Case Study - 390 MVA GSU Transformer [1]

Background [2] 

On the 15 May 2005 this 390 MVA GSU transformer failed. During the previous unit outage the transformer oil was purified to remove moisture. The transformer was initially commissioned in October 1971. In 1996 the transformer was involved in a coupling transformer incident with another unit and was sent to the repair facility for refurbishment in April 1997. It was reinstalled in February 2000 where it started to gas and was removed for an internal inspection revealing loose flexible connections. This was repaired on-site and returned to service.

The gassing continued slowly and the transformer was removed from service for an internal inspection at the transformer repair facility which revealed a fault on the HV winding crossover. This was repaired. The impulse test although acceptable revealed the absolute value of the C-phase (300 pico-coulombs). PD test results being higher than that for the A and B phases (200 pico-coulombs).

The transformer was then installed on 14 June 2004. Between September-December 2004 the on-line DGA indicated an upward trend. The unit was then shut down for the generator replacement project and the last sample was taken on 22 December 2004. The transformer was returned to service on 25 March 2005 before undergoing oil purification where it failed on 15 May 2005. After the failure the internal inspection revealed damage to the C-phase HV winding as depicted in figure 5.

Analysis

From Figure 6, the LEDT for the period 1991 to 1994 the oil samples remained in the Normal region until the first trigger on the 21 October 1994 with hydrogen being 12 ppm, methane 56 ppm and carbon monoxide 245 ppm. After the coupling transformer incident in 1996 the sample taken on the 11 September 1996, moved into the T2 region with hydrogen being 59 ppm, methane 197 ppm and carbon monoxide 322 ppm. Subsequent samples were all in the defective region T1/T2 and the trend was along the %CH₄ axis. The transformer was then taken out of service in April 1997 for repairs. 

The LEDT in figure 7 represents samples for the period 2000 to 2005. The trend recorded at the top of the triangle was in the second period (4 January 2000 to 13 March 2000) after the transformer was repaired. As can be seen from this LEDT the transformer was already in a bad state in the T3 fault region. The sample values at commissioning were 225 ppm hydrogen, 434 ppm methane and 112 ppm carbon monoxide. On-site internal inspection revealed loose flexible connections, which were repaired on-site and returned to service. The trend still progressed within the T3 fault region and finally the transformer was taken out of service on 13 March 2000 where the recorded values for hydrogen, methane and carbon monoxide were 166 ppm, 530 ppm and 27 ppm respectively. The transformer was sent to the transformer repair shop for repairs.

The transformer was then installed on the 14 June 2004 with the oil sample results in the Normal region but soon after as observed in figure 7 the trend started to move into the defective region. The oil was then purified in an outage in December 2004 where the sample taken before this on the 22 December 2004 revealed hydrogen, methane and carbon monoxide to be 14 ppm, 79 ppm and 226 ppm respectively. After the oil purification these values were 2 ppm, 2 ppm and 38 ppm accordingly moving the transformer state back into the Normal region. The next trigger was received with the oil sample taken on the 13 May 2004, a day before the failure.

Figure 8 indicates the R-value trends with the initial samples below the 0.13 limit. The first trigger out of this limit was 21 October 1994. The next trigger was on the on 11 September 1996 after the coupling transformer incident. The R-values thereafter remained in the defective region (T2), even after the transformer was reinstalled in January 2000 after the on-site repair. When the transformer was sent to the repair shop in 2004 the samples started in the Normal region but soon immediately moved into the defective region (T1). This was once again noted after the purification of the oil where it moved into the defective region almost immediately thereafter. 

Summary

In this case study the initial oil results recorded were in the Normal region and then progressed to T1 and T2 of the LEDT shortly after the coupling transformer incident. The LEDT was consistent with capturing these changes. An interesting observation is made in the period when the transformer was installed in early 2000 where the oil sample results were focused on the top of the triangle. These were for high levels of methane and moderate levels of carbon monoxide. However after the on-site repair and with the reconditioning of the oil the actual ppm values of the gases decreased but the oil samples still remained in the top region of the triangle indicating that the ratio of these gases was still the same but in lower ppm values. This interesting observation provides some indication that the LEDT maybe insensitive to fluctuations in the actual ppm values.