• Author / Uploaded
• Deptiranjan Mohapatra

Citation preview

F

E

A

T

U

R

E

A

R

T

I

C

L

E

Dissolved Gas Analysis of Alternative Fluids for Power Transformers Key Words: oil/paper insulation, DGA, esters, overheating, low and cold corona-type discharges

Introduction

C

onventionally, the insulation system of power transform ers consists of mineral oil, cellulose paper, and pressboard. In recent years, there has been an increase in the use of environmentally-friendly fluids such as synthetic esters and natural esters in place of mineral oil. This has been particularly prevalent at distribution voltage levels [1], but it now also is taking place at transmission voltage levels [2]. The main driver for the use of alternative fluids from the electrical supply utility perspective is the environmental benefits that result from their use with no compromise on safety or reliability [3]. Mineral oil is a mixture of hydrocarbons and is refined from crude oil. Mineral oil has a low biodegradability (20% of mineral oil will biodegrade within 28 days) resulting in the need to construct bund walls around large transformers preventing escape into the environment should a leak occur. In the case of an oil leak, significant financial penalties would be imposed by environmental enforcement agencies. In contrast, esters are very biodegradable (95% or more of esters will be biodegraded within 28 days [4],[5]), and they conform to the readily biodegradable definition according to the OECD 301 series of tests [6],[7]. In addition to their better environmental performance, esters have higher flash and fire points than mineral oil. This is desirable from a fire safety perspective, particularly for a transformer operating underground or offshore. Esters have been used in distribution transformers for several decades without fires being reported [2]. Esters also are far more hygroscopic than mineral oil as the ester group (COOR) in the molecular chain structure has a higher ability to participate in hydrogen bonding. The high moisture saturation level means that, for the same moisture content expressed in absolute parts per million, esters will have a lower relative humidity in comparison with mineral oil. This means that moisture has less of an impact on the dielectric strength of esters than mineral oil. When esters are used in conjunction with cellulose paper and pressboard, cellulose is kept in a drier

Imad-U-Khan, Zhongdong Wang, and Ian Cotton Electrical Energy and Power System, University of Manchester, Manchester M60 1QD UK

Susan Northcote TJ/H2b Analytical Services Ltd., Chester CH1 6ES UK

Ester-based transformer fluids have the same DGA fingerprints as mineral oil. However, with lower volumes of gas produced, they will demand more precise dissolved gas measurements and modified or new ratio criteria to allow fault detection and diagnosis.

condition and the rate of cellulose degradation consequently is slower than in mineral oil [8]. This article examines the impact of alternative fluids on dissolved gas analysis (DGA). DGA has been used for many years as an effective and reliable tool to detect incipient faults in mineral oil-filled transformers. The information provided by DGA analysis is extremely important to the asset managers with electricity supply companies. Therefore, it is essential to ensure that traditional DGA analysis techniques still can be used if alter-

September/October 2007 — Vol. 23, No. 5 0883-7554/07/$25/©2007IEEE

5

native oils are used in transformers. In order to clarify that DGA diagnostic techniques still yield the correct result when applied to ester filled transformers, it is necessary to determine if the same types of fault gases are generated, to identify the generation rate, and their concentration in the alternative fluids against a mineral oil benchmark. This article gives the results of experiments that have simulated a number of faults that can be found in power transformers and looks at the DGA analysis results in each case for a number of oil types. Table 1 gives the types of dissolved gases evolved during transformer faults and their indicative relationships with types of faults in mineral oil. A number of diagnostic methods are available to identify the types of faults and their intensities. These include the IEC, IEEE standard and Duval triangle diagnosis methods [9]–[11]. Two broad categories of faults in a transformer can be detected by DGA: thermal faults and electrical faults. The various DGA standards then subdivide these basic types of fault in different ways. In IEC 60599 [9], thermal faults are represented as being in three temperature bands, 300ºC (T2) and >700ºC (T3). Electrical faults can be further classified as partial discharges of the cold plasma (corona) type (PD), low energy discharges (D1), and high energy discharges (D2). The triangular graphical representation of Duval, which is used to visualize a DGA fault diagnosis, uses the same subdivisions as the IEC standard. However, instead of using the concentration of five gases and three ratios to define the type of fault, the relative percentage of three gases is used for the analysis in the Duval diagnosis technique [12]. The IEEE analysis method uses the concept of key gases [10]. The key gas for each type of fault is identified, and the dominating percentage of this gas to the others is used to diagnose the fault. For example, low intensity PD or corona produces mainly H2. Similarly, the key gas C2H2 is for arcing, C2H4 for overheating oil, and CO is for overheating of cellulose.

