1

C57.104 – IEEE Guide for the Interpretation of Gases Generated in Oil – Immersed Transformers Tuesday, March 13, 2012 Nashville, Tennessee, USA Minutes of WG Meeting

The meeting was called to order by Chair Rick Ladroga at 3:15pm. Vice Chair Claude Beauchemin and Secretary Susan McNelly were also present.

There were 47 of 83 members present. There were 44 guests, and 7 guests requesting membership. A membership quorum was achieved. Guests attending the WG meeting for the first time who request membership will be deferred until the next meeting attended.

Guests requesting membership were (those identified with an asterisk (5 of the 7) will be added as WG members):

Jagdish Burde Anthony McGrail* Frank Damico* Nicholas Perjanik* Shawn Galbreath* Pugal Selvaraj Rowland James*

Agenda 1. Welcome & Introductions 2. Quorum Check 3. Approval of Minutes from fall 2011 Boston meeting. 4. Status 5. Presentation by Claude Beauchemin on Data 6. New Business 7. Adjourn

The minutes from the fall 2011 Boston, Massachusetts meeting were approved as written.

Review of recent activities:

Rick gave a summary of recent activities and indicated that offsite meetings/webinars will be held between TR Committee meetings. He is tentatively looking at the 3rd week in May.

The framework, case work, and bibliography have been done or are in progress. The intent is to provide recommendations at the fall 2012 meeting in Milwaukee for the WG to discuss.

Rick requested case study information from utilities.

Presentation by Claude Beauchemin - Analysis Preview - Review of results to date from analysis of DGA database

Claude extended a thank you to the following people for their efforts:

• Michel Duval • Norman Field • Luiz Cheim • Lan Lin - for the tremendous work done to date on data analysis • All anonymous data suppliers - To give us the opportunity to answer old questions

2

C57.104 Table1 What was the choice for limits?

• Personal Experience ? • One user database analysis ? • Consensus from early users ? • Lab recommendation ? • Early mention in 1978 of 90% “probability norms” for some levels (now limit

condition 1) • 1991 mention for table 1 “Consensus values based on the experience of many

company”

• Condition 1: < 90% of DGA population? • Condition 2: 90% to 95% ? • Condition 3: 95% to 99% ? • Condition 4: > 99% ?

We are using these values for analysis purpose only

Process of data analysis: • Database filtered to remove inconsistent entries

– Obvious error – Missing important information – Non transformer

• Population curve computed for each gas and each studied condition – 90% to 99.5% population value used for evaluation

3

Source of data (479,191Samples)

Data Analysis: • Values proposed need to be sound from a statistic point of view • Original data used to set table 1 is unavailable • Comparison between table 1 and actual data indicate a mix of good and poor

correlation using the 90, 95 and 99% hypothesis • CAUTION: LARGE DISPERSION OF RESULTS

Table 1 VS Percentile, All data

D, 0.1%A, 2.2% C, 6.0%E, 2.2%

G, 0.9%

J, 7.0%

H, 11.5%

I, 60.9%

B, 7.8%

F, 1.5%

A UtilityB LabC UtilityD Industrial UserE UtilityF Insurance Co.G UtilityH LabI LabJ Utility

Delta % H2 CH4 C2H2 C2H4 C2H6 CO CO2 TDCG90 ‐7% ‐29% 0% 12% 42% 105% 200% 44%95 ‐69% ‐60% ‐50% 24% 91% 60% 156% ‐26%99 ‐5% ‐13% 123% 462% 300% ‐1% 84% 17%

Percentile H2 CH4 C2H2 C2H4 C2H6 CO CO2 TDCG90 93 85 1 56 92 717 7491 103495 215 162 5 124 191 912 10223 142999 1706 869 78 1124 600 1386 18435 5439

Condition H2 CH4 C2H2 C2H4 C2H6 CO CO2 TDCG 1 ‐ 2 100 120 1 50 65 350 2500 720 2 ‐ 3 700 400 10 100 100 570 4000 1920 3 ‐ 4 1800 1000 35 200 150 1400 10000 4630

4

Example of data dispersion

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

4,500

5,000

90 91 92 93 94 95 96 97 98 99 100

TDCG vs Data Source

                    All data C

F

 J

I

H

A

E

D

B

G

Problematic of data analysis: • Dispersion between sources is large

– Different Network? – Different History? – Different Utilisation? – Different Laboratories?

• This fact must be taken into account during the analysis process

What parameters influence DGA levels ? • Age ? • Size ? • Voltage Class ? • Sealed / open ? • Energized TC VS Non-Energized TC ? • GSU / Transmission / Distribution ? • North / South (Weather) ? • Utility / Industrial ? • Laboratories used ? • Other?

