How paralleled generators divide load
When sets of different sizes run in parallel, knowing how many kW each one carries is what decides which combination to start. With governor droop matched and the bus at steady state, real power divides in proportion to each machine's rating, so ratings plus total load are all the inputs needed.
The relation is share = total load × (that set's rating ÷ sum of all ratings), which means every set ends up at the same load rate, namely total load ÷ total rating. Apparent power per set is its kW share divided by power factor, and total reactive power is load × tan(cos⁻¹ PF).
Put a 500 kW and a 300 kW set on the bus with 600 kW of load at 0.8 PF: A takes 375 kW, B takes 225 kW and both sit at 75%, comfortably inside the normal band. Total apparent power is 750 kVA with 450 kVAR of reactive demand. Drop the load to 200 kW and both sets fall to 25%, which is wet stacking territory, so running one machine alone is the better call.
Real installations also have to balance reactive load (AVR cross current compensation), synchronizing conditions, reverse power protection and harmonic interaction. Treat this as a first pass for set selection and loading, and have a licensed engineer or the packager confirm the paralleling switchgear settings and protection coordination.
Frequently asked questions
Paralleled sets all run at the same frequency, so when every governor carries the same droop setting each machine picks up real power in proportion to its own rating. Mismatched droop makes one set hog the load, which is why droop is matched before closing the breaker.
Below about 30% of rating a diesel burns too cool and unburned fuel collects in the exhaust, the condition known as wet stacking. It costs output, makes smoke and damages injectors, so light load periods are usually met by shutting a set down rather than idling both.