Most standby sets are exercised monthly. The controller starts the engine, runs it for fifteen or twenty minutes off load, and stops it. The log says the test passed. On the day the mains actually fails, some of those sets trip within a minute of picking up the building.
What an off-load run actually tests
It tests the battery, the starter motor, the fuel solenoid and the fact that the engine will fire. That is worth having, and it is why the exercise run exists. It also keeps seals wet and stops the fuel system gumming up.
What it does not test is everything downstream of the engine starting. Off load, the engine burns a fraction of its rated fuel, the coolant may never reach normal operating temperature, the alternator carries almost no current, and the cooling system is barely working. Nothing in the machine is under the stress it was sold to survive.
What a stepped load test finds
Put a resistive load bank on the set and step it up to full rating, holding at each step, and a specific list of faults reveals itself.
Cooling restriction. A radiator that is fine at 20 per cent load will run to high-temperature shutdown at 100 per cent. Blocked fins, a recirculating hot air path in a plant room, an intake louvre with less free area than the drawing assumed: all of these are invisible off load and all of them stop the set within minutes of a real transfer.
Fuel restriction. A partially blocked filter, a lift pump losing suction on a long run, or a tank vent that cannot breathe fast enough will pass an off-load run and starve the engine at full fuel demand.
Exhaust back pressure. An exhaust run longer or more convoluted than the engine allows shows as lost power and high exhaust temperature under load, and as nothing at all off load.
Voltage and frequency regulation. The AVR and governor only have to do real work when load is applied and removed. Steady-state regulation and the recovery time after a step are numbers, and numbers can be recorded and compared year on year.
Alternator temperature rise. Insulation class is a temperature rise limit, and you cannot verify a rise limit without current in the windings.
Wet stacking. A stepped run to full rating burns off the deposits that months of light-load exercising have left behind.
Resistive, reactive, or both
A resistive load bank loads the engine at unity power factor. It proves the prime mover, the cooling and the fuel system. It does not fully load the alternator, because the alternator’s field is worked hardest by reactive current.
A reactive bank, used together with a resistive one to reach the rated power factor, loads the excitation system and the windings the way a real building does. On a large or critical installation it is worth having. On a small standby set, a resistive test to full kW is a great deal better than the off-load run it replaces.
Record the numbers, not the outcome
A test record that says pass is nearly useless a year later. Record the load at each step, the voltage and frequency at each step, the recovery time after each step change, coolant and exhaust temperature, oil pressure and fuel consumption. Trends in those figures are how a cooling problem gets found while it is still a trend and not a shutdown.