Improvement of DGA Interpretation with Scoring

Index Method
Norazhar Abu Bakar1* , Huize Cui2, A. Abu-Siada2, and Shengtao Li3
1
Universiti Teknikal Malaysia Melaka, 76100 Melaka, Malaysia
2
Curtin University, Perth, WA6102, Australia
3
State Key Laboratory of Electrical Insulation and Power Equipment, Xi’an Jiaotong University, China
*[email protected]

Abstract- Dissolved gas analysis (DGA) is the most effective produces C2H4 and C2H6 gases [2]. Hence, the fault nature can
method to detect incipient faults in power transformers. It uses be determined based on the gas type. However the analysis is
the concentrations of various gases dissolved in the transformer not always straight forward since there is possibility of more
oil due to decomposition of the oil and paper insulation. Gases than one fault present at the same time [7].
such as hydrogen, methane, acetylene, ethylene, and ethane are
generated as a results of oil decomposition, while carbon
Various DGA interpretation techniques have been developed
monoxide and carbon dioxide are generated due to the paper such as key gas method (KGM), Doernenburg ratio method
degradation. In the meantime, faults such as partial discharge, (DTM), Rogers ratio method (RRM), IEC ratio method (IRM),
overheating, and arcing produce a range of gases, hence the Duval triangle method (DTM), and Duval pentagon method
concentrations of which can be used to identify faults and (DPM). All of these methods rely on accumulated knowledge
estimate their severity. Various techniques have been developed and personnel experience rather than mathematical
to interpret DGA results such the key gas method, Doernenburg formulation, and not necessarily lead to the same conclusion
ratio, Rogers ratio, IEC ratio, Duval triangles method, and the for the same oil sample [7].
latest Duval Pentagon method. However, each of these techniques KGM considers the individual concentration of gases
relies on the accumulated knowledge and experience of various
experts rather than quantitative scientific models, therefore
produced during faults and interprets DGA results by referring
different diagnoses may yield for the same oil sample. Hence the to the four sets of charts shown in Fig.1. Four common
accountability of the judgments can be query if only rely on one transformer faults; partial discharge (PD), arcing, overheated
interpretation technique’s. This paper proposed the used of of oil and overheated of cellulose can be analyzed based on
scoring index method to improve the accountability of DGA these charts. Even though the KGM charts look simple,
interpretation results. DGA result of each samples are analyzed however studies show that this method leads to a large number
with four existing interpretation techniques; Doernenburg ratio, of wrong diagnoses [5],[7]. Only 42% of diagnosis using
Rogers ratio, IEC ratio, and Duval triangles method which will KGM is correct, while the rest is under misinterpretation [5].
generate the scoring point for all fault types. Each techniques also
been assigned with specific weightage, where the summation of
scoring point times weightage will represents the scoring index
for particular faults. The proposed method is validated through
DGA historical data of various transformers of pre-known health
condition and life span. Results show that the proposed method
can increase the accuracy of the DGA diagnoses method.

