Dissolved gas analysis and the Duval Triangle
by Michel Duval
-DGA is for Dissolved Gas Analysis. -Still today, DGA is probably the most powerful tool for detecting faults in electrical equipment in service. -Over one million DGA analyses are performed each year by more than 400 laboratories worldwide.
. i. -for example. CH2* and CH* radicals. -some of these bonds may break and form H*.. C-C).e. as a result of faults or chemical reactions in the equipment. linked by chemical bonds (C-H. containing hydrogen and carbon atoms. CH3*. oil is a molecule of hydrocarbons.-Gases in oil always result from the decomposition of electrical insulation materials (oil or paper).
All these radicals then recombine to form the fault gases observed in oil:
CO and H2O. because of the presence of oxygen atoms in the molecule of cellulose:
. the decomposition of paper produces CO2.-in addition to these gases.
The main gases analyzed by DGA
Hydrogen Methane Ethane Ethylene Acetylene Carbon dioxide Oxygen Nitrogen H2 CH4 C2H6 C2H4 C2H2 CO2 O2 N2
Carbon monoxide CO
will form mainly H2 and CH4.
. low energy faults such as corona partial discharges in gas bubbles. or low temperature hot spots.-some of these gases will be formed in larger or smaller quantities depending on the energy content of the fault. -for example.
-and finally. to form large amounts of C2H2. -by looking at the relative proportion of gases in the DGA results it is possible to identify the type of fault occurring in a transformer in service. such as in electrical arcs.-faults of higher temperatures are necessary to form large quantities of C2H4. it takes faults with a very high energy content.
. 1. as a result of poor drying or poor oil-impregnation.Partial discharges of the corona-type (PD). -typical examples are discharges in gas bubbles or voids trapped in paper.6 basic types of faults detectable by DGA have thus been defined by the IEC and other organizations.
2. -or low-energy arcing. inducing carbonized perforations or surface tracking of paper.
. inducing pinholes or carbonized punctures in paper.Discharges of low energy (D1) -typical examples are partial discharges of the sparking-type. or carbon particles in oil.
Discharges of high energy (D2) -typical examples are high energy arcing. tripping of the equipment or gas alarms . with power followthrough. resulting in extensive damage to paper.3.
. metal fusion. flashovers and short circuits. large formation of carbon particles in oil.
4. stray flux in beams
. -black or carbonized (> 300 °C). Typical examples are overloading. blocked oil ducts.Thermal faults of temperatures < 300 °C (T1) Faults T1 are evidenced by paper turning: -brown (> 200 °C).
Typical examples are defective contacts or welds.Thermal faults of temperatures between 300 and 700°C (T2) Faults T2 are evidenced by : -carbonization of paper.5. -formation of carbon particles in oil. circulating currents.
Thermal faults of temperatures > 700°C (T3) Faults T3 are evidenced by : -extensive formation of carbon particles in oil. -metal coloration (800 °C) or metal fusion (> 1000 °C). short circuits in laminations.
. Typical examples are large circulating currents in tank and core.6.
The first one was the Dornenburg method in Switzerland in the late 1960s.Several diagnosis methods have been proposed to identify these faults in service. Variations on these methods have later been proposed by the IEC (60599) and IEEE. then the Rogers method in UK in the mid 1970s.
because they fall outside the defined zones. C2H2/C2H4 and C2H6/C2H4). codes or zones are defined for each type of fault. Depending on the values of these gas ratios. One drawback of these methods is that no diagnosis can be given in a significant number of cases.All these methods use 3 basic gas ratios: (CH4/H2.
Another method used by IEEE is the so-called keygas method.g. e.
. One drawback of this method is that it often provides wrong diagnoses. C2H2 for arcing. which looks at the main gas formed for each fault.
C2H4 and C2H2) corresponding to the increasing energy levels of gas formation.
. with a low percentage of wrong diagnoses. and is based on the use of 3 gases (CH4.Finally. One advantage of this method is that it always provides a diagnosis. which was developed empirically in the early 1970s. there is the Triangle method.
% Unresolved diagnoses 0 33 % Wrong diagnoses 58 5 3 8 4 % Total 58 38 29 23 4
Key gases Rogers
Dornenburg 26 IEC Triangle 15 0
.Comparison of diagnosis methods.
so I will try to explain that in more detail today. However.The triangle representation also allows to easily follow graphically and visually the evolution of faults with time.
. many people are not quite familiar with the use of triangular coordinates.
The triangle method.
plus a DT zone (mixture of thermal and electrical faults).The triangle method plots the relative % of CH4. C2H4 and C2H2 on each side of the triangle.
