Paper P502

21 rue dArtois, F-7508 PARIS http://www.cigre.org

Cigr 2009 6th Southern Africa Regional Conference

PREDICTING TRANSMISSION TRANSFORMER CONDITION STATUS USING DGA SIGNATURES

A. SINGH Eskom Holdings Limited (South Africa)

L. MOODLEY Doble Engineering Africa (South Africa)

K. GOVENDER Doble Engineering Africa (South Africa) S. GOVENDER Trans Africa Projects (South Africa)

Y. CHINABOO Doble Engineering Africa (South Africa)

SUMMARY
This paper presents the results from a DGA study into failed Eskom transformers to identify typical failure signature patterns. Historical DGA data of failed transformer units was analysed to identify typical signature patterns of the failures. From the results it is evident that careful analysis of DGA patterns can predict dielectric, thermal and overloading problems. The methodology was developed by analyzing the DGA signatures and comparing the failure causes obtained after tear downs of the failed units. For the cases that were analyzed the failure modes predicted using DGA signatures verified the findings after the tear downs. It is important to note that the DGA signatures gave clear indication of a fault in its early stages of development. The trends of the DGA signatures, over time, have proved to be a key component in the detection of faults.

KEYWORDS
Dissolved Gas Analysis, Thermal Failure Mode, Dielectric Failure Mode, Overheating Failure Mode

E-mail: Areno.Singh@eskom.co.za Tel: +2731 2796320

INTRODUCTION The current DGA methods generally identify faults (abnormalities) prior to failure, which supports the idea of developing a prediction model. Furthermore, the classification of the faults (post failure) by the different analytical methods is strongly correlated. It has been shown that the historical DGA signatures, in general, can give early warning indicators of an imminent fault, the effects of a heavily loaded transformer, and the effects of intervention and the lack thereof. Furthermore, it can also be stated that many of the dielectric failures were initiated by thermal defects. This study performed thus far documents the application of the existing DGA techniques on transformer oil condition data of failed transformers and determining their failure prediction success rate. The best fit solutions for the various transformer faults are provided. This work will aid in the development of DGA signatures to predict unhealthy/abnormal conditions or provide alerts for impending failures. The final research objective is to use transformer condition data to assess its current state and based on the results, the application of suitable interventions to manage the life of the transformer. Several failed units for which the DGA signatures were obtained were analysed and the results were compared to the findings from the failure investigation reports made available from Eskom. FAILURE MODES AND MECHANISMS Historical DGA of failed units was performed to identify typical signature patterns of the failures. The DGA failure signatures were divided into 3 categories, namely dielectric, thermal and overheating/overloading. The trends of the DGA signatures proved to be a key component in the detection of these faults. Figure 1 shows the typical signatures of the failure modes considered. Signatures A and B are thermal faults, Signatures C and D are dielectric faults and Signatures E and F are overheating/overloading faults.
Failure Modes
100 90 80 70

TCG (%)

60 50 40 30 20 10 0

B
CO

C
H2 CH4 C2H4

D
C2H6 C2H2

Thermal Failure Mode (A and B) It can be seen that the dominant gases (Most prominent, as in position in relation to other gases), is typically ethylene and methane.

Dielectric Failure Mode (C and D) It can be seen that the dominant gases, is typically hydrogen and acetylene. Overheating/Overloading Failure Mode (E and F) It can be seen that the dominant gases, is typically methane and ethane. Furthermore, each of the above failure modes were then refined into the failure mechanisms listed below. Thermal Fault a) Circulating Current (Core and Earth (Tank), Core to Frame to Earth (Tank), Circulating current in windings) b) Bad Connection with no cellulose involvement c) Covered conductor Dielectric Fault a) Discharge from cellulose b) Arcing typically involving regions of high electric stress c) Floating Potential possibly due to a compromised earth connection Overheating Fault a) Overheating due to overloading b) Severe Overheating due to inadequate or ineffective cooling The severity of the fault can be determined by placing the fault into 1 of the 3 stages (developing, developed or advanced). For example, for a circulating current type fault, the dominant gases shall be ethylene followed by significant levels of methane and ethane. If the fault is increasing in terms of severity the levels of hydrogen shall increase. Several failed units for which the DGA signatures were compiled were compared to investigation reports made available from Eskom. The results are presented in this paper.

CASE STUDIES 1. Thermal Fault - Circulating Current

100

90

80

Start of Fault

Percentage TCG

70

60

50

40

30

20

10

10/02/93

03/11/93

18/02/94

10/11/94

05/04/95

24/10/95

21/11/95

25/03/96

18/09/96

09/04/97

09/03/99

18/02/01

18/06/01

25/04/02

20/05/02

23/01/03

24/04/03

15/08/03

09/01/04

16/01/04

20/07/04

24/01/05

17/02/05

23/07/05

24/09/05

CO

H2

CH4

C2H4

C2H6

C2H2

Figure 1: Historical DGA trend for transformer Figure 1 gives the historical DGA signature. The failure mechanism was a core circulating current whose characteristic trend is high levels of ethylene followed by significant levels of methane and ethane. The DGA signature does not show a significant change over the history of the fault. The sample on the 10/02/93 (start of fault) was identified as a thermal failure mode. Intervention on this date may have prevented failure of the transformer. The severity of the fault is defined by increasing levels of hydrogen and acetylene. The failure sample is indicative of a dielectric type fault. This can be very misleading if the if the DGA signature is not analysed to identify the root cause of the fault.