The generation of various forms of fault within three dielectric fluids and the analysis of dissolved gas produced are under investigation in this article. The fluids used are a mineral oil — Nynas Nytro 10GBN; a synthetic ester, Midel 7131; and a natural ester — FR3. The synthetic ester, Midel 7131, consists of four ester groups with saturated chains as shown in Figure 1(a), i.e., there are no double bonds between the carbon atoms in the chain. FR3, the natural ester, is a natural triglyceride ester with a mixture of saturated and unsaturated fatty acids. The triglyceride ester molecule may be represented as in Figure 1(b), the glycerol backbone in blue and the fatty acid parts in red [13].

Fault Gases Evolved by Simulated Thermal Faults Thermal tests of both oil and oil/paper mixes have been carried out. For the oil/paper mix, the oil/paper ratio was 20:1 by weight. For the tests involving only oil, all the fluids were preprocessed by drying at 85°C under a vacuum for 72 hours. For the thermal tests involving both oil and paper, the Kraft papers were preprocessed by drying at 105°C in an air circulating oven for 24 hours followed by further drying at 85°C under vacuum for 24 hours. The fluids were preprocessed by drying at 85°C under a vacuum for 48 hours. Following the drying of both the oil and the paper, the paper was impregnated in the preprocessed fluid and dried under vacuum for an additional 24 hours. At the end of preprocessing, moisture contents in the samples were measured by Karl Fischer titration method using Metrohm KF 786 coulometer and KF Thermoprep 832 (Metrohm, http://www.metrohm.com). The average moisture contents of mineral oil, Midel 7131 and FR3 were 6 ppm, 24 ppm, and 16 ppm, respectively. The paper samples had moisture contents of less than 0.6% by weight. All of the fluid and fluid/paper samples then were sealed in glass bottles and heated uniformly in an air circulating oven at temperatures of 90°C, 150°C or 200°C for periods up to 14

Table 1. Fault indicator gases. Fault gases

Key indicator

H2 (hydrogen)

Corona

Secondary indicator Arcing, overheated oil

CH4 (methane)

Corona, arcing, and overheated oil

C2H6 (ethane)

Corona, overheated oil

C2H4 (ethylene)

Overheated oil

Corona, arcing

C2H2 (acetylene)

Arcing

Severely overheated oil

CO (carbon monoxide)

Overheated cellulose

Arcing if the fault involves cellulose

CO2 (carbon dioxide)

Overheated cellulose, arcing if the fault involves cellulose

O2 (oxygen)

Indicator of system leaks, over-pressurization, or changes in pressure or temperature.

N2 (nitrogen)

Indicator of system leaks, over-pressurization, or changes in pressure or temperature.

TDGC: The total concentration of the six combustible gases (H2, CH4, C2H2, C2H4, C2H6, CO) in ppm. CO2/CO:Trending Ratio used to determine severity of cellulose degradation. O2/ N2 : Trending Ratio used to determine system leaks, over-pressurization, or changes in pressure or temperature.

6

IEEE Electrical Insulation Magazine

Figure 1. Chemical structure of synthetic and natural esters. days. This testing was intended to simulate the maximum top oil temperature found in a transformer during operating conditions (90°C) and two cases of low intensity thermal faults (150°C and 200°C). As with all of the other tests described in this article, a number of control samples of fluid and fluid/paper were kept for DGA tests to provide a benchmark. All the DGA results presented are the average of three samples to improve accuracy.

A. DGA Results from Thermal Tests of Oil at 90°C, 150°C, and 200°C At the 90°C maximum top oil temperature found under operating conditions, both mineral oil and esters should be stable for a long period of time; therefore, no significant dissolved gas should be evolved. This should not be the case for the temperatures of 150°C and 200°C at which chemical decomposition will take place. Table 2 compares the concentration of fault gases found in the three oils under the different test conditions. In terms of gas volume, Midel 7131 generated the smallest amount of fault gases. In contrast, FR3 generated a significant amount of ethane and hydrogen, particularly in the case of the 90°C test. At 90°C, none of the three types of fluid produced ethylene, which usually taken is to be a characteristic of high energy thermal faults (see Table 1). This is a positive result as the oils are expected to be stable at this temperature. For the test carried out at 200°C, the gases ethylene, ethane, and methane were generated with approximately the same ratios seen at 150°C. Figure 2 shows the relative percentages of fault gases in the total dissolved combustible gases (TDCG) at 90°C and at 150°C. Ethylene, the primary indicator for high energy thermal faults, is negligible in concentration at 90°C. It is more dominant in the results relating to the 150°C test (see Figure 3 for clarity, which is plotted without H2 and CO). In FR3, Ethane is generated in significant quantities. In Midel 7131 and mineral oil this is not the case. In all cases, methane is not present in significant quantities.

