DGA Monitoring Systems

D1/A2
Technical Brochure
DGA monitoring systems

Reference: 783
October 2019
WG D1/A2.47
Members
M.DUVAL, Convenor CA I.BOCSI, Secretary HU

E. ALZIEU FR O. AMIROUCHE IT
C. BEAUCHEMIN CA S. BHUMIWAT NZ
P. BOMAN US S. BRAUER CA
T. BUCHACZ PL M. CYR CA
A.M. DALE NO K. DEVILLÉ BE
S. EECKOUDT BE R. FROTSCHER DE
M. GRISARU IL J. HINSHAW US
H.J. KNAB CH S. LEIVO FI
M.A. MARTINS PT C. MICHELLON FR
X. MINNITTI FR B. NEMETH HU
C. PERRIER FR F. SCATIGGIO IT
J. SCHUEBEL DE M.SZEBENI HU
Copyright © 2019
“All rights to this Technical Brochure are retained by CIGRE. It is strictly prohibited to reproduce or provide this publication in any
form or by any means to any third party. Only CIGRE Collective Members companies are allowed to store their copy on their
internal intranet or other company network provided access is restricted to their own employees. No part of this publication may
be reproduced or utilized without permission from CIGRE”.
Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any responsibility, as to the
accuracy or exhaustiveness of the information. All implied warranties and conditions are excluded to the maximum extent permitted
by law”.
WG XX.XXpany network provided access is restricted to their own employees. No part of this publication may be
reproduced or utilized without permission from CIGRE”.
ISBN : 978-2-85873-485-6
Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any
ISBN : 978-2-85873-485-6
Executive summary
This Technical Brochure is a continuation of the work done on the accuracy of on-line gas monitors by
CIGRE TF D1.01.15 (published in 2010 as TB 409)[B4], and on the accuracy of DGA results using the
head space principle by CIGRE TF 15-01-07 (published in 2003 at IEEE) and CIGRE WG D1.32 (published
in 2010 as TB 443)[B9].
This Technical Brochure describes:

 the different types of on-line gas monitors available on the market.
 technologies on which they are based.
 number of gases each of them can detect and measure.
 faults in transformers each of them can identify.
 pros and cons of single and multi-gas monitors.
 comparative costs and benefits of DGA monitoring by laboratory analysis and on-line gas
monitors.
 faults in on-load tap-changers that can be identified by each type of on-line gas monitor.
 where to install on-line gas monitors on transformers and on-load tap-changers.
 the effect of poor accuracy of DGA results coming from on-line gas monitors and laboratories
on DGA diagnosis, gassing trends and actions on electrical equipment.
 accuracy of the on-line monitors on the market evaluated by WG D1/A2.47.
 improved procedure for evaluating the accuracy of on-line gas monitors and laboratories, using
gas-in-oil standards (in Appendixes).
 on-line gas monitors using infrared and gas chromatographic detectors (also in Appendixes).
5
Contents
Executive summary ............................................................................................................. 5
1. Introduction................................................................................................................ 9
2. Types of on-line gas monitors on the market ........................................................ 11

2.1 Types of on-line gas monitors versus technology used ................................................................. 11
2.2 Types of on-line gas monitors versus faults possible to identify in transformers ....................... 12
2.3 Detection of hydrogen with types of monitors M1, M1* and M2 in transformers .......................... 13
2.4 Detection of carbon monoxide with monitors of type M2 in transformers .................................... 14
2.5 Fault identification in transformers with on-line gas monitors ...................................................... 15
2.6 Detection of fast-occurring faults in transformers with on-line gas monitors .............................. 15
2.7 Costs and benefits of various DGA monitoring techniques for transformers .............................. 16
3. On-line gas monitors for on-load tap-changers .................................................... 19
4. Installation of on-line gas monitors........................................................................ 21

4.1 Installation on transformers .............................................................................................................. 21
4.2 Installation on on-load tap-changers9............................................................................................... 21
5. Accuracy of DGA results coming from on-line gas monitors and laboratories .. 23
5.1 Definitions ........................................................................................................................................... 23
5.2 Accuracy of DGA results coming from transformers ...................................................................... 23
5.3 effect of poor accuracy on DGA diagnosis ...................................................................................... 23
5.4 effect of poor accuracy of monitor readings on gassing trends .................................................... 24
6. Evaluation of accuracy of on-line monitors ........................................................... 25
7. Conclusions ............................................................................................................. 29
APPENDIX A. definitions, abreviations and symbols ..................................................... 31
APPENDIX B. links and references .................................................................................. 31
APPENDIX C. DGA on-line gas monitors using infrared gas measurement methods .. 33
APPENDIX D. DGA monitors using Gas Chromatography (GC) measurement methods .

.................................................................................................................................. 37
APPENDIX E. Gas-in-oil standards .................................................................................. 41
APPENDIX F. Procedure of WG D1/A2.47 for evaluating the accuracy of gas monitors

and laboratory results ....................................................................................................... 45
6
APPENDIX G. Examples of monitor accuracy calculations6,14 ....................................... 49
APPENDIX H. Diffusion of helium in oil ........................................................................... 51
Figures and Illustrations

Figure 2.1 Sudden Change in Gas Formation and Type of Fault ...................................................................... 15
Figure 4.1 Oil Sampling Loop for On-Line Gas Monitors ................................................................................. 22
Figure 5.1 Effect of Poor Accuracy on Uncertainty of Diagnosis of Faults in Transformers [B14, B15] ................. 24
Figure 5.2 Effect of Poor Accuracy of Monitor Readings on Gassing Trends [B14] ............................................ 24
Figure 6.1 Accuracy of monitors units tested by CIGRE TF15 and WG47 ......................................................... 26
App Figure C.1 Schematic picture on IR light absorption by molecules shifting to excited state. ........................ 33
App Figure C.2 Schematic picture of optical module. ..................................................................................... 33
App Figure C.3 Mid-IR light absorption of CO2, CO, C2H2, C2H4, C2H6 and CH4 .................................................. 34
App Figure D.1 A gas chromatography (GC) measurement system for DGA ..................................................... 37
Tables
Table 2.1 General Types of On-Line Gas Monitors Available on the Market [B3] ............................................... 11
Table 2.2 Definition of the Types and Sub-Types of Faults or Stresses ............................................................ 12
Table 2.3 Faults Possible to Identify in Transformers with On-Line Gas Monitors [B4] ...................................... 12
Table 2.4 Monitors Suitable for Different Types of Transformers [B2, B4] ........................................................ 14
Table 2.5 Costs/Benefits of Various DGA Monitoring Techniques..................................................................... 17
Table 3.1 On-Load Tap-Changers Faults Possible to Identify with On-Line Gas Monitors 8 ................................. 19
Table 6.1 List of On-Line Gas Monitors Evaluated by WG D1/A2.47 ................................................................ 25
Table 6.2 Accuracy of On-Line Gas Monitors Tested by Members of WG D1/A2.47 ........................................... 26
App Table E.1 Accuracy of Laboratories Using Gas Extraction Methods ........................................................... 41
App Table G.1 Example 1 of Accuracy Calculations Performed for Test No 1 of Table 6.26. ............................... 49
App Table G.2 Example 2 of Accuracy Calculations Performed for Test No 1 of Table 6.214 ............................... 50
7
8
1. Introduction
The scope (terms-of-reference) of Working Group D1/A2.47, as approved by Study Committees D1 and
A2 in June 2011, was “to investigate the following items, all related to transformers and their
accessories:
 case studies and suggestions for typical problems and actions initiated by DGA interpretation.
 significance of CO, CO2 and CO2/CO for detecting paper involvement in faults.
 review progress in the field of DGA”.
Work on the first two items is reported in another TB of the WG [B1].
Particular items concerning progress in the field of DGA mentioned by former members of Working
Group D1.32 and members of Working Group D1/A2.47 were:
 update on on-line gas monitors on the market since 2010.
 accuracy of on-line monitors as compared to laboratory results.
In CIGRE Technical Brochure 409 (2010) [B2], on-line gas monitors on the market up to 2008 were
evaluated for accuracy and reliability by CIGRE TF D1.01.15. These included EMH’s Hydrocal, Energy
Support’s Mobile GC, Gatron’s TGM, GE’s Transfix, Transport-X and Hydran M2, Morgan Schaffer’s
Myrkos and Calisto 2, Serveron’s TM8 and Unisensor’s E200.
Since then, a large number of new monitors have appeared on the market, particularly hydrogen
monitors using solid state technology, but also multiple gas monitors measuring from 2 to 11 gases
depending on manufacturers and models. Among the tasks of Working Group D1/A2.47 was to evaluate
their accuracy and reliability for the benefit of users and manufacturers alike. Some of the previous
models were plagued with frequent recalibration problems and contamination from interfering chemical
species in the oil and from ambient atmosphere, particularly those of the infrared type.
Indeed, large differences have been reported by1, 2, 3 between their gas monitors and laboratory results,
hence the interest in accuracy measurement being evaluated independently by CIGRE 4.
This document is strictly informational4 and not intended to steer the end user to any specific
technology, but rather provide insight into various conditions of testing, which may be useful when
selecting the appropriate on-line monitor for a specific application. It is up to the end user to explore
fully their options and make an informed decision, based on their specific needs.
Superscripts in this document refer to unpublished contributions by individual members indicated in the
Acknowledgments section of 0.
9
10
2. Types of on-line gas monitors on the market

2.1 Types of on-line gas monitors versus technology used
Technologies used in the various types of on-line gas monitors on the market for electrical equipment
are summarized in Table 2.1. [B3].
Table 2.1 General Types of On-Line Gas Monitors Available on the Market [B3]
Extraction Gas measurement Advantages Limitations

Method technology
Membrane, Gas Very similar to standardized -Carrier gas (He) and calibration gas
Vacuum Chromatography techniques (IEC, ASTM) need to be replaced ~every2-4 years
Separation of signals to avoid - Depending on models, GC columns
interferences must be replaced every 3 to 5 years
Automated recalibration with and there is a potential for carrier gas
on-board calibration gas to leak into transformer4
-Requires management of
compressed gas cylinders
Membrane, Infra Red No consumable gases -Other sensors required to measure
Vacuum, (direct absorption No compressed gas cylinders H2, O2, N2
Direct or photo-acoustic) -Some models sensitive to
headspace contamination by oil vapours, leading
to inaccuracy over time and the need
for recalibration
- In some models, accuracy is
degraded by interfering compounds
present in the oil and/or the ambient
air
Membrane, Thermal -H2 and CO only
Direct conductivity cell
contact Electrochemical cell No consumable gases -Composite gas signal
with oil Metal-oxide -H2 and CO only, limited accuracy
sensors
Metal film sensors -H2 only
Monitors using membranes for gas extraction from the oil are based on the head space principle. In all
monitors, partition coefficients are used, which are dependent on temperature and should be
compensated for internally in terms of both electronic signals and partition coefficients used. This is
usually done by manufacturers of the monitors, so monitor users do not have to perform these
adjustments.
Partition coefficients are also dependent on other factors such as type and aging of oil, which often are
not compensated for by manufacturers (except for some ester oils). This in some cases may explain
differences with laboratory results.
More and more on-line and portable gas monitors of the infrared-type are or will be available on the
market, e.g., CoreSense 8, HydroCal1008, LumaSense 9, Optimus, Totus, TM8-F, Transfix, Transport-
X, etc[B3].
Guidelines5,6,7 for the use of infrared detectors and monitors for DGA are given in APPENDIX C. These
guidelines will be examined by Maintenance Team MT25 of IEC Technical Committee TC10 and proposed
as an Appendixto IEC 60567 (presently containing only guidelines for GC detectors).
Guidelines for the use of GC detectors and monitors are also given in APPENDIX D5,6,7.
11
2.2 Types of on-line gas monitors versus faults possible to identify in

transformers
Table 2.2indicates thesix basic types of faults (PD, D1, D2, T3, T2 T1) detectable with the IEC 60599,
Rogers, Duval Triangle 1 and Pentagon 1 and other methods and thefive sub-types of faults (T3-H, C,
O, S, PD) detectable only with Duval Triangles 4, 5 and Pentagon 2, which are possible to identify in
transformers[B1][B21]with the different types of on-line gas monitors.
Table 2.2 Definition of the Types and Sub-Types of Faults or Stresses

Identifiable by DGA [B1]
Fault or Stress Definition

T3 Thermal T > 700 °C
T2 Thermal 300 < T < 700°C
T1 Thermal < 300 °C
PD Partial Discharges of the Corona-Type
D1 Discharges of Low Energy
D2 Discharges of High Energy
S Stray Gassing of Oil < 200 °C
O Overheating < 250 °C
C Possible Carbonization of Paper > 300 °C
T3-H T3 in Oil Only
Table 2.3indicates the faults possible to identify in transformers with the different types of on-line gas
monitors, including the six basic types of faults (PD, D1, D2, T1, T2, T3) and the five sub-types of faults
(T3-H, C, O, S, PD).
Table 2.3 Faults Possible to Identify in Transformers with On-Line Gas Monitors [B4]
Application Type of Gases Faults Possible to Identify Faults not Possible to

monitor Measured Identify
M8/(M9) H2, CH4, C2H6, -none
C2H4, C2H2,
CO, CO2, O2, (N2) -all 10 faults in Table 2.2,
Fault M6/(M7) H2, CH4, C2H6, at an early stage -faults in paper very often
Diagnostic C2H4, C2H2, CO, are not detected correctly
(CO2) with CO only with M6, M5
and M2
M5 H2, CH4, C2H4,
C2H2, CO -the 6 basic faults only -the 5 sub-types of faults.
M3 CH4, C2H4, C2H2
M2 H2, CO -none of the 10 faults can -may not detect faults D1,
Fault M1 H2 be identified. D2 in their early stages,
Detection M1* Composite only in their late,
reading of H2 and sometimes catastrophic
other gases stages
In Table 2.3, several of the monitors can also measure moisture in oil, but this is outside the scope of
this TB.
When inaccurate readings of an individual monitor unit are suspected, its accuracy may be verified
using the procedures described in APPENDIX F.
12
2.3 Detection of hydrogen with types of monitors M1, M1* and M2 in

transformers
Monitor types M1, M1* and M2 can measure the concentrations of hydrogen in transformers and detect
increases in hydrogen concentrations with time (trends of hydrogen).
A challenge with these monitors is the choice of threshold limits of hydrogen concentrations and rates
(trends). Since they cannot identify the fault responsible for hydrogen formation, the limit used by
default in them is the typical concentration value of hydrogen (around 100 ppm in [B1]) and/or the
typical rate of hydrogen increase (around 20 ppm/year, also in [B1].
The detection with these monitors of electrical and thermal faults in their early stages is thus often
difficult. For instance, in case of an arcing fault D1 or D2 in windings, which is potentially the most
dangerous fault in transformers, the IEC/CIGRE typical value (or condition 1 value of IEEE) of acetylene
that should be detected is ~ 2 ppm. Only about 6 ppm of hydrogen will be formed together with these
2 ppm of acetylene [B1]. So in order to detect arcing in its early stage, a warning signal should be sent
to the operator of the hydrogen monitor when there is a rapid increase of hydrogen of 6 ppm, asking
for a sample of oil to be sent to the laboratory for a full DGA analysis.
However, an increase of 6 ppm cannot be detected accurately and reliably with monitor types M1, M2
using metal film or metal oxide sensors or electrochemical cells, because of their relatively high detection
limits (> 25 ppm)6.
Also, this 6 ppm of hydrogen may be due to a more benign condition such as stray gassing of oil, the
typical value of which is in the 400 ppm range [B1], so the monitor will send a false alarm. Such false
alarms due to stray gassing will be much more frequent than real alarms of arcing, which occurs only
in a very small number of cases in transformers. So after repeated such false alarms, the operator may
come to disregard them. And when a real alarm of arcing in its early stages occurs, it may be missed
by the operator.
Arcing may be picked up only in its late (condition 3 or 4) values, when larger spikes of acetylene and
hydrogen are formed over short periods of time, sometimes too late to avoid a catastrophic failure if
arcing is in windings [B4].
In the case of thermal faults T3, T2 and T1, much less hydrogen is formed than of the other hydrocarbon
gases[B1]. One such example can be found in[B5]. If the hydrogen warning signal is set by default at
100 ppm or even 40 ppm. These faults will be picked up only in their late stages [B1]. And if it is set at
10 ppm to catch the faults in their early stages, it will send lots of false alarms (e.g., for faults S). as in
the case of arcing faults, and will lose its credibility with the operators.
Multi-gas monitors are therefore more reliable for the detection of arcing in its early stages (with
acetylene around 2 ppm), and for the detection of thermal faults in their intermediate stages. However,
monitor types M1 and M2 are efficient for identifying hydrogen formations above the typical
concentration and rate values of CIGRE, IEEE, or established in-house in relation to monitor readings
from preceding periods of months or years.
M1, M1* and M2 monitors, even if they catch faults later in their progression, can be installed on a
larger number of transformers because of their lower cost, however, increasing the likelihood of
detecting faults across the transformer fleet. And they tend to be simpler. There is a wide range of M1
and M2 monitors available, and those with high sensitivity and stability are more likely to detect
successfully hydrogen contributions from thermal and discharge faults.
13
The recommendation in CIGRE TB 409 [B2][B4] concerning the use of monitor types M1,M1* and M2
and the other multi-gas monitors is indicated in Table 2.4. For monitor types M1, M1* and M2, a possible
compromise could be to replace them by multi-gas monitors in gassing or critical transformers, when
hydrogen in them increases above typical values of hydrogen, hoping that these increases were not due
to arcing or thermal faults.
Table 2.4 Monitors Suitable for Different Types of Transformers[B2], [B4]
Type of monitor Gases Transformers

M7/M8/M9 H2, CH4, C2H6, C2H4, C2H2, -all transformers, including critical ones
CO, CO2, (O2), (N2) (GSU, nuclear) and those already gassing
abnormally, requiring immediate action
M6 H2, CH4, C2H6, C2H4, C2H2, -critical and gassing transformers not
CO requiring immediate action (oil sample
M5 H2, CH4, C2H4, C2H2, CO required to confirm paper involvement).
M3 CH4, C2H4, C2H2
M2 H2, CO -not critical transformers and not already
M1 H2 gassing transformers (oil sample required).
M1* Composite reading of H2
and other gases
2.4 Detection of carbon monoxide with monitors of type M2 in

transformers
These monitors can measure the concentrations of carbon monoxide in transformers and detect
increases in carbon monoxide with time.
Here too, the main challenge is the interpretation of carbon monoxide values. For instance, in the case
of sealed transformers, it is increasingly common to observe very high concentrations of carbon
monoxide in oil (e.g., 1000 ppm), which are not due to a fault in the paper but rather to oxidation of
the oil under conditions of low oxygen content [B1].
If the typical value of carbon monoxide (700 ppm in [B1]) is used in such sealed transformers as a
default alarm threshold in monitors of type M2, a warning signal may be sent to the operator of the
carbon monoxide monitor, asking for a sample of oil to be sent to the laboratory for a full DGA analysis,
which will indicate that no fault is present. After a few such false alarms, the operator may come to
disregard carbon monoxide alarms from the monitor.
In cases where a fault actually occurs in the paper, if only a relatively small volume of paper is involved,
which occurs in the majority of cases of thermal faults in paper (for instance in windings or in leads),
not enough additional carbon monoxide will be formed to indicate reliably a fault in the paper against
the usually large background of carbon monoxide in the oil [B1]. So monitors M2 may not detect them
in time.
In addition, very few if any cases of faults in paper have been reported which could reliably be detected
by significant increases of carbon monoxide alone, without knowledge of carbon dioxide values and/or
of the other gases. Multi-gas monitors are therefore much more reliable than monitors of type M2 for
the detection of faults in paper, using not only carbon monoxide but also carbon dioxide and the other
hydrocarbon gases (e.g., with the triangle and pentagon methods).
14
2.5 Fault identification in transformers with on-line gas monitors
Monitor typesM1, M1* and M2 cannot identify any type of fault on-line, and monitor types M3, M5, M6
can identify only some of them. These monitors can be used to follow the evolution with time of one or
several gases, but an oil sample and a laboratory DGA analysis are needed to get full fault identification
using [B1].
The only on-line gas monitors that can identify on-line all types of faults and sub-types of faults are
monitor types M7 to M9.
See [B1]for gas concentration levels (threshold limits) to be used for the specific type of fault and sub-
type of fault identified. Also for rate threshold levels to be used for on-line monitors (10 times those
used for laboratory results). Also for actions on the equipment and use of health indexes.
See also [B1] for concentration levels (threshold limits) to be used by default when the type of fault is
not known.
2.6 Detection of fast-occurring faults in transformers with on-line gas

monitors
Fast-occurring faults and sudden changes in type of fault between two oil sampling intervals, in most
cases cannot be not picked up by manual DGA but only with on-line gas monitors17.
Such a case is shown in Figure 2.1, where a hot spot T2 suddenly evolved into an arcing fault D1 on a
loose bolted clamp in oil [B3]. In some cases, faults may be evolving so rapidly that even on-line gas
monitors will not detect them in time, but only Buchholz relays and sudden pressure valves [B2].
Figure 2.1 Sudden Change in Gas Formation and Type of Fault

Picked up by an On-Line Gas Monitor[B3]
Such cases of very low gas levels followed by fast-occurring faults and failure occur in less than 0.3%
of DGA cases, based on annual failure rates of transformers[B8], but they may be avoided with on-line
gas monitors when not too fast[B1].
15
Other examples of sudden failures occurring without advance warning by DGA or other means are
transformers containing corrosive sulphur in oil, although this is still debated [B6], and when the
insulation system is compromised at the end of its useful life 8.
2.7 Costs and benefits of various DGA monitoring techniques for

transformers
CIGRE WG A2.44 has published a general guide on transformer monitoring “architecture” systems [B7],
which, however, does not provide any practical information on gas monitors, including their costs and
benefits.
Table 2.5summarizes the costs and benefits of various monitoring DGA techniques for transformers
[B5].
The analysis is based on a population of 2000 large power transformers having a failure rate of 0.3%
per year, as commonly observed on many networks worldwide[B8], therefore with 6 failures per year
of the main tank (not of accessories such on-load tap-changers or bushings), requiring repairs or
replacement. This includes a catastrophic failure with fire or explosion in 0.04% of cases[B8] (0.8 such
failure per year).
Calculations are based on the following considerations:
 average cost of these 6 transformers is 6 M$ each, their total cost 36 M$/y.
 average cost is 6 M$ for replacing each of them; 2 M$ for repairing them (4 M$/y on average);
cost in case of catastrophic failure with collateral damage, lost revenues and penalties, is 30
M$.(30x0.8=24M$/y).
 laboratory DGA is able to avoid 2 failures due to slow developing faults (2x4=8 M$/y), while
the cost of DGA analysis of all transformers, including oil sampling is ~ 1 M$/y. Avoided costs:
(8-1) =7 M$/y.
 monitors M1, M1* or M2 are installed on the 2000 transformers, at a cost of 10,000$ each
amortized during 10 years (cost = 2 M$/y).
They are able to avoid 2 additional failures due to medium developing faults ((2x4)-2=6) M$/y),
in conjunction with laboratory DGA (7 M$/y). Avoided costs: (7+6)=13 M$/y.
 monitors M3 to M9 are installed on critical transformers and severely gassing transformers (30
and 50 transformers, respectively), at a cost ~50,000 $ each on average, amortized 10 years
(cost < 1M$/y).
They are able to avoid in time 3 additional failures due to medium-to-fast developing faults
((3x4)-1) = 11 M$/y, in conjunction with laboratory DGA (7 M$), and 0.8 catastrophic failure
(24 M$/y). Avoided costs: (7+11) = 18 (+24) M$/y.
 avoided costs in Table 2.5include cost of monitors but not of their installation, maintenance and
communications, which depend very much on each type and brand of monitor and should be
evaluated with monitor manufacturers then deduced from the avoided costs of Table 2.5.
Values in Table 2.5only intend to provide a rough comparison of the costs/benefits of different DGA
monitoring techniques. Actual values will need to be verified and improved by manufacturers and users.
16
Table 2.5 Costs/Benefits of Various DGA Monitoring Techniques
DGA Monitoring Number of Avoided Avoided Costs

Technique Used Failures per Year in M$ per Year
None None None
Laboratory DGA 2 7
Monitors M1, M2 4 13
Monitors M3 to M9 5 (+0.8) 18 (+24)
17
18
3. On-line gas monitors for on-load tap-changers
On-line gas monitors can also be applied to on-load tap-changers. As arc-breaking-in-oil tap-changers
(often also called ‘arcing type on-load tap-changer’ or ‘oil-switching type load tap-changer’) heavily
deteriorate the oil by their normal arcing activity, the carbonized particles suspended in the oil can cause
monitor reliability problems, particularly for the oil pump, the gas extraction membrane, and any in-line
filter used by the monitor. Some DGA monitors have been successfully used to monitor such on-load
tap-changers by installing the monitor not directly on the on-load tap-changer but rather on an oil
filtration system that circulates and cleans the on-load tap-changer on a continuous or regular basis.
Depending on the type and design of on-load tap-changer, the actual load current and number of
operations, the gas levels observed inside the tap-changer oil compartment may vary over several orders
of magnitude. Arc-breaking-in-oil tap-changers can produce huge amounts of gases during normal
service and cause great variations of the normal gassing rate. This, and the fact that the tap-changer
oil must be exchanged regularly, makes trend analysis difficult. Incipient faultsmay be masked by the
normal fluctuations of gas levels. Some monitors 6 (multi-gas and few-gas) are designed to read
accurately at the high concentration levels that may arise in on-load tap-changers.
On-line monitoring of vacuum type tap-changers is more straightforward, with the switching arcs
encapsulated inside vacuum interrupters. Some sparks may occur, due to current commutation on the
main or by-pass contact, and gases may be generated by some heating of the transition resistors
(depending on the on-load tap-changer model). The conditions for vacuum on-load tap-changers are
roughly comparable to the conditions of the main tank transformer oil, including the gas levels and
compositions.
In general, the following faults can occur on on-load tap-changers:

 X1: Overheating of current-carrying parts (contacts, leads; T<300°C),
metallic parts on floating potential (unwanted low-energy arcs or sparks)
 T2: Coking of contacts, overheated transition resistors, partially broken braids (300<T<700°C)
 T3: Heavy coking of contacts and severely damaged braids(T>700°C)
 D1: Worn or broken (switching) contacts, broken braids or leads (abnormal arcing)
 X3: Overheating in progress, increased contact resistance, unwanted arcing (combined faults).
Depending on the type of on-loadtap-changer, these faults can be detected by DGA. The
absolute gas values and trends can be evaluated, and Triangle 2 can be applied in combination
with the extended “normal zones”.
For more detailed information about DGA fault interpretation in on-load tap-changers, see [B9][B10],
[B20].
Table 3.1 On-Load Tap-Changers Faults Possible to Identify with On-Line Gas Monitors 8
Number of gases Faults to Identify

measured Gases Arcing type OLTCs Vacuum type OLTCs
19
H2, CH4, C2H6, C2H4, - all 5 faults, also at early stage

- all 5 faults
M8/ (M9) C2H2 - degree of oil aging (CO, CO2)
- clogged breather
CO, CO2, O2, (N2) - clogged breather (N2/O2 ratio)
H2, CH4, C2H6, C2H4, - all 5 faults, also at early stage
M6/ (M7)
C2H2, CO, (CO2) - degree of oil aging (CO, CO2)
H2, CH4, C2H4, C2H2, - all 5 faults - all 5 faults, also at early stage
M5
CO - degree of oil aging (partly, via CO)
M3 CH4, C2H4, C2H2 - all 5 faults, also at early stage
M2 H2, CO - abnormal arcing D1
- none
M1 H2 - unwanted low-energy arcs/sparks
In Table 3.1, for vacuum type on-load tap-changers, an evaluation of the thermal stress of the tap-
changer oil is possible via a trend analysis of carbon monoxide and carbon dioxide, either by tracking
the absolute values or the carbon dioxide/carbon monoxide ratio (see [B1]). With this, unusual heating
can be revealed, when correlated with relevant operating parameters, such as oil temperature, actual
load, and frequency of operation. Trends of hydrogen can be used in the same wayto detect unusual
arcing or sparking when correlated with relevant operating parameters (same as above).
The degree to which on-load tap-changers benefit from on-line DGA monitoring on an hourly or daily
basis rather than yearly or monthly, is still open to question.
20
4. Installation of on-line gas monitors
4.1 Installation on transformers
As in the case of oil sampling for laboratoryDGA, monitors can be installed in the bottom oil, top oil,
middle tank oil or in pipes going to or coming from radiators.
In cases of low-to-moderate gas formation, i.e., below the threshold rate levels of [B1], which allow
gas equilibrium to be reached throughout the tank oil, DGA results and monitor readings will be the
same at all locations [B2].
In cases of fast gas formation, the main flow of hot oil containing the fault gases is drawn into the pipes
going to the radiators. This is where the monitor response can be expected to be least delayed, as well
as in the return pipes of coolers10.
Some secondary flow of hot oil will remain in the top oil and will be mixed and diluted with it. However,
since the top oil valve is usually relatively far from the main flow of oil, some delay in the monitor
response can be expected if the monitor is installed on the top oil valve.
Hot oil drawn into the radiators will go to the bottom oil, where it will be mixed and diluted in the whole
tank oil[B2]. This will take time, however, so the monitor response can be expected to be most delayed
there, especially if the bottom oil valve is not directly in the flow of oil coming from the radiators.
Furthermore, whether connected to the top or bottom oil, some monitors are installed with long oil lines
which can further delay the monitor response. The response time of a monitor will therefore depend on
the monitor design, oil circulation to the monitor and oil circulation in the transformer. Some monitor
manufacturers specify the response time of the monitor in typical installation conditions.
Tests done on a transformer1 have indicated a 2-hour delay for gas injected on one side of the bottom
oil to reach the other side of bottom oil and the top oil.
A general recommendation is to install the monitors as close as possible to the sampling valve 6. In the
US, IEEE has replaced the half-inch requirement for bottom sampling valves by a quarter-inch one8. To
improve reliability, some monitors measure oil circulation through the monitor and provide notifications
if the oil flow is too low in the oil lines. Three-way adapter valves or pipes can be used on the oil
sampling valve6.
No preference has been given by WG members and end-users as to where to install gas monitors. It
has been mentioned that 90% of DGA values obtained by oil sampling were taken from the bottom oil.
When installed in the top oil, maintenance of the monitor and manual oil sampling are more difficult
because of the proximity of high voltage.
4.2 Installation on on-load tap-changers9

Concerning the installation of on-line gas monitors to on-load tap-changers, it is important to distinguish
between in-tank type models and compartment type models. For compartment type on-load tap-
changers, the on-line gas monitor can easily be attached to an appropriate pipe flange (or, for monitors
with built-in sampling loop, to two small pipe flanges, e.g. 3/8”) at a side wall or bottom of the on-load
tap-changer compartment. The exact location must be defined so that monitor parts extending into the
on-load tap-changer compartment do not violate existing insulating arrangements.
For in-tank type on-load tap-changers, an oil sampling loop is necessary to move some oil out of the
on-load tap-changer compartment for analysis. On-line gas monitors with built-in sampling pump can
easily be connected to two pipe flanges on the on-load tap-changers head, with the monitor input
connected to the on-load tap-changer suction pipe and the monitor output to another pipe flange which
ends closely below the head cover. (Usually, these flanges are used for attaching an oil filter unit –
which is not commonly applied to vacuum type on-load tap-changers). Monitors without a built-in
sampling pump (like membrane sensors) require a separate oil circulation system including an oil volume
21
to which the monitor may be attached; see Figure 4.1. The tap-changer manufacturer should be
consulted to find and develop an appropriate solution.
OLTC
oil sampling loop
on-line
gas monitor
Figure 4.1 Oil Sampling Loop for On-Line Gas Monitors

Attached to In-Tank Type On-Load Tap-Changers9
22
5. Accuracy of DGA results coming from on-line

gas monitors and laboratories
5.1 Definitions
Accuracy, as defined by IEC [B11]ISO and CIGRE TB 409 [B2], is thedifference in % betweengas values
measured by a laboratory in a sample of gas-in-oil standard (GIOS) and gas values introduced in this
GIOS.
The bias of a DGA analytical method11 (e.g., Method C), as defined by ASTM [B12], is the difference
between the prepared content of one gas in a GIOS sample, and the average of values for this gas
measured by several laboratories using this analytical method and a sample of this GIOS.
The difference between accuracy and precision is explained in [B22],
Figure 5.1.
5.2 Accuracy of DGA results coming from transformers

Significant and sometimes very large differences between DGA results coming from different
laboratories, and between on-line monitors and laboratories, have been reported by several WG
members1,2,3, also in [B2].
This is particularly true for results using the head space principle of gas extraction from the oil [B13],
[B14], used by a majority of laboratories and by most (but not all) on-line gas monitors.
This has been confirmed by numerous round robin tests (RRT) and “Proficency Test Programs” (PTP)
organized by IEC, CIGRE, ASTMand the Institute for Interlaboratory Studies (IIS) in NL, respectively.
The RRTs of IEC and CIGRE all used GIOS samples and were thus able to report large differences in
the accuracies of participating laboratories. The PTPs of ASTM and IIS do not use GIOSs, or use them
but do not calculate and report their accuracies, so they only evaluate the dispersion of DGA results
between different laboratories, not their accuracies.
5.3 effect of poor accuracy on DGA diagnosis

The effect of poor accuracy of DGA results coming from both laboratories and monitors on the
uncertainty of diagnosis of faults in transformers is illustrated in
Figure 5.1[B14], [B15]. The uncertainty of fault diagnosis in Duval Triangle 1 (left) is represented by
the red and blue polygons.
When the polygons of uncertainty cross several fault zones, identification of the fault is uncertain and
may result in inadequate actions on the equipment. When accuracy of DGA results is > 50%, fault
diagnosis is totally impossible (Triangle on right). This is very often the case at low gas concentrations
in oil (<10 ppm), where inaccuracies of 75% or more are quite common at concentrations < 5 ppm.
However, even at routine concentrations > 10 ppm, inaccuracies of 30% are also quite common,
particularly when using head space methods for gas extraction from oil (see App Table E.1 Accuracy of
Laboratories Using Gas Extraction MethodsApp Table E.1in APPENDIX E).
23
Accuracy of 15% (in red), 30% (in blue) Accuracy of 15, 30, 50 and 75%
(in green, blue, violet and red, respectively)
Figure 5.1 Effect of Poor Accuracy on Uncertainty of Diagnosis of Faults in Transformers[B14],[B15]
5.4 effect of poor accuracy of monitor readings on gassing trends

Inaccurate monitor readings will also provide inaccurate monitor gassing trends, contrary to a common
belief.
This is illustrated schematically in

Figure 5.2[B14][B9]. If the actual initial gas concentration in the transformer is 100 ppm, a monitor
with a systematic inaccuracy of + 50% for this gas, because of e.g., an internal calibration problem,
will provide a reading of 150 ppm. One year later if the actual gas concentration is 200 ppm, the monitor
reading will be 300 ppm.
So the gassing trend provided by the monitor will be 150 ppm/y instead of 100 ppm/y and if the rate
limit is 100 ppm/year, the monitor will send a false alarm.
Reversely, if the monitor inaccuracy is -50% for this gas, the gassing trend provided by the monitor will
be 50 ppm/y, and the monitor will miss a potentially dangerous fault.
Inaccuracies > 50% are quite common at low gas concentrations (see section5.3).
Figure 5.2 Effect of Poor Accuracy of Monitor Readings on Gassing Trends [B14]
The effect of poor accuracy of laboratory results on gassing rates and trends 12 has been discussed in
detail in [B15].
For sources of inaccuracy of DGA results and readings, especially when using the head space principle,
see Appendix0.
24
6. Evaluation of accuracy of on-line monitors
A dozen new on-line gas monitors have become available on the market since 2010 and have been
evaluated for accuracy in transformers in service by the WG, and are listed in Table 6.1, by decreasing
number of gases analyzed by the monitors then by alphabetical order of their manufacturers. Some are
still under evaluation by the WG at the time of publication of this TB.
Table 6.1 List of On-Line Gas Monitors Evaluated by WG D1/A2.47
Manufacturer Model Number Evaluated for

of Gases accuracy by
Camlin Totus 9 Terna (IT)13
GE Transfix EdF (FR)1
LumaSense LumaSense 9 Terna (IT)13
EdF (FR)1
Morgan Schaffer Calisto 9 Elektropomyar-
Elektryka (PL)14
Serveron TM8 EdF (FR)1
HydroCal HydroCal1008 8 EdF (FR)1
Terna (IT)13
Serveron TM8-F Terna (IT)7,13
Vaisala Optimus 7 Laborelec (BE)15, 13
National Grid (GB)16,14
Messko MSense 2 Elektropomyar-
Elektryka (PL)14
ABB Core Sense 1 Terna (IT)13
Terna (IT)7,13
Serveron TM1 Elektropomyar-
Elektryka (PL)14
The procedure used to do this evaluation is similar to the one described in TB 409 [B2], but clarified
further by WG D1/A2.47 in APPENDIX F1,6,14,17. Using the standardized procedures of [B2] and APPENDIX
F allows comparing the accuracies of monitors available before 2010 to those available from 2010 to
2018 and those that will be available after 2018.
The accuracy of monitors evaluated by WG D1/A2.47 is indicated in Table 6.2. The lower the absolute
values of accuracy in Table 6.2, the more accurate the monitor for this gas. (-) values in Table 6.2means
values below detection limits of the laboratory, or no values reported. Negative values in columns 3 to
9 of Table 6.2means values lower than expected. Positive values means values higher than expected.
Values in column 10 are the average of values in columns 3 to 9, expressed in absolute values.
25
Table 6.2 Accuracy of On-Line Gas Monitors Tested by Members of WG D1/A2.47
Accuracy in %
Test No Monitor H2 CH4 C2H4 C2H6 C2H2 CO CO2 Average
1 A 14 11 - 3 - 2 3 6
2 A -19* 0 - -4 - 21 0 5
3 B -25* -5 19 4 - -16 0 9
4 C -35* -8 -15 -7 - 18 0 10
5 D -20* -23 0 23 - 8 -20 15
6 D 0 4 16 31 - 16 0 11
7 E 11 17 42 48 51 45 28 35
8 F 11 - - - - - - 11
9 G 12 - - - - - - 12
10 H 0 7 25 78 8 14 16 21
11 I -26 0 3 0 - 10 10 8
Values of Test No 1 (monitor A) are the average values obtained with two reference laboratories
(Elektropomyar-Elektryka and Laborelec).
In Table 6.2, during Tests No 2 to 5, four different monitors had been installed during one year on a
test loop of oil attached to a transformer. Results of these Tests in Table 6.2were taken at the end of
the test period.
Hydrogen concentration readings from these four monitors, marked with an * in Table 6.2, wereall
significantly lower than expected, by 19* to 35*% when compared to corrected laboratory results.
Possible reasons for these unexpectedly low readings were investigated by the WG 15and are discussed
in APPENDIX H.
The accuracy of monitors tested by TF D1.01-15 in 2008 in TB 409 [B2]and by WG D1/A2.47 in 2015
in the present TB is illustrated in
Figure 6.1. The accuracy value indicated is the average accuracy AC av for all gases except oxygen and
nitrogen, expressed in absolute values (±). Each dot corresponds to one monitor model or unit tested
in 2008 (green dots), or in 2015 (red dots). Dots are plotted by increasing values of inaccuracy (from
4% to more than 50%).
Some results plotted in
Figure 6.1(2015) are not indicated in Table 6.2as agreed with the monitor manufacturers.
Figure 6.1 Accuracy of monitors units tested by CIGRE TF15 and WG47
Figure 6.1shows that a majority of monitor models on the market meet the accuracy requirement of
IEC (<15%) [B2], [B11]. However, a minority of them do not, with accuracies as bad as 50%. The
procedure of APPENDIX F should be followed to evaluate the accuracy of a monitor in service suspected
to have a poor accuracy, particularly if the readings are not in agreement with laboratory DGA results,
and if the monitor data are being used to prioritize and plan maintenance decisions on the transformer.
26
Some monitor manufacturers specify the monitor measurement performance under ideal factory
conditions. Others specify the measurement performance that should be expected throughout the
specified monitor operating life conditions. Users should be aware of this important distinction.
Some monitors are designed in such a manner that, as part of the installation procedure, site-specific
offsets are programmed to align the monitor readings with recent laboratory DGA results. Users should
therefore be aware of this distinction between monitors designed to provide readings of absolute
accuracy versus those that require site-specific offsets. The latter ones should be discussed with the
manufacturer.
Accuracy values in Table 6.2and
Figure 6.1were measured in newly installed monitors. After some time in operation, the accuracy of
some monitor models and some units of these models may, and have indeed been observed to drift
and become poorer6. This may sometimes require recalibration of the complete gas extraction/ analysis
system of these monitors and the replacement of some monitor components (e.g., columns, membranes
or detectors). It is indicated in Appendix0when to suspect such accuracy changes and how to verify
them with gas-in-oil-standards.
27
28
7. Conclusions
Tests done since 2003 by CIGRE TF D1.15-01-07, TF D1.15.01.15 and WG D1/A2 47, on ~20 on-line
gas monitors on the market, show that a majority of them meet the accuracy requirement of IEC
(±15%). However, a minority of them do not, possibly resulting in wrong fault identifications and actions
on electrical equipment.
When such inaccurate monitor readings are suspected, for instance if they are markedly different from
laboratory DGA results and this results in uncertainty on the type of fault and action on the equipment,
the accuracies of the monitor unit and the laboratory may need to be verified, using the procedure
provided inAPPENDIX F, and gas-in-oil standards.
Among the various types of on-line gas monitors examined by WG D1/A2.47, only those of the multi-
gas type (M7 to M9) can provide full DGA diagnosis on line and immediate action on the equipment.
Those of the single-gas type (M1 to M2) indicate abnormal formation of these gases and require taking
an oil sample before deciding on appropriate actions.
Another issue in the industry is the inaccurate DGA results provided by some laboratories. To address
this problem, it is recommended inAPPENDIX E.0that all laboratories follow the new accuracy
requirements of ISO 17025 and 17034. Also, that during bids for DGA analysis contracts, a requirement
for the accuracy of DGA analysis provided by bidding laboratories is included in the contracts, verifiable
by their customers with gas-in-oil standards.
29
30
APPENDIX A. definitions, abreviations and symbols
APPENDIX B. links and references

[B1] CIGRE Draft Technical Brochure (2019) Advances in DGA Interpretation, WG D1/A2. 47, www.e-
cigre.org
[B2] CIGRE Technical Brochure 409 (2010) Report on Gas Monitors for Oil-Filled Electrical Equipment,
TF D1.01.15, www.e-cigre.org
[B3] Identifying and Analyzing Quick-Developing Faults with DGA, SIGAT Conference, Cartagena,
Colombia, 2014, M.Duval
[B4] 100 Years of Gas Monitoring in Transformers, EPRI-TSUG Conference, San Antonio, 2018,
M.Duval
[B5] Case Study: Transformer Fault Detection and Repair Followed by Degassing Monitored with Online
DGA, 2016, S.Leivo and J.Mertanen, https://www.vaisala.com/vaisala/documents/CEN-G-Power-
TEchCon 202016_casestudy_Vaisala_Fingrid-2016
[B6] A Critical Review of Potentially Corrosive Sulphur, EuroTechCon Conference, Strattford-on-Avon,
2014, A. de Pablo
[B7] CIGRE Technical Brochure 630(2015), Guide on Transformer Intelligent Condition Monitoring
(TICM) Sysems, WG A2.44,www.e-cigre.org
[B8] CIGRE Technical Brochure 642 (2015), Transformer Reliability Survey, WG A2.37,www.e-cigre.org
[B9] CIGRE Technical Brochure 443 (2010),DGA in Non-Mineral Oils and Load Tap-Changers and
Improved DGA Diagnosis Criteria, WG D1.32, www.e-cigre.org
[B10] Application of Duval Triangles 2 to DGA Analysis in Load Tap-Changers, EuroTechCon Conference,
Glasgow, 2012, M.Duval
[B11] IEC Publication 60567 (2012), Oil-filled electrical equipment – Sampling of gases and analysis of
free and dissolved gases – Guidance, https://www.iec.ch/
[B12] ASTM Test Method D3612, Annual Book of ASTM Standards, American Society for Testing and
Materials, Philadelphia, PA, USA
[B13] New Techniques for Dissolved Gas-in-Oil Analysis, IEEE EI Magazine, Vol. 19, No 2, pp.6-15,
2003, M.Duval and CIGRE TF D1.15-01-07
[B14] Accurate On-Line Gas Monitors are Needed for both DGA Diagnosis and Trends, Qualitrol Webinar,
2016, M.Duval
[B15] Improving the Reliability of Transformer Gas-in-Oil Analysis, IEEE EI Magazine, Vol.21, No. 4,
pp.21-27, 2005, M.Duval and J.Dukarm
[B16] IEC Publication 60475 (2012) Method of sampling insulating liquids, https://www.iec.ch/
[B17] ISO/IEC 17025, General requirements for the competence of testing and calibration of laboratories
[B18] ISO/IEC 17034: 2016, General requirements for the competence of reference material producers
[B19] CIGRE Technical Brochure 607 (2015), Contracts for Outsourcing Utility Management Work, WG
B3.06, www.e-cigre.org
[B20] IEEE C57.139 (2015), IEEE Guide for Dissolved Gas Analysis in Transformer Load Tap-Changers,
IEEE Transformer Committee
[B21] IEEE C57.104 (2019), IEEE Guide for the Interpretation of Gases Generated in Mineral Oil-
Immersed Transformers, IEEE Transformer Committee
[B22] Dissolved Gas Analysis – It Can Save Your Transformer, IEEE EI Magazine, Vol.5, No.6, pp.22-27,
1989, M.Duval
Acknowledgments
1
E. Alzieu, 2 S. Bhumiwat,3 B. Nemeth,4C . Perrier, 5S. Leivo, 6 S. Brauer,7 J. Hinshaw,8P. Boman,9 R.
Frotscher,10 S. Ryder,11 M. Cyr, 12 A.M. Dale, 13 F. Scatiggio, 14 T. Buchacz, 15 S. Eeckoudt, 16 A.
31
Fieldsend-Roxborough, 17 M. Duval, 18 M. Grisaru, 19 H.J. Knab, 20

B. Noirhomme, 21
J. Wang, 22
I.
Atanasova-Hoehlein, 23 M. Banovic, 24 C. Michellon,25 X. Minniti
32
APPENDIX C. DGA on-line gas monitors using

infrared gas measurement methods
There are various on-line gas monitors commercially available that use infrared light (IR) to detect and
quantify fault gases from transformer oils5,6,7. These monitors vary significantly by their design and even
by their basic operating principles depending on the manufacturer. The following sections discuss the
IR-based monitors and their use in DGA application.
Gas extraction methods
In general, the same methods to extract gases from oil are used as with other online monitors e.g.
headspace, membrane and vacuum. The oil volume with these monitors is usually larger than with gas
chromatographs and thus also the gas volume used for the analysis.
It is important to know that partition (Ostwald) coefficients of the gases are dependent on oil type,
temperature of gas extraction and to a lesser degree the oxidation condition of the oil. The manufacturer
should declare how that is taken into account in the monitor, as those may affect monitor accuracy.
Factors affecting the conversion of gas-in-gas values measured by the detector to gas-in-oil values in
the oil sample when using the head space principle are described in Appendix0.3.
In some models there may be risk of contamination due to aerosols and heavier hydrocarbons extracted
from oil, as well as contamination from ambient air if measured gases are released.
Principle of IR-based gas analysis
When extracted gas molecules in a monitor are exposed to infrared light, they absorb energy as they
shift into an excited molecular state (App Figure C.1). For each gas, absorbed wavelengths are unique
forming gas specific “fingerprints”, which can be used to identify the gas components in the extracted
gas mixture.
Infrared-based analysis is a fundamental method where the absorption wavelengths and absorbances
of the fault gases do not change over time. This enables calibration-free operation over a long period,
provided other possible drift mechanisms are known and compensated within the monitor.
The analyzing part of an IR-based monitor typically consist of a light source, a wavelength selection
system such as band-pass filters, gas cell, mirror, and detectors (App Figure C.1). Note that gas
molecules like oxygen, nitrogen, hydrogen do not absorb light in the IR-range meaning that other
sensing technologies are needed to measure those gases.
App Figure C.1 Schematic picture on IR light App Figure C.2 Schematic picture of optical
absorption by molecules shifting to excited state. module.
IR absorption of different gases

App Figure C.3 presents the absorption spectra of key fault gases, carbon oxides and hydrocarbons.
33
App Figure C.3 Mid-IR light absorption of CO2, CO, C2H2, C2H4, C2H6 and CH4
The measured wavelengths are selected with band-pass filters that allow only a certain wavelength
band to pass through the filter. Both narrow band and tunable filters are used. Usually the final gas
analysis is based on signals gathered from multiple wavelengths (i.e. band-pass filters). IR wavelengths
used for detecting the gases vary depending on the monitor manufacturer.
Main IR-based technologies used in online DGA monitors

Most IR-based monitors available on the market operate either on the basis of non-dispersive infrared
(NDIR) or photoacoustic spectroscopy (PAS). The main difference in their operating principle is the
actual detection of IR light absorption(near or mid IR). Fourier Transform Interferometers (FTIR) are
also used.
 PAS
 Detectors typically used are highly sensitive microphones that detect the pressure
(sound) waves caused by molecule movement (heating)due to light absorption. The
detected acoustic signal increases with the number of absorbing gas molecules within
the gas cell. PAS system usually use band pass filters.
 NDIR
 The intensity of IR light that passes through the gas cell is measured with and without
gas absorption. The ratio of these two intensities, attenuated and unattenuated, gives
the amount of absorption. This is proportional to the number of gas molecules in the
gas cell, so that the IR detected light intensity decreases with the number of absorbing
gas molecules. NDIR detectors are for example thermopile, pyroelectric detector, and
photodiode.
 FTIR
 FTIR interferometers and algorithms convert the signal into a complete spectrum with
high accuracy wavelength reference.
Factors affecting gas analysis with IR

Although infrared-based analysis is a fundamental method where the absorption characteristics of the
fault gases do not change over time, measurement signals may be affected by other factors which
depend on the type and technology used in the monitor. Absorption spectra of gases are affected by
the total gas pressure. This is caused by pressure broadening of individual spectral lines. If the total
gas pressure in the IR measurement cell is different from the calibration pressure, this effect has to be
taken into account in order to have accurate readings.
Used transformer oils are rather complex in their chemical composition, thus other gases and volatile
compounds are extracted as well. Those may have overlapping IR absorption spectrum with the fault
gases, which may cause cross sensitivity (interference) if they enter the gas cell where analysis takes
34
place. Interfering species that may be present in aged transformer oil include moisture, heavier
hydrocarbon gases, sulphur hexafluoride, and other components. To provide accurate DGA readings,
the monitor must have a method to minimize or eliminate these effects. Methods used for this vary by
manufacturer.
Gas extraction temperature affects how much of each gas is extracted, because gas solubility values
(Ostwald coefficients) are temperature dependent. Thus, the component where gases are extracted
should be temperature controlled or the extraction temperature compensated.
IR light sources and detectors are typically sensitive to temperature variation. The IR module of the
monitor should be temperature controlled. In NDIR, a reference measurement is preferred to eliminate
temperature dependencies. How the effect of temperature is compensated and on which range, varies
between monitor manufacturers.
In PAS, the detector (microphone) may be sensitive to vibration, pressure, mechanical stresses, etc.
The manufacturer should specify if this needs to be considered when selecting an installation location
for the monitor.
Possibly occurring drift can be seen as changes in long term gas trends compared to oil sample
laboratory analysis provided that oil sampling and laboratory analysis are done properly. Drift can be
either offset or gain type.
Drift may occur just in one gas or many gases, depending on the initial cause. Reasons for drift may
include:
1. Contamination of optical components.

2. As oil ages, the type and amount of other volatile compounds having overlapping IR absorption
may also change.
3. Changes in oil composition due to ageing may change gas solubility (Ostwald) coefficients.
Some gas extraction methods may be more sensitive to that.
4. In case the sensing part of the monitor has a connection to ambient air (e.g. venting measured
gases out), traces of sulphur hexafluoride from the substation environment may enter the
device and affect its measurement. Sulphur hexafluoride has strong IR absorption and in case
it is overlapping with fault gas detection wavelengths, even trace amounts of sulphur
hexafluoride may have a significant effect. Sulfur hexafluoride may be dissolved in the oil due
to leaks from sulfur hexafluoride insulated equipment on the transformer. The same may occur
if there are solvent vapors in a lab or industrial environment.
5. Change of the light source intensity.
6. Change of the detector sensitivityor in the detector electronics.
7. Shift in the band pass filter transmission spectrum.
8. Material deterioration of optical surfaces.
Some monitors offer the possibility for the user to adjust the readings based on laboratory sample
results to compensate drift. However, it must be taken into account that such adjustment is usually only
the offset type at the adjustment gas concentration and does not affect gain drift. In addition, if the
sample is not representative or laboratory analysis is not with good accuracy, the performance of the
monitor may also become less accurate after such offset adjustment.
If contamination or drift occurs, the affected components need to be changed or recalibrated at factory.
It may then become necessary to perform such corrective action in frequent intervals, unless
contamination and drifts are prevented by the monitor technology itself. The monitor manufacturer shall
declare, if there are components that may drift/contaminate and how often such components should be
changed.
35
Calibration
Typically, factory adjustment and calibration of the IR-based technologies is done with certified
reference gases during manufacturing.
Some monitors may come also with a test report from a factory gas-in-oil test. However, that applies
as such only for the oil used in that factory test, because gas solubility coefficients vary between oils.
This applies also even within different mineral oil brands, depending on their base oil being paraffinic
or naphthenic, and what is their aromatic content, for example.
Use of gas-in-oil standards might be possible only at the factory, because IR-based online monitors
often require a large oil volume for their operation, for example a few liters. As IR technologies tend to
use significantly more oil volume than GC, commercial gas-in-oil standards cannot be used.
Some manufacturers may require the use of site-specific offsets, based on laboratory DGA oil sample
measurements, to achieve their published accuracy specifications.
How to evaluate an IR-based online monitor in operation

Things to consider when comparing results between an on-line monitor and a laboratory:
 A reference oil sample for laboratory analysis should be taken as close to the monitor oil intake
as possible. The sampling procedure should follow the best practices of IEC, ISO or ASTM.
 Samples should be stored and transported carefully.
 Sample analysis in the laboratory should be executed following the best practices and
procedures of IEC, ISO or ASTM.
 The gas solubility coefficients used in the laboratory may be different than those used in the
monitor. This may cause a few percent difference between the readings, depending also on the
gas extraction methods applied.
 One should also consider what standard is used in the laboratory versus in the monitor in
question (IEC or ASTM) and especially at which temperature the gas volume is calculated (0°C
or 20°C).
 Uncertainty of laboratory results should also be taken into account.
Detailed instructions on how a proper evaluation of monitor readings vs laboratory analysis can be made
are indicated in APPENDIX F.
Validation of IR-based DGA monitors in laboratory

 IR-based DGA monitors often require relatively large gas-in-oil volume to operate, minimum a
few liters usually. Thus, a laboratory must have a method to prepare a large-volume gas-in-oil
sample of 10 liters or more for example and maintain the concentrations for a longer period.
Gas-in-oil values can be measured by GC laboratory analysis of the large-volume gas-in-oil
sample using the procedures in TB 409 [B2] and APPENDIX F of this document.
 Some monitors may require a certain minimum hydrostatic pressure for their proper operation.
This specification should be provided by the monitor manufacturer.
 Monitors releasing measured gases to ambient are ‘consuming’ gases from test oil, which
becomes visible as a continuous decrease in the measured gas concentrations. The smaller the
sample oil volume, the bigger the effect will be.
 Preparing gas-in-oil samples with carbon dioxide concentrations below ambient air
concentration (about 400 ppm) may be challenging. Ambient air gets easily into contact with
the prepared oil sample and carbon dioxide will dissolve in the oil.
 Performance validation of the monitors at low dissolved gas total pressures is recommended as
well in order to simulate sealed new transformers and transformers after degassing.
36
APPENDIX D. DGA monitors using Gas

Chromatography (GC) measurement methods
In gas-chromatography (GC)6,7, a sample of gas is analyzed by first separating the molecular compounds
that make up the sample, and then measuring the concentration of each compound one-at-a-time. The
separation is done using an inert carrier gas, a mobile phase, to flush a small volume of the sample gas
through one or more gas-separating columns. The columns are coated on the inside or filled with various
porous materials, called the stationary phase. The stationary phase interacts with the sample
components and selectively retains them to varying degrees against the flow of carrier gas. The rates
at which sample components pass through the column(s) are determined by each component’s vapor
pressure and chemical nature in relation to the stationary phase, gas flow, and column temperature.
Each component passes through the column and exitsat a characteristic retention time after injection.
A detector at the end of the column records a signal peak as each compound exits the column at a well-
defined “retention” time that is characteristic for that compound. The concentration of each compound
is determined by comparing the amplitude or integrated area of each detected peak with that obtained
when a known “calibration gas” mixture is injected into the same system.
App Figure D.1 A gas chromatography (GC) measurement system for DGA
Advantages of GC methods for DGA

GC methods are the cornerstone of laboratory DGA measurements around the world. Many on-line DGA
monitors and portable DGA analyzers also use GC methods. Various stationary phases have been applied
to separate the DGA gases, including porous polymers and zeolite or carbonaceous molecular sieves.
Several possible GC configurations are described in the reference standards [B11], [B12]. In general,
porous polymer columns can be used to separate CO 2 and the hydrocarbons up to C3, while molecular
sieves separate the lighter gases H2, O2, N2, CH4, and CO. In another configuration, a single column can
be employed.
The highly selective separate-and-measure approach is well suited to DGA applications because
transformer liquids, and gas samples extracted from them, often contain compounds other than the
gases of interest. These other compounds can include heavier hydrocarbon gases, alcohols, acids, and
sulphur hexafluoride that may leak into the dielectric fluid from gas insulated bushings, among others.
As depicted in red in App Figure D.1, a well-designed GC system for DGA will separate the signals of
the gases of interest from the signals of other potentially interfering compounds, so that the accuracy
of the DGA measurements is not compromised by the presence of these other compounds.
Calibration gas for lifetime accuracy

GC methods maintain their accuracy over the life of the product through the use of calibration gas.
Nearly all gas measurement systems are prone to drift with time and with changes in the operating
environment (e.g. temperature, pressure, humidity). Calibrating the GC system to a known standard on
a regular basis (typicallyevery one to four days) serves to compensate for these drifts.
37
In addition, the calibration results themselves are useful to verify that the gas measurement system is
functioning as designed. If the calibration results are not within normal bounds, an online DGA monitor
can avoid reporting erroneous data which could trigger a false gas alarm condition for the transformer.
Rather, the system can send a maintenance alert to investigate the malfunction on a non-urgent basis.
Gas chromatographs base their zero measurement on the signal level generated by high purity carrier
gas. A gas chromatography analyzer is not affected by the presence of traces of DGA target gases such
as CO2 or CO, or of interfering gases such as SF6, in the surrounding air.
Sensitivity
DGA on-line gas monitors with low gas detection limits are advantageous for early fault detection and
interpretation, particularly in new transformers. As an example, it is important to investigate the first
time a new transformer develops as little as 2 ppm of acetylene, especially if this occurs early in the
service life as it may indicate a manufacturing defect or transportation damage. Some monitors using
GC methods have sub-ppm detection limits for acetylene (and other gases) and are accurate in the low
ppm range. This sensitivity allows such faults to be detected and tracked at their earliest onset, so that
informed corrective actions may be taken as soon as possible.
Gas extraction membranes
The volume of gas needed for a GC measurement is very small, generally less than 1 cm 3. Such small
gas samples can be extracted from the transformer fluid within a typical one-hour measurement cycle
using a semi-permeable membrane gas extractor. Membrane extractors are effective in preventing oil
droplets and vapors from entering the gas measurement system during system operation. They also
provide a robust means to prevent liquid oil entering the gas measurement system during shipping and
handling. In GC DGA monitors, membrane extractors keep the gas valves and gas-separating column(s)
clean, and help ensure long-term performance and reliability.
Factors affecting the conversion of gas-in-gas values measured by the detector to gas-in-oil values in
the oil sample when using the head space principle are described in Appendix0.3.
Column lifetime
Separation of the DGA target gases by GC requires one or more GC columns. The columns may be
subject to gradual retention time reduction due to absorption of traces of water present in the oil or
carrier gas, or from other contaminants that can accumulate over long time periods. Column lifetime is
dependent upon individual installations and conditions. GC-based systems can monitor these effects via
periodic calibration and validation and provide diagnostic information to indicate when column
replacement will be necessary.
Back-flushing with reversed carrier-gas flow as well as reconditioning by baking out the columns at
elevated temperatures helps to remove trace contaminants such as water or heavier hydrocarbons that
may build up over time. However, column bake-out is generally limited to laboratory GC due to
temperatures in excess of 300 °C that may be required.
DGA monitor deployment
Unlike laboratory DGA systems, online DGA monitors of all types are subject to extremes of temperature
and environmental conditions. Thus, online DGA systems are deployed in weatherproof enclosures that
withstand a wide ambient temperature range of about -40 to + 55 °C, direct sunlight, precipitation,
dust, and humidity. In addition, top-oil temperatures greater than 100 °C are not uncommon, so the
temperature of oil reaching the monitor must be well understood and controlled. Generally, on-line DGA
monitors make use of specialized heating, cooling, and oil-management systems to stabilize the sensitive
analytical instrumentation against environmental extremes.
As they are installed in electrical generation, transmission, and distribution locales, DGA monitors must
be capable of safely withstanding applied voltages and electrical fault conditions well beyond
requirements for laboratory systems, in addition to emission of and susceptibility to radio frequencies.
Such systems are subject to a variety of locale-specific regulations concerning their safety under normal
operating conditions as well as after introduction of various fault conditions. Purchasing specifications
for online DGA monitors should encompass all the required environmental, electrical, and workplace
standards for the targeted installation.
38
Evaluation of GC-based DGA monitors in the field and laboratory is covered extensively in this document,
see e.g.APPENDIX E and APPENDIX F.
Maintenance costs
All DGA on-line monitors require some amount of maintenance during their service life. Some models
require scheduled on-site maintenance to clean or change oil filters, or to ensure that external air flow
paths are unobstructed. Some models require a major overhaul after 3-5 years in service, which can be
done in the field in some cases. Firmware upgrades and periodic “health check” downloads are good
practice, and the associated costs can be minimized if remote communications have been configured
for these tasks. Naturally, more reliable models reduce the costs associated with unplanned breakdowns.
For monitors that use GC methods, the carrier gas and calibration gas cylinders need to be replaced
during the scheduled maintenance each 3-4 years. To simplify gas cylinder management, some models
remotely communicate gas pressures, gas consumption rates, and the time remaining until the cylinders
need to be replaced. When comparing the lifetime cost of different DGA monitors, the cost of all service
visits, materials, and repairs/overhauls should be considered.
39
40
APPENDIX E. Gas-in-oil standards
Sources of inaccuracy of DGA results from laboratories and on-line gas

monitors
E.1.1. Oil sampling
If oil sampling from transformers does not strictly follow the standardized procedures of IEC 60475
[B16], ASTM, or others, and is not done by properly trained personnel, this may affect the accuracy of
DGA results, whatever the DGA analysis method used by laboratories [B11], [B12], e.g., because of
contamination by ambient air or gas losses from syringes. On-line gas monitors do not require taking
an oil sample manually, so they are not affected by sampling errors.
E.1.2. Oil analysis
If DGA analysis in the laboratory does not strictly follow the standardized QA/QC analytical procedures
of IEC [B11] or ASTM [B12], and is not done by properly trained personnel, this may also affect the
accuracy of DGA results. Procedures used inside on-line monitors should also be well designed.
E.1.3. Differences between DGA results
Large differences in DGA results coming from different laboratories and from on-line gas monitors for
the same transformer are often observed, and have been reported by CIGRE in 2003 [B13] and 2010
[B2] among many others.
Differences between laboratory methods used are illustrated in App Table E.1[B14], based on DGA
results reported by large numbers of laboratories participating to Round Robin Tests of IEC and CIGRE,
and to Proficiency Test Programs of the Institute for Inter-Laboratory Studies in the Netherland and
ASTM in the US. App Table E.1 shows that it is more difficult for laboratories to obtain accurate results
with the head space method C [B13].
For the definition of accuracy in App Table E.1, see section 5.1. “Average accuracy” in App Table E.1
means the average of accuracies for each gas in the measured oil samples. “Number of inaccurate labs”
in App Table E.1 means the number of labs not meeting the IEC requirements for accuracy (e.g., < ±15%
above 100 ppm).
App Table E.1 Accuracy of Laboratories Using Gas Extraction Methods

A (Partial Degassing), B (Stripping) and C (Head Space)17
Reasons for these differences have been attributed by TF D1.01.15 and WG D1.32 11,17[B9], [B13], and
confirmed by tests done within WG D1/A2.47, to the large number of factors and manipulations that
can affect results when extracting gases and converting gas-in-gas concentration values in the
headspace of vials or monitors to gas-in-oil values in the oil.
This for example requires exact knowledge of the oil/gas ratio in the vial and of the Ostwald coefficients
K of the oil sample, which are affected in a semi-logarithmic way by the temperature of gas extraction,
also by the type of oil (mineral or non-mineral), brand and chemical composition of the oil used, degree
of oil oxidation, total content of air or nitrogen in oil, and very low or very high concentrations of the
fault gases in oil11, 17.
To circumvent these potential sources of errors, it has been recommended in 2010 by CIGRE [B9] and
IEC 60567 [B11] to calibrate the head space method with gas-in-oil standards (GIOS) rather than with
41
only gas-in-gas standards as in method C. With the IEC method, exact knowledge of K values and
oil/gas ratios is not needed any more. It has indeed been observed that laboratories using the IEC
method tend to provide on average more accurate DGA results 15.
When using vacuum gas extraction methods, e.g., Method A of ASTM [B12]and the corresponding IEC
method [B12], only second-order corrections with K values are needed to calculate gas-in-oil
concentrations, so these K values do not need to be known very precisely but only approximately such
as in publication [B11]. By contrast, with the head space method C, K values must be known very
precisely because of the first-order corrections needed that may affect the accuracy of results.
Verification of accuracy of DGA results with Gas-in-Oil-Standards (GIOS)
As mentioned above, the IEC recommends using GIOS samples to calibrate the head space procedures
used by laboratories and monitors, and to regularly verify them.
Some WG members have thus reported that they successfully recalibrate their DGA laboratory
equipment with GIOS every 4 months for accreditation purposes, and every month internally, to ensure
a very good accuracy of their DGA results15.
Others6 verify their calibration curves with GIOS every day, and use 6 different concentrations of GIOS
to prepare their DGA calibration curves. This is particularly relevant for laboratories using head space
gas extraction17[B14]. Not all laboratories, however, are using GIOS as they should for their calibration
curves5.
Some others have asked CIGRE D1 and A2 to recommend using GIOS for the accreditation of
laboratories19.
It has indeed been mentioned that ISO 17025 [B17] now requires a traceable (GIOS) external standard
to prove to ISO representatives that their results are precise and accurate 15. Laboratories must also
participate in inter-laboratory round robin tests (RRTs) prepared by an independent organization, for
each of the tests performed, at regular intervals (at least every year)5,11,15.
A certification using GIOS6 for DGA analysis is also now required in ISO Guide 17034 [B18].
GIOS are also needed to verify the accuracy of DGA laboratory results and on-line gas monitor readings
coming from transformers. Such a verification is needed when results provided for the same transformer
by different laboratories, or by laboratories and monitor readings, differ by more than 15% (see
Figure 5.1), and when this results into different fault identifications and actions on the transformer[B2].
GIOS can be purchased, or prepared using one of the following in-house methods. They should indicate
the temperature at which gas concentration values have been calculated (e.g., 0°C or 20°C), see
Appendix 0.
Methods for preparing in-house samples of GIOS

The following methods can be used to prepare in-house GIOS samples and verify the accuracy of DGA
laboratory results and on-line gas monitor readings:
E.3.1. Procedure 1
Volumetric method of IEC 60567 (section 6.3) [B11], using a precision gas-tight syringe of 1 mL or less
and a 100 mL glass syringe of oil, or a 30 mL vial in Note 3 of the method.
This method was initially developed in Belgium 15 then adopted without changes by ASTM D 3612
Appendix0.1.[B12]. It is used by several laboratories to verify on a semi- regular basis (e.g., every 3
months) the calibration procedure and overall accuracy of their DGA results (including the gas extraction
stage)13,15.
A slightly improved version of Procedure1, first developed in Canada 20 and used in China21, consists in
placing a reversible red rubber septum on the tip of the 3-way valve of the oil syringe, allowing
42
placement of the oil syringe in a vertical position, tip down, when injecting gas from the precision gas-
tight syringe. This allows bubbles of injected gas to rise to the top of the oil syringe, and prevents
trapping of bubbles of air in the valve.
Still another version of Procedure 1, developed in Israel18, consists in preparing high gas concentrations
in oil first, in order to obtain better precision and reduce the risk of gas loss from syringes. Fill a 250
mL syringe with 200 mL of oil, then introduce 20 mL of a gas-in-gas standard mixture (e.g., with 10%
of each of the gases in carbon dioxide). When gas is dissolved, introduce the oil in a large volume of
degassed oil (~ 1 L), which will be used as a batch of gas-in-oil standard.
E.3.2. Procedure 2:
Gravimetric method proposed in Canada (see TB 443, section 3.1)[B9]. Bubble pure gases, in a batch
of degassed oil, up to equilbrium, at a constant temperature (e.g., 25°C) 11. Inject a weighed sample of
this oil into the vessel of the Partial Degassing Method of IEC 60567 (or of Method A of ASTM), extract
gases completely with multiple cycles of vacuum extraction, then measure the total volume of gas
extracted. Calculate the partition coefficients of each gas, after having converted the weight of oil to its
volume, using the density of oil11.
GIOS samples can then be prepared by bubbling a gas-in-gas standard mixture in oil, and calculating
gas-in-oil values using the partition coefficients calculated above.
A variation on Procedure 220 consists in using one cycle of vacuum only, then calculating the amounts
of gases remaining in the oil, using IEC 60567 APPENDIX C[B11]. The other steps are the same.
E.3.3. Procedure 3:
Procedure 3 using glass vials in Germany22. Two needles are inserted in the vial septum. Pure gases are
bubbled in the oil in the vial through one needle up to equilibrium, at a constant temperature (e.g.,
25°C), and the overpressure of gas used for bubbling it into the oil is measured (bubbles can form only
if there is a slight overpressure11).
The weight of oil is measured before bubbling the gas and is converted into volume of oil, using the oil
density. The total weight of oil and dissolved gas is measured after bubbling and the weight of dissoved
gas obtained by difference11.
The weight of dissolved gas is converted into volume of gas, at the overpressure and temperature
used11, then converted to normal pressure. The volume of gas dissolved is calculated in ppm per volume,
then the partition coefficient of this gas at normal pressure and temperature.
Standard mixtures of gas-in-nitrogen or in another gas are then bubbled in oil until saturation, then the
ppm values of each gas in oil is calculated, using the partition coefficients of each gas.
E.3.4 Commercial Procedures:

Two commercial procedures are available5,6 but are proprietary and have not been disclosed for the
preparation of in-house GIOS.
In the first commercial method6, known amounts of pure gases or gas-in-gas standard mixtures, are
weighed and introduced in a large volume of degassed oil measured by weight, using a deformable
system preventing any gas loss to the atmosphere. GIOS samples thus prepared are NIST-tracable
and have an accuracy better than 2% [B2](NIST = National Institute of Standards and Technology).
In the second commercial method5, known amounts of a gas-in-gas mixture with very little air or
nitrogen in it, are introduced in a large volume of oil measured by weight, using another proprietary
deformable system. These GIOS are also NIST-tracable and accurate to ± 2%.
Commercial GIOS samples are now available for non-mineral oils such as FR3 or silicone oils6.
The type or brand of oil used for preparing GIOS samples should be the same as in the transformer.
43
New commercial GIOS are expected to appear on the market17.
A comparison of GIOS purchased commercially and prepared in-house following methods described in
IEC 60567 [B11] indicates that there is a good agreement between them, up to their expiry dates 15.
A preference for in-house Procedure 1 (section E.3.1) has been indicated by members of the WG. It has
been mentioned that very few DGA labs if any can achieve the accuracy of ± 2 % of a GIOS 6, 11.”
Use of GIOS Samples
As indicated in E.2, GIOS can be used[B2] to:
 calibrate the whole DGA analysis process of laboratories, including its gas extraction stage, as
recommended in IEC 60567, particularly for those using the head space principle. This applies
also to gas monitors using this principle.
 verify the accuracy of laboratory results, as also recommended in IEC 60567 (sections 9.1 and
9.3.4) [B11], and the accuracy of gas monitors [B2], particularly those using the head space
principle.
 verify that laboratories meet the IEC accuracy requirement of ±15% when bidding for DGA
analysis contracts17, and are not selected only on their cost per analysis. It has been
mentioned23 that presently DGA contracts are given by accountants (based on cost only), not
by engineers and chemists (based also on quality of DGA results).
It is also strongly recommended in CIGRE TB 607 [B19] that in contracts for outsourcing utility
maintenance work, suppliers of such services should follow ISO standards for quality assurance. This
means that laboratories providing DGA analysis should now follow ISO 17025 and ISO 17034, and use
GIOS to verify the quality and accuracy of the results they provide to their customers (see Appendix 0).
Several WG members are in favor of such an accuracy requirement for accreditation of laboratories and
when bidding for DGA analysis contracts5,15.
Storage of GIOS samples
High quality glass syringes should be used for preparing and storing GIOS samples, especially at very
high and very low gas concentrations6.
Repeated tests by one manufacturer of commercial GIOS6 samples have shown that the loss of hydrogen
one week after they have been prepared is less than 1% (<2.5% per month).
It is recommended to use GIOS rapidly, however, not more than one or two weeks after they have been
prepared. Those planning to wait a longer time (e.g., one month) before using them should order “long
life” GIOS samples6. GIOS that have been opened should be used within one or twodays only14.
GIOS samples are very reliable and accurate when prepared and used properly 15. The same was
observed with gas-in-oil samples prepared in-house, following Procedure 1 of section E.3.115.
More generally when using glass syringes for DGA, what should be looked at is the absence of gas
bubbles in oil that may have formed during sampling or transportation, since these bubbles may increase
the loss of hydrogen from oil in the syringe. This has been investigated in CIGRE TB 443 (section 5.1)
[B9], however, and was found to be really significant only when very large bubbles of gas are formed.
Special enclosures are available for commercial GIOS samples transported in airplanes.
44
APPENDIX F. Procedure of WG D1/A2.47 for

evaluating the accuracy of gas monitors and
laboratory results
Outline of the procedure1,6,14,17:
First purchase a sample of gas-in-oil standard from an appropriate vendor, or prepare one according to
IEC 60567 or ASTM D3612 (see 0)
Take a reading on one or several on-line gas monitors installed on a transformer in service.
Take four duplicate samples of oil from the sampling point of the monitor(s), immediately after having
taken the reading. For portable gas monitors, take five duplicate samples of oil from the transformer.
Using one of the five samples, take a reading of the portable monitor.
Send all the above samples of oil to the DGA laboratory including the GIOS samples.Analyse all samples
on the same day (or over no more than a few days), using the same analytical equipment for all samples.
Perform calculations indicated in 0.
Symbols used
GIOS = gas-in-oil-standard
S = prepared gas values in GIOS, in ppm
Ls = values measured by Laboratory in GIOS, in ppm (or average of two analysis of the same GIOS
sample)
i = oil sample i taken from the transformer
Li = uncorrected lab value for sample i, in ppm
Lic = corrected lab value for sample i, in ppm
n = number of oil samples i or s analyzed (usually 4)
A = average of n corrected lab values Lic, in ppm

R = repeatability of laboratory for the n samples in %
M = monitor reading in ppm
D = difference between M and A in %
AC = maximum accuracy of the monitor

ACav = average maximum accuracy of the monitor for all gases
m = number of gases used for ACav
Calculations
45
All measured ppm (µmoles/l) values for S, Ls, Li and M should be converted first to the same unit,
either 20°C (IEC) or 0°C (STP) or 25°C, depending on preference.
Lab accuracy = ((Ls-S)/S)x100, in %, for each gas and on average of gases.

Lic = Li (S/Ls) in ppm
A =∑𝑛𝑖=1 𝐿𝑖𝑐 / n, in ppm, (with L = Li or Ls)
R = ((∑𝑛𝑖=1 𝐿