You are on page 1of 59

10th IndiaDoble Power Forum, 2012

Oil Tutorial – Dissolved Gas


Analysis

Richard Heywood – Doble UK

pa ny
om
10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA

ng C
ine eri
n g
ob le E
D
© in a Transformer
Fluids

Why use fluids in a transformer?


• It is an efficient cooling/heat transfer medium
• It provides sufficient dielectric strength (in conjunction
with paper)
• Quenches arcs
• Properties remain stable over lifetime
• It can be used to transfer easily accessed diagnostic
information as it passes around the transformer.

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 2


Fluids in use- but we will concentrate on mineral oils

• Mineral oil
– Naphthenic
– Paraffinic
– Inhibited/ uninhibited

• Silicones for fire risk reduction in buildings

• Esters
– Synthetic for high temperature use
– Natural for environmental impact reduction

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
3

C om
ng
ine eri
n g
o b le E
D
© content- Oil diagnostics
Tutorial

Chapter 1
Oils and their important properties
Detecting deterioration of oil. The impact of deteriorated
oil on transformer performance

Chapter 2
Detecting transformer deterioration- dissolved gases
and Furans

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 4


Chapter 1- Oils

• What are the oils in use

• Optimisation of oil properties

• Oil deterioration – the problems for a transformer

• Detection of oil deterioration

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
5

C om
ng
ine eri
n g
ob le E
D
© Use Mineral Oils
Why

• Price and availability good- or has been!


• Efficient cooling/heat transfer medium
• Provides sufficient dielectric strength
• Fills voids in the core and coil assembly thus providing
better insulating properties
• Restricts oxygen and moisture permeation into paper and
other materials
• Stable for many years
– no sludging and low acidity
– retains essential electrical properties

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 6


Why paper and oil?

• The dielectric strength of oil is:


» 12 kV per mm
• The dielectric strength of paper is:
» 40 kV per mm
• The dielectric strength of oil impregnated paper is:
» 64 kV per mm

There is a 12kV synergy effect per mm which represents a


23% increase in dielectric strength for the paper/oil
insulation system

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
7

C om
ng
ine eri
n g
ob le E
© D
Naphthenic and paraffinic crude oil

Mineral oils are a mix of molecules- it is


just the proportion that means one is
referred to as naphthenic or
paraffinic

Paraffinic content (IR)* Considered as

CP 42-50% Naphthenic
CP 50-56% Intermediate
CP 56-65% Paraffinic

*These levels are not exactly defined but more


intended as a guideline

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 8


Specifications

Basic International Standards for mineral oil for electrical


use:

IEC 60296, and ASTM D 3487

BUT these standards state the minimum requirement for electrical


grade oils
Some other standards seek to enhance the requirements
for large transformers
IEC60422, Doble TOPS

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 9

C om
ng
ine eri
n g
o b le E
D
© 2 Oil Testing
Chapter

Oil Testing to diagnose problems with a


transformer

1. Dissolved gas analysis (DGA)

2. Furans
3. Oil Quality Will discuss if there is sufficient time

4. Sulphur Corrosion

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 10


Difficulties of using DGA

• Deciding on an appropriate sampling interval

• Sampling is important

• Laboratory procedures

• Deciding when a result is abnormal

• Deciding on seriousness of a fault

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
11

C om
ng
ine eri
n g
ob le E
© D
Dissolved Gas Analysis - History

1920s
Users were concerned about safety and whether Buchholz
gases were combustible- so used flame at vent.
Essentially looking for Hydrogen.

1930-50
Able to identify H2, C2H2 and CO in relay gas

1950-60- development of chromatography


Realised better to test the oil rather than relay gas.
Able to detect all combustible gases
10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 12
Why Measure Dissolved Gases in Oil (DGA)

• Ideal condition monitoring tool

– Sensitive, broadband
• Sampled by site staff and analysed by oil laboratory

• Or by Online Gas Sensors

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
13

C om
ng
ine eri
n g
ob le E
© D
Dissolved Gas Analysis

• Lab Analysis involve gases extracted from the oil and the
analysed through a Gas Chromatograph (GC)
• Method A - Vacuum extraction/GC Analysis

• Method B - Stripper Column/GC Analysis

• Method C - Headspace/GC Analysis

• Online methods use different detectors but still require the


gases to be separated from the oil.

