earth and battery records | Any value outside the manufacturer or project limit |
| First start and no-load run | That the set starts, builds voltage and frequency, and runs without leaks or alarms | Start record, no-load readings, alarm and shutdown test record | Unstable voltage, abnormal noise, leak or unproven shutdown |
| Load acceptance | That the set carries the real load profile without exceeding its limits | Load steps, current, power, temperature, fuel and exhaust records | Overload, derating exceedance or inability to hold the block load |
| Transfer and recovery | That the ATS, UPS and generator work as one sequence | Time-stamped transfer record for both directions and failure modes | Transfer before the set is ready, or a failed retransfer |
| Defect close-out and handover | That every open item is closed or formally accepted | Defect log, retest records, as-built documents, training record | Open critical defect or missing handover document |
Start with the documents and the site boundary
Commissioning begins with a document review, not a toolbox. The site team needs the approved single-line diagram, the generator and ATS layout, the protection settings, the ATS and AMF logic, the load schedule and the manufacturer’s commissioning data. If any of those is missing, the first start becomes an experiment. A drawing marked for construction is not the same as an approved as-built, and a protection schedule that names the breaker but not the setting is not a schedule.
The boundary review is equally important. Someone has to own the cable from the generator terminals to the ATS, the neutral and earth arrangement, the fuel supply, the exhaust route, the cooling air path and the interface to the building management system. On a split contract, each party can believe the other is responsible for the final connection. The commissioning checklist should name the owner of every interface before the set is energised, and the name should appear in the handover file.
Site conditions belong in the same review. The planning reference for this site is 40 C ambient, and the generator room may be hotter than the outdoor air. The actual room temperature, the ventilation path, the exhaust back pressure and the fuel system layout have to be compared with the assumptions used for the rating. If the room is hotter than the reference, the available output is reduced by the manufacturer’s derating data, not by an assumed margin. The same logic applies to altitude, humidity and any enclosure restriction.
Fuel autonomy is part of the design boundary. A base tank is commonly arranged to cover eight to twelve hours at full load, with a bulk tank and transfer pump where longer autonomy is required. Confirm the usable volume, suction, return, filling and drain before the first start.
| Document or interface | What it must show | Why it blocks the start |
|---|---|---|
| Approved single-line diagram | Generator, main breaker, ATS, distribution, neutral and earth arrangement | Defines what is isolated, what is live and where the protection zones end |
| Protection schedule | Device, setting, curve, CT ratio and trip action | Without settings, a trip cannot be distinguished from a fault |
| ATS and AMF logic | Start command, transfer condition, retransfer, cool-down and failure modes | An inverted or incomplete sequence can transfer load before the set is ready |
| Load schedule | Connected load, demand load, largest motor and step changes | Sets the load points and the block load that must be proved |
| Ventilation and exhaust layout | Air inlet, air outlet, radiator discharge, exhaust route and silencer support | Recirculation or excess back pressure can derate or stop the set |
| Fuel system drawing | Tank, fill point, vent, suction, return, transfer pump and drain | Air ingress or a closed valve will stop the set under load |
| Earthing and neutral drawing | System earthing, neutral link, bonding and separately derived source | A wrong link is a safety issue, not a performance issue |
| Commissioning test sheet | Every reading, limit, instrument and signature | A test without a record cannot be reviewed or repeated |
Dead checks: what must be true before cranking
A generator is commissioned dead before it is commissioned alive. The set is isolated and locked off, the upstream and downstream breakers are proven open, and the terminals are tested for absence of voltage. That sounds obvious until a parallel source, a UPS output or a control supply backfeeds a panel that everyone believed was dead. The lock-off and test-for-dead procedure is the first record in the commissioning file, and it should cover every source that can energise the generator side of the ATS.
The mechanical checks come next. The foundation and anti-vibration mounts are inspected, the exhaust bellows are free to move, the silencer is independently supported, the fuel line is connected and primed, the coolant level is correct, the oil level is correct, the battery is charged and the ventilation openings are clear. A flexible exhaust section that is stretched, a fuel hose that is kinked, or a radiator that is blocked by packaging are common enough to justify a written check rather than a glance.
