Transformer temperature readings are rather imprecise. I have seen adjacent temperature wells on a transformer read several degrees different using the same sensor. Including sensor drift and winding temperature compensation could result in more than a 15 C error.
I turn on fans at lower temperature than the MFR recommends. Most of my transformers have very low fan run times because the transformers typically carry N-0 loads but are sized for N-1 loads. The suggestion might be different for a transformer with constant load where the energy used by the fans is non-trivial or the fans might not last the whole transformer lifetime. I use 80C/90C winding temperature. If the fans are based on oil rather than winding temperature, the settings would be 10-15 C lower.
I would allow at least a 10 C margin trip between the calculated winding temperature at the transformer MVA rating and the trip setting. Although IEEE transformers are based on 30 C ambient + 65 C rise + 15 C winding rise, keep in mind that the 30 C ambient value is a 24 hour average. The standards actually allows a peak ambient of up to 40 C, leading to a meaning an expected normal life expectancy winding temperature of up 120 C. If the transformer is going to be loaded beyond nameplate, the trip would be set even. If excess loading will bring the wind temperature near the 140 C, the risk from high temperature changes from cumulative paper damage to bubble formation, so a different approach may be needed at high levels of allowed overloads.
In general PRC-023-4 requires a 15% margin between relay settings and the Facility Rating. Although high temperature trips work slightly differently, I assume best practice would be to allow a 15% margin in terms of MVA. For certain BES transformers, PRC-023-4 R1.10 prohibits setting top oil supervision below 100 C and prohibits setting winding temperature supervision below 140 C.
My comments are based on the assumption that system operators will be monitoring alarms and will make adjustments to keep the transformer with the published Facility Ratings. Decades ago substations were monitored by customer outage calls and monthly station checks instead of real-time SCADA monitoring. Without active monitoring, I can see why old installations had winding trips set at 110 C.
One other item to consider is whether there needs to be a few second time delay on the trip to prevent inadvertent trips. Such trips could be caused by power cycling a temperature monitor with contacts programmed as failsafe, a loose wiring connection to the RTD, or a momentary fault current that tricks the temperature monitor.