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


Typical Standard Gas Chromatogram

NITROGEN

METHANE
CO2
pA

5.927 -
1750

12.150 -
4.694 -
1500

TCD1 B, (121202ST\004F0401.D) FID2 A, (121202ST\004F0401.D)


1250

ETHYLENE
1000

ACETYLENE

ETHANE

11.722CO
750
HYDROGEN

8.879 -
OXYGEN
500

9.816 -
7.933 -
250
3.455 -

4.144 -

ny
2 4 6 8 10 12

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa
C om
ng
ine eri
n g
ob le E
© D
Minimum Detection Limits
(MDL) – Lab Analysis

Gas MDL MDL MDL MDL


Method A Method B Method C Method
(ppm) (ppm) (ppm) IEC 60567
(ppm)
Hydrogen 5 20 0.6 2
Carbon 25 2 0.09 10
Monoxide
Hydrocarbons 1 1 0.04-0.06 0.2 –1

Online devices vary depending on technology used

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


What Effects the Gas in Oil Level

¾Oil Type – Stray gassing!


¾Oil Preservation Systems Systems
¾Loading
¾Historic Issues
¾Bushing faults
¾Tap changer faults
¾Was Oil Degassed
¾Partition of the Gas between headspace and the oil

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
17

C om
ng
ine eri
n g
ob le E
D
© of oil type on DGA
Effect

• Oil type can change symptoms exhibited


• Different gassing characteristics
– temperature and time effects
• May also affect head-space gas extraction

• How can we allow for this ?


– need to know oil type
– need information on gassing properties
– tailor scoring/interpreation parameters for oil type

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 18


Oil Preservation Systems

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
C om
ng
ine eri
n g
ob le E
© D
Oil Partition Coefficients

Gas Ostwald Coefficient


Oxygen 0.138
Nitrogen 0.0745
Carbon Dioxide 0.900
Carbon Monoxide 0.102
Hydrogen 0.0429
Methane 0.337
Ethane 1.99
Ethylene 1.35
Acetylene 0.938

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


Gas Partitioning

Gas Gas-in-Oil Estimated Estimated


ppm ppm %
(vol/vol) gas-in-gas gas-in-gas
space** space
Oxygen 28,400 206,000 20.6
Nitrogen 59,000 792,000 79.2
Carbon Dioxide 1,000 1,110 0.11
Carbon Monoxide* 100 980 0.10
Hydrogen* 100 2,330 0.23
Methane* 100 297 0.03
Ethane* 100 50 0.01
Ethylene* 100 74 0.01
Acetylene* 100 107 0.01
*Combustible gases
**Estimated value under equilibrium conditions at 25°C and 1 atm

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
C om
ng
ine eri
n g
o b le E
D
© of Gases in Oil
Solubility

• Hydrogen, Nitrogen, CO, and Oxygen


– increase with temperature
• CO2, Acetylene, Ethane, Ethylene
– decrease with temperature
• Methane
– essentially unchanged
• All increase proportionally with Pressure
– doubling the pressure doubles the gas
concentration in oil, in atmospheres
• All modestly increase with decreasing oil density
10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA
Solubility of Gases in Oil

• Hydrogen, Carbon Monoxide, (Methane to lesser


degree)
– proportionally higher in closed conservator than gas
blanketed
– slowly lost to atmosphere in open conservator
– Leak

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
C om
ng
ine eri
n g
ob le E
© D
Homogeneity of Gases in Oil

• Rate of generation
• Access of fault area to flow
• Rate of mixing
• Presence of a gas blanket
• Diffusion

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


Oxygen and Nitrogen

• Oxygen
– increase-leak
– Decreases with overheating
– consumed in chemical reactions
• Nitrogen
– pressure and operating temperature dependent in
gas blanketed systems

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
C om
ng
ine eri
n g
ob le E
D
© Acceptable Limits Various Sources
DGA

Hydrogen CO Methane Ethane Ethylene Acetylene CO2 TCG

DOBLE 100 250 100 60 100 5 -- 610

100 350 120 65 50 35 2500 720


*IEEE 101-700 351-570 121-400 66-100 51-100 36-50 2500-4000 721-1920
701-1800 571-1400 401-1000 101-150 101-200 51-80 4001-10000 1921-4630
>1800 >1400 >1000 >150 >200 >80 >10000 >4630
IEC
60599 60-150 540-900 40-110 50-90 60-280 3-50 5100-13000
Typical
Range

*IN THE PROCESS OF BEING REVISED

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


Why DGA – where do the gases come from?

Heat is generated in a transformer by-


¾ Normal current flow -but should not exceed the permitted hot
spot of 98 or 110ºC
¾ Partial discharge around a defect or arcing at a major fault
¾ Hot spots from design, manufacturing or in service fault
- stray flux heating in the winding
- at poor connections, shunts, multiple earth paths
Oil is a coolant medium and should pass over all heated
areas. The oil will also overheat locally and breakdown
to create dissolved gases.