I. INTRODUCTION

Dissolved gas analysis (DGA) of transformer oil is used to

examine the condition of power transformers. It uses the
concentration of various gases dissolved in transformer oil
which generates due to decomposition of the oil and paper
insulation [1]. Gases such as hydrogen (H2), methane (CH4),
acetylene (C2H2), ethylene (C2H4), and ethane (C2H6) are
generated due to oil decomposition, while carbon monoxide
(CO) and carbon dioxide (CO2) are generated due to paper
Figure 1. Key Gases Method Chart [1]
decomposition [2]-[4]. Faults such as overheating, arcing, and
partial discharge also produce a range of gases. Each fault DRM is developed based on thermal degradation principles
evolves particular characteristic gases which can be used to which utilize the ratio of gas concentration to indicate fault
identify the faults and estimate their severity [5],[6]. The low types. Predefined set limit for CH4/H2, C2H2/CH4, C2H2/CH4
level energy, partial discharge produces H2 and CH4 gases, and C2H6/C2H2 ratios is used to interpret the DGA results [2].
while the high level energy, arcing is capable of producing all Table I shows the ratio interpretation table of DRM. The main
gases including C2H2. On the other hand, high thermal fault drawback of DRM is that the rate of unresolved diagnoses is
high due to the incomplete ratio ranges [1]. According [8],[9], various temperatures ranges from partial discharge of low
more than 25% of DGA results are unresolved. energy density up to thermal fault more than 700°C has been
RRM follow the similar procedure initiated by DRM, but proposed in original IRM [11]. In 1999, IRM is revised and
with some modification and improvement in order to fill in the only six classification of faults are remained which referred to
gap of DRM [11]. RRM diagnose is valid when any of partial discharge, discharges of low or high energy, thermal
individual gases exceeds its normal limit and does not rely on faults of temperature less than 300°C, temperature between
specific gas concentrations [2]. An original RRM utilizes four 300°C to 700°C, or temperature more than 700°C as shown in
ratios; C2H6/CH4, C2H2/C2H4, CH4/H2, and C2H4/C2H6 that Table III[2]. Although IRM diagnosis offers better consistency
leads to twelve proposed diagnosis. However, the ratio than DRM and RRM, however the number of unresolved case
C2H6/CH4 can only indicate a limited temperature range of is still high [7].
decomposition and do not assist in further fault interpretation TABLE III
IRM DIAGNOSIS INTERPRETATION BASED ON IEC60599 [1]
[11],[12]. Therefore in the IEEE Standard C57.104-1991, C2H2/ CH4/ C2H4 /
RRM is revised and the ratio of C2H6/CH4 is omitted from the Case Characteristic Fault
C2H4 H2 C2H6
RRM code. Only six diagnosis interpretations remain in
revised RRM; normal, low-energy density arcing- PD, arcing- PD Partial Discharges NS <0.1 <0.2
high-energy discharge, low temperature thermal,
D1 Discharges of low energy >1 0.1-0.5 >1
thermal<700°C, and thermal>700°C (Table II) [1]. This
method does not consider dissolved gases below normal Discharges of high
D2 0.6-2.5 0.1-1 >2
concentration limits, also certain ratio values are inconsistent energy
with the diagnostic assigned and lead to invalid codes [7]. T1
Thermal faults not
NS >1 but NS <1
TABLE I exceeding 300°C
RATIO INTERPRETATION TABLE OF DRM [2]
Thermal faults
Ratio 1 (R1) Ratio 2 (R2) Ratio 3 (R3) Ratio 4 (R4) T2 exceeding 300°C but not <0.1 >1 1-4
Suggested CH4/H2 C2H2/C2H4 C2H2/CH4 C2H6/C2H2
fault exceeding 700°C
diagnosis Gas Thermal faults
Gas Gas Gas
Oil Oil Oil Oil Spac T3 <0.2 >1 >4
