Dissolved gas analysis is a nonintrusive tool used to determine the internal condition of power transformers. Over the years, Southern Company Transmission has made tremendous progress in refining its initial program. Managers who were once skeptical of our recommendations to remove a large generator step-up unit (GSU) from service based solely on dissolved gas analysis (DGA) results now rely on us to routinely make these decisions. Quite often, we have been able to repair critical units that eventually would have failed if left unattended. Although we don't catch all the problems, as some units seem destined to fail in spite of our efforts, the ability to obtain a rush sample and have results within an hour would definitely give us an upper hand.

Alabama Power (Birmingham, Alabama, U.S.), a Southern Company, currently has some 4000 power transformers ranging in size from 46/12-kV distribution type to large 1000-MVA generator step-up units and 500/230-kV autotransformers. Our normal sampling interval is annual with the exception of GSUs, which are sampled quarterly. We also sample 900 load-tap changers twice a year.

The conventional method for obtaining this information requires that an oil sample be pulled into a glass syringe and transported to a lab for analysis using a gas chromatograph (GC). Several hours or even days can elapse before results are obtained. However, a new method allows DGA results to be obtained in the field, eliminating lengthy delays.

GOOD AND GETTING BETTER

Alabama Power began using DGA in the late 1970s to detect incipient faults within power transformers. Samples were pulled on large units only when internal problems were suspected. However, because of limited experience with results interpretation, small developing faults were often overlooked, eventually leading to unit failure. Based on the experience of others in the industry and published papers, Alabama Power implemented a program in 1984 to use DGA on a scheduled interval and maintain historical results for comparison and trending on large critical units. Over the past 20 years, we have refined the program to include all power transformers with at least annual sampling. In 1996, Alabama Power implemented a program using DGA for load-tap changers. Both programs have been successful in detecting faults before failure and extending maintenance intervals with load-tap changers.

Alabama Power's in-house DGA lab is located in the geographical center of the state and runs approximately 6500 samples per year. Six regional maintenance centers pull routine samples and send them via courier in boxes of eight. There is often as long as a two-week delay between sampling and getting the results back. If a sample exceeds established flag-point values, a confirmation sample is requested; on critical units or with high levels of fault gases, we require a rush sample. This process can take as long as two days to get the results depending on the location of the suspect unit. If necessary, the samples are delivered to the lab directly from the job site. But, from remote locations this still can take as long as six hours.

Our operating procedures call for a DGA sample before reenergizing a GSU that has tripped off-line. Loss of revenue alone for an 800-MW unit is astronomical when you're waiting on lab results to determine if a unit can be placed back in service. On-line DGA is the preferred method, but until budgets allow installation of such devices on critical units, we still would prefer a faster method of obtaining DGA results.

NEW TECHNOLOGY

Alabama Power is continually searching for new ideas and products that will enhance its maintenance process and increase reliability. Over the years, we have experimented with several ideas and products of which some were successful and others were eventually scrapped. This leads us to be cautious and often skeptical when approached by a manufacturer with a new technology.

When Kelman (Lisburn, Northern Ireland) first demonstrated the Transport X prototype to us in late 2003, we were skeptical that DGA results could be obtained from such a small device using technology with which we were not familiar. Furthermore, to use it in the field under harsh conditions seemed impossible. We were familiar with our lab GC using vacuum extraction in a controlled environment.

However, after some discussions, we realized there might be potential for this technology and agreed to be one of two U.S. utilities to beta test the Transport X. We determined that duplicate samples taken from the same syringe ran on both our GC and the Transport X was the best method for comparing beta tests. We initially located the Transport X at our lab to allow the chemist to compare the two. After lab testing, the unit would be taken to the field to test its performance in real operating conditions.

PHOTO-ACOUSTIC SPECTROSCOPY

Transport X uses infrared photo-acoustic spectroscopy rather than conventional GC to measure the fault gases dissolved in an oil sample. The figure on page 41 shows the conceptual design of a practical photo-acoustic spectroscopy measurement module. A simple hot-wire source produces broadband radiation across the infrared range that is focused into the measurement cell using a parabolic mirror. The chopper wheel rotates at a constant speed giving a stroboscopic effect to the radiation. Before reaching the measurement cell, the radiation is passed through one of several optical filters. These filters are designed to transmit specific wavelengths chosen to excite one of the compounds under investigation.

The sample is introduced into the measurement cell, and the acoustic signal level is recorded at the chopper frequency from the microphones as each optical filter is indexed into the light path. The series of readings produced then give the concentration of the desired compounds in the sample.

