- The average age of power transformers is >42 years and increasing 0.6 years per year.
- Transformer failure rates, both catastrophic and non-catastrophic, continue to increase.
Funding the cost to replace enough power transformers to reduce or flatten the growth of the average age is not an alternative for most utilities. This situation demands the best asset management and condition assessment approaches available to garner the most value from the existing fleet while maintaining reliability to ever increasing standards. Dissolved Gas Analysis (DGA) of transformer insulating oil is considered the single best indicator of a transformer’s overall condition and is practiced universally today. However, interpretation of DGA data through the use of DGA diagnostic tools is not always “state of the art”.
The use of appropriate DGA diagnostic methods can improve the conclusions from the DGA process which results in improved service reliability, avoidance of transformer failure and deferred capital expenditures for new transformer assets.
This is one of the most frequently used diagnostic tools and, unfortunately, one of the weakest in our arsenal. The combination of frequent use and poor diagnostic capability unite with the result being the source of a significant number of misdiagnoses in the field. Why is the Key Gas Procedure so frequently used when it is not accurate? Because it is a good “shortcut” or “approximation” technique that can be implemented quickly.
Key gases are defined in the IEEE guide as “gases generated in oil-filled transformers that can be used for qualitative determination of fault types, based on which gases are typical or predominant at various temperatures.”
The IEEE guide Key Gas Method offers diagnosis through calculating the relative proportions (in percent) of these key gases to the rest of the gases in the transformer. The proportions indicate the general fault type and these fault types with their relative proportions of gases (in percent) are identified in Table 1.
There are a few issues that contribute to the Key Gas Method’s poor diagnostic accuracy:
- There are only four generalized fault types named while other diagnostic methods offer more detailed fault type identification
- Transformers will typically not exhibit the exact relative proportions of gases outlined by the IEEE guide and users need to make a “judgment call” as to which fault type is being indicated
- Users frequently mistake the IEEE-defined qualitative nature of this diagnostic to be more absolute as in the nature of a quantitative method
Studies based on the IEC data bank of inspected transformers have shown the Key Gas Method to arrive at an incorrect diagnosis 58% of the time. This is a significant error and suggests that this method should be subordinated or eliminated in favor of more accurate approaches to DGA diagnostics.
The problem gets still larger when a “modified” version of the Key Gas Method is practiced. This version, not found in any guide, nor supported by any empirical evidence, matches the change in a single gas to a general fault type. There are no normal, caution or warning levels defined; only the judgment of the practitioner determines the level of the problem. The method is based on a number of assumptions shown below along with an analysis of the assumptions:
- Assumption No. 1: Acetylene in a transformer is caused by an arc; therefore it is possible to diagnose an arcing condition solely by looking at acetylene.
Analysis:
This assumption does not enable the practitioner to understand the nature of the arc. Is it a harmless case of sparking partial discharge in oil from a poorly grounded part or, is it the early stages of a dangerous high energy discharge? Could it be acetylene formed in a localized high temperature fault in oil rather than in an electrical arc? Maybe it is the result of communication between the oil of the main tank and LTC tank? None of these questions can be answered without taking the ratio of acetylene to other gases into account. The lack of an accurate diagnosis could mean that either a harmless situation is over-treated (de-energizing, draining and inspecting the tank and finding no evidence, because partial discharges typically cannot be visually identified) or, allowing a serious problem to worsen.
- Assumption No. 2: CO in a transformer indicates overheated cellulose, therefore it is possible to diagnose cellulose problems solely by looking at CO
Analysis:
Based on the IEC data bank of inspected cases in service when using the formation of CO only to detect paper involvement in a fault, a wrong diagnosis will be provided in about 65% of cases .
Furthermore, increasing amounts of CO in service do not necessarily mean that there is a fault involving paper. This very much depends on the corresponding amounts of CO2. Indeed, in a large number of cases, CO increases are related to oil oxidation only, as a result of overheating, even in transformers equipped with air-preservation systems, where some oxygen is always present because of leaks in these systems.
CO ppm by itself is not a reliable indicator of localized paper-insulation damage because:
a) the level is usually reduced by dilution in a large quantity of oil,
b) the level is affected by oil-temperature (absorption & de-sorption by paper insulation) caused by load and/or ambient-temperature changes and
c) its tendency to escape depending upon the type of oil expansion system and how tightly the transformer is sealed.
