In short. Commissioning proves that the installed generator system does what the design says, under defined conditions, with evidence. It is a planned sequence of pre-start checks, no-load and loaded tests, protection and alarm proofs, and a full automatic transfer and retransfer, each judged against acceptance criteria agreed before the test. A generator that starts has not been commissioned; a generator whose records show it carried its load steps and transferred correctly has.
Define the boundary and the criteria first
Before anyone presses start, the plan states what is in scope and what counts as a pass. Scope might be the generator alone, the generator with its ATS, or the complete power path from utility failure through transfer to load restoration and return to normal. Acceptance criteria come from the design and the manufacturer’s data: voltage and frequency bands, transient dip and recovery time on defined load steps, start time, transfer and retransfer delays, protection pick-up values and alarm set points. A test without criteria produces numbers without a decision, and criteria invented after the readings are known are not criteria. The plan also states who witnesses each test, what load will be available and whether a load bank is hired; the pre-test list is at generator commissioning checklist.
Pre-start and baseline
Commissioning begins with the installation record, confirming that the set and its interfaces match the drawings and that the installation checks were completed and signed. The engineer then records a baseline: oil and coolant levels and grades, fuel level and condition, battery voltage and charger output, insulation resistance of the alternator and cables, earth resistance, controller firmware and settings, protection settings against the settings sheet, and the mechanical freedom of the exhaust, mounts and flexibles. A machine showing a serious pre-existing problem, such as low alternator insulation resistance or a fuel leak, is not loaded in the hope that the test will explain it. The baseline gives context if anything abnormal appears later in the day.
The test sequence
| Step | What is done | Typical acceptance basis |
|---|---|---|
| Manual start, no load | Crank, fire, idle and rated speed, listen and look for leaks | Starts within the specified cranking cycles, oil pressure and charging within limits |
| Voltage and frequency setting | Confirm 415 V, 50 Hz at no load, phase sequence and phase-to-neutral balance | Within the design tolerance, correct rotation |
| Protection and shutdown proofs | Simulate low oil pressure, high coolant temperature, overspeed, emergency stop, overcurrent and earth fault where testable | Each trips as designed and is logged by the controller |
| Load steps | Apply 25, 50, 75 and 100 percent of rating on a load bank or building load, and the largest design step | Voltage and frequency dip and recovery within the ISO 8528-5 performance class specified |
| Sustained load run | One to two hours at a high load fraction, logging temperatures, pressures, exhaust colour, backpressure and electrical values every 15 minutes | Stable readings within manufacturer limits, no leaks, no alarms |
| Load rejection | Remove full load in one step | Overshoot and recovery within specification, no overspeed trip |
| Automatic transfer | Simulate utility failure, observe start, transfer, retransfer on utility return and cool-down stop | Timers as designed, neutral state correct in each mode, interlocks proven |
| Alarms, monitoring and remote signals | Each alarm in the matrix reaches the panel, BMS or remote system | Alarm matrix fully ticked |
Loaded testing: what the steps reveal
A no-load run proves that the engine turns and the alternator excites. It does not prove full-load fuel flow, heat rejection in a 44 degree room, exhaust backpressure, AVR behaviour under a motor start or termination temperature at rated current. Load steps prove all of these. ISO 8528-5 defines performance classes G1 to G3 for transient voltage and frequency behaviour, and the class specified for the project sets the acceptance band for the dips. Where the building cannot supply a controllable load, a resistive or resistive-reactive load bank is used; the choice and sizing are described in load-bank sizing and the service at load-bank testing. The largest design step, typically the biggest motor or the UPS inrush, is applied against the realistic background load, not against an empty bus, because that is the event the set was sized for.
Transfer and interface testing
A set can run perfectly in isolation while an external permissive, a wrong timer or a mis-wired sensing circuit prevents the system from working. The transfer test simulates utility failure at the sensing point, not by pressing a button on the controller, and observes the whole sequence: start delay, cranking, warm-up, transfer, load acceptance, utility return detection, retransfer delay, retransfer, cool-down and stop. The neutral state in each mode is checked where a four-pole transfer is installed. Interlocks are proven by attempting the prohibited operation, with the set stopped, and confirming it is blocked. Where UPS systems, synchronising or multiple sources are involved, each normal, backup and bypass arrangement has its own agreed check. The control logic is explained at AMF panels.
Protection settings and fault behaviour
Generator fault current is far lower than utility fault current, so protection assumptions from utility operation do not automatically hold. Commissioning confirms the generator breaker and feeder settings against the settings sheet from the design, proves overcurrent and earth-fault operation by secondary injection where relays are fitted, and confirms that the controller’s electrical protections, including over and under voltage, over and under frequency and reverse power where applicable, are set and functional. A failed protection test never leads to raising the setting until the test passes; it leads to finding out why.
Defect management and records
Every test produces a record with the configuration, settings reference, instruments, test condition, readings and result, distinguishing observed results from assumptions and completed checks from deferred ones. Defects are classified as a condition to monitor, a punch item to close before handover, or a stop item that halts testing. Punch items are assigned, dated and closed with evidence. The commissioning record, the settings sheet, the alarm matrix and the as-installed drawings together form the acceptance package that the owner signs and that start-up and handover builds on. Years later, that package is what a diagnostics engineer reads at two in the morning to know what normal looked like, and it is the reference for recommissioning after a relocation, an overhaul, a switchboard change or an emission retrofit.
Related photographs and videos
Common mistakes
- Accepting a no-load start as commissioning; fuel delivery, cooling, exhaust and terminations are never proven and the first real outage is the test.
- Deciding what pass means after the readings are in; the record cannot support acceptance and disputes follow.
- Testing the transfer from the controller’s test button instead of at the mains sensing point; the sensing circuit and timers are never exercised and fail in service.
- Raising a protection setting to get a failed test to pass; the protection is defeated and the next fault damages the alternator or cables.
What to send us
Send the set details and nameplate photos, the approved single-line diagram and transfer scheme, the protection settings sheet, the load bank availability or the building load that can be used, the shutdown window for the transfer test, and the installation completion record with its open items. For recommissioning, add the original commissioning records and a description of what changed. Use the service and AMC enquiry and we will issue a commissioning plan with the test sequence and acceptance criteria before the visit; installation prerequisites are listed under generator installation.
Questions engineers ask
How long should a generator load test last during commissioning?
Long enough for temperatures to stabilise and for the step tests to be completed, which is typically one to two hours at high load plus the step and rejection tests. There is no universal duration; the objective and the manufacturer’s recommendations set it, and the record should state the load and duration achieved.
Is a load bank necessary for commissioning?
It is necessary when the building load cannot be applied in controlled steps up to the set’s rating, which is the case for most new installations. A resistive bank proves engine and cooling capability; a resistive-reactive bank also proves alternator and AVR behaviour at the rated power factor.
What does ISO 8528-5 performance class mean?
It classifies how much voltage and frequency may deviate and how fast they must recover when load is applied or removed. G1 suits simple resistive loads, G2 roughly matches utility quality for general loads, and G3 is for sensitive loads such as telecommunications and UPS systems. The class specified sets the acceptance band for commissioning step tests.
What happens if a test fails?
The failure is recorded with its readings, the cause is investigated, the defect is corrected and the test is repeated under the same conditions. Settings are not changed to make a result pass. If the failure makes further testing unsafe, commissioning stops until the item is resolved.
Technical content reviewed by Srinivas Reddy, Managing Partner, Reddy Generators. Last updated 19 October 2026.