Space Space Space exceeding 700°C
e
1. Thermal NS = Non-significant whatever the value
<0.
decomposi >1 >0.1 <1 <0.3 <0.1 >0.4 >0.2
75
tion
2. Partial
discharge
Duval Triangle method is an innovation developed from an
Not existing IEC 60599 Ratio Method and IEC TC10 databases
(low- <0.1 <0.01 <0.3 <0.1 >0.4 >0.2
significant
intensity which represent the DGA results in more user-friendly
PD)
3. Arcing
graphical form [5],[8],[10]. DTM faults interpretation is based
>0.01 on the values of three gases, CH4, C2H2 and C2H4, and their
(high- >0.1
to >0.75 >1 >0.3 >0.1 <0.4 <0.2
intensity to <1
<01 location on triangle developed by Duval [8] as shown in Fig. 2.
PD)
Duval Triangle is characterized by six faults zones, which
covered partial discharge, thermal fault at various
TABLE II
RATIO INTERPRETATION TABLE OF RRM [2] temperatures, and electrical arcing [5].
Case R2 R1 R5 Suggested Fault
C2H2/ CH4/ C2H4/ Diagnosis PD Partial Discharges
C2H4 H2 C2H6 D1 Discharges of low
energy
0 <0.1 >0.1 to <1 <1 Unit normal D2 Discharges of high
energy
Low-energy density DT Combination of
1 <0.1 <0.1 <1 thermal faults and
arcing- PD
discharges
Arcing – High-energy T1 Thermal faults not
2 0.1 to 3 0.1 to 1 >3 exceeding 300°C
discharge
T2 Thermal faults
Low temperature exceeding 300°C but
3 <0.1 >0.1 to <1 1 to 3 not exceeding 700°C
thermal T3 Thermal faults
Thermal less than exceeding 700°C
4 <0.1 >1 1 to 3
700°C
Thermal exceeding
5 <0.1 >1 >3
700°C
Fig.2. Coordinates and fault zones of the DTM[1]
The IEC Ratio method was developed in 1978 as an evolution
of RRM [6]. Three IRM ratios are similar to those used in the Among existing interpretation methods, DTM provides the
revised RRM (IEEE Standard C57.104-1991) but with most accurate and consistent analysis. In [7], consistency
different code ranges [2]. Nine classifications of faults with analysis conducted on 338 transformers shows that DTM
diagnose can achieve 72% consistency. However, technique to provide more reliable and consistent decision on
inconsiderate implementation may leads to wrong diagnoses. the fault condition.
Moreover, Duval triangle does not encompass fault-free zone, TABLE V
ACURACY AND SUCCESSFUL RATE COMPARISON OF VARIOUS DGA INTERPRETATION
therefore this method cannot be used to detect incipient faults. METHODS
This paper proposed the used of scoring index method to Fault Actual Prediction Successful Successful Accuracy
Method
improve the accuracy and accountability of DGA Code Number Number Prediction Rate (%) (%)
F1 60 64 54 84.4 90.0
interpretation results. Four existing interpretation techniques; F2 33 14 11 78.6 33.3
DRM, RRM, IRM, and DTM are used to generate the scoring DRM F3 25 22 18 81.8 72.0
point for all fault types. F4 32 20 19 95.0 59.4
F5 0 30 0 0.0 0.0
II. DGA ACCURACY ANALYSIS F1 60 60 55 91.7 91.7
F2 33 17 10 58.8 30.3
RRM F3 25 18 15 83.3 60.0
DGA analysis was performed on 150 oil samples collected F4 32 10 7 70.0 21.9
from various transformer rating and life span. All the 150 F5 0 45 0 0.0 0.0
DGA data were analyzed using four existing interpretation F1 60 67 54 80.6 90.0
methods; DRM, RRM, IRM, and DTM, and accuracy of each F2 33 15 11 73.3 33.3
method was calculated based on its successful prediction with IRM F3 25 27 23 85.2 92.0
respect to the pre-known fault type for each sample. All F4 32 7 5 71.4 15.6
F5 0 34 0 0.0 0.0
diagnostic methods were grouped according to the faults type F1 60 60 56 93.3 93.3
as shown in Table IV. F2 33 35 31 88.6 93.9
TABLE IV DTM F3 25 22 20 90.9 80.0
FAULT TYPES
F4 32 33 28 84.8 87.5
F5 0 0 0 0.0 100.0
Types F1 F2 F3 F4 F5