The electrical dead tests prove that the conductors are continuous, insulated and connected in the right order. Insulation resistance, continuity, phase sequence, CT ratio and polarity, earth continuity and neutral arrangement are recorded before the first start. The actual test voltage, minimum insulation value and acceptable deviation are determined by the equipment manufacturer’s data and the project specification; they are not universal numbers to be copied from another site. What matters here is that the value is written down, compared with the correct reference and signed.
The control system deserves its own pass. Test the emergency stop, battery charger, fuel solenoid, louvres or dampers and every alarm channel. A panel with the wrong input mapping can pass a visual inspection and fail the first real transfer.
| System | Check | Record |
|---|---|---|
| Isolation | Prove every source open, lock off, test for absence of voltage | Source list, lock numbers, tester and result |
| Alternator and cables | Insulation resistance and continuity, phase by phase | Test voltage, reading per phase, reference limit |
| Phase sequence | Confirm generator, ATS and building sequence agree | Rotation meter result at each interface |
| Earth and neutral | Confirm the system earthing and neutral link match the approved drawing | Continuity result and visual record of the link |
| CT and metering | Confirm ratio, polarity and direction of each CT | Injection or comparison result, meter reading |
| Battery and charger | Confirm electrolyte or charge state, charger output and start circuit | Voltage, charger current, load test result |
| Fuel system | Confirm usable volume, valves open, priming complete and no leak | Tank level, valve position, leak inspection |
| Cooling and exhaust | Confirm air path clear, louvres open, exhaust free and supported | Photograph-free written check and back-pressure reading where required |
| Controls and alarms | Test emergency stop, fuel solenoid, dampers and each annunciation | Channel list with pass or fail |
First start and no-load running checks
The first start is a cold start, and it is worth treating as a test in its own right. The engine should crank for a time within the manufacturer’s limit, catch without excessive smoke, build oil pressure and settle at its governed speed. The alternator should build voltage smoothly, the AVR should hold the set point, and the frequency should be stable before any load is applied. A set that starts but hunts, surges or emits abnormal mechanical noise is telling the commissioning team something that a load test will only amplify.
The no-load run is used to check the things that load can hide: a small coolant weep, a loose exhaust joint, a fuel leak at a fitting, a vibrating panel, a charger that is not charging, or a radiator fan running in the wrong direction. The set should be observed from the control panel and from the room, with the doors and louvres in their final operating position. A machine that is quiet with the room open can be noisy or hot once the room is closed, which is the condition the building will actually see.
Shutdown tests belong here, before the set is loaded. Low oil pressure, high coolant temperature, overspeed and emergency stop should be proven by the method the manufacturer allows, and each result should be recorded with the action taken. A shutdown that is simulated only in software is not the same as a shutdown that opens the fuel valve and stops the engine. The commissioning team should know which tests can be done for real and which require a simulation, and the test sheet should say so.
| Check | What is observed | Evidence |
|---|---|---|
| Cranking and start | Start time, sound, smoke, repeated attempts | Start record against the manufacturer limit |
| Oil pressure | Pressure builds promptly and holds at no load | Gauge or controller reading at a stated ambient |
| Voltage build-up | Smooth rise to set point, no oscillation | Voltage per phase and AVR set point |
| Frequency | Stable governed speed at no load | Frequency reading and governor response note |
| Phase balance | Phase voltages within the project tolerance | Three phase readings and the reference limit |
| Leaks and vibration | No fuel, oil or coolant weep; no loose panel or pipe | Inspection record with location and action |
| Cooling and exhaust | Fan direction, airflow, exhaust joints and room condition | Room ambient and back-pressure reading where required |
| Shutdowns | Low oil pressure, high temperature, overspeed, emergency stop | Test method, simulated or real, and result |
| Alarms | Each channel annunciates and resets correctly | Channel list with pass or fail |
Load acceptance on the real load
Load acceptance is the point where the set stops being a machine and becomes part of the building. The ideal test uses the real load, in the steps the project specification defines, with the largest motor or the largest block load applied at the point where it will actually start. If the building is not ready, a load bank is used to reproduce the load points; the load bank is covered in load bank testing. A load bank proves the set, but it does not prove the site wiring, the ATS, the protection settings or the real load profile.
