From the field · Case 04

Generator Temperature Alarm — Is the Fault in the Sensor?

A Vestas V126 was alarming at 116 °C on generator temperature and derating itself, but everything else about the machine looked healthy — no unusual noise, no vibration, cooling running. When the readings and the behaviour disagree like this, one of them is wrong, and it is worth finding out which before you start dismantling anything.

Symptom

• SCADA: generator temperature 116 °C, against a nominal limit below 100 °C
• Consequence: the control system was derating output to protect the machine
• But: no unusual noise, no vibration, cooling fan running, radiator clean
• Trend: the reading had climbed by roughly 2 °C per day over the preceding three days

First assessment

A generator genuinely at 116 °C does not behave politely. You would expect it to smell, to change its sound, and for the cooling circuit to show the load somewhere. None of that was present. That mismatch — a severe number with none of its consequences — is the classic shape of an instrumentation fault rather than a machine fault.

The other clue was the trend. A real thermal problem tracks load and weather; it does not climb by a tidy 2 °C every day regardless. A steady daily drift looks like something in the measurement chain moving, not something in the machine heating up.

Step-by-step diagnosis

1. Took a physical reading with an IR thermometer. This is the step that decides the case. The generator housing measured around 75 °C where SCADA claimed 116 °C — a gap of roughly 41 °C. The machine was not hot; the reading was.
2. Took several readings around the housing. To be sure the IR reading was not a local artefact. The distribution was even and consistent with a healthy machine under that load.
3. Checked the cooling circuit. Fan running, radiator clean, airflow normal. Nothing to explain a real 116 °C, which reinforced the instrumentation hypothesis.
4. Measured the RTD sensor resistance. The element read about 68 Ω where roughly 110 Ω was expected at that temperature. The sensor was reporting a resistance that corresponds to a much hotter machine than the one in front of us.
5. Checked the wiring and terminals. To rule out a loose connection or corroded terminal adding resistance, which can mimic this. Connections were sound, so the drift was in the element itself.
6. Reviewed the sensor's service age. The element had been in service about four years — well within the range where RTD drift becomes common.

Root cause

Calibration drift in the generator RTD temperature sensor. After roughly four years in service the element's resistance characteristic had shifted, and SCADA was faithfully converting that wrong resistance into a wrong temperature. The control system then did exactly what it was designed to do — protect the machine — and derated a generator that was in perfectly good health.

The cost of this fault was not damage. It was lost production, plus the risk of someone opening up a healthy generator chasing a number.

Fix

1. Immediate: SCADA offset applied (30 minutes): a −30 °C correction was entered so the displayed value matched physical measurement, restoring full output while the replacement was arranged. Applied as a documented temporary measure, not a permanent setting.
2. Threshold reviewed: the alarm limit was temporarily raised in a controlled way so the machine would not keep derating on a known-bad reading.
3. Sensor replaced (planned, 3 hours): the drifted RTD element was replaced and the new reading verified against the IR thermometer across several load points.
4. Offset removed: the temporary correction was taken out once the new sensor was confirmed accurate — leaving it in place would have been the dangerous outcome of this repair.

Distinguishing a real thermal fault from a sensor fault

Preventive maintenance

• On any temperature alarm, take a physical reading first. An IR thermometer costs minutes and settles the question of whether you have a machine problem or a measurement problem.
• Treat RTD sensors as consumables. They drift with age. Include a resistance check in the periodic schedule rather than waiting for a nonsense reading.
• Log temporary offsets and put an end date on them. An offset left in place permanently will one day hide a real fault.
• Read the trend shape, not just the value. A tidy, load-independent daily climb is almost never a thermal problem.