A genset step load acceptance test asks a set to take a defined block of load in one movement, and records how far the voltage and the frequency fall before both recover. Tesla Power reads that record in the test bay, and the figures below are what a supplier needs before a quotation is fixed for an export order.
- Write the step as a share of the rating rather than as a kilowatt figure, because the same block lands differently on two frames.
- Record the trough and the recovery time together, since a shallow dip lasting four seconds fails a load that a deep one of 200ms passes.
- Check the class printed on the plate against the 400kW frame on the Weichai WP13 platform, where the class sits beside the rating.
- Ask for the record at the ambient the site will see, because a set tested at 20 ℃ accepts a larger step than the same set at 45 ℃.
What does genset step load acceptance decide in a quotation?
It decides whether the frame is large enough, and it decides that in a way a load total cannot. A set can carry 320kW all afternoon and still fail the first two seconds of a chiller restart, because the block arrives faster than the governor can add fuel. The acceptance figure is the one line that separates a frame that works from a frame that trips on transfer. A genset for standby duty takes the whole block at transfer, and a genset in parallel operation splits it between two frames, so the block lands on each of them at a different depth. Tesla Power reads that figure on the site sheet before the plate is printed.
Two buyers can order the same frame and receive two different machines. One order states the step, the class and the ambient; the other states the rating alone. The second order is filled to the rating, and the difference appears at the first transfer rather than on the invoice. Most genset manufacturers publish a reference figure and leave the class to the buyer. A genset for a hospital adds a G3 step beside the imaging load, and the two are not the same test.
How do the ISO 8528-5 classes set the step size?
The standard sorts sets into four performance classes by the disturbance they may show while taking a step. G1 covers general plant load, G2 covers lighting and small motors, G3 covers equipment connected to the bus when the step lands, and G4 covers data processing and electronic load. Each class fixes a frequency deviation, a voltage deviation and a recovery time, and the three move together.
The practical reading is that a G1 set may drop about 10% of frequency and 15% of voltage before it recovers, a G3 set may drop less on both counts and recover in about three seconds, and a G4 set is asked to hold a band narrow enough that a switch mode supply does not reset. The percentages matter less than the pairing, because a step accepted with a 10% frequency drop is a different machine from one holding 4%. Tesla Power reads that figure on the site sheet before the plate is printed.
| Class | Step the test applies | Frequency drop allowed | Recovery to the steady band |
|---|---|---|---|
| G1 | Up to half the rating in one block | About 10% | Loose, several seconds |
| G2 | Up to half, then a second block after recovery | About 7% | Around five seconds |
| G3 | About a third, applied at the worst moment | About 5% | Around three seconds |
| G4 | About a quarter, with load already on the bus | About 4% | Under two seconds |
The class belongs on the enquiry beside the step, and a set quoted without one is quoted against an assumption. Most tenders in our export file name G2 by default, which suits a factory and fails a hospital running imaging equipment through a transfer.
Which step does a 400kVA frame accept at 0.8 power factor?
Work the figure on the frame most export enquiries start from. A 400kVA alternator at 0.8 power factor carries 320kW, and a step of half that rating is 160kW arriving in about 150ms. In per unit terms 160 ÷ 320 = 0.5, so the set is asked to double its output in one movement of the fuel rack.
That frame meets the block on a G2 test and shows a frequency drop near 8% before the governor catches it, recovering in about four seconds. The same 160kW step against a 250kVA frame, which carries 200kW, works out at 160 ÷ 200 = 0.8 per unit, and no governor class recovers that inside the band. The second case is what we see when a buyer sizes on a running total and leaves the block out.
Where is the step measured, and with what instrument?
Measure at the set terminals, on the outgoing side of the breaker, with a recorder that samples at least every half cycle. A controller trend log at one second resolution shows the recovery and hides the trough, and the trough is the figure that trips a contactor. Clamp the voltage leads at the busbar and the current leads around the cable feeding the stepped load.
Three readings belong in the record: the pre-step voltage, the lowest voltage before recovery, and the time from the step until both figures sit inside the steady band again. Add the frequency trough, because a governor that holds voltage and drops frequency has a fuel or rack problem rather than an excitation problem, and the two need different fixes.
How does the governor limit the first two seconds?
A mechanical governor reacts to a speed error and moves the rack through a linkage, which takes about 300ms from the step to the first fuel increase. An electronic governor reads speed fifty times a second and moves an actuator in under 50ms, which is why two sets with the same engine can show a frequency trough a full point apart. The engine still has to burn the fuel before the alternator sees the result.
A turbocharged engine at light load has little exhaust energy to spin the compressor, so the first second after a step is close to a naturally aspirated response. Tesla Power Equipment writes the turbo lag into the test note for that reason, and we take the step from 10% load rather than from idle, because a site that transfers at three in the morning transfers onto a cold, lightly loaded set.