The load schedule should include the sustained demand, the largest step change and the light-load condition. The annual maintenance test is specified elsewhere on this site as one test a year at 50% to 100% of rating, and the light-load range is 10% to 20%; commissioning is the first time those points should be demonstrated on the actual installation. The hold time at each point, the permitted voltage and frequency deviation, and the recovery time after a block load are determined by the manufacturer’s data and the project specification, not by a generic rule of thumb.
Fuel consumption is a useful cross-check during the loaded run, not an acceptance criterion on its own. The planning figures are 0.24 to 0.29 litres per delivered kWh, with a 500 kVA set using about 101 litres an hour at full load, 80 litres an hour at 75% and 60 litres an hour at 50%. A 1000 kVA set is planned at roughly 200 litres an hour and about 4800 litres over 24 hours. If the measured fuel rate is far from the expected value at a steady load, the cause may be a load reading error, a fuel leak, a return-line problem or a set that is not operating at the assumed conditions; the commissioning record should show the evidence rather than a guess.
For a data centre or another critical site, load acceptance must include the UPS and the cooling load. The UPS covers the seconds between mains loss and the generator carrying load, and the PUE 1.4 example shows why the generator cannot be sized for the IT load alone: cooling and losses are part of the demand. The test should reproduce that sequence, or prove the load-shedding logic that keeps it inside the generator’s capability.
| Test point | What it proves | Record |
|---|---|---|
| Step load | The set accepts the specified step without collapsing voltage or frequency | Step size, voltage dip, frequency dip and recovery time |
| Largest motor or block load | The starting method and the set can absorb the worst single event | Starting current, duration, voltage dip and recovery |
| Sustained load | The set holds the expected demand without overheating | kW, kVA, current per phase, coolant and oil temperature |
| Full load or the specified maximum | The installation can reach the rating claimed for it | Reading at the rating, ambient and duration |
| Light load | The set is stable in the 10% to 20% range and the maintenance implication is understood | Load, duration, exhaust condition and oil interval note |
| Fuel consumption check | The measured rate is consistent with the load and the expected figure | Fuel used, run time, average kW and calculated L/kWh |
| UPS and cooling sequence | The generator carries the protected load after the battery bridge | Time-stamped sequence and the load present at transfer |
Transfer, UPS and the seconds that matter
The transfer test is where the generator, the ATS, the UPS and the building controls meet. It should be run in both directions, from mains to generator and from generator back to mains, and it should include the failure modes the site can realistically see: a mains loss while the set is at rest, a mains loss while the set is already running, a failed start, a transfer that does not complete, and a retransfer that is blocked. Each scenario needs a time-stamped record, not a description.
The critical interval is the time between the loss of mains and the generator carrying load. The UPS covers that interval for the protected load, but the battery has a finite autonomy and the generator has a finite start and acceptance time. The commissioning team should measure the actual start command, crank, voltage build-up and transfer times, and compare the total with the UPS autonomy and the project requirement. A sequence that works on paper but leaves the UPS close to exhaustion is a defect, even if the transfer eventually happens.
The ATS logic should be tested with the generator in automatic mode, not only in manual. The retransfer should be checked for the specified delay and cool-down, and the set should return to standby without an alarm. The control panel should show the same state as the ATS, and the building management system should receive the status it expects. A mismatch between the panel, the ATS and the monitoring system is a commissioning defect because it will slow every future response.
| Step | What to observe | Evidence |
|---|---|---|
| Mains loss detection | ATS recognises the loss and issues the start command | Time stamp and signal at the ATS input |
| Engine start | Crank, start and run-up to rated speed | Start time and engine speed trace |
| Voltage and frequency ready | The set reaches the acceptance window before transfer | Voltage, frequency and ATS ready signal |
| Transfer to generator | Load is transferred without a prolonged interruption | Time stamp, current step and UPS status |
| Load acceptance | The set carries the protected load and the cooling load | kW, kVA, current and temperature after transfer |
| Retransfer to mains | The return is delayed, synchronised or interlocked as specified | Time stamp, voltage and frequency at retransfer |
| Cool-down and stop | The set runs its cool-down and returns to standby | Cool-down time, stop event and alarm state |
| Failure modes | Failed start, failed transfer and blocked retransfer are annunciated | Scenario, alarm and operator action |
Defects, retest and handover evidence
Commissioning produces defects, and the way they are closed is part of the product. Each defect should have an owner, a severity, a proposed correction and a retest record. The severity scale is set by the project specification, but the principle is simple: a defect that affects safety, transfer, protection or the ability to carry the design load cannot be left open at handover. A cosmetic item can be listed and accepted with a date, provided the acceptance is written and the responsible person signs it.