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
27

C om
ng
ine eri
n g
ob le E
D
© are the “key” Gases
What
• The 5 “key” gases are used to determine the type of fault
– Acetylene – C2H2
– Ethylene – C2H4
– Ethane – C2H6
– Methane – CH4
– Hydrogen H2
• Other gases
– Carbon Monoxide
– Carbon dioxide
• Lesser extent depending on preservation system
– Oxygen
– Nitrogen

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 28


Energy Required to Break Bonds and Form Gases
Various gasses are created during oil decomposition depending on the
type of fault
Low Hydrogen H H 338 kJ/mole
Temperature H
About 120 °C
Methane H C H 338 kJ/mole

H
H H
Ethane H C C H 607 kJ/mole

H H
Ethylene H H 720 kJ/mole
High C C
Temperature H H
About 700 °C Acetylene 960 kJ/mole
H C C H
10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA
pa ny
C om
ng
ine eri
n g
o b le E
D
© Gas Analysis (DGA)- Halstead 1973
Dissolved

HYDROGEN
Normal and overload temperatures plus PD

METHANE

ETHANE

ETHYLENE

Hot Spots High temperature thermal ACETYLENE


faults
Temperature

200°C 500°C 700°C


TEMPERATURE
10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 30
Key Gases - Arcing

100
90
Combustibles, % 80
70
60
50
40
30
20
10
0
Hydrogen

Methane

Ethylene

Acetylene
Ethane
CO

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
C om
ng
ine eri
n g
ob le E
© D
Key Gases - Overheating, Oil

100
90
80
Combustibles, %

70
60
50
40
30
20
10
0
Hydrogen

Methane

Ethylene

Acetylene
Ethane
CO

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


Key Gases - Partial Discharge

100
90
Combustibles, % 80
70
60
50
40
30
20
10
0
Hydrogen

Methane

Ethylene

Acetylene
Ethane
CO

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
C om
ng
ine eri
n g
ob le E
© D
Key Gases - Overheating, Paper

100
90
80
Combustibles, %

70
60
50
40
30
20
10
0
Hydrogen

Ethylene
Methane

Acetylene
Ethane
CO

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


Summary Key Gases and particular faults

Fault Primary Secondary


Partial Discharge Hydrogen
Dielectric Hydrogen Ethylene
(Arcing/sparking) Acetylene
Dielectric Hydrogen Methane
(Power arc) Ethylene
Acetylene
High Operating Methane Carbon Monoxide
Temperature Ethane
Thermal hotspot Ethylene Methane
Ethane
Overheated Carbon Monoxide Carbon Dixoxide
Cellulose

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 35

C om
ng
ine eri
n g
o b le E
© D
Interpretation of dissolved gas analyses

¾Seems Easy...?
¾But its not always easy to interpret results so a
number of methods are used.
¾Lab software?
¾IEC/IEEE or normal/ abnormal values?
¾Trends?
¾Engineering experience of design and
application?

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 36


Interpretation of dissolved gas analyses

• The science behind gas production and temperature was


published in J. Inst Petroleum 1973 by WD Halstead of
CEGB R&D - and the previous slide was based on this.
• Ratio methods- Rogers, IEC and Duval
• Trending Gas levels
• Action levels
• Expert systems
• Key gas method of IEEE
• DGA algorithm

We will look at each in turn, then at some case studies

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
37

C om
ng
ine eri
n g
ob le E
D
© Important Ratios
Some
• Acetylene/Ethylene
– <10% thermal problems
– > 10 but <20% thermal problems often developing into
electrical fault – beginning of discharge activity
– >20% arcing
• Carbon dioxide/Carbon monoxide
– <3:1 indicative of high temperature localized
overheating of paper
– >10:1 general heating of the winding

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


Important Ratios
• Ethylene/Ethane
• >4:1 indicative of heating of bare metal
• <4:1 indicative of heating of paper-wrapped
conductor
• Not always correct but can provide indication
that paper is involved even when carbon oxides are
low
• Sometimes not enough paper overheated to
increase carbon oxides in large oil volume beyond
normal concentrations
• Sometimes paper burned away at some point and
little further deterioration in paper
10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA
pa ny
C om
ng
ine eri
n g
ob le E
D
© methods- Rogers and CEGB workers
Ratio

Going back to the Halsted work- his engineering colleagues


realised that temperature affected the ratio between gases It
was developed into the ratio method attributed to Rogers in
1974 and eventually built into IEC.