What does a motor start add to an acceptance test?
A motor start is a step with a shape of its own. The first two cycles draw six times running current at a low power factor, the current then falls as the rotor accelerates, and the block lasts two to four seconds instead of 150ms. A set that accepts half its rating in a resistive step can still fail an 80kW direct on line start, because that start asks for roughly five times the motor rating.
Test the two separately. Run the resistive block to prove the governor, then run the motor start to prove the alternator and the excitation. A combined test tells a buyer that something failed without telling which part, and the answer changes the remedy: a governor setting for the first case, a larger alternator or a soft starter for the second.
Which failures show up on the test bed rather than on site?
Three failures appear in the bay and stay hidden until the first real transfer. The first is a governor that reaches its fuel limit before the step is met, which shows as a frequency falling for two seconds and then climbing slowly. The second is excitation running out of forcing, where voltage recovers to 88% and sits there. The third is a breaker whose short time delay opens on the step.
We have seen the third on a coastal order where the breaker was set to a motor protective curve rather than a generator curve. The set passed the bay test into a resistive bank and tripped on site into a pump, because a bank has no starting current and a pump does. A test that uses only a resistive bank proves the engine and leaves the protection unproven.
How should a buyer write the acceptance step into an order?
Five lines close the argument. Name the class, the step as a share of rating, the power factor at which the block is applied, the ambient at the test, and the instrument that will record it. A sixth line costs nothing and helps: the largest motor on site with its starting method, so the supplier can decide whether a motor start forms part of the run. The reference condition is defined in the ISO 8528-1 rating definitions, which names the tolerance and the test point.
Write the step as a block that arrives in one movement, not as a ramp. A ramp of 30 seconds is a load test rather than an acceptance test, and a set that passes a ramp can fail the block by a wide margin. Where the site cannot say how fast the block arrives, take the closing time of the transfer switch as the answer. The same frame appears in the 120kW set on the WP6 platform, where the plate states the same basis.
Which records prove the step was accepted?
Three documents carry the proof. The factory test sheet names the class, the step and the ambient, and carries the trough values beside them. The load bank printout holds the waveform at half cycle resolution. The commissioning record repeats the test on site at the ambient of the day, and the comparison shows whether the installation changed the answer. The same frame appears in the 1000kW container build for standby duty, where the plate states the same basis.
Keep all three in the order file, because a warranty question about a trip is read against them. Conformity assessment for the assembly then checks the class on the plate against the class on the test sheet. The engineering notes on site planning set out what a site should hold at handover, and the ISO 8528-5 transient limits define the envelopes the test sheet is read against.
Yangzhou Tesla Power sends the three as one file rather than as three emails, and the factory and the test bay keep the waveform beside the plate record. A buyer who reads the trough, the class and the ambient on one page can compare two suppliers; one who reads a rating alone can only compare two prices. The same practice holds on the 500kVA silent set with the ATS option.
Frequently asked questions about genset step load acceptance
Half the rating in one block suits most standby sites, provided the plate says G2 or better. A site with a large chiller or a fire pump should ask for the motor start as a separate line, because that block is longer than a resistive one. Tesla Power reads that figure on the site sheet before the plate is printed.
It takes a bigger block in proportion to its rating, and the proportion is what matters. A 500kVA frame at half its rating is the same shape of block as a 250kVA frame at half of its own, and both sit at the same per unit figure.
Because the engine has to burn more fuel before the alternator can deliver more current, while the excitation responds inside one cycle. The frequency trough therefore tracks the governor, and the voltage trough tracks alternator impedance and the forcing ceiling.
Around three seconds on a G3 site and under two on a G4 site, measured to the moment both figures sit inside the steady band. A set that reaches the band in one second and leaves it again has not recovered, and the record needs the whole trace. Tesla Power reads that figure on the site sheet before the plate is printed.
Often it can, because it turns one block into a ramp and cuts apparent demand by about half. The trade is start time and harmonics, and a site with sensitive load on the same bus weighs the second one before choosing a starter.
It changes the kilowatt figure behind the percentage, because the engine gives up capacity as intake air warms. A set tested at 45 ℃ takes a block about 15% smaller than the same set tested at 20 ℃, at the same class.
G3 covers most hospital load, and G4 suits the imaging suite and the server room behind it. The choice belongs in the enquiry, since the class changes the alternator and the governor rather than the enclosure. Tesla Power prints the class on the plate for that reason.
Some do and many do not, and the ones who do will put the class, the step and the ambient on the same line. A quotation without those three cannot be compared with a second one, and the gap shows at transfer.