Optimum Transformer Cooling Control with Fiber Optic Temperature Sensors

by J. N. Bérubé, B. L. Broweleit and J. Aubin

Cooling control of power transformers is traditionally provided by a winding temperature indicator (WTI) that is based on a measurement of the top-oil temperature and a simulation of the winding hottest spot temperature. This method has some drawbacks and utilities are now considering the application of fiber optic temperature sensors for this critical function. Assessing winding temperature from directly measured top oil temperature can lead to significant errors depending on the cooling mode and shape of load profile. This situation is reflected in the current revision of the IEEE Loading Guide. The industry consensus leans toward a method based on bottom oil temperature and a proper representation of temperature evolution in the cooling duct. The various hot spot temperature calculation methods are reviewed for a real 240MVA transformer where fiber optic sensors have been selected for controlling the cooling banks. It can be shown that during rapid load change, methods used by classic WTI’s can indicate a lower temperature by more than 10°C even if they are properly adjusted for steady state conditions. Our conclusion is that given the dependability of modern fiber optic sensors, long-term performance of transformer cooling can be better achieved with these more accurate monitoring devices.

Winding temperature determination
Winding temperature is a prime concern for transformer operators. This variable needs to be known under all loading conditions, and especially during rapid dynamic load changes. Accurate knowledge of the winding hottest spot temperature is a critical input for calculation of insulation aging, assessment of the risk of bubble evolution, and short term forecasting of overload capability. It is also critical for efficient control of the cooling banks to ensure that the transformer is cooled effectively. Today, several methods are available for on-line determination of the winding hottest spot temperature.

WTI approach
Cooling control of power transformers is traditionally provided by a winding temperature indicator (WTI). These devices typically rely on a measurement of the top-oil temperature and a simulation of the winding hottest spot temperature rise. This method has drawbacks and many utilities are now considering the application of fiber optic temperature sensors for cooling control, because they provide for faster and more appropriate response to sudden load increases.

WTI’s come in several variations. The classical instrument involves a bulb inserted in a thermowell surrounded by insulating oil. To simulate the winding temperature, the thermowell is additionally fitted with a heater element which is fed by a current proportional to transformer loading. The bulb is filled with a gas or liquid with a large coefficient of thermal expansion, and is also connected through a capillary tube to a spiral-wound Bourdon tube in the measurement device. The Bourdon tube will unwind slightly when the gas expands in response to increasing temperature and the torque generated is transmitted to a pointer moving in front of a calibrated scale.

In some variations, the heating element is located directly in the measuring instrument. In more recent devices, the top oil temperature and load current signals are fed to a numerical microprocessor where the winding hot spot temperature is calculated using a standard algorithm based on inputs from the top oil temperature plus an increment for winding hot spot rise.

This simulation method is consistent with the traditional equation used in loading guides since 1945.  It implies that the winding hot spot temperature can be continuously simulated (or calculated) by measurement of the top oil temperature plus a factor proportional to the load, as follows:

This calculation method is based on the thermal diagram shown in Figure 2. This model assumes that the oil temperature at the top of cooling duct is the same as the top oil temperature.

The rated winding hot spot rise above top oil is usually derived from the temperature difference between average winding and average oil temperatures; this difference is then multiplied by a hot spot factor to take into account additional losses toward the winding ends. This hot spot factor is normally estimated using factory heat run test results and must be provided by the manufacturer. For a given load current, the difference between average winding and average oil temperature is assumed to remain the same, operating under any cooling stage. This method was the basis of IEEE and IEC loading guides since 1945 and has been well accepted by the industry.

Now that transformers are operated closer to their rated values and overloading is more frequent, the hot spot calculation method needs improvement to cover occurrence of a step load increase. It was observed that the model was not adequate to simulate the winding temperature during severe dynamic loading conditions, because:

  • Oil temperature in the winding cooling duct can be much higher than top oil temperature
  • The time constant of oil in the cooling duct is much shorter than the top oil time constant
  • Under sudden load variations, oil circulation does not respond immediately.

Due to the above factors, the winding temperature transient value can thus be significantly underestimated.

Direct measurement of winding temperature
In order to rectify these short comings, IEEE is developing a more sophisticated calculation method based on additional parameters extracted from transformer characteristics.  Fiber optic (FO) temperature sensing can be seen as an alternative approach for the determination of winding hot spot temperature under dynamic loading conditions. The use of fiber optic sensors removes several uncertainties in the process, including:

  •  Calculation of eddy and stray losses in the hot spot area
  •  Estimation of oil flow and temperature in the hot spot area
  •  Estimation of winding temperature rise above local oil in hot spot area
  •  Adjustment of these values for various tap positions and cooling stages
  •  Modeling of winding(s) time constant(s)
  •  Modeling of top oil (or top of cooling duct) time constant as a function of oil viscosity

The main constraints with fiber optic sensors are that they must be installed at the winding manufacturing stage and the locations of the hottest spots must be known to the manufacturer. This last requirement also applies to the other methods since determination of winding hot spot rise above oil implies that the location of this hottest spot is known.