The handover pack is the evidence that the site can be operated without the commissioning team. It should include the approved as-built drawings, the protection settings, the commissioning test sheets, the defect log with closures, the manufacturer’s O&M manuals, the training record, the spare parts list and the fuel test result. The warranty start date and the service contact should be stated in writing, and the site team should know which readings are the baseline for future maintenance.
| Item | What it contains | Owner |
|---|---|---|
| As-built drawings | Single-line, layout, ATS, earthing, fuel and exhaust routes | Design and installation contractors |
| Protection settings | Device list, settings, curves and CT ratios as installed | Protection engineer |
| Commissioning records | Dead checks, first start, load steps, transfer sequence and ambient | Commissioning engineer |
| Defect log | Open items, closures, retest evidence and accepted exceptions | Project manager |
| O&M documents | Manufacturer manuals, spare parts list and service schedule | Supplier |
| Training record | Operators trained, topics covered, attendance and date | Commissioning team and site |
| Fuel and fluid records | Fuel test result, tank condition, coolant and oil specification | Site and supplier |
| Warranty and contacts | Warranty start, service contact and escalation route | Supplier and owner |
What is the difference between commissioning and a site acceptance test?
Commissioning is the whole site process: document review, dead checks, first start, load acceptance, transfer tests, defect close-out and handover. A site acceptance test is one formal stage inside that process, where the installed system is tested against the agreed criteria and the result is accepted or rejected. A site can pass a narrow acceptance test and still have open commissioning defects if the test scope did not cover them.
Can a generator be commissioned without a load bank?
Yes, if the real load is available and the project specification allows it. The real load is the better proof because it includes the building wiring, the ATS and the actual load profile. A load bank is used when the real load cannot be applied safely or at the required points. The annual load test at 50% to 100% of rating still has to be demonstrated by an approved method.
How long should generator commissioning take?
There is no universal duration. The time depends on the number of sources, the ATS arrangement, the load steps, the availability of the real load, the number of defects and the witness requirements. The hold time at each load point and the acceptance tolerances are determined by the manufacturer’s data and the project specification. A short commissioning period is not evidence of efficiency if the records are incomplete.
What load should be used for commissioning?
Use the load schedule and the project specification. The test should cover the largest step or motor start, the sustained demand and the light-load condition, with the annual test range of 50% to 100% of rating as the planning reference. The 10% to 20% light-load range is also relevant because prolonged running there has a maintenance implication. The exact steps and hold times belong to the approved test procedure.
What should be in the handover pack?
At minimum: approved as-built drawings, protection settings, commissioning test sheets, the defect log with closures, O&M manuals, the training record, the spare parts list, fuel and fluid records, and the warranty start with service contacts. The commissioning readings should be kept as the baseline for the annual load test and the first service visit.
Send the single-line diagram and the load schedule
Tell us the generator rating, the ATS arrangement, the largest step load, the site ambient and the handover date. The reply sets out the commissioning sequence, the records to prepare and the points that should be witnessed, in writing.
Written and maintained by the KVA Grid sales engineering desk. The planning figures on this page are the ones published elsewhere on this site: the annual load test at 50% to 100% of rating and the shortened oil interval for 10% to 20% light-load running from maintenance scheduling; the fuel figures of 0.24 to 0.29 litres per delivered kWh, with 101, 80 and 60 litres an hour for a 500 kVA set at full, 75% and 50% load, from fuel consumption; the UPS bridging the seconds between mains loss and the generator carrying load, the PUE 1.4 example and the 1000 kVA case at roughly 200 litres an hour and about 4800 litres for 24 hours from data center backup power; the 40 C reference ambient and derating above it from hot and humid climates; the base tank covering eight to twelve hours at full load from fuel tank and autonomy; and repeatable load points from load bank testing. The factory test and site acceptance sequence is covered in factory test and site acceptance. Test voltages, hold times, load steps, transfer times and acceptance tolerances are planning values only; the equipment manufacturer’s data and the project specification govern the system that is built. Nothing on this page is a quotation for a particular project. Reviewed 3 November 2026.