III. PROPOSED METHOD

Normal

Thermal Out of
DRM PD Arcing Figure 3 shows the flow chart of the proposed approach
decomposition code
which begins from the results of routine DGA analysis. Each
DGA results are examined with existing DGA interpretation
Low
Low
methods; DRM, RRM, IRM, and DTM and the fault type of
temperature each method are identified. However, in the case of invalid
Normal

energy Out of
RRM thermal fault, Arcing code (F5), the scoring point is given based on the condition
density code
Thermal<700°C,
and >700°C
PD stated in Table VII. Total scoring index for each fault types are
obtained by applied equation (1) whereas W is the weightage
Thermal(T) fault Low energy
value as in Table VI, and S is the scoring value as in Table
Normal

T<300°C , discharge, Out of VII. Table VI is developed based on the previous study done
IRM PD by [5] and [7]. Four scoring index range are proposed to
300<T<700°C, High energy code
and T>700°C discharge represent the finding; 0 to 0.25 (Insignificant); 0.25 to 0.50
(Less Significant); 0.5 to 0.75 (Significant); and 0.75 to 1.0
Thermal fault Low energy
(Very Significant).
Normal

<300°C , discharge, Out of

DTM PD
300<T<700°C, High energy code
and T>700°C discharge

The accuracy and success rate of each method in identifying

different types of faults is summarize in Table V. It can be
concluded that DTM are successfully predicted all fault types
and the most consistent method with average accuracy and
successful rate more than 80%, followed by DRM then IRM
and RRM. Even though DRM, RRM and IRM are successful
predicted faults F1, F3, and F4, however the number of
unresolved code (F5) is high. On the other hand, the results in
Table V confirmed the previous studies done in [5],[7]
whereas none of the existing DGA interpretation techniques
promise 100% of successful prediction rate. Therefore, this
paper proposed a method that incorporates these four existing
Figure 3. Flow chart of the proposed approach
methods and then associates the results using scoring index
 TABLE VI 3 and 4 which indicates a unity mark (1) for F4 and F2
WEIGHTAGE TABLE
Fault DRM RRM IRM DTM respectively.
Code (W1) (W2) (W3) (W4) For sample 2, code F4 get the highest scoring index (0.764)
F1 0.25 0.25 0.25 0.25 which indicates very significant of arcing. In contrast with
F2 0.2 0.2 0.2 0.4 sample 1, codes F1, F2, and F3 of sample 2 shows some value
F3 0.2 0.2 0.2 0.4
which indicates one or more of diagnose methods are represent
F4 0.2 0.2 0.2 0.4
invalid code (F5).
TABLE VII The highest mark for sample 5 is code F2 (0.866) which
SCORING TABLE indicates very significant sign of thermal fault, while others
Fault DRM RRM IRM DTM code indicates 0.066 marks which reflects “Insignificant”
Code (S1) (S2) (S3) (S4)
All code
indication. Meanwhile, sample 6 indicates “Significant”
valid
1 1 1 1 indication for code F2 (0.55) and “Less Significant” indication
1 code
0.75 0.66 0.66 N/A
for code F3 (0.4). It is notice that for sample 6, IRM and RRM
invalid predicts fault type F3, while DTM predict code F2 and DRM
2 code is out of code (F5). Sample 7 and 9 shows “Significant”
0.5 0.33 0.33 N/A
invalid
3 code indication of partial discharge (F3) while others fault types are
0.25 0 0 N/A irrelevant.
invalid
All code
0 0 0 N/A
Table VIII also shown the output of the proposed method
invalid reveal a high agreement with the transformer actual fault type.
Overall results shows that the proposed method is able to
Fx   ii 14,, jj14 (Wi  S j ) (1) accurately match 49 out of 50 transformer actual conditions,
which represents 98% overall agreement with the transformer
actual condition.
IV. VALIDATION OF THE PROPOSED METHOD
V. CONCLUSION
To examine the accuracy and successful rate of the proposed
This paper introduces a new approach to improve the
method, DGA results of 50 power transformers of different
accountability of existing DGA interpretation methods. The
rating and life span with pre-known health conditions from
approach integrates between four existing DGA diagnose;
previously published research papers [5],[7],[12]-[14] are
DRM, RRM, IRM, and DTM, using predefined weightage and
tested using the proposed method. Table VIII shows the results
scoring index. The highest scoring index represents the output
of the proposed method for 9 samples along with the actual
of the proposed method. The proposed method is validated
condition of the transformer.
through historical data of fifty transformers of different ages,
The scoring index for sample 1 shows a highest mark for
ratings and pre-known health condition. Results of the
code F4 (1) which indicates a very significant indication of
proposed method reveal a good agreement with the real
arcing. It’s also point out that all existing methods are agreed
transformer fault type.
with the same diagnose. The same results obtained for sample