- Assumption No. 3: Hydrogen indicates partial discharge as well as other faults, therefore it is important to measure hydrogen
Analysis:
Hydrogen appears in almost all fault conditions (see Table 1) and is therefore an indicator and not a diagnostic gas. It must be combined in a ratio-based analysis with other gases in order to begin to diagnose an incipient fault. Unfortunately, the three gases listed here in the “modified” key gas method do not combine into any meaningful ratios with the exception of a H2/C2H2 ratio indicating LTC communication with the main tank. In fact, when gas ratios are used, methane offers better diagnostic capability than hydrogen. This is because hydrogen is the least soluble gas in oil and also has a high diffusion rate (escapes easily from the transformer or laboratory oil sample) making the exact quantification of hydrogen difficult.
This “modified” key gas approach has not been formally evaluated regarding its accuracy due to the fact that it is an undocumented “de-facto” approach. The documented Key Gas Method offers the worst diagnostic record of any approach evaluated here and the “modified” version has elements in it that would lead us to believe it is inferior even to the Key Gas Method. Further, with the gaining popularity of on-line DGA monitors the pitfalls of this approach using on-line monitors are exacerbated. This is because a monitor with this three gas combination cannot support any diagnostics, as explained above, and would simply offer more frequent misdiagnoses.
Ratio-based Diagnostic Tools
The Doernenburg Ratios, Rogers Rogers, Basic Gas Ratios diagnostic tools have a more effective diagnostic accuracy rate. They involve more calculation and therefore aren’t always the first choice. However, these tools can offer superior results and there are now more automated means of calculating these results.
The ratios used by three of the methods are listed below. The process for each method uses a subset of these ratios with diagnosis of fault type based on the fit of each ratio result to a specific range of values.
Ratio 1 (R1) = CH4/H2
Ratio 2 (R2) = C2H2/C2H4
Ratio 3 (R3) = C2H2/CH4
Ratio 4 (R4) = C2H6/C2H2
Ratio 5 (R5) = C2H4/C2H6
Another ratio-based method is the Duval Triangle found in the IEC guide Annex B.3. The Triangle method was developed empirically in the early 1970’s. It is based on the use of 3 gases (CH4, C2H4 and C2H2) corresponding to the increasing energy levels of gas formation. One advantage of this method is that it always provides a diagnosis, with a low percentage of wrong diagnoses.
The accuracy of the main diagnostic methods used has been evaluated, using the IEC data bank of inspected transformer failures and other reports. Table 2 shows the results of this effort:
Table 2 summarizes many of the main points developed in this article. There are a variety of diagnostic tools available to the DGA practitioner and it is important to understand which ones to apply and when.
On-line DGA and Diagnostic Tools
On-line DGA is gaining in popularity worldwide and, because of its frequency and accuracy of measurements, provides the best data set for the ratio-based diagnostic tools. The advent of on-line DGA has enabled the DGA practitioner to go from infrequent snapshots in time of transformer condition to understanding the dynamic behavior of gases over the operating cycles of the transformer. On-line DGA data is delivering new insights previously unavailable. The trending of the diagnostic ratios rather than just the basic gas data is the breakthrough. The value is in rate of change of diagnostic ratio indicators, not in static snapshots. On-line DGA monitors are available now with different combinations and counts of the 8 diagnostic gases. Software and services also accompany these monitors that automate the calculation and display of some of the diagnostic tools.
However, let the buyer beware: To deliver on the new insights, on-line DGA monitors must offer the right combinations of gases measured that support the diagnostic tools you prefer to use. As an example: If the gases measured by the on-line monitor only support a key gas type of approach, (or worse, the “modified” key gas method highlighted earlier) then the promise of trending diagnostic ratios and delivering correct diagnoses is not fulfilled. It would just become a more expensive way to misdiagnose transformer faults.
Summary
Progress has been made by the DGA community in developing and refining diagnostic tools. Some tools are proving to perform better than others and DGA practitioners can benefit from reviewing the latest information and incorporating it into their DGA procedures. The imperative for transformer asset managers in the current environment of an aging transformer fleet that needs to perform more reliably than ever under increasing loads: Be as effective as possible in transformer condition assessment through a robust DGA diagnostic program incorporating laboratory as well as on-line DGA approaches.