Assessment of winding insulation temperature under any loading condition is a critical step for efficient management of a power transformer. The traditional model used for hot spot temperature determination has shown serious limitations, especially under transient conditions. New models based on bottom oil temperature could provide more accurate results but they require that 12 to 15 parameters be determined accurately for each cooling stage.

Direct measurement with fiber optic temperature sensors bypasses these difficulties and uncertainties. These sensors provide dependable information for each winding, under any loading condition, and are sufficiently rugged now to provide cooling control during the entire transformer life.

Field experience with direct temperature measurement
The Priest Rapids Project is located on the Columbia River in Washington State and is operated by the Grant County Public Utility District (GCPUD). The generating station has recently been refurbished with new GSU transformers, 3-Phase, 240 MVA, 13.2/13.2/230kV, with ONAN/ONAF cooling (Figure 3).

The transformers are located on a concrete powerhouse deck that exposes the transformers to solar heat during most of the day. The ambient temperature is high in summer when the power delivery is most critical, and the concrete structure retains heat long after direct exposure to the sun. Since each increase of 7°C in the winding temperature doubles the solid insulation aging rate, Fiber Optic Temperature Sensors were selected for cooling control. They provide the most dependable tool to optimize cooling control and also allow accurate, long term estimation of insulation aging.

The new GSUs have dual-primaries. In the coil assembly, the two LV 13.2 kV windings are erected one above the other as shown in Figure 4a, as “X” and “Y”. Each unit was fitted with 18 fiber optic sensors in the winding plus three more on the core legs (1/ph, not shown here).

The winding sensors are located as shown below:

  • 6 near top of HV winding in spacers between disks or in oil duct (2/ph)
  • 6 near top of upper primary oil duct (X-winding, 2/ph)
  • 6 near bottom of lower primary oil duct (Y-winding, 2/ph)

The data collected during one ONAN heat run test with constant losses is shown in Figure 4b. All 18 sensors operated normally and measurements are very consistent. Fiber optic sensors on similar locations on phase A, B and C give almost identical results indicating a uniform oil circulation in each phase. The differences between phases may be attributable to oil-flow patterns influenced by asymmetrical tank geometry – the no-load tap-changer compartment shares the oil space on one end of the transformer – and the influence from many leads routed to the NLTC.

Design requirements for long life fiber optic probes
Controlling the cooling banks with fiber optic sensors implies that these sensors must be designed to outlast the transformer itself, which normally means periods as long as 50 years. When users started to install fiber optic sensors 20 or 25 years ago, they were happy to see them survive the heatrun test, as their main goal was to confirm the thermal model of their transformers. Today, however, transformer users are demanding more.

Ideally, long-life probes should have the following characteristics:

  • The sensing technology must be absolutely stable with time, immersed in harsh chemical environments, and at elevated temperatures
  • The probes must be designed so they can mechanically survive steady vibrations and frequent thermal cycles
  • The technology that converts the light signal into temperature information must be insensitive to light intensity variations. This is required to avoid problematic compensation for any aging effects in the optical fiber link and any interconnections
  • Fiber connectors must be leak-proof, reliable and maintenance-free. Oil in optical path should not interfere with signal.

These essential points are discussed below:
Neoptix equipment relies upon the well proven GaAs sensing technology. This technology has been used for temperature sensing inside power transformers for almost 15 years, and is based on a phenomenon that any semiconductor material exhibits: the band gap wavelength or absorption shift. As it is a fundamental characteristic of the crystalline structure of GaAs material, it is always the same and thus never requires re-calibration. Furthermore, this absorption shift is absolutely stable as a function of time.

The band gap wavelength technology has a big advantage of being completely insensitive to signal or light intensity. Any aging effects that might happen through the optical link between the electronic module and the GaAs chip will have no effect on temperature reading accuracy. Because temperature reading rates for transformers are relatively slow (e.g. seconds, not milliseconds), the Neoptix electronic module can be set to operate with de-rated light level, which gives extraordinary light source lifespan, calculated at more than 300 years for a typical system.

Neoptix probes are constructed with the GaAs chip being freely mounted at the end of the fiber optic cable. Thus, the sensor “floats” in a very small volume of transformer oil. This insures that the constant mechanical vibration and numerous temperature cycles inside the transformer will not “fatigue” the bond between the GaAs chip and fiber, while also exhibiting virtually no partial discharge.

Recommendations for efficient instal­lation of fiber optic temperature sensors
The 240 MVA transformer studied in this paper has been fitted with a large number of fiber optic probes, i.e., a total of
21 probes. Normally, 8 to 12 probes are considered sufficient, providing some spares in case some probes are broken during installation.

A good discussion of sensor placement is provided in section 4 of the IEEE P1538 Guide for the Determination of Maximum Winding Temperature rise in Liquid Filled Transformers [1].

