It is Not Uncommon for Bus Fault-Current Levels to Exceed 30 kA. Large power transformers, especially older ones with windings not braced for this fault-current magnitude, can experience significant winding deformations and dielectric degradation with repeated fault exposure.

Fault data compiled from four National Electric Energy Testing, Research & Applications Center (NEETRAC; Atlanta, Georgia, U.S.) member utilities indicates that 5% of distribution buses and 8.8% of transmission buses have fault levels exceeding the 30-kA mark. Continued growth of the power grid will exacerbate this situation even more. A practical and consistent on-line winding deformation and dielectric degradation monitoring technique would be valuable, especially for large, essential power transformers.


Many utilities are already using some form of off-line (transformer de-energized and switched out of service) frequency-response analysis (FRA) on transformers, because it is known to be sensitive to winding distortion. In fact, FRA is the only technology that is sensitive to significant winding deformation (in coils, layers, turns and leads) in power transformers. A significant amount of deformation can occur before imminent winding failure and transformer relay operations occur.

After the initial winding deformation occurs, the voltage stress changes in the insulation structure, which leads to partial discharge and gassing over time. However, by the time partial discharge and subsequent gassing appear, degradation of the transformer has already occurred. Winding deformation is one of the first and fundamental precursors to a decline in transformer condition.

Previously, there was no way to perform FRA on an energized and in-service transformer. On-line FRA data would provide the utility with an up-to-date condition assessment of large, essential transformers. On-line FRA adds a new dimension of data to analyze the physical structure of the coils and their dielectric surroundings in a sensitive and definitive fashion while the transformer remains in service.

NEETRAC's on-line FRA method uses normal system-switching operations, such as capacitor bank and reactor operations, along with lightning from local thunderstorms for the FRA test-signal source. This patented technology can perform FRA signatures on transformer windings using a variety of input waveforms with different time and amplitude characteristics.


Georgia Power Co.'s (GPC; Atlanta) O'Hara substation was the location of the first on-line FRA installation and the continuing development work. The initial installation took place in October 2003, and several upgrades have occurred since then. Currently, the hardware and software are being upgraded to a commercial package: a smaller, more portable, lower-cost unit. The on-line FRA system block diagram is shown in Fig. 1.

The O'Hara autotransformer installation consists of three 500/230-kV, 672-MVA single-phase units. The 500-kV and 230-kV capacitive bushing taps are coupled with purpose-designed high-pass filters to enable transient data for FRA and 60-Hz data to be gathered simultaneously. The FRA signature of a winding is affected to some degree by bushing condition, so bushing power factor (PF) is monitored. In addition, there are certain bushing indicators from portions of the winding FRA data.

One of the effects of premature bushing aging or the beginning of bushing failure is the increase in the difference in PF with bushing temperature. Figure 2 shows this effect using a screen shot from On-Line Monitoring Inc.'s PF Live relative PF equipment, which works with the on-line FRA from the same bushing taps. The red and blue traces represent bushings that change PF with a bushing operating temperature change. Figure 2 shows that the difference in relative PF is increasing with time, a trend that should be monitored closely.

In addition to using the high-voltage and low-voltage bushing taps for on-line FRA, a high-pass filter is installed directly across the neutral bushing with a physical mounting on the side of the transformer. This is mounted just after a 35-kV cutout for maintenance access and BIL integrity, as shown in the Fig. 3 inset. Figure 3 shows the FRA trailer setup at O'Hara substation. Figure 4 shows the digitizers and computer inside the trailer.

Figure 5 shows data from the on-line summer 2006 versus fall 2006 for the O'Hara phase-three high-voltage winding for a bandwidth of about 2.5 MHz. The peak amplitude differences are insignificant for detecting winding deformation and usually indicate a difference in temperature, in dielectric loss or in magnitude of input energy. Therefore, there was no significant change for the phase-three high-voltage winding from summer to fall of 2006. In addition, the characteristic similarity of the two transfer functions demonstrates test repeatability using two separate sets of on-line data.

The off-line comparison of all three high-voltage windings is illustrated in Fig. 6. The useful bandwidth for the off-line test stops at about 2.5 MHz; therefore, the on-line data has the same useful frequency content as the off-line test. Note the similarities in the characteristic shape and the frequencies for trace peaks in comparing the on-line data in Fig. 5 with the off-line data in Fig. 6. Also note the same similarities across phases for all three high-voltage windings for the 500-kV sister transformers, from the off-line traces of Fig. 6.


Florida Power & Light Co. (FPL; West Palm Beach, Florida, U.S.) is the location of the second on-line FRA installation, but it is the first attempt of on-line FRA on a 3-phase autotransformer with all three phases in one tank with one external neutral bushing. FPL began recording on-line data on April 19, 2005, for the Plumosus 230/138-kV, 400-MVA, 3-phase autotransformer with tap change under load.

The O'Hara development work was on an installation made up of three single-phase units with access to all three external neutrals. Due to inherent single-phase isolation, the phase source and neutral current contribution for the corresponding phase are automatically known for O'Hara. In contrast, the FPL installation must be able to separate the 3-phase influences from a single composite-neutral waveform, determine which phase is the source of the input pulse, and determine whether the input source is the high- or low-voltage winding.

It is also important to demonstrate that when an FRA measurement is repeated at a later date and the winding conditions have not changed, the same FRA signature can be repeated on the follow-up test. The graph in Fig. 7 represents the two transfer-function magnitudes from the phase-two low-voltage winding of the Plumosus transformer.