• Each individual parameter have to be studied to see if it has an influence • Each influence has to be properly isolated • Quantification of influence has to be statistically sound and documented

Example of a possible influential parameter: Age

5

100

1,000

10,000

90 91 92 93 94 95 96 97 98 99 100

TDCG vs Age

                    All data

0‐10 

60‐70

70‐80

50‐60

40‐50 20‐30

30‐40

10‐20

TDCG all 0‐10 10‐20 20‐30 30‐40 40‐50 50‐60 60‐70 70‐8090 1034 747.3 993 1061 1123 1179.3 1177 1391.1 1062.891 1087 783 1033.9 1107 1169 1233 1207.9 1438.7 1133.192 1148 820 1086 1154 1220 1292.6 1266.3 1458.2 1173.893 1222 865 1141.9 1212 1271 1350.6 1307.7 1495.1 120594 1311 920.8 1212 1276 1337 1430 1371.2 1528.4 134695 1429 980.6 1309.4 1367.6 1415 1525.6 1432 1569.8 1403.296 1602 1071 1445 1498 1521.8 1665.6 1512.5 1671.8 1447.497 1904 1193.4 1661 1724.5 1669.2 1856 1641.9 1834.8 148298 2656 1391.3 2147.9 2266.7 1924 2181.7 1925.2 2071.5 1568.399 5439 2239.7 4061.9 3418.3 2848.9 3261.9 2902.5 2282.2 1975.3

99.5 11386 4481.3 7501 5177.7 4295.7 6376.1 3803.1 2471.5 2723.8

22.8%24.2%

22.3%

17.9%

8.4%

3.5%

0.6% 0.2% 0.0% 0.0% 0.0%

0.0%

5.0%

10.0%

15.0%

20.0%

25.0%

0 ‐ 10 10‐20 20‐30 30‐40 40‐50 50‐60 60 ‐ 70 70 ‐ 80 80 ‐ 90 90 ‐ 100 100 ‐110

Years in Operation

6

TDCG 90%, 95% and 99%

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

0‐10 10‐20 20‐30 30‐40 40‐50 50‐60 60‐70 70‐80

Age

PPM

Condition 1

Condition 4

Condition 3

Condition 2

90%

95%

99%

TDCG 90%

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

0‐10 10‐20 20‐30 30‐40 40‐50 50‐60 60‐70 70‐80

Age

PPM

Condition 1

Condition 4

Condition 3

Condition 2

7

H2 90%, 95% and 99%

0

200

400

600

800

1000

1200

1400

1600

1800

2000

0‐10 10‐20 20‐30 30‐40 40‐50 50‐60 60‐70 70‐80

Age

PPM

Condition 1

Condition 4

Condition 3

Condition 2

90%

95%

99%

TDCG 90%, 95% and 99%

0.00

0.50

1.00

1.50

2.00

2.50

0‐10 10‐20 20‐30 30‐40 40‐50 50‐60 60‐70 70‐80

Age

Actual / Con

dition

 Lim

it90%

95%

99%

8

CO2 90%, 95% and 99%

0

2000

4000

6000

8000

10000

12000

14000

16000

18000

0‐10 10‐20 20‐30 30‐40 40‐50 50‐60 60‐70 70‐80

Age

PPM

Condition 1

Condition 4

Condition 3

Condition 2

90%

95%

99%

CO 90%, 95% and 99%

0

200

400

600

800

1000

1200

1400

1600

1800

0‐10 10‐20 20‐30 30‐40 40‐50 50‐60 60‐70 70‐80

Age

PPM

Condition 1

Condition 4

Condition 3

Condition 2 90%

95%

99%

9

C2H6 90%, 95% and 99%

0

100

200

300

400

500

600

700

800

0‐10 10‐20 20‐30 30‐40 40‐50 50‐60 60‐70 70‐80

Age

PPM

Condition 1

Condition 4

Condition 3

Condition 2 90%

95%

99%

CH4 90%, 95% and 99%

0

200

400

600

800

1000

1200

0‐10 10‐20 20‐30 30‐40 40‐50 50‐60 60‐70 70‐80

Age

PPM

Condition 1

Condition 4

Condition 3

Condition 2

90%

95%

99%

10

C2H2 90%, 95% and 99%

0

10

20

30

40

50

60

70

80

90

100

0‐10 10‐20 20‐30 30‐40 40‐50 50‐60 60‐70 70‐80

Age

PPM

Condition 1

Condition 4

Condition 3

Condition 290%

95%

99%

C2H4 90%, 95% and 99%

0

200

400

600

800

1000

1200

1400

0‐10 10‐20 20‐30 30‐40 40‐50 50‐60 60‐70 70‐80

Age

PPM

Condition 1

Condition 4

Condition 3

Condition 2 90%

95%

99%

11

Influence of Rating:

Influence of voltage class:

TDCG 90%, 95% and 99%

0

2000

4000

6000

8000

10000

12000

14000

16000

18000

<34.5KV 34.5‐69KV 69‐230KV >230KV

KV

PPM

Condition 1

Condition 4

Condition 3

Condition 2

90%

95%

99%

TDCG 90%, 95% and 99%

0

5000

10000

15000

20000

25000

<1MVA 1‐5MVA 5‐10MVA 10‐20MVA 20‐50 MVA 50‐100 MVA 100‐500 MVA >500 MVA

Power class

PPM

Condition 1

Condition 4

Condition 3

Condition 2 90%

95%

99%

12

Open or Closed:

Suspicious VS All

Rate of rise (ppm/day)

Discussion:

Question: Fredi Jakob – Regarding Table 1 vs Percentile slide – He indicated he wonders that if Table 1 was from late 80s and 90s, they were pretty young. If still in service, twenty years later, is the difference due to age? Certainly on the CO and CO2 values. Response: Beauchemin - Age is likely influencing the difference. If this is the case, it will show up in the slide on age. If an influence is seen, it will be identified.

Question: Jin Sim – Utilities have started measuring DGA on smaller transformers such as layer type transformers. This also could be influencing the data. Response: Beauchemin - Yes, this could be influencing the data.

Question: Juan Castellano – Was the type of TR compared? Response: Beauchemin – It was not. A very small percent of the data population included this information and what we have we will look at.

Question: Fredi Jakob – In his opinion Table 1 should only be used to give an idea of when a next sample should be taken. He recommends that Table 1 provide direction on what to do in this regard. Response: Beauchemin – There are instructions to this effect already there, but unfortunately, it is often not read. Ladroga – Whether the table will be kept or

ppm/day H2 CH4 C2H2 C2H4 C2H6 CO CO2 TDCG90 0.13 0.08 0.00 0.04 0.08 0.60 6.6 1.0195 0.43 0.23 0.00 0.14 0.23 1.25 14.1 2.599 6.9 3.3 0.22 3.1 2.0 6.3 69.6 26.3

90% H2 CH4 C2H2 C2H4 C2H6 CO CO2 TDCG

All 93 85 1 56 92 717 7491 1034Suspicious 782 912 32 1255 452 738 7749 4305

500

1,000

1,500

2,000

2,500

3,000

90 91 92 93 94 95 96 97 98 99 100

TDCG vs Oil Preservation System

                    All data

Open

Closed

Unknown

13

not is being looked at. The challenge is make the guide simple and useful. The intent is to gear the guide more toward how things are really done.

Question: Jin Sim – Does the core group feel the values in Table 1 should be erased.? Depending on the volume should there be correction? Response: Beauchemin – He indicated that the statistics will dictate, not the core group. Sim – Disagreed, indicating that there are many of the data that are not valid. Response: Beauchemin – That is why there is statistical analysis done to remove some of these outliers. He indicated he also would like to see a resolution to this. Luiz Cheim – We expect that the data is representative. Outliers and cases that could confuse the data needs to be removed, however this is not simple. Better tools and people with time to analyze the data are needed. One thing that may be looked at is making the table more of a matrix to look at the level along with the rate of increase. The goal is to come up with something helpful to the industry.

Fredi Jakob – Paper in IEEE Journals for Power Delivery – There is emphasis on TCGs, which doesn’t make much sense. Rick Ladroga requested a copy of the paper.

Question: Anthony McGrail – Indicated he is disturbed that we are having this conversation at all. He indicated that we need to be very careful that the 99 percentile does not indicate a condition. Response: Ladroga – It is very much indicative of the data distribution. The goal is to determine if we can correlate.

Question: - Indicated that the Table is used by his insurance company to tell them what maintenance needs to be done.

Question: Doug McCullough – Have we asked the manufacturers to give a table on the gas concentrations on materials used in the transformers. This may help to draw correlations. Response: Ladroga – That is a good suggestion and if the manufacturers can provide this information, it will be reviewed.

Question: Leon White – Samples were not always taken properly. Is there any thought on using only samples taken in the last 10 years now that people are more aware of how to properly take the samples? Response: Beauchemin – Yes, the data could be reviewed based on the date of samples to see if there is an evolution in this regard. Mel Wright - Looking at the total dissolved gas and the ratio of oxygen and nitrogen can tell you if the sampling is consistent and if it was properly obtained.

Rick indicated that there has been a concern raised about the quality of the data and the security of the data. He is hoping to keep the data with IEEE for future use and limit the access to the data.

The meeting was adjourned at 4:30 pm.

Rick Ladroga WG Chair

Claude Beauchemin WG Vice-Chair

Susan McNelly WG Secretary