Methane/Hydrogen if >1 score 1 – if <1 score 0


Ethylene/Methane if >1 score 1 – if <1 score 0
Ethylene/ Ethane if >1 score 1 – if <1 score 0
Acetylene/ Ethylene if >1 score 1 - if <1 score 0

This produces scores 0000 (normal) , 1000, 0100, 0010…etc

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 40


Ratio methods- Rogers and CEGB workers

A- Methane/Hydrogen if >1 score 1 –overheating <150C


B- Ethylene/Methane if >1 score 1 – same 200-300C
C- Ethylene/ Ethane if >1 score 1 – conductor overheating
D- Acetylene/ Ethylene if >1 score 1 – Power flashover

If A and B scored 1- overheating -150- 200C


If A and C scored 1- circulating currents and overheated joints
If B and D scored 1 – tap changer current breaking
If C and D scored 1 – Arc with power though or persistent
sparking

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
41

C om
ng
ine eri
n g
ob le E
D
© Methods
Ratio
• Advantages
– quantitative
– independent of oil volume
– can be computer programmed
• Disadvantages
– don’t always yield an analysis
– not always correct
– dependence of preservation system
• Solid insulation handled separately

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


Duval’s triangle

The Rogers method is binary- just a simple question of the


ratio being > or < 1.
But does it matter if the ratio is just over 1 or as high as 10x?
Duval tries to make more out of the ratios by using the actual
values and produced a triangle.

He recognises the use of the key gas method- that the %TCG
proportions depend upon the fault type and builds that in.

But like the ratio method he is trying to identify the type of


fault- not to identify a malfunction at its outset.

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 43

C om
ng
ine eri
n g
o b le E
D
© use of Key Gases -
Duval’s

100
90
80
Combustibles, %

70
60
50 Arcing
40 Heating Oil
30 PD
Heating Paper
20
10
0
Acetylene
Methane
Hydrogen

Ethane

Ethylene
CO

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 44


Classical Duval Triangle – Mineral Oil Filled
Transformers

PD = Corona Partial Discharges


D1 = Discharges of low energy
D2 = Discharges of high energy
T1 = Thermal faults, T < 300 °C
T2 = Thermal fauts, 300 °C < T < 700 °C
T3 = Thermal faults, T > 700 °C
DT = Mixtures of electrical, thermal faults

Note there is no
area in the triangle
for a transformer
that doesn’t have a
problem.

Provided by Michel Duval

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
C om
ng
ine eri
n g
ob le E
© D
Use of the triangle

PD = Corona Partial Discharges


D1 = Discharges of low energy
D2 = Discharges of high energy
T1 = Thermal faults, T < 300 °C
T2 = Thermal fauts, 300 °C < T < 700 °C
T3 = Thermal faults, T > 700 °C
DT = Mixtures of electrical, thermal faults

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 46


Comment on Rogers and Duval methods

• Good for their Thermodynamic basis


• Good for diagnosing fault after a trip
• Not good at deciding if DGA result is normal

– e.g. Rogers ‘Normal Deterioration’ (0000)


code
• Errors at low gas levels- dividing one small number by
another can get infinity, when levels are at lab sensitivity

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
47

C om
ng
ine eri
n g
ob le E
D
© Analysis
Trend

• Look at Key gases


• Total combustible gas and not all the gases
• Rate of gas generation - important
• Fingerprints – DGA Signature
– Does it look like another transformer/same design?

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


Gassing Rates – Note

• Thermal and PD – general rule


– <1 ppm per day - normal, depending on loading,
previous behavior, time interval
– 1-3 ppm- indications of developing problem
– 3 + ppm increasing serious problem
– 100 ppm - immediate removal from Service

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
C om
ng
ine eri
n g
ob le E
D
© Analysis
Trend

Increasing gas levels considered indication of developing


fault, and gas levels constant for inactive fault but ...

• Degassing removes gases and ruins trend


• There is increase rate after de-gassing to equilibrium
levels
• solubility and dispersion effects
– different interpretation for H2 and C2H2 trends
• may not have enough results for good trend

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 50


A good trend –but when to de-energise?

3- 1996 Rapid
increase phase

2- Gas evolution
increase but
take out now?

1- 1994 Something
clearly happening

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 51

C om
ng
ine eri
n g
o b le E
D
© unit- Degassing loses the trend
Another

80 T5533 Dissolved Gas levels


Oil was
Degassed
CH4
70 C2H4
C2H6
60 C2H2
H2
Gas level (ppm)

50

40

30

20

10

0
14/11/84 11/08/87 07/05/90 31/01/93 28/10/95 24/07/98 19/04/01
Date

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 52


Normal and Action Levels

• What level is significant ...


• Statistical approach is problematic:
– what exceedance should be used - 90, 95 or
97% ?
– should all gases be treated as equally
important ?
• ‘abnormal’ results are ‘unusual’ rather than
‘unhealthy’

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 53

C om
ng
ine eri
n g
ob le E
D
© levels in Standards- OK if levels below…
Action

Gas IEEE CIGRE JWG Case A Case B


C57.104 15/12-01-11 What about these 2 units?
H2 <100 50 – 150 110 88

CO <350 400 – 600 120 227

CH4 <120 30 – 130 250 17

C2H4 <50 60 – 280 250 9

C2H6 <65 20 – 90 120 4

C2H2 <2 2 - 20 10 9

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 54


Case A- LV winding fault caused by
overheating

Fault was at a transition point in


innermost layer of LV.
Leakage flux in LV-tertiary gap
when tertiary loaded would have
circulated current through inter-
strand fault.