. from 0% to 100%. The 6 main zones of faults are indicated in the triangle.
CH4 is formed in smaller amounts than H2 (typically 10 to 20 times less). can be identified in the Triangle without using this gas ? Answer: in such faults. which form a lot of H2.Question: how corona PDs. but it can still be measured easily by DGA.
A possible explanation (?): H2 diffuses much more rapidly than hydrocarbon gases from transformer oil. why not use H2 rather than CH4 to represent low energy faults ? Answer: because CH4 provides better overall diagnoses for all types of faults. This will affect gas ratios using H2 but not those using hydrocarbon gases.Another question: in the Triangle.
.So. how to use the triangle ? If for example the DGA lab results are: CH4 = 100 ppm C2H4 = 100 ppm C2H2 = 100 ppm First calculate: CH4 + C2H4 + C2H2 = 300 ppm.
.Then calculate the relative % of each gas: relative % of CH4 = 100 / 300 = 33. and should correspond to only one point in the triangle. To verify that the calculation was done correctly.3 % relative % of C2H4 = 100 / 300 = 33. the sum of these 3 values should always give 100%.3 % These values are the triangular coordinates to be used on each side of the triangle.3 % relative % of C2H4 = 100 / 300 = 33.
The zone in which the point falls in the Triangle will identify the fault responsible for the DGA results.Each DGA analysis received from the lab will always give only one point in the triangle.
so check it first with the free software available from duvalm@ireq. or with the help of a small algorithm or software. Errors are often made when developing such an algorithm.ca.
.The calculation of triangular coordinates can easily be done manually.
. Several commercial software are available for that purpose. Kelman or Delta-X Research in Canada. from Serveron. e.
.g. it is also possible to develop a software displaying the point and the fault zones graphically in the triangle.For those familiar with computer graphics.
. allows to easily follow the evolution of faults with time.. being a graphical method. for instance from a thermal fault to a potentially much more severe fault such as D2.
Fault zones in the triangle have been defined by using a large number of cases of faulty transformers in service which had been inspected visually.
Cases of faults PD and D1
Cases of faults D2
laminations . bad contacts
.Cases of thermal faults in oil only
circulating currents .
carbonized paper . not mentioned
.Cases of thermal faults in paper
brownish paper .
If the CO2 / CO ratio is < 3.A fault in paper is generally considered as more serious than a fault in oil only. either a hot spot or electrical arcing. this is a strong indication of a fault in paper. A popular ratio used for that purpose is the CO2 / CO ratio. because paper is often placed in a HV area (windings.
The amounts of furans in oil may also be used in some cases to confirm paper involvement. however. because it is also affected by the background of CO2 and CO coming from oil oxidation. the interpretation of results is often difficult. is not very accurate.The CO2 / CO ratio. however.
3 in breathing transformers). -O2/ N2: a decrease of this ratio indicates excessive heating (< 0. -C2H2/ H2 : a ratio > 3 in the main tank indicates contamination by the LTC compartment
.Other useful gas ratios: .
. -“Stray” gassing of oil: the “unexpected gassing of oil at relatively low temperatures (80 to 200 °C)”.
-Catalytic reactions on metal surfaces: formation of H2 only.Gassing not related to faults in service:
Oil Non-stray gassing Strongly stray gassing H2 3 1088 CH4 1 172 C2H4 11 C2H6 27 C2H2 CO 3 500 CO2 43 1880
.Stray gassing after 16hours of test at 120°C. in ppm :
. or under overloading conditions.It has been found at CIGRE that stray gassing:
.will not interfere with diagnoses during factory tests.
-may interfere with DGA diagnoses in service only in the case of the most stray gassing oils.
Now. Like everyone else they will sometimes make mistakes. Laboratory accuracy. DGA labs are not perfect. and some are not as accurate as we expect them to be.
. however. a critical look at DGA results coming from the laboratory. has a direct effect on diagnosis accuracy and on diagnosis uncertainty.
and ± 100% near the detection limit. following approximately the equation: ±15% ± 2 ppm (detection limit).
. Accuracy decreases rapidly as gas concentration decreases.The accuracy of the “average” lab has been found by CIGRE to be ± 15% at medium (routine) gas concentration levels (> 10 ppm for hydrocarbons). Accuracy will thus fall to ~ ± 30% at 6 ppm.
. respectively) on DGA diagnosis uncertainty.Effect of laboratory accuracy (±15% and ±30%.
When an area of uncertainty crosses several fault zones in the triangle.