Typically, 2 probes will be mounted at the hot spot location in each phase. Usually, conductors near the top of a winding are exposed to the maximum leakage field and the highest surrounding oil temperature; however, experience has
shown the hottest spots are down a few conductor layers from the top. Furthermore, at least one additional probe should be used for top oil temperature. ALL probes should be monitored during the heatrun test, and the hottest ones should be selected for long term monitoring. Depending on the winding design parameters, it is often the case that the highest temperatures are found in the low voltage windings.

For long-term reliability, ensure that probes are solidly attached in their intended position. Teflon® material tends to soften as temperature increases, so it is important that the probe cable be well supported at the installation point and along its entire length.

Mountings probes on conductors or in radial spacers?
Although it may seem more accurate to mount fiber optic probes directly on the conductors, experience has shown that it is much simpler to mount them in radial spacers, as shown in Figure 5. Also, it appears that the temperature [3] under the spacer is higher by 1 or 2°C than the surrounding conductor.

Furthermore, when the probes are installed in spacers, they can be installed much later during the transformer manufacturing cycle, reducing the chances of probe breakage. Due to these considerations, most fiber optic sensors today are installed in radial spacers.

Numerous papers have shown that older transformer loading equations were quite imprecise. The new equations presented in recent versions of loading guides are quite complex, and assume knowledge about parameters that are not readily available. A large number of assumptions must be made in order to use them. As demonstrated in this paper, an easier and more accurate solution is to use fiber optic sensors embedded into the transformer windings, both for heatrun tests and long-term transformer cooling control. Conventional control technology (WTI, etc.) is progressing too, so we expect to see more and more transformers in the future where fiber optic sensors and conventional measuring means will coexist. FO probes have been installed in transformers for more than a quarter of a century. The technology has advanced rapidly so we can now envision using them to replace WTI devices to control cooling of newly built transformers. Installation procedures for probes have also matured, resulting in an installation success rate approaching 100%.

Neoptix’ FO sensors have been designed to outlast the life of the transformers in which they are installed. By utilizing FO cooling control and the new loading guide methodology, it is possible to minimize the solid insulation aging during operation, thereby reducing chances of premature equipment failure by this mechanism and making a 50+ year transformer lifetime a realistic expectation. Conversely, FO technology can allow utilities to manage overload conditions more effectively, making better-informed economic decisions about trading Megawatts-hours for transformer life.

The installation of 5 new HHI (Hyundai Heavy Industries, Ulsan, Korea) GSU transformers at Priest Rapids Dam, GCPUD (Grant County Public Utility District) in Washington State has shown that the GaAs sensors can be reliably used to control the cooling fans. Some of these transformers have now been controlled by Neoptix T/Guard+ systems for more than 3 years, with no problems reported.


  1. IEEE P1538-2000 “Guide for Determination of Maximum Winding Temperature Rise in Liquid Filled Transformers”.
  2. IEEE Draft 2006. “Requirements of Winding Temperature Indicators for the Rapid Response and Accurate Measurement of the Hottest Spot Winding Temperature in Liquid Filled Power Transformers”.
  3. Bérubé, Aubin and McDermid. ”Transformer Winding Hot Spot Temperature Determination”, Electric Energy T&D Magazine, March-April 2007.
  4. IEEE C57.91 1995 “IEEE Guide for Loading Mineral-Oil-Immersed Transformers”

About the Authors

Jean-Noël Bérubé is VP Technology at Neoptix Inc. A founding member and partner at Neoptix, since its inception in 2004, he is heavily involved in design and application of fiber optic temperature probes used for hot spot temperature monitoring in power transformers. Bérubé travels worldwide to train transformer manufacturers on how best to install and use fiber optic temperature probes. An electrical engineer and IEEE member with more than 35 years of experience, his field of expertise has been instrument design and applications where optic, electronics and software are applied together.

Bruce Lee Broweleit was educated in Electrical Engineering at Washington State University, has been employed with Grant County PUD for almost 27 years, and is currently pursuing professional licensure in the State of Washington. He is the designated “Subject Matter Expert” for transformer maintenance at the PUD’s Hydro Division.  His transformer-related responsibilities include: factory witness and field testing, electrical and oil test analysis, engineering design, procurement and installation technical specification preparation, on-line DGA monitoring, internal inspections, and making recommendations about maintenance and continued operation. Bruce recently completed replacement of 5 GSUs at Priest Rapids Dam, and presently is Project Manager of a contract to supply six replacement GSUs for Wanapum Dam from 2011 through 2016.

Jacques Aubin is a consultant to the electrical industry on power transformers. He is mainly involved in development and testing of advanced monitoring systems for power transformers. Aubin has spent 30 years with Hydro-Quebec where he was initially involved with transformer design, specifications and acceptance tests. He later directed research activities related to overload, short circuit and other acceptance tests on power transformers. Aubin is a former member of the IEEE transformer committee and he has lead several Working Groups at CIGRE and IEC.

Most consulted news