September/October 2007 — Vol. 23, No. 5

B. DGA Results from Thermal Tests of Oil/Paper at 90°C, 150°C, and 200°C Table 3 compares the concentration of fault gases of mineral oil, Midel 7131, and FR3 in the presence of Kraft paper (these tests were carried out for 14 days at 90°C/150°C and 1 hour for 200°C). The inclusion of paper has caused an increase in the concentration of carbon monoxide and carbon dioxide for the 90°C temperature. These gases are key indicators for cellulose degradation, in both mineral oil and esters. The concentrations of CO and CO2 are the highest in mineral oil, lower in Midel 7131, and the least in FR3 at this temperature. At 150°C, the inclusion of paper also increases the level of carbon monoxide and carbon dioxide significantly. This is expected as Kraft paper generally starts to be thermally degraded at temperatures above 105°C. The generation of CO is less in esters than in mineral oil suggesting that they may be protecting the paper in some way. Figure 4 shows the relative percentage of fault gases in the TDCG for the oil/paper mix at 90°C and at 150°C. In the case of mineral oil and Midel 7131, the dominant gas is carbon monoxide; however, for FR3, the concentration of carbon monoxide is similar to that without paper, indicating paper integrity may be preserved [14]. Figure 5 shows the same results with H2 and CO excluded to allow viewing of the other fault gases more prominently.

Fault Gases Evolved by Simulated Electrical Faults A. DGA Results from Low-Energy Arc Tests A 220 V/40 kV, 8 kVA test transformer was used to generate a breakdown across needle to plane electrode configuration with an oil gap distance of 15 mm. When the breakdown occurs, it will degrade the oil locally as the energy being dissipated causes the molecular structure of the oil to be disrupted. This allows the formation of fault gases like acetylene. The fault gases then dif-

7

Table 2. Dissolved gas content in parts per million (ppm) for thermal tests of oil only at 90°C aAnd 150°C for 3 and 14 days and at 200°C for 1 hour. The values highlighted in bold italics are considered to be significant shifts in dissolved gas values (only combustible gases examined). Oil type

Mineral oil

Test temp Test time

90°C

C

3d

150°C

14d

3d

200°C

14d

1h

Synthetic ester (Midel) 90°C

C

150°C

3d

14d

3d

Natural ester (FR3)

200°C

14d

1h

90°C

C

3d

150°C

14d

3d

200°C

14d

1h

H2

5

16

38

14

16

21

7

9

7

14

14

8

8

64

253

59

19

17

CH4

1

2

4

48

194

95

0

1

1

7

40

16

1

1

4

7

23

7

C2H6

0

1

2

28

125

48

0

1

0

2

49

4

2

18

103

88

179

177

C2H4

1

1

1

7

14

9

1

0

1

3

34

3

1

0

1

5

16

4

C2H2

1

0

1

0

0

5

0

0

0

0

0

0

6

0

0

0

0

0

152

533

74

6

16

53

171

540

68

1073 3514

521

82

129

430

1586

5359

914

102

24

98

414

330

777

273

CO

18

25

98

262

592

148

9 17

60

CO2

73

165

502

1976

3354

1006

111 89

283

TDCG

26

45

144

359

941

326

17 28

69

fuse from the local fault location to the bulk volume. To ensure a sufficient concentration of fault gases, a total of 20 breakdowns was produced in oil that had previously been preprocessed in a similar way to that described for the thermal tests. There was at least a 1-minute interval between each breakdown. For each breakdown test, the voltage was steadily ramped up until the oil gap broke down. The current was interrupted by the operation of an over-current relay. This relay, on the low voltage side of the power supply, was set at a 3-A limit to ensure rapid interruption of the current following formation of the arc. It normally operated within 20 ms after the formation of the breakdown, but it could reach 100 ms in certain cases.