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 55

C om
ng
ine eri
n g
ob le E
© D
Case B- hoop buckling and damage to tertiary
underneath

Note this hadn't failed!


10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 56
CASE A- But there had been a long duration DGA
trend and 2 Buchholz alarms
250

When do you
Second Buchholz alarm
take this unit
at next attempt to load tertiary
200

- thereafter banned
out? Here at
CIGRE /IEEE
150
CH4 limit?

100 First Buchholz

50

0
25/08/95 24/09/95 24/10/95 23/11/95 23/12/95 22/01/96 21/02/96 22/03/96 21/04/96 21/05/96 20/06/96

H2 CH4 C2H4 C2H6 C2H2

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 57

C om
ng
ine eri
n g
o le E
bAction from DGA results
© D

Can you rely upon DGA?

Would you expect DGA to indicate risk in these two units?


When could you have ascribed a higher risk rating on these
two? In case B –never and in case A only very late- well after
2 Buchholz alarms!!!

So you must make sure you understand what criteria your oil
lab uses for its interpretation

– or analyse results yourself

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 58


Expert systems?

• attempt to classify similar results


• search for hidden information
• identify changes of state
• Kohonen net cluster analysis technique
• rely on human expert to ascribe significance

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
59

C om
ng
ine eri
n g
ob le E
© D
Alternative Interpretation Schemes DGA Scoring
Algorithm?

• Method based on IEEE Key Gas Method C57-104 1991


• Combustible gas signature - bar chart of percentage
total combustible gas (%TCG), CO, H2, CH4, C2H4,
C2H6, C2H2
• Used in IEEE to distinguish faults ie like IEC ratios used
after a trip- too late!
• Based of examples of faults/failures - to differentiate
normal and abnormal units for risk assignment.

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 60


Based on the key gas method

High [ CO ] ( 60-90 %TCG ) = normal

– ( Higher levels may be abnormal )

• High [ C2H4 ] ( > 10 %TCG ) = abnormal

• Significant [ H2 ] and [ C2H2 ] = serious

• Significant [ CH4 ] and [ C2H6 ] = not serious

Use this ratio system and add in a gas level factor to

build an algorithm

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 61

C om
ng
ine eri
n g
o b le E
D
© transmission transformers
Normal

100

80
Percentage TCG

60

40

20

0
1 2 3 4 5
Transformer

CO H2 CH4 C2H4 C2H6 C2H2

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 62


Normal generator transformers

100

Percentage TCG 80

60

40

20

0
I II III IV V
Transformer

CO H2 CH4 C2H4 C2H6 C2H2

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 63

C om
ng
ine eri
n g
ob le E
© D
Faulty transformers

100

80

Overheating Discharge/arcing
Percentage TCG

60

40

20

0
A B C D E F G
Transformer

CO H2 CH4 C2H4 C2H6 C2H2

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 64


Several with Bare metal Thermal Faults

100

90
Now with Ethylene as dominant gas and > methane
80 >ethane
70
Percentage combustible gas

60

50

40

30

20

10

0
V W X Y

CO H2 CH4 C2H4 C2H6 C2H2

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 65

C om
ng
ine eri
n g
ob le E
Several units with Winding Thermal Faults © D

100

90
Now have ethylene, together hydrogen dominating
80
some times acetylene coming through
70
Percentage combustible gas

60

50

40

30

20

10

0
A B C D E F G H

CO H2 CH4 C2H4 C2H6 C2H2

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 66


Dielectric problems - DGA

Typical DGA results for transformers with PD faults


90

80
H2
70

60
%CO
Transformers

50 %H2
%CH4
%C2H4
40 %C2H6
%C2H2
30

20

10

0
1 2 3 4
Percentage TCDG

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 67

C om
ng
ine eri
n g
o b le E
© D
Dielectric problems - DGA

DGA Results for a family with low energy arcing/sparking faults

100

90

80

70

60 %CO
Transformers

%H2
%CH4
50
%C2H4
%C2H6
40
%C2H2

30

20

10
Acetylene
0
1 2 3 4
Percentage TCDG

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 68


Case A The thermal damage unit

So when did this problem start and is there only one


problem? The next slide shows that this should have
been identified between 1991 and 1994 and unit put on
a risk list for investigation and life management…. But
the Scoring Algorithm wasn’t being used then.
250

200

150

100

50

0
25/08/95 24/09/95 24/10/95 23/11/95 23/12/95 22/01/96 21/02/96 22/03/96 21/04/96 21/05/96 20/06/96

H2 CH4 C2H4 C2H6 C2H2

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 69

C om
ng
ine eri
n g
ob le E
© D
Case A - key gas method determines fault type
and when it started.