When using the Transport X, the oil sample is drawn from the transformer in the conventional manner. It is then introduced into the measurement container directly from the sampling syringe. The oil is stirred while the air in the headspace is recirculated though the sampling loop and oil to extract the dissolved gases. Once equilibrium has been established, the headspace gases are analyzed using the photo-acoustic spectrometer and the results are presented on the integrated display.

LABORATORY RESULTS

We discussed our beta-test ideas with our in-house chemist, and he suggested taking the comparison test to a higher level than we had originally planned. The methodology was initially to generate a large number of identical oil samples, each spiked with a known cocktail of the typical transformer fault gases. These samples were then analyzed using three different GC instruments and two Transport X units (prototype and preproduction units). The tests were performed over a two-month period in late 2003.

Although the sample size was relatively small, the use of identical samples throughout the test meant that some interesting conclusions could be drawn from this study. It was clear that in this study the results from the Transport X system exhibited less variability than the spread between the three GC instruments. The repeatability of the Transport X results were never worse than the GCs and, in many cases, were significantly better, especially with carbon monoxide and carbon dioxide.

In this sample, the accuracy of the Transport X measurements appears at least equal to that of the GCs and, in some cases, it is better than the GC average. In general, it is fair to conclude that the Transport X and GC measurements are comparable and that other effects, such as sample preparation and handling, could easily mask the differences between the two techniques.

FIELD RESULTS AND APPLICATION

Our lab testing gave us a high degree of confidence in the results from the Transport X. Examination of subsequent test data from other laboratories over a much wider range of operating conditions and gas levels confirmed our laboratory findings, showing a very strong correlation for all gases.

The results of controlled lab experiments and practical field trials demonstrate that the photo-acoustic spectroscopy and modified headspace gas extraction used in the Transport X is directly comparable to the conventional techniques of GC used in a wide variety of labs. In many cases, the variability of the Transport X results is significantly better than that displayed by some of the GC results.

In general, we concluded that the two techniques would give directly comparable DGA results over a wide range of typical transformer and load-tap changer oils, and that any differences between the two may be attributed to normal experimental errors.

After completion of the beta testing, two commercial products were purchased, and we began using the Transport X in the field. The units are used whenever a confirmation or rush sample is required. As of 2006, we now have six units located in each of the six maintenance areas and one unit kept within the Equipment Test Group. The Generation Group has also expressed interest in using the device for sampling auxiliary transformers located inside the plants.

We do not plan to replace our current in-house lab program. Rather, we will complement it with the field units. Over the past three years, these have been used for emergency testing, quite often at night and on weekends. We have seen both lab overtime and callout reduced by having the Transport X units strategically located in the maintenance areas. We have found the units easy to operate, requiring little expertise other than the ability to pull a good sample. The actual gas values are easily obtainable, and the test takes approximately 20 minutes to run.

THE TOOL OF CHOICE

Kelman has made improvements to the Transport X including allowing communication with a PC or laptop, which provides the means to download records, upload information and export into third-party programs. This development allows users who are unfamiliar with the interpretation of DGA results to run analysis programs, define threshold values, trend results and generate reports.

If we only had one tool in our toolbox to determine the condition of our transformer fleet, Alabama Power would choose DGA as that tool. We have made much progress since our program was first implemented 20 years ago. Although we can't catch all imminent failures using DGA, our success rate far exceeds our failure rate. We feel that with the addition of the Transport X, DGA in a Box, we now have the ability to make critical decisions concerning problem units quicker than before, thus improving our reliability.

---

Danny Bates has been the equipment test team leader at Southern Company Transmission since 1985. The Equipment Test Group provides maintenance and testing support to the maintenance regions and generating plants, and researches and implements on-line monitoring and on-line test devices. DEBATES@southernco.com

JUSTIFICATION

During beta testing of the Transport X, Alabama Power changed high-side taps on a unit that had low levels of arcing gases: 5-ppm acetylene. Two days after placing the unit back in service, a sample was run using the Transport X. Results indicated that severe arcing was going on inside the unit: 788-ppm acetylene. Preparations were made to switch out the unit immediately. After the unit was taken out of service, we took another sample to determine the rate of rise. This new sample contained 992-ppm acetylene; results were confirmed by the lab we sent a sample to. Internal inspection revealed a disk-to-disk failure in the high-voltage tap winding. If left in service, this unit would likely have had a violent failure. This case alone was enough to justify our efforts during beta testing of the Transport X.