177

670

Oil samples for DGA testing were taken from the bottom valve of a sealed test vessel. The oil was naturally forced into glass syringes according to the BS EN 60567 standard [15]. As the test vessel is sealed, a homogeneous distribution of fault gases can be expected when enough time is left after the tests for the fault gases to diffuse into the bulk of the oil. Table 4 shows the results of these tests. Acetylene should be one of the key gases produced during lowenergy arc discharge faults; and it therefore, is a primary indicator for this type of fault. This is found in the largest concentration in all samples. Hydrogen and ethylene also usually are evident in significant amounts. Although the same level of low energy

Figure 2. Relative percentages of dissolved combustible gases for mineral oil and esters at 90°C and 150°C (oil only).

8

IEEE Electrical Insulation Magazine

Figure 3. Relative percentages of dissolved combustible gases (without H2 and CO) for mineral oil and esters at 90°C and 150°C (oil only).

discharge took place in the three oils, the acetylene concentration in mineral oil is about 5 to 10 times higher than that seen in the esters. Midel 7131 has the lowest amount of dissolved gas as a result of this test; this is the same result as that seen in the thermal tests. Figure 6 shows the relative percentages of fault gases in the total dissolved combustible gases (TDCG).

down. The partial discharge tests used a standard PD detection circuit. The partial discharge inception voltages were 27.9 kV for mineral oil, 15.5 kV for Midel 7131, and 12 kV for FR3. The PD level measured during the tests was less than 100 pC for all of the types of oil. Table 4 shows the DGA results of three types of fluid for PD activity with a normalized duration of 1 hour (owing to the different generation rates, the test on mineral oil was carried out for half an hour, on Midel 7131 for 4 hours, and on FR3 for 1 hour). As shown in Figure 5, hydrogen is the key indicator for low-energy discharges, and this was found significantly in all of the oils. Mineral oil had the highest generation rate, and Midel had the lowest.

B. DGA Results from Partial Discharge Tests The fluids used for the partial discharge test were preprocessed as earlier described. The electrical circuit and test electrodes used were the same as the ones in the arcing test with the addition of a water resistor to limit the current in case of inadvertent break-

Table 3. Dissolved gas content in parts per million (ppm) for thermal tests of oil and paper at 90°C and 150°C for 14 days and at 200°C for 1 hour. The values highlighted in bold italics are considered to be significant shifts in dissolved gas values (only combustible gases examined). Oil type Test temp

Mineral oil

Control

Synthetic ester (Midel)

90°C

150°C

200°C

Control

Natural ester (FR3)

90°C

150°C

200°C

Control

90°C

150°C

200°C

H2

8

46

34

19

7

13

24

14

8

244

26

23

CH4

1

10

259

90

1

3

40

15

1

6

31

10

C2H6

0

2

187

43

1

0

33

4

1

116

179

171

C2H4

1

2

25

5

1

1

16

4

1

2

19

7

C2H2

1

1

1

0

1

1

0

1

1

0

0

1

CO

6

590

9187

890

5

307

3815

541

6

88

5472

1330

CO2

108

3407

101167

19603

45

2212

56508

9524

82

1354

60675

18717

17

654

9693

997

16

325

3928

579

18

456

5727

1542

TDCG

September/October 2007 — Vol. 23, No. 5

9

Figure 4. Relative percentages of dissolved combustible gases for mineral oil and esters at 90°C and 150°C for 14 days (oil and paper).

DGA Diagnosis The results obtained were taken as inputs into DGA diagnosis based on the various standards. In doing this, it has to be noted that the laboratory measurement error expected during the DGA measurements is ±10% for the 50 ml (minimum) oil samples used in these tests. Figure 7 shows the results of the first analysis technique to be assessed, the Duval triangle DGA fault diagnosis method. Results from the thermal tests and the low-energy, arc discharge test were analyzed. The results of the corona test were not analyzed for reasons that will be detailed later. In terms of the results for mineral oil, use of the Duval triangle method diagnosed all faults correctly as being either in the T1 region (thermal fault of less than 300°C) or in region D1 (lowenergy discharges). This result is to be expected if the laboratory tests are assumed appropriate. For Midel 7131, the Duval triangle places the oil used in the 200°C thermal test into the T1 region (i.e., overheating < 300°C). The oil subjected to the 150°C thermal tests was placed incorrectly into the T2 region (i.e. overheating temperature T, 300°C< T 98% CH4 and