Evidence of thermal fault years before Buchholz alarm #1

Buchholz alarm #1
Buchholz alarm #2

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 70


One Scoring system – Lapworth Algorithm

• ‘Quality’ function

– importance of gas signature

• ‘Strength’ function

– importance of gas levels

The limitation of the IEEE key gas method is it relies


upon the signature and doesn’t use level information
And its important to look at signature and gas level
10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA
pa ny
71

C om
ng
ine eri
n g
ob le E
D
© NG algorithm for DGA interpretation
DPT-

• Normal < 60

– all normal results can be ignored


• Suspect > 60

– only suspect results need to be investigated


• Serious > 100

– winding and dielectric faults


– some can be discounted

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 72


Benefits of scoring system

• Normal results can be screened out (ignored)

• Consistent, objective assessment of mass of routine


data- a first stage screen

• Based on actual failures/faults

• Resource then can be focussed on deserving cases

• Also can ignore ‘unusual’ cases

• New insights into significance of results

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 73

C om
ng
ine eri
n g
ob le E
D
© the onset of abnormality- 765 kV Reactor
Detecting
400

350

100 300

250
DGA Score

90 200

150

80
100

50
Percentage combustible gas

70
0
31/01/93 01/02/95 01/02/97 02/02/99 02/02/01 03/02/03 03/02/05

60 Date

50

40

30

20

10

0
26/7/93 22/4/99 24/10/02 16/5/05 26/1/06
Date

CO H2 CH4 C2H4 C2H6 C2H2

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 74


Case B- So how did we handle this one?

Note this hadn't failed!


10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA
pa ny 75

C om
ng
ine eri
n g
ob le E
D
© B- Managing DGA Data and its Interpretation
Case

3-
120

1- Test 100

80
investigate
60
high scores
40

120
20

100

0
18/02/1982 17/02/1985 17/02/1988 16/02/1991 15/02/1994 14/02/1997 14/02/2000 13/02/2003 12/02/2006

80

2- Apply Score ALL 60

results
40

20

0
18/02/1982 17/02/1985 17/02/1988 16/02/1991 15/02/1994 14/02/1997 14/02/2000 13/02/2003 12/02/2006 11/02/2009

H2 CH4 C2H4 C2H6 C2H2

20
5 Look at past history
18

16

14

12

10

8 4 Look at gas
pattern for
6

0
18/02/1982 17/02/1985 17/02/1988 16/02/1991 15/02/1994 14/02/1997 14/02/2000 13/02/2003 12/02/2006 11/02/2009
fault type
C2H4 C2H2

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 76


Case B- The first step all DGA scored - here >50

120 The DGA score is a composite


value calculated from ratios
100
(weighted average of %TCG’s),
trained on past failures.
80
(2002 Doble Conference paper)

60

Only faulty units go above about 50


40

20

ny
18/02/1982 17/02/1985 17/02/1988 16/02/1991 15/02/1994 14/02/1997 14/02/2000 13/02/2003 12/02/2006

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa 77

C om
ng
ine eri
n g
o b le E
D
©B- Look at DGA history for incidents/transients
Case
120

In this case the jump in DGA score was


100 due to a significant increase in H2.
This coincided with a CT failure
80
(through fault) on the HV side of the
transformer.
60

40

20

0
18/02/1982 17/02/1985 17/02/1988 16/02/1991 15/02/1994 14/02/1997 14/02/2000 13/02/2003 12/02/2006 11/02/2009

H2 CH4 C2H4 C2H6 C2H2

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 78


Case B- And the DGA history investigated
20

18

16
2003
CT failure
14

12

10

0
18/02/1982 17/02/1985 17/02/1988 16/02/1991 15/02/1994 14/02/1997 14/02/2000 13/02/2003 12/02/2006 11/02/2009

ny
C2H4 C2H2

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa 79

C om
ng
ine eri
n g
ob le E
© D
Case B- But still took us 2y to get the outage to test
T4971 LV to N 12/10/05
0

-10

-20
Amplitude in dB

-30

-40

-50

-60
0 0.5 1 1.5 2
Frequency in MHz
T4971
A phase B phase C phase A phase B phase C phase

Main windings 4,246 pF 3,368 pF 4,273 pF


to earth, CH 0.75 % 0.51 % 0.62 %
Clearly winding
movement and Main windings 8,277 pF 5,775 pF 9,030 pF
inspection justified. to tertiary, CHT 0.42 % 1.91 % 0.54 %

Expect hoop Tertiary winding 20,668 pF


buckling. to earth, CT 0.46 %

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 80


Case B- Drop oil and look in- significant
clamping plate damage

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
81

C om
ng
ine eri
n g
ob le E
© D
Case B- And so to scrapping….

