Sie sind auf Seite 1von 7

F E A T U R E A R T I C L E

Calculation of DGA Limit Values


and Sampling Intervals in Transformers
in Service
Key Words: Transformer, dissolved gas analysis, gas limits, sampling intervals

Introduction Michel Duval


D issolved gas analysis (DGA) is widely used to detect
incipient faults in oil-filled electrical equipment. IEC
Standard 60599–1999 [1] and IEEE Standard C57.104–1991 [2]
IREQ, Varennes, QC, Canada

provide guidance for the interpretation of DGA results in service.


Several diagnosis methods, such as the Rogers and Duval Triangle Methods presented in this paper allow
methods [3], are available to identify the type of fault involved individual networks or transformer
(e.g., arcing, partial discharges, or hot spots), while the level of
gas formation is used to evaluate the severity of the fault. users to calculate their own specific
For that latter purpose, CIGRE [4] and the IEC [1] recommend DGA gas limits and required
the use of 90% typical values of gas concentration and rate of gas sampling intervals as a function of
increase. An amendment to the IEC Standard published in 2007
[5] provides the ranges of typical values observed worldwide. gas concentrations and rates of gas
However, individual networks are encouraged to calculate their increase in service.
own typical values, because these may vary depending on operat-
ing conditions (e.g., vs. load, climate, types of transformers used,
etc.). Above typical values, CIGRE and the IEC recommend that results. 90% typical values are cumulative percentile values, i.e.
gas limit values are based on the probability of failure of the 90% of DGA values observed in service in a given transformer
equipment in service. population are below the typical values, and 10% are above.
The IEEE Standard [2] recommends 4 levels of individual Below 90% typical values, where little or no gas formation
gas concentrations (“Conditions 1-4”) and 3 levels of rates of is observed, it is considered that transformers are operating nor-
increase for the total combustible gases. Recommendations for mally. Actually, there is always a small probability that a fault
oil sampling intervals and operating procedures are given based may occur without warning but it would be too costly to suspect
on total combustible gas values and rates of increase. The IEEE all transformers at these relatively low gas levels. Also, below
Transformer Committee is revising this standard with the inten- typical values, DGA results are less accurate and reliable, so it
tion of providing the user a methodology to calculate individual has been found reasonable to concentrate maintenance efforts on
gas concentrations and rates of increase levels for subsets of a the 10% of transformer populations with the highest gas levels
transformer population based on transformer type and operating and presumably most at risk.
conditions. Below typical values, the sampling intervals recommended by
The present paper proposes methods for calculating these the IEC are those used for regular maintenance of the equipment
individual gas limits and sampling intervals, based on the recom- (e.g., every year for transmission transformers or every 6 months
mendations and observations published by CIGRE and the IEC. for nuclear transformers).
The ranges of typical values (gas concentrations and rates
Typical Values of gas increase) observed by CIGRE [4] and the IEC [5] are
CIGRE and the IEC recommend the use of 90% typical values indicated in Tables 1 and 2. “Communicating OLTC” in Tables
as a starting point above which the frequency of sampling for DGA 1 and 2 means that in these transformers, some oil and/or gas
should be increased and more attention should be given to DGA communication is possible between the OLTC compartment and

September/October 2008 — Vol. 24, No. 5 0883-7554/07/$25/©2008IEEE 7


Table 1. 90% Typical Gas Concentration Values Observed by CIGRE and IEC for Power Transformers,
in ppm.

C2H2 H2 CH4 C2H4 C2H6 CO CO2

All transformers 50–150 30–130 60–280 20–90 400–600 3800–14000

No OLTC 2–20

Communicating
60–280
OLTC

the main tank, or between the respective conservators. Values Calculating Your Own 90% Typical Values
from a large number of networks worldwide are included in these Use your whole DGA data bank to evaluate your entire trans-
tables. They mean, for example, that hydrogen concentrations in former population, or separate from the data bank the analyses
90% of the transformers surveyed were below 150 ppm or 50 related to the segment of your transformer population that you
ppm depending on the network considered. Values on other indi- want to evaluate (e.g., power transformers without a communicat-
vidual networks may differ. Values in Table 2 should not be used ing OLTC). Transfer all DGA analyses related to the population
to re-calculate values in Table 1, because the cumulative curves chosen into a calculation program sheet (e.g., Excel).
used to calculate typical gas concentrations and typical rates of
gas increase are very different (for example, rates of gas increase A. Calculation of 90% Typical Gas
may be negative at times).
Relatively large ranges of values are observed in some cases, Concentration Values
reflecting the differences in operating practices and transformer Copy the concentrations of one gas into a separate column and
populations of different networks. It has been shown by CIGRE rank them by increasing order of values. Note the total number of
[4] that typical values and gas levels in service are: lines (n) in the calculation program (e.g., Excel) column. Multiply
• lower in younger equipment, suggesting that there are some n by 90% (= 0.9n). The gas concentration corresponding to line
weak chemical bonds in new oil and paper which are broken 0.9n in the Excel column is the 90% typical concentration for this
in the first years of operation and then stabilize. They are gas. Repeat the same procedure for all the other gases.
a bit higher in very old equipment.
• higher in shell-type power transformers and shunt reac- B. Calculation of 90% Rates of Gas Increase
tors than in core-type power transformers, and lower in Time intervals between 2 dates of analysis indicated as YY/
instrument transformers, probably because these types MM/DD in the data bank can be expressed in days with the cal-
of equipment operate at higher and lower temperatures, culation program (e.g., Excel). Divide these time intervals by 365
respectively. to get them in fractions of years. For each individual transformer
• not affected significantly by oil volume, suggesting that and each gas, calculate the difference between 2 successive DGA
larger faults are formed in larger equipment. analyses, in terms of gas concentrations (in ppm) and time inter-
• relatively similar in air-breathing and in sealed or nitrogen- vals (in fractions of year). The ratio of these 2 terms gives the rate
blanketed equipment. of gas increase (RGI) in ppm/year. Extend this calculation of RGI
Because of these differences, typical values should preferably values to all gases, DGA analyses, and transformers.
be calculated by each individual network, on its whole population Copy the RGI values of each gas into a separate calculation
of transformers or on more specific segments of it if desired (e.g., program (e.g., Excel) column and rank them by increasing order of
with or without a communicating OLTC, sealed or air-breathing, values. Note: using this method, the first analysis performed on a
core- or shell-type, GSU or transmission, geographic location or transformer, and the transformers for which only one analysis was
transformer manufacturer, with fault involving paper or not, etc.). performed, provide values divided by zero which are rejected at
The only limitation of such a segmentation is the number of DGA the end of the calculation program (e.g Excel) column and should
analyses available in each segment, which should be large enough
to be representative (e.g., > 100).

Table 2. 90% Typical Rates of Gas Increase Observed by CIGRE and IEC for Power Transformers, in ppm/year.

C2H2 H2 CH4 C2H4 C2H6 CO CO2

All transformers 35–132 10–120 32–146 5–90 260–1060 1700–10,000

No OLTC 0–4

Communicating OLTC 21–37

8 IEEE Electrical Insulation Magazine


typical values. For DGA data banks where such information on
failure-related events is available, the PFS for each gas can be
obtained by calculating the number of DGA analyses followed by
a failure-related event, in a given concentration range, then divid-
ing this number by the total number of DGA analyses in the data
bank in the same concentration range [6]. An example of such a
calculation is shown in Figure 1, where the PFS is expressed as a
function of the concentration of acetylene in power transformers
of Hydro Quebec. On the upper horizontal axis are indicated the
corresponding 90% and 99% typical values.
It can be seen that the PFS remains constant at ~10% as con-
Figure 1. Probability of having a failure-related event in centration increases above the 90% typical value (~3 ppm in this
service (PFS) in %, vs. the concentration of acetylene in ppm. T example), until an inflexion point is reached on the curve, which
= 90% typical value; P = pre-failure value. has been called the pre-failure value (~350 ppm in this example),
above which the PFS increases dramatically.
The PFS vs. gas concentration of the other gases is indicated
be deleted. Negative values should NOT be deleted. in Figure 2, and vs. the rate of gas increase in Figure 3, as well
Note the total number of remaining lines (n) in the Excel col- as the corresponding pre-failure values. Pre-failure concentration
umn. Multiply n by 90% (= 0.9n). The value corresponding to line values were found by CIGRE [4] to be surprisingly similar on
0.9 n in the Excel column is the 90% typical rate of gas increase different networks, suggesting that failure occurs when a criti-
for this gas. Repeat the same procedure for the other gases. cal amount of insulation is destroyed. CIGRE pre-failure values
can therefore be used as a reasonable estimate of the levels of
Pre-Failure Values gas formation observed just before failure. They are indicated
Calculating the probability of having a failure-related event in Table 3 for gas concentrations (in ppm) and for rates of gas
in service (PFS), such as fault gas alarm, tripping, failure, fire increase (in ppm/day).
or explosion, has been recommended by CIGRE [4] and the It may be noted from Figures 2 and 3 that the PFS does not
IEC [1] to determine how much at risk a transformer is above significantly increase even at the highest concentrations of CO2

Figure 2. Probability of having a failure-related event in service (PFS) in %, vs. the concentration of all gases in ppm. T = 90%
typical value; P = pre-failure value.

September/October 2008 — Vol. 24, No. 5 9


Table 3: Average Pre-Failure Values Observed at CIGRE.

H2 CH4 C2H4 C2H6 C2H2 CO CO2

Gas concentrations ppm 725 400 800 900 450 2100 N/R

Rates of gas increase ppm/ day 3 5 5 11 0.5 N/R N/R

N/R = Pre-failure value never reached in service.

and at the highest rates of increase of CO and CO2 observed in using on-line gas monitors, in the event that the equipment has
service. This means that the corresponding pre-failure values are not been removed from service yet (for instance, in cases where
never reached in practice and that very high levels of formation heavy gassing is occurring without prior warning).
of these two gases can generally be tolerated in service without The sampling intervals required between typical and pre-failure
increasing the risks to the equipment. Because CO and CO2 are values can easily be calculated as illustrated in Figure 4, using
related to solid insulation or oil oxidation aging, this could also as an example the average typical concentration values observed
mean that insulation aging may contribute much less to failure in the US by Weidmann in California (reported in [4]) and GE-
events than electrical or thermal faults. Energy [7], the pre-failure concentration values of Table 3, and a
sampling interval of 1 year for regular maintenance.
Calculating Your Own Sampling Intervals Gas concentrations in ppm are indicated on the logarithmic
for DGA vertical axis of Figure 4 (pre-failure values P and typical values T
Typical sampling intervals are those used for regular main- for each gas are indicated on the left and right axis, respectively).
tenance of the transformers (e.g., every year for transmission Sampling intervals in hours are indicated on the logarithmic hori-
transformers and every 6 months for nuclear transformers in the zontal axis (1 hour for pre-failure values, 1 year or 8760 hours
US, or your own intervals). for typical values, and in-between values expressed in day, week
For the purpose of calculations in this paper, the sampling in- and month at 24, 168, and 720 hours, respectively).
terval corresponding to pre-failure values has been chosen as one Concentrations requiring monthly, weekly, and daily sampling
hour. This is the shortest interval that is feasible in service when intervals can be obtained from the curves drawn between typical
and pre-failure concentrations values for each gas in Figure 4. For

Figure 3. Probability of having a failure-related event in service (PFS) in %, vs. the rate of increase of all gases in ppm/ year. T =
90% typical value; P = pre-failure value.

10 IEEE Electrical Insulation Magazine


and the curve for CO.
The values obtained when using the average typical concentra-
tions observed by CIGRE/ IEC in Table 1 are indicated in Table 5
(the average of actual values observed has been used rather than
the average of the range).
Curves similar to those shown in Figure 4 can be obtained
by using rates of gas increase rather than concentrations on the
vertical axis. The values obtained when using the average rates
of gas increase observed by CIGRE/ IEC in Table 2 are indicated
in Table 6 (the average of actual values observed has been used
rather than the average of the range).
An example of calculation on a segmented transformer popula-
tion of Hydro Quebec is illustrated in Table 7. Only the ~10% of
all DGA results where paper was involved in the fault (as indicated
by CO2/CO ratios <3 [1]) were kept in this example. It appears
that pre-failure and other limit values are lower when paper is
involved, particularly for C2H4 and CO2.
For more precision, values in Tables 4–7 have been calculated
using the following equation, which mathematically represents the
curves shown for each gas in Figure 4 and in similar figures:
Figure 4. Determination of sampling intervals as a function
of gas concentration in service for an average US power
log G = –a log t + log b, (1)
transformer.
where G = the gas concentration or rate of gas increase of a
example for C2H2, typical and pre-failure concentration values given gas in service, and t = the corresponding sampling interval
are 2.5 and 450 ppm, and concentrations requiring monthly, required (expressed in hours).
weekly and daily sampling intervals are 11, 29, and 97 ppm, The values of constants a and b for each gas in equation (1)
respectively. can be obtained by using numerical values of G and t correspond-
The numerical concentration values for the other gases are ing to the transformer population (or segment of population)
indicated in Table 4. Also indicated in Table 4 are the concentra- considered:
tion values obtained when using a regular sampling interval of 6 GT = the typical value T for this transformer population
months (instead of 1 year) for transformers in nuclear plants. and this gas (typical concentration or typical rate of
In Table 4, gas concentrations requiring monthly, weekly and increase), and tT = the sampling interval for regular
daily sampling intervals have been called levels 2, 3, 4. They had maintenance (e.g., 1 year or 6 months, expressed in
been called conditions 2, 3, 4 in the IEEE Standard of 1991 [2], hours).
and caution, warning, and alarm levels in its Draft of revision GP = the pre-failure value P for this gas (concentration or
D11d-2004 (dropped). The curve and pre-failure value for CO2 rate of gas increase) indicated in Table 3, and tP = 1
in Figure 4 have been extrapolated from the typical value of CO2 hour.

Table 4: Examples of Sampling Intervals and Gas Concentration Limits in ppm Calculated for an Average US Power Transformer.

Sampling intervals

Concentration H2 CH4 C2H4 C2H6 C2H2 CO CO2 TDCG Nuclear Transmission

Typical 88 69 65 53 2.5 780 5990 1058 6 months Yearly

Level 2 143 104 117 102 8 983 9829 1546 Monthly

164 116 137 123 11 1048 11274 1717 Monthly

Level 3 214 145 188 175 22 1186 14695 2104 Weekly

237 158 212 201 29 1245 16294 2277 Weekly

Level 4 366 226 355 360 84 1524 25161 3178 Daily

388 237 380 388 97 1566 26661 3322 Daily

Pre-failure 725 400 800 900 450 2100 50000 5380 Hourly Hourly

September/October 2008 — Vol. 24, No. 5 11


Table 5: Examples of Sampling Intervals and Gas Concentration Limits in ppm Calculated for an Average CIGRE/ IEC Power Transformer.

Sampling intervals

Concentration H2 CH4 C2H4 C2H6 C2H2 CO CO2 TDCG Nuclear Transmission

Typical 100 80 170 55 3 500 8900 908 6 months Yearly

Level 2 158 116 244 105 10 699 13314 1375 Monthly

180 129 270 126 13 766 14885 1542 Monthly

Level 3 231 158 327 179 25 917 18466 1926 Weekly

254 170 352 205 32 983 20084 2101 Weekly

Level 4 381 237 484 364 89 1319 28600 3024 Daily

403 248 505 393 102 1372 29980 3175 Daily

Pre-failure 725 400 800 900 450 2100 50000 5380 Hourly Hourly

The gas concentration or rate of gas increase (Gx) in service values GT of their own transformer population or segment of
requiring a given sampling interval tX (e.g., monthly, weekly, or population (as calculated in this paper), their regular sampling
daily, expressed in hours), can then be calculated using equation intervals tT, the pre-failure values of Table 3, and equation (1). Free
(1). software is available for that purpose from duvalm@ireq.ca.
The sampling intervals indicated in Tables 4–7 are required if Gas levels and sampling intervals thus calculated for a given
at least one (not necessarily all) of the corresponding gas concen- transformer population apply to most transformers in this popula-
trations or rates of gas increase have been reached. tion. However, the values may be different for some segments of
If the sampling intervals required by gas concentrations and the population (e.g., families of transformers of a specific type
rates of gas increase are different (e.g., weekly for concentrations or design built by some manufacturers during a period of years).
and yearly for rates), an intermediate interval (e.g., monthly) may Engineering judgment should be exercised to detect such excep-
be chosen as a compromise. Rates of gas increase are generally tions, and more specific sampling intervals calculated, if possible,
considered as more important than gas concentrations, because for those segments of the population.
they indicate that the fault is active. However, high backgrounds In Eq. (1), typical values based on a different percentile value
of gas concentrations, even if the fault has become inactive, may (e.g., 85% or 95% rather than 90%) may be chosen if considered
indicate permanent damage done to the insulation. more appropriate for regular maintenance (depending on the
Tables 4–7 may be used by default. However, typical values maintenance budget available and on the strategic importance
and, therefore, gas levels 2, 3, 4 and corresponding sampling of transformers). The 90% value, however, is the most widely
intervals may be different for individual networks; so it is recom- used worldwide. Typical or “normal” values based on statistical
mended that individual networks and transformer users calculate methods other than percentile values may also be used.
their own sampling intervals in service, using the 90% typical

Table 6: Examples of Sampling Intervals and Rate of Gas Increase Limits in ppm/year Calculated for an Average CIGRE/ IEC Power Transformer.

Sampling intervals
Gassing
rate H2 CH4 C2H4 C2H6 C2H2 CO CO2 TDCG Nuclear Transmission

Typical 83 65 89 47 2 660 5850 946 6 months Yearly

Level 2 151 141 180 132 6 1408 12472 2049 Monthly

179 175 218 176 7 1737 15382 2539 Monthly

Level 3 247 266 319 308 13 2607 23067 3841 Weekly

280 313 369 382 17 3054 27012 4513 Weekly

Level 4 475 620 686 951 42 5940 52492 8896 Daily

509 679 745 1074 47 6491 57351 9738 Daily

Pre-failure 1095 1825 1825 4015 182 17000 150000 26000 Hourly Hourly

12 IEEE Electrical Insulation Magazine


Table 7: Examples of Sampling Intervals and Gas Concentration Limits in ppm Calculated for Power Transformers where Paper is Involved in the Fault.

Sampling intervals

Concentration H2 CH4 C2H4 C2H6 C2H2 CO CO2 TDCG Nuclear Transmission

Typical 70 50 60 40 5 950 3000 1175 6 months Yearly

Level 2 115 69 74 68 12 1003 3208 1432 Monthly

132 75 79 79 15 1018 3268 1513 Monthly

Level 3 173 89 88 105 24 1049 3388 1683 Weekly

193 96 92 118 28 1060 3435 1754 Weekly

Level 4 299 128 111 190 60 1112 3644 2088 Daily

317 132 114 202 67 1120 3673 2137 Daily

Pre-failure 600 200 150 400 200 1200 4000 2750 Hourly Hourly

Conclusions [3] M. Duval, “A review of faults detectable by gas-in-oil analysis in


Methods are presented in this paper allowing individual net- transformers,” IEEE Electr. Insul. Mag., vol. 18, no. 3, pp. 8–17,
2002.
works or transformer users to calculate their own specific DGA
[4] M. Duval et al., Joint Task Force D1.01/ A2.11 of CIGRE, “Recent
gas limits and required sampling intervals as a function of gas developments in DGA interpretation,” CIGRE Brochure # 296,
concentrations and rates of gas increase in service. This calcula- 2006, available from publications@cigre.org. Summary in Electra,
tion can be done on the entire transformer population or on more no. 226, pp. 56–60, June 2006.
specific segments of it. [5] Mineral oil-impregnated electrical equipment in service - guide to the
The purpose of this paper is only to provide a tool for calcu- interpretation of dissolved and free gases analysis, IEC Publication
lating gas levels requiring more frequent sampling intervals for 60599-Amendment 1 (2007–04), April 2007.
DGA. The actions to be taken on the equipment at any of these gas [6] M. Duval et al., “Dissolved gas analysis: it can save your transformer,”
levels and sampling intervals (e.g., removal or not from service, IEEE Electr. Insul. Mag., vol. 5, no. 6, pp. 22–27, 1989.
[7] C. Beauchemin, “Evaluation of proposed gas-in-oil limits of Table
additional testing, installation of on-line gas monitors) are outside
1 and section 7.3 of C57.104 Draft D11d”, presented to the IEEE
the scope of this paper. These actions depend on a large number Transformer Committee, 2006.
of other parameters (e.g., experience and maintenance practices
of individual users, strategic importance and type of equipment
used, type and location of the fault, fault active or not). Michel Duval graduated from the
The actual sampling intervals to be used also remain the deci- University of Toulouse in 1966 with a
sion of maintenance personnel, based on their best operational B.Sc. degree in chemical engineering,
practices. and received a Ph.D. degree in polymer
chemistry from the University of Paris in
Acknowledgments 1970. Since then, he has worked for Hydro
The contribution of B. Noirhomme (IREQ) to the calculations Quebec’s Institute of Research (IREQ)
made in this paper is gratefully acknowledged. on electrical insulating oils, dissolved gas
analysis, and lithium polymer batteries. A
References Fellow member of IEEE and the Chemical Institute of Canada,
[1] Mineral oil-impregnated electrical equipment in service - guide to the he holds 15 patents; has authored over 70 scientific papers, book
interpretation of dissolved and free gases analysis, IEC Publication chapters, or international standards; and is very active in several
60599 (1999–03), March 1999. CIGRE, IEC, and IEEE working groups. He may be reached at
[2] Guide for the Interpretation of Gases Generated in Oil-Immersed
IREQ, 1800 boul. Lionel Boulet, Varennes, J3X 1S1, Canada, or
Transformers, IEEE Std C57.104–1991, Institute for Electrical and
Electronics Engineers, 1991.
at duvalm@ireq.ca.

September/October 2008 — Vol. 24, No. 5 13

Das könnte Ihnen auch gefallen