The most common misconception in the engine room is that 'load sharing' is one thing. It is not. Active load, kW, is shared by the governors. Reactive load, kVAr, is shared by the AVRs. Two independent systems, two independent failure modes, two different fixes.
Before touching anything, one measurement is required: kW and kVAr per generator, simultaneously, at no fewer than three load points. Without it, everything else is guesswork.
Case A — unequal kW sharing
If the machines show correct, balanced kVAr but unbalanced kW, the problem is on the engine and governor side.
- Different droop settings between machines. In droop, two generators share proportionally only if their droop percentage matches. 3% against 5% gives permanent imbalance.
- One machine in isochronous and one in droop. The isochronous machine absorbs every load change until it reaches its limit.
- A speed-setpoint offset. A few hundredths of a Hz translate into tens of kW.
- A worn mechanical governor or an actuator with hysteresis — the command arrives, the engine does not follow it.
- A genuine power limitation on one machine: fuel supply, turbocharger, exhaust temperature. Here the governor is doing its job correctly and the problem is mechanical.
Case B — unequal kVAr and circulating currents
If kW shares correctly but one machine runs strongly lagging while the other runs leading, there is a circulating current between them. Both heat up, and neither produces useful power for it.
- Cross-current compensation disabled or wrongly connected. This is the most common cause.
- Wrong polarity or wrong phase on the AVR paralleling CT — the system corrects in the wrong direction.
- Different voltage setpoints between the AVRs. The machine with the higher setpoint takes the reactive load.
- An AVR of a different type or generation fitted as a replacement, with settings never aligned.
- One AVR left in manual excitation mode without anyone knowing.
Case C — sharing that changes with load
When sharing is correct at low load and degrades at high load — or the reverse — this is not an offset error but a slope error. Check:
- The droop percentage of each machine, measured rather than read off a plate.
- The calibration of the power-measurement CTs. A CT with the wrong ratio reads correctly at one point and wrongly everywhere else.
- Actuator linearity across the full travel.
- Load-sharer gain and stability settings — too much gain gives oscillation, too little gives permanent offset.
What 'correct' means in practice
For generators of equal rating, a deviation of about ±5% of rated load is accepted on most systems; the exact requirement follows from the manufacturer's specification and the vessel's class rules. For generators of different rating the target is not equal kW but equal percentage loading — and that is a setting, not something that happens by itself.
In every case the adjustment closes with a test: stepped loading, recorded results, and a load rejection to confirm that no machine goes into reverse power during the transient.