It is likely that the hoop


buckling to the common
winding led to crushing
damage to inner tertiary
(right) and it was PD
from here that caused
the DGA results. But the
levels were so low we
would never have
spotted the problem
from other methods.

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 82


CASE C – 240MVA Main Tank
400/132kV 240 MVA
Manufactured 1966.

Removed from
service and
replaced because of
overheating.

Originally leading
ethylene, but in later
life hydrogen
following and CO.
Furans only 0.8ppm

DGA Score follows


trend, as expected.

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
83

C om
ng
ine eri
n g
ob le E
D
© C – 240MVA Main Tank
CASE
400/132kV 240 MVA
Manufactured 1966.

Removed from
service and
replaced because of
overheating.

Originally leading
ethylene, but in later
life hydrogen
following and CO.
Furans only 0.8ppm

DGA Score follows


trend, as expected.

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 84


CASE C – 240MVA Main Tank
400/132kV 240 MVA
Manufactured 1966.

Removed from
service and
replaced because of
overheating.

Originally leading
ethylene, but in later
life hydrogen
following and CO.
Furans only 0.8ppm

DGA Score follows trend, as expected. But fault not clear over entire history

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
85

C om
ng
ine eri
n g
ob le E
D
© C – 240MVA Main Tank
CASE

Ethylene/Hydrogen increasing relatively to each other in latter life

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 86


CASE C – 240MVA Main Tank

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
87

C om
ng
ine eri
n g
ob le E
D
© C – 240MVA Main Tank
CASE

Overheated tie bars –


Leakage flux.

Caused charring of
wood package blocks
10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 88
DGA Scoring Conclusions

• Scoring system has been valuable


– more consistent and objective assessments
– resources focussed where deserved
– Still need to apply engineering knowledge
• Applicable to wider class of transformers ?
– further work required to improve
• Oil type needs to be taken into account
– further work to understand effects on stray gassing
• Vehicle for improving understanding...maybe?

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
89

C om
ng
ine eri
n g
ob le E
D
© D - Selector
CASE

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 90


CASE D Selector Inspection

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
91

C om
ng
ine eri
n g
ob le E
D
© E - DGA Selector
CASE

180
Stalybridge SGT1May
Inspected Blue 09
Phase
andSelector
Sep 10
160
Nothing found

140 H2
CH4
120 C2H4
) C2H6
m
p 100
p
(
l
C2H2
e
v
e
L
s 80
a
G

60

40

20

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 92


CASE E - DC Winding Resistances

Problem here but not here?

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
93

C om
ng
ine eri
n g
ob le E
D
© E Spring Tension Issues
CASE

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 94


Case F McGraw Edison 550C LTC
Gas 1/19/2005
Hydrogen 177
Ethane 1,454
Methane 1,800
Ethylene 3,295
Acetylene 14.6
Carbon Monoxide 178
Carbon Dioxide 750
Oxygen 25,490
Nitrogen 67,800
Total Combustible Gas 6,919
Methane/Acetylene 123
Ethylene/Acetylene 226
(H2+Acet)/TCG-CO 0.03
Doble Condition Code 1

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
C om
ng
ine eri
n g
ob le E
© D
Case F Transformer and LTC

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


Case F Infrared Thermography

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
C om
ng
ine eri
n g
ob le E
© D
Case F Coking on Reversing Switch

Coking on reversing switch

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


CASE G Selector DGA

Leading Acetylene....low
levels of other gases

Inspection no arcing within selector....however oil was leaking

ny
from diverter into selector!
10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA
pa 99

C om
ng
ine eri
n g
o b le E
D
© 3 - Online DGA
Chapter

• Laboratory testing is generally a very cost effective


solution and served the industry well.
• But on-line devices are coming down in price and are
getting more reliable. Also better communications!.
• Early ones were simple hydrogen detectors, now multi
gas systems are becoming more common.
• But mass of data needs careful handling- many rely
upon Rogers and Duval- not much good in early stages…

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 100


Online DGA Tools

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
C om
ng
ine eri
n g
o b le E
D
© Online DGA
Bespoke

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


CASE H – DGA Selector

More gassing, inspected, nothing,


online monitor installed, Aug 11
Silver Corrosion, oil
replaced in selector Oil replaced in sel
and mtnk, more
Reclaimed corrosion, next slide!
and
Inhibited
Passivated 2008, still positive to
CuS, further oil processing.

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 103

C om
ng
ine eri
n g
ob le E
D
© H – Selector Inspection
CASE

2008 2011

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 104


CASE H – Online Monitoring
Red Phase Selector – Acetylene

Note No PD detected. Therefore believed to be chemical?

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny
105

C om
ng
ine eri
n g
ob le E
D
© I – Online DGA Monitoring
CASE

• 400/132kV 240MVA
• Manufactured 1997
• Removed from service in 2009
– After some investigation!
– Ethylene leading fault
– Lower amounts of hydrogen
• Design had been under suspicion
– Sister units overheating faults
– Core/frame insulation/circulating current issues

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 106


CASE I – 240 MVA – Main Tank DGA
Diagnostic Gas History for SEABABK SGT2 (T6991)_Main Tank
350

300

Gas concentration (ppm)


250

200
H2
150 CH4
C2H4
100
C2H6
C2H2
50

DGA Score History for SEABABK SGT2 (T6991)_Main Tank


Sample date
100
90
80
70
60
50
Score

40
30
20
10
0

ny
Sample date

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa 107

C om
ng
ine eri
n g
ob le E
D
© I Core/Frame Insulation Resistor Installed
CASE
Parts Insulation Resistance (Mohm)

19/12/1996
25/10/2008
(commissioning)
Core – earth (frame earthed) 966 15 000
Frame – earth (core earthed) 839 14 000
Core and frame – earth 1830
Core – frame 1120 5000

Lead to core/frame

Spark Gap

Lead to earth mat

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 108


CASE I DC Winding Resistances - 2008
LV-N Winding Resistance for Seabank SGT2
Measurements at Works 24/10/96
0.15

Factory Data
0.14
Winding Resistance (ohm)

0.13

0.12

0.11

LV-N Winding Resistance for Seabank SGT2


0.1 0.15

0.145
0.09
1 3 5 7 9 11 0.14
13 15
Tap
0.135

Winding Resistance (ohm)


A phase B phase C phase
0.13

0.125

0.12

No clear evidence LV-N. 0.115

0.11

0.105

Selector required maintaining. 0.1


1 3 5 7 9 11 13 15
Tap

A phase B phase C phase

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 109

C om
ng
ine eri
n g
ob le E
D
© I Online DGA Fitted 2008/Selector Maintained
CASE

Gas
continued to
rise. Unit
switched
out during
meeting

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 110


CASE I Repeated Offline DC Winding Resistances
LV-N Winding Resistance for Seabank SGT2
Measurements at Works 24/10/96
0.15

0.14

Clear evidence of high


Winding Resistance (ohm)

0.13
resistive joint on C
0.12 Phase LV-N.
0.11
Note Autotransformer
0.1
LV-N Winding Resistance for Seabank SGT2
Measurements at Site 25/10/08
0.09
0.15
1 3 5 7 9 11 13 15
Tap

A phase B phase C phase


0.14

Bad joint - C phase


Winding Resistance (ohm)
0.13
common winding or tap 15
lead or possible part of the 0.12

LV line lead assembly.


0.11

Repairable at site only if a


0.1
bad joint on lead to selector.
Whereas fault in common 0.09
winding or the LV line lead 1 3 5 7 9 11 13 15
Tap

ny
assembly would probably A phase B phase C phase

a
only be repairable at works.

p
10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 111

C om
ng
ine eri
n g
ob le E
D
© I Internal Inspection – Repairable?
CASE
Paper removed
and no fault found!

C Phase Riser
(Top)

C Phase Top Series


Connection

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 112


CASE I Scrapped - Fault Location

Worst DP was
approx 600
80% life left!
2-FAL 0.54ppm

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 113

C om
ng
ine eri
n g
ob le E
D
© I 240MVA – Conclusion
CASE

• Two design faults now identified


– Core/frame issues
– Weak joint in conductor

• Careful Judgement with existing knowledge


• Leading ethylene not always bare metal faults
• Online monitoring allowed instant decision on switch out
• Which eliminated the risk of catastrophic failure

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 114


Conclusions

• DGA is only part of the story


• Other oil tests provide Condition information
– Furan Analysis
– Oil Quality
– Including Corrosive Sulphure

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 115

C om
ng
ine eri
n g
ob le E
© D
We can learn a lot from oil testing- but

Remember- this is
not a
transformer –it
is only an oil
sample

don’t
forget the
engineering
This is ONLY a starting position

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA 116


Condition assessment is more than oil testing

On-line,
Identify SURVEY
diagnostics to & non Outage
match relevant invasive investigative
failure modes testing testing

FMECA & Assess


Current Maintenance
Condition
Obtain OEM Work, future plans &
& RLA
operational
design environment
performance
Identify
priority
actions

10th IndiaDoble Power Forum, 2012, Vadodara, Gujarat, INDIA


pa ny 117

C om
ng
ine eri
n g
ob le E
© D

You might also like