A large hospitality facility lost utility power for a few minutes when a breaker was opened in error at the serving substation. The onsite emergency generators started as designed. Within a couple of seconds, two of the three paralleled generators tripped OFF and the emergency bus was not fully supplied. The generators had run for years without a problem, passed their routine exercise, and had recently received upgraded digital voltage regulators.
The utility event started the outage. The investigation had to explain why the backup system then failed to carry the emergency load.
Simplified Explanation
Each generator's voltage regulator uses a current transformer (CT) to measure how much current the generator is delivering and in which direction. The CTs on the paralleled generators were installed backwards (reverse polarity). As a result, each regulator read its own reactive power with the wrong sign. When the generators were connected together and picked up load, the regulators "saw" a condition that looked like a loss of excitation and shut their generators down to protect them.
A secondary, contributing factor was the voltage regulator settings. The regulators were set with no voltage droop and slightly different voltage set points, so the paralleled generators worked against each other and reactive current circulated between them.
How it works
Paralleled generators share real power (kW) through their engine governors and reactive power (kVAR) through their voltage regulators. A generator that is overexcited exports kVAR. A generator that is underexcited absorbs kVAR. When a generator absorbs too much reactive power it is at risk of losing synchronism, so voltage regulators and generator protection include a loss-of-excitation (reverse VAR) function that trips the unit.
Reversing a CT flips the sign of the current the regulator measures. The voltage is still correct, so the regulator still computes a power value, but the reactive power reads with the wrong sign. The failure sequence is:
- Utility power is lost and the generators start and close onto a common bus.
- Small differences in voltage set point cause reactive current to circulate between the paralleled units. With no droop programmed, nothing in the regulators acts to correct the difference.
- The generators with the higher voltage set points are actually exporting kVAR, and their field voltage rises. Because their CTs are reversed, their regulators read that export as kVAR being absorbed.
- Those units read a large reverse-VAR condition and trip on "Excitation Loss (Reverse VAR)," even though they were overexcited, not underexcited.
- The remaining generator is overloaded or isolated from the bus it was meant to supply, and the emergency load is not carried.
Reversed CT polarity can go unnoticed for years. A single generator running alone on a load bank at unity power factor may look normal. The fault only shows itself when generators are paralleled and the reactive load is shared, which is exactly the condition during an actual outage.
What the evidence looked like
- The regulator alarm logs showed the tripped units operated on reverse VAR (loss of excitation).
- Functional testing reproduced the event. Lowering the voltage reference on any one generator caused the other two to trip OFF. The unit with the lowered reference stayed ON and read a large positive kVAR, which is the opposite of what the physics required.
- Inspection of the CT secondary wiring at each generator found each CT installed in reverse polarity relative to the regulator's current-sensing input. Correcting the wiring and repeating the test with a reactive load bank produced normal load sharing.
- The as-found voltage droop was set to 0% on the affected units.
What it means for owners
- It is recommended that owners verify CT polarity at every generator whenever a voltage regulator, governor, or paralleling control is replaced or upgraded.
- Confirm that droop, cross-current compensation, or the paralleling controller's reactive-sharing method is set consistently on every unit that is designed to parallel, using the target value from the manufacturer and the system designer.
- Perform a periodic full-load test with the utility source actually disconnected, so the generators parallel and carry the real emergency load profile. A monthly exercise with a single unit on a resistive load bank does not test reactive sharing.
- Include a reactive (inductive) load bank in commissioning and recommissioning. A resistive-only load bank can hide a reactive-sharing problem.
- After any corrective wiring change, recommission the system. A wiring correction that fixes one circuit can disturb another protection or metering circuit that shared the same CT.
- Install relays or controllers that capture oscillography and event logs, and keep the clocks synchronized. Without event data, some questions about the sequence cannot be answered afterward.
What is still unknown or commonly misread
One question remained unanswered: why one generator stayed online when the other two tripped. Several hypotheses fit the data. It may have been lightly loaded, it may have been the slowest to trip, or its own reversed CT may have made an underexcited condition read as overexcited. Without event records captured during the outage, answering that question would have required extensive modeling, which is why recommendation 6 matters.
The utility is often blamed in these outages. A utility outage is an expected event, and the emergency power system is designed for it, so the liability question is usually why the backup system did not perform as designed.
Standards referenced
- NFPA 110, Standard for Emergency and Standby Power Systems (testing and maintenance program)
- NFPA 70E, Standard for Electrical Safety in the Workplace (arc flash labeling was also absent in this facility)
- Generator and regulator manufacturer instructions for droop and paralleling settings
Details have been generalized to protect client confidentiality.