The two transfer functions are very similar and do not need to be direct overlays to indicate no change in the low-voltage winding from past (before 2005 hurricanes) to present (after 2005 hurricanes). The blue trace is the transfer function for the first time period and the red trace is the transfer function for the second time period. The similarity of the transfer functions before and after magnitude traces indicate that there was no significant phase-two low-voltage winding or insulation damage during the 2005 hurricane season. In addition to indicating no significant hurricane damage, the results also indicate good data replication over the June-to-October hurricane season.


The FPL autotransformer has “plus 8" (raise) to “neutral” to “minus 8” (lower) tap positions, which only ranged from 2R to 2L during the last year of FRA monitoring. The tap position is checked and recorded in the database immediately before the digitizer waveform downloads, because a change in tap position results in an FRA signature change.

When FPL began FRA monitoring on this transformer, it was assumed that the tap position would be unchanged throughout the test. However, after the first hurricane in 2005, the tap changer operated about 200 times during one month. After the hurricane season, FPL installed a tap-position monitor to record the tap position in the database along with the other data.

Although experiencing some unusable data with unknown tap-position information, the installation of the tap-position monitor proved a valuable step in determining the detection sensitivity of the on-line FRA technique. For example, after the tap monitor installation, FPL was able to detect the change of one tap position from the on-line FRA traces. Of the 3000 pulse group with tap positions recorded, about 2000 were on tap 1L, 900 were on tap 2L and 100 were on 2R. The most significant differences for tap 1L versus 2L data were the frequency shifts in the resonance peaks in the magnitude plot along with the absence of a peak at about 1.65 MHz on tap 2L, compared to the transfer-function data for tap 1L (Fig. 8).

This demonstrates the good sensitivity of the on-line FRA method to a relatively small change in the overall low-voltage winding.


The technology has been developed and demonstrated to perform on-line FRA on a 3-phase autotransformer with three winding assemblies in three separate single-phase tanks. The technology also has performed on-line FRA on a 3-phase autotransformer with three winding assemblies in one tank with one external neutral bushing.

Very good on-line test repeatability is demonstrated up through 2 MHz over a period of almost one year at the FPL installation. Very good on-line test repeatability is demonstrated from season to season, as well as a good match from off-line to on-line data at the Georgia Power installation.

The ability to detect a one-position change on a 17-position tap changer was an unexpected bonus to the development of on-line FRA, and it demonstrates good sensitivity of the method to a relatively small change in overall winding and insulation structure.

The On-Line Monitoring Inc. bushing relative PF equipment continues to work in concert with the on-line FRA equipment to monitor transformer-bushing condition.


This on-line tool, capable of quantifying additional transformer aging parameters, will become an extremely valuable component of future predictive-maintenance programs. This system, combined with other tools like furanic analysis and future software, may be able to predict imminent transformer failures.

There is nothing inherently expensive or maintenance intensive about the NEETRAC on-line FRA method. The method works because of its ability to perform FRA using various transient pulses that normally occur on all power systems, and to produce repeatable and sensitive on-line FRA test results.

NEETRAC plans to install commercial prototype on-line FRA equipment at additional locations among its member utilities. The prototype will be physically smaller, more transportable and will enable on-line FRA monitoring of multiple transformers using one set of equipment. NEETRAC is also seeking commercialization partners.


The authors would like to thank Jim McBride of JMX Services Inc. for his contribution as a software and high-voltage consultant to NEETRAC for the duration of this project. The authors also would like to thank Jeff Benach of On-Line Monitoring Inc. for his engineering contribution to enable the simultaneous operation of the PF Live relative PF system with the NEETRAC on-line FRA equipment.

Larry Coffeen is a research engineer for NEETRAC, an application research group of Georgia Tech with a membership of power equipment manufacturers and electric utilities. Coffeen's work includes development of technology and equipment to perform off-line and on-line transformer frequency-response analysis (FRA). He was with Georgia Power Co. for 29 years in various engineering positions in transmission substation and high-voltage testing. He continued work in the high-voltage laboratory after the formation of NEETRAC in 1996. He holds a BSEE degree from Georgia Tech and three U.S. patents on transformer FRA testing.

Donald Cantrelle, a principal engineer in the substation support group of Georgia Power, provides technical support for substation major equipment. During his 30-year career with Georgia Power, he has worked in distribution engineering, substation test engineering, system protection and substation support. Prior to joining Georgia Power, Cantrelle worked for two years in nuclear plant construction for TVA. He has a BSEE degree from Louisiana State University and is a registered professional engineer in Georgia.

Joe Mango is the transformer and technology manager at Florida Power & Light Co., where he has been employed in various capacities for 27 years. He is currently responsible for all substation and power plant transformers as well as on-line monitoring equipment. Mango has a BSEE degree from Southern Illinois University and a master's of engineering management from the University of South Florida. He is a registered professional engineer in the state of Florida.


As the site of the first on-line frequency-response analysis (FRA) installation, Georgia Power Co.'s O'Hara substation has been closely monitored. These are some of the key results realized from the FRA monitoring:

  • Proved the concept of on-line FRA and that it can be performed in a practical and economical manner, with a bandwidth of at least 2 MHz

  • Demonstrated that on-line FRA can be performed at the 500-kV level and the equipment can survive system switching and local lightning storms

  • Demonstrated that on-line relative power factor of the transformer bushings can be performed simultaneously with on-line FRA

  • Showed a season-to-season match with on-line FRA traces using separate data

  • Obtained a frequency-peak match for off-line FRA versus on-line FRA trace magnitudes using the 12-bit digitizer upgrade.