Introduction

This case study investigates the failure of a 700 MVA GSU transformer to understand what the generally accepted dissolved gas analysis methods identified and any learnings for future application.

This 700 MVA GSU transformer first failed in late 2003. It was then sent to the repair shop for a total rewind and then installed on 28 May 2004. It then failed on the 4 January 2007. The following analysis is based on the generally accepted and contemporary dissolved gas analysis methods such as combustible gas trend, Duval’s Triangle 1, Duval’s Triangle 4, Duval’s Triagnle 5, Duvals Pentagon, LEDT, and R-Value methods to identify some experience with how the fault progressed and what the different analysis methods were able to identify.

Combustible Gas Trend

The first step is to always get an indication of the history of the oil samples and to identify any trends. Figure 1 is a representative of the Combustible Gas Trend for the period 28 May 2004 (Installation after repair) to when it failed on 4 Jan 2007.

It is observed that the Carbon Monoxide (CO), Methane (CH₄), and Ethane (C₂H₆) levels immediately started to increase. The Ethylene (C₂H₄) and Acetylene (C₂H₂) levels remained low suggesting that this may not be a high-energy thermal fault or an electrical discharge fault. The Hydrogen (H₂) levels fluctuated at 50 ppm. The constant increase in the Methane levels was an indication of increasing thermal fault in the oil and the increasing level of Carbon Monoxide was an indication of the involvement of paper cellulose insulation.

From the initial few samples, this transformer required closer monitoring.

Duval’s Triangle 1

The next step in the process was to use Duval’s Triangle 1 method to identify the fault types. Figure 2 below provides a representation of the oil samples in the Duval’s Triangle 1.

It is observed that most of the oil samples were in the T1 region suggesting a Thermal Fault of less than 300° C with some progression into the T2 region Thermal Fault greater than 700° C. The failure on 4 Jan 2007 was in the D1 region – discharges of low energy (sparking).

LEDT Method

The LEDT Method is the Low Energy Degradation Triangle method and is a new contemporary method based on low energy degradation of the insulation. Figure 3 provides the plot of the oil sample results.

The first sample taken was in the normal region. From the second oil sample, there was movement into the abnormal region of the T1 indicating a thermal fault of temperature less than 300° C. All other samples then progressed into the T1 and T2 regions until the final failure was plotted in the D2 region indicating discharges of high energy.

R-Value Trend

The R-value trend (Figure 4) is derived from the LEDT Method. It is an indicator of the abnormal degradation from the normal operation to the degradation state. From the trend, it becomes very clear that there is a definite progression of the fault with each sample taken over the period. This method provides a very good indication of the fault progression over time.

Duval’s Pentagon 1

Duval’s Pentagon 1 (Figure 5) was derived to combine the hydrogen and the 4 main hydrocarbon combustible gases.

The first batch of oil samples were in the Stray gassing region after which there was progression into the T1, T2, T3, and with the eventual failure in the D2 region.

Duval’s Triangle 4

Duval’s Triangle 4 (Figure 6) was derived to identify more information pertaining to low-energy faults. Most of the initial oil samples were in the overheating region with some carbonization.

Duval’s Triangle 5

Duval’s Triangle 5 (Figure 7) was derived to assist in the identification of medium to high-level energy faults and to provide more information on thermal faults in paper and oil. Duval’s Triangle 5 should be used only for faults identified first with Duval’s Triangle 1 as faults T2 or T3. From this case study, Duval’s Triangle 1 only suggested minor T2 with no T3 influence it was used to provide some detail on the fault characteristic. It was noticed that most of the samples indicated an overheating condition with some carbonization.

Conclusion

The failure was identified as a possible inter-turn fault at a cross-over point, with a resultant short circuit taking place between winding discs on the R phase HV winding.

From the combustible gas trend, it is interesting to note that the ethane levels throughout the period have increased significantly with low levels of hydrogen and ethylene. Although the fault was present it was found that there was a possible external source from the overhead power line fault which could have initiated a conductor winding short circuit.

This case study provides an indication of the effectiveness of the LEDT in detecting changes in transformer state from normal to defective. The defective state trigger was received at least two and a half years before the failure.

Due to the higher levels of ethane, the samples in the Duvals Pentagon were mainly in the stray gassing region. Both Duval’s Triangle 4 and 5 supported the overheating thermal fault condition diagnosis.

Learnings

The following learnings can be gained from this case study:

1. DGA is an effective tool for identifying internal faults in power transformers
2. Combustible gas trends provide a global indication of a developing fault
3. The Duval’s Triangle 1 provides a good indication of the fault type
4. The LEDT Method also provides a good indication of low-energy developing faults
5. The R-value trend provides a clear visual trend for the fault progression