2. Thermal Fault- Winding Tracking Fault

100

90

80

Percentage TCG

70

Start of Fault

60

50

40

30

20

10

11/02/01 02/06/01 11/12/01 13/06/02 12/12/02 19/06/03 26/11/03 10/02/04 16/03/04 22/04/04 22/07/04 23/07/04 09/10/04 17/11/04 17/12/04 04/03/05 06/05/05 25/05/05 30/05/05

CO

H2

CH4

C2H4

C2H6

C2H2

Figure 2: Historical DGA trend for transformer Figure 2 gives the historical DGA signatures. The root cause for the failure was a thermal fault of the winding tracking type, which began on the 26/11/03, whose characteristic trend is ethylene and methane being the dominant gases. They are at similar levels and significantly high. A good indication of the severity of the fault is given by a decrease in carbon monoxide and an increase in ethylene and methane.

3. Dielectric Fault - Discharge

100

90

80

Start of Fault
70

Percentage TCG

60

50

40

30

20

10

0 04/06/97 13/03/98 09/03/99 21/10/99 25/02/00 11/12/00 30/07/02 04/08/04 06/08/04 16/08/04 02/09/04 19/10/04 18/11/04 06/12/04 08/12/04 20/12/04 21/12/04

CO

H2

CH4

C2H4

C2H6

C2H2

Figure 3: Historical DGA trend for transformer

Figure 3 gives the historical DGA signatures. The root cause for the failure was a discharge type fault. The fault began on the 04/08/04. The failure report indicated that it failed due to tracking, causing a flashover to earth and damaging the windings of the C-phase. The characteristic trend is high levels of hydrogen followed by significant levels of methane. The severity of the fault is defined by increasing levels of methane. The fault sample is indicative of a dielectric type fault.

4. Overheating Fault - Severe

100 90 80

Percentage TCG

70 60 50 40 30 20 10 0
14/01/98 04/08/98 02/12/98 08/05/99 06/01/00 28/03/00 14/06/00 02/06/01 11/12/01 24/04/02 13/06/02 18/12/02 19/06/03 26/11/03 10/02/04 01/04/04

CO

H2

CH4

C2H4

C2H6

C2H2

Figure 4: Historical DGA trend for transformer

Figure 4 gives the historical DGA signatures. The DGA signature does not show a significant change over the history of the fault. The root cause of the problem was an overheating/overloading fault, which existed throughout the DGA history. It was concluded that ethane leading methane is of a more severe problem than vice versa. This trend often leads to transformer failure. The total combustible gasses did give some indication of a fault.

5. Stray Gassing (Normal Condition)

100 90 80

Percentage TCG

70 60 50 40 30 20 10 0
18/12/02 30/06/03 26/11/03 10/02/04 20/04/04 22/04/04 17/11/04 30/05/05 14/07/05 16/11/05 17/05/06 14/05/07 23/05/07 01/02/08

CO

H2

CH4

C2H4

C2H6

C2H2

Figure 5: Historical DGA trend for transformer Depending on the type of oil used, stray gassing may be mistaken as a discharge, overheating, and/or catalytic gassing on metal surfaces type of fault. It is therefore important to identify such gassing patterns and avoid misinterpreting stray gassing for a fault. Figure 5 gives the historical DGA signatures. Stray gassing has a similar signature to that of an Overheating/Overloading DGA trend. The difference is the abnormal levels of hydrogen. The levels of hydrogen, methane and ethane are seen to fluctuate throughout the DGA history.

CONCLUSIONS The following conclusions are relevant to the studies conducted: (1) The oil condition data of the failed transformers provided DGA signature trends of the various faults, namely: Circulating current, bad Connection with no cellulose involvement, Covered conductor, Discharge, Arcing, Floating Potential, Overheating, Severe Overheating and Stray Gassing. This confirms that historical DGA data of suspect transformers contain information relating to imminent faults. (2) None of the existing DGA methods can identify stray gassing. The fluctuating levels of hydrogen, methane and ethane are a good indication of stray gassing. (3) For the population of failed transformers that were studied it is evident that the historical DGA signatures provided a trigger for an impending failure. Suitable interventions could have been employed in order to prevent the failure. The intervention to employ can also be determined by conducting more frequent oil sampling (and analysis), online testing (IR and PD scanning) and electrical testing of the transformer. (4) The final research objective is to use develop an algorithm which shall use the transformer condition data to assess its current state and based on the results, the application of suitable interventions to manage the life of the transformer. REFERENCES [1] A Singh, L Moodley, K Govender, S Govender Analysis of Failed Eskom Transmission Transformers with DGA Signature Patterns, Doble Africa Conference, 2008. [2] S Govender, A Singh, L Moodley, K Govender, Y Chinaboo Application of Existing DGA Techniques on Failed Transformer Oil Condition Signatures, Trans-Africa Projects, December 2008. [3] S Govender, A Singh, L Moodley, K Govender, Analysis of DGA Signatures of Failed Transformers, Trans- Africa Projects, March 2008. [4] J Lapworth A Novel Approach for Integrating Dissolved Gas Analysis Results into A Life Management System, Doble Engineering Conference, 2002.