TABLE VIII
SOME DGA SAMPLES (PPM) AND THE CORRESPONDING PROPOSED METHOD OUTPUT

Sample Highest
CH4 C2H6 C2H4 C2H2 H2 CO CO2 F1 F2 F3 F4 Actual Condition
No. Score
1 28 8 31 32 90 1380 11700 0 0 0 1 F4 Arcing in windings
2 5 2 21 21 60 188 2510 0.05 0.132 0.182 0.764 F4 Arcing in oil
3 77 22 170 240 220 1800 13800 0 0 0 1 F4 Arcing in windings
Thermal runaway inside
4 610 259 260 9 360 12000 74200 0 1 0 0 F2 thick insulation with
puncture
Thermal runaway in
5 27 49 4 1 1 53 254 0.066 0.866 0.066 0.066 F2
thick paper insulation
Low temperature
6 149 20 3 0 2031 556 3008 0.05 0.55 0.4 0.05 F2
overheating
Low energy partial
7 2397 157 0 0 32930 313 560 0.066 0.132 0.732 0.132 F3 discharge and x-wax
formation
Corona and x-wax
8 1740 249 8 8 37800 56 197 0.066 0.132 0.932 0.066 F3
formation
9 995 60 7 6 9340 60 620 0.132 0.132 0.732 0.066 F3 Partial discharge
 REFERENCES
[1] N. A. Bakar, A. Abu-Siada, and S. Islam, "A review of dissolved gas
analysis measurement and interpretation techniques," IEEE Electrical
Insulation Magazine, Vol. 30, No. 3, pp. 39-49, 2014.
[2] "IEEE Guide for the Interpretation of Gases Generated in Oil-Immersed
Transformers - Redline," IEEE Std C57.104-2008, pp. 1-45, 2009.
[3] "IEEE Guide for the Interpretation of Gases Generated in Oil-Immersed
Transformers," IEEE Std C57.104-1991, p. 0_1, 1992.
[4] "IEEE Guide for the Detection and Determination of Generated Gases in
Oil-Immersed Transformers and Their Relation to the Serviceability of
the Equipment," ANSI/IEEE Std C57.104-1978, p. 0_1, 1978.
[5] M. Duval and A. dePabla, "Interpretation of gas-in-oil analysis using
new IEC publication 60599 and IEC TC 10 databases," IEEE Electrical
Insulation Magazine, vol. 17, pp. 31-41, 2001.
[6] H.-C. Sun, Y.-C. Huang, and C.-M. Huang, "A Review of Dissolved Gas
Analysis in Power Transformers," Energy Procedia, vol. 14, pp. 1220-
1225, 2012
[7] A. Abu-Siada and S. Islam, "A new approach to identify power
transformer criticality and asset management decision based on dissolved
gas-in-oil analysis," Dielectrics and Electrical Insulation, IEEE
Transactions on, vol. 19, pp. 1007-1012, 2012.
[8] M. Duval and J. Dukarm, "Improving the reliability of transformer gas-
in-oil diagnosis," Electrical Insulation Magazine, IEEE, vol. 21, pp. 21-
27, 2005.
[9] S. Corporation, "Serveron White Paper : DGA Diagnostic Methods,"
2007.
[10] M. Duval, "New techniques for dissolved gas-in-oil analysis," Electrical
Insulation Magazine, IEEE, vol. 19, pp. 6-15, 2003.
[11] R. R. Rogers, "IEEE and IEC Codes to Interpret Incipient Faults in
Transformers, Using Gas in Oil Analysis," Electrical Insulation, IEEE
Transactions on, vol. EI-13, pp. 349-354, 1978.
[12] S. A. Ward, "Evaluating transformer condition using DGA oil analysis,"
in Electrical Insulation and Dielectric Phenomena, 2003. Annual Report.
Conference on, 2003, pp. 463-468.
[13] N. A. Bakar and A. Abu-Siada, “Fuzzy logic approach for transformer
remnant life prediction and asset management decision”, IEEE Trans.
On Dieectrics and Electrical Insulation, vol. 23, pp. 3199-3208, 2016.
[14] A. Abu-Siada, S. Hmood, and S. Islam, "A new fuzzy logic approach for
consistent interpretation of dissolved gas-in-oil analysis," Dielectrics and
Electrical Insulation, IEEE Transactions on, vol. 20, pp. 2343-2349,
2013.