. a reliable diagnosis cannot be given. Lab accuracies worse than 30% in general will provide unreliable or totally wrong diagnoses.
Diagnosis uncertainty corresponding to lab accuracies of ± 15. 30. 50 and 75 %:
Accuracy of laboratories at medium gas concentrations
Accuracy of laboratories at low gas concentrations
to be able to calculate the uncertainty on the diagnoses.Users should ask their DGA labs to indicate the accuracy of their DGA results.
To verify the accuracy of routine DGA analyses. users should also from time to time send to the lab a “blind” sample of gas-in-oil standard.
. e.Such gas-in-oil standards are now available commercially. following procedures or concepts described in IEC 60567 or ASTM D3612.. from Morgan Schaffer in Canada They can also be prepared by the laboratory.
are a waste of money since they cannot be used reliably. low or high. with possibly serious consequences for the equipment.Inaccurate DGA results. they may lead to wrong diagnoses. whatever their cost.
A similar investigation is presently underway at CIGRE TF15 to evaluate the accuracy of on-line and portable gas monitors.
. Low gas levels may be due to contamination or aging of insulation.Gas levels in service A recommendation of CIGRE and the IEC is that DGA diagnosis should be attempted only if gas concentrations or rates of gas increase in oil are high enough to be considered significant. not necessarily to an actual fault.
and it would not be economically viable to suspect all pieces of equipment.
. It is better to concentrate on the upper percentile of the transformer population with the highest gas levels.Also. there is always a small level of gases in service.
A lot of work has been done recently at CIGRE and the IEC in these areas.This is the philosophy behind the use of 90% typical concentrations and 90% typical rates of increase.
. in order to concentrate maintenance efforts on the 10% of the population most at risk. and a consensus reached on typical values observed in service worldwide.
Ranges of 90 % typical concentration values for power transformers. in ppm:
C2H2 All transformers No OLTC Communicating OLTC 2-20 60-280 H2 50150 CH4 30130 C2H4 60280 C2H6 2090 CO 400600 CO2 380014000
. in ppm/year:
C2H2 All transformers No OLTC Communicating OLTC 0-4 21-37 H2 35132 CH4 10120 C2H4 32146 C2H6 590 CO 2601060 CO2 170010.Ranges of 90 % typical rates of gas increase for power transformers.
etc. with some differences related to the individual loading conditions. equipment used.
.90% typical values are within the same range on all networks. weather.
-A bit higher in very old equipment.Influence of some parameters on typical values: -Typical values are significantly higher in young equipment (suggesting there are some unstable chemical bonds in new oil and paper ?). -Significantly lower in instrument transformers.
. -Not affected by oil volume (suggests that larger faults are formed in larger transformers ?). -Higher in shell-type and shunt reactors (operating at higher temperatures ?).
-the equipment should not be considered at risk.When DGA results are above typical values: -a diagnosis may be attempted to identify the fault producing these gases.
. the equipment should be monitored more frequently by DGA. -however.
for each gas. the cumulative number of analyses should be drawn as a function of concentration. To calculate typical concentration values.
.The typical values surveyed by CIGRE are ranges of values observed worldwide on a large number of networks. Each individual network should preferably calculate its own specific typical values.
Cumulative number of DGA analyses. in ppm
T = the 90% typical concentration value
. in % vs. gas concentration.
To evaluate how much at risk a transformer may become above typical values.As long as DGA values in service remain relatively close to typical values. there is no reason to be concerned by the condition of the transformer. etc). divided by the total number of analyses. fire. the probability of failure in service (PFS) has to be examined. fault gas alarm. PFS has been defined as the number of DGA analyses followed by a failure-related event (e.. tripping.g.
. at a given gas concentration.
% ) vs. in %
.Probability of having a failure-related event ( PFS. in %
99 Norm. the concentration of C2H2 in ppm
until it reaches an inflexion point on the curve (pre-failure value).The PFS remains almost constant below and above the 90% typical value. DGA monitoring should be done more and more frequently as gas concentrations increase from typical to pre-failure value.
700460 990 7501800 310600 9843000
This suggests that failure occurs when a critical amount of insulation is destroyed.
H2 2401320 CH4 C2H4 C2H6 C2H2 CO 270.Pre-failure values were found by CIGRE to be surprisingly close on different networks.
also on the maintenance budget available. higher alarm values may be used when the maintenance budget is low.
. and lower alarm values in the case of strategic equipment.In-between typical and pre-failure values. depending on the tolerance to risk of the maintenance personnel. specific alarm values can be defined. For example.
Summary of typical. alarm and pre-failure values: