Dissolved Gas Analysis (DGA) is widely used to monitor the condition of mineral oil-immersed transformers. It can help identify abnormal gas generation, track changes over time and provide valuable clues about developing faults.
However, before interpreting DGA results, there is an important step that should not be overlooked: checking whether the data itself is reliable.
According to IEEE Std C57.104-2019, IEEE Guide for the Interpretation of Gases Generated in Mineral Oil-Immersed Transformers, a DGA data quality review should be performed before interpreting the results. The purpose is to identify errors or inconsistencies that could lead to a misleading diagnosis.
So, what should engineers look for when reviewing DGA data?
Why DGA data quality matters
DGA interpretation relies on the assumption that the measured gas concentrations accurately represent what is happening inside the transformer.
In practice, however, errors can occur at almost every stage of the DGA process, including equipment identification, sampling, transportation, laboratory analysis, data entry and reporting.
IEEE C57.104-2019 notes that data corruption can occur throughout the data management, work management, sampling, analysis and reporting stages.
A data quality problem can therefore make a normal transformer appear abnormal, or potentially mask a genuine developing problem.
This is why IEEE recommends reviewing the quality of the DGA data before moving on to fault interpretation.
What is a DGA data quality review?
A DGA data quality review is essentially a check to determine whether the results are credible and consistent with the transformer’s history.
IEEE C57.104-2019 states that the review most often involves comparing the current test information with previous DGA results. For an initial sample, where there is no historical data, the reviewer is instead limited to basic reality checks. If an error cannot be corrected, IEEE recommends resampling the equipment.
The standard identifies several types of data quality problems that should be considered.
1. Check for transcription and typographical errors
Simple data-entry errors can have a significant effect on DGA interpretation.
IEEE identifies several common patterns, including:
- Values shifted into the wrong column
- Missing or skipped values
- Two values being swapped
- Digits being added or lost
- Digits being transposed
- Incorrect equipment identifiers
These errors can appear as an abrupt or unusual change in an individual gas concentration. An incorrect equipment identifier can be even more serious because the DGA result may be associated with the wrong transformer.
What to check:
Compare the latest DGA report against previous reports and confirm that:
- The transformer identification is correct
- The sample date is correct
- Gas values are recorded in the correct columns
- Units are consistent
- There are no unexplained sudden changes caused by an obvious data-entry error
2. Look for missing or duplicated data
A missing value may occur because a result was accidentally omitted or deleted.
IEEE also highlights the possibility that an earlier sample may be reported again instead of the latest sample. One warning sign is when all gas concentrations are exactly the same as the previous sample, particularly for gases such as CO₂, N₂ and O₂ where some variation would normally be expected.
What to check:
Look for:
- Missing gas measurements
- Duplicate results
- Identical results across consecutive samples
- Gaps in the DGA history
- Unexpected changes in sampling frequency
A clean and complete historical record is particularly important when DGA is being trended over time.
3. Confirm the sample and equipment identification
A DGA result is only useful if it is associated with the correct equipment and sampling point.
IEEE C57.104-2019 identifies incorrect serial numbers, equipment identifiers, sampling points and sample dates as potential sources of error. Samples can also be accidentally swapped.
For example, a sample from an LTC compartment could potentially be incorrectly identified as coming from the transformer main tank.
What to check:
Verify:
- Transformer serial number
- Asset or equipment ID
- Location
- Sampling point
- Main tank vs. LTC or other compartment
- Sample date
- Reason for sampling
IEEE also recommends that correct transformer information accompany the sample, including details such as the sampling compartment, voltage and MVA ratings, preservation type and insulating liquid type.
4. Check for sample mishandling and air exposure
Sampling quality can directly affect the reliability of DGA results.
IEEE identifies air exposure as one possible source of data quality problems. A leaking syringe, poor sampling technique or improper handling can expose the sample to air, potentially causing hydrogen loss while increasing oxygen and nitrogen levels.
Air can also become entrained during sampling, handling or laboratory processing. This can alter oxygen, nitrogen, the O₂/N₂ ratio and total gas concentration, making calculated changes and gas generation rates questionable.
This is why sampling procedures are an important part of DGA data quality.
IEEE recommends that DGA samples be taken in accordance with ASTM D923 and notes that a gas-tight glass syringe is the preferred container. The sampling process should minimise air contamination and protect the integrity of the sample.
5. Look for cross-contamination
Cross-contamination can occur during either sampling or laboratory processing.
For example, contaminated sampling equipment, reused tubing or contaminated storage equipment can introduce gases into a sample.
According to IEEE C57.104-2019, cross-contamination may appear as:
- The appearance of a trace gas
- The appearance of a new apparent fault
- A change in fault identification
When a significant change occurs, IEEE recommends confirmation through a follow-up test.
This is particularly important when a new DGA result suddenly suggests a fault that has not appeared in previous samples.
6. Investigate unusually inconsistent results
A single DGA sample that is dramatically different from the historical results deserves closer attention.
IEEE notes that a drastically different result could indicate a genuine fault, an incorrectly identified sampling point or a sampling or measurement problem. Confirmation by resampling may therefore be necessary.
If several consecutive samples show large inconsistencies, this may point towards a sampling or measurement issue rather than a change in transformer condition.
This is an important distinction. A sudden change in gas concentration does not automatically mean that the transformer has developed a new fault.
7. Check for chronically low hydrogen
Hydrogen is one of the key gases considered in transformer DGA.
IEEE C57.104-2019 highlights chronically extremely low hydrogen as another potential data quality issue, particularly when other combustible gases are present or when hydrogen is consistently low across other transformers.
Possible causes include sampling problems, leaking syringes or measurement issues. IEEE notes that a quality control standard with known gas concentrations may be used to check whether the laboratory or portable analyser is measuring hydrogen correctly.
This demonstrates why individual gas values should not always be interpreted in isolation.
8. Examine the O₂/N₂ ratio
The oxygen-to-nitrogen ratio can provide another indication of potential sample quality problems.
IEEE states that an isolated large increase in the O₂/N₂ ratio, particularly when accompanied by a decrease in hydrogen, may indicate air exposure.
Mistyped oxygen or nitrogen values can also produce unusual or impossible O₂/N₂ ratios. For mineral oil with air dissolved in it, IEEE notes that the O₂/N₂ ratio is normally between approximately 0.4 and 0.5.
The interpretation also depends on the transformer preservation system. For example, air exposure can present differently in sealed transformers, bladder-equipped conservators and transformers with open breathers.
DGA consistency matters when trending data
DGA is often most useful when results are viewed as a trend rather than as isolated numbers.
However, IEEE cautions that DGA measurements naturally have some variation. Two identical samples will not necessarily produce exactly the same result. The guide distinguishes between repeatability, which relates to repeated analysis of the same sample under the same laboratory conditions, and reproducibility, which relates to differences between laboratories, operators, equipment or times.
This means that a small difference between two DGA results does not necessarily represent a real change inside the transformer.
IEEE also notes that consistently large fluctuations between samples can indicate sampling or analytical errors. Such results should not be used for fault identification or severity assessment until the reason for the fluctuations has been established.
Be careful when comparing results from different sources
Historical DGA data may come from different laboratories, analytical methods or online monitoring systems.
IEEE warns that factors such as insulating liquid temperature, different DGA methods and sample quality can affect rate-of-change calculations. Data from incompatible sources can potentially create misleading alarm conditions.
Therefore, when trending DGA results, it is important to understand how each result was obtained, not just what the reported gas concentrations were.
What should you do when a result looks alarming?
An unexpected DGA result should not automatically trigger a major maintenance decision.
IEEE C57.104-2019 specifically states that DGA data should undergo a quality review before interpretation. If surprising or alarming results are obtained, the guide recommends collecting and processing another sample to confirm the result.
The confirmation sample is particularly important when the latest result represents a significant departure from the historical trend.
After data quality and confirmation issues have been addressed, IEEE recommends considering previous DGA results together with other relevant information, such as:
- Test records
- Maintenance records
- Transformer loading
- Environmental conditions
- Data from similar or “sister” transformers
This broader context can help engineers understand whether an abnormal result represents a developing issue or another source of variation.
Data quality comes before diagnosis
DGA can provide valuable insight into transformer condition, but the quality of the conclusion depends on the quality of the underlying data.
The message from IEEE Std C57.104-2019 is clear: perform a data quality review before interpreting DGA results. Data should be checked for identification errors, missing or duplicated values, sampling problems, air exposure, contamination, inconsistent results and measurement issues.
Just as importantly, an alarming result should be confirmed before significant action is taken. DGA is a diagnostic tool that can help track the evolution of an existing condition, but it should be considered alongside historical data, transformer operating information and other relevant diagnostic evidence.
A disciplined data quality review can therefore help engineers distinguish between a genuine change in transformer behaviour and a problem with the DGA data itself.
Need More Visibility into Transformer Condition?
Insulect provides DGA monitoring solutions designed to help asset owners monitor dissolved gases and identify changes in transformer condition.
Talk to our team about DGA monitoring for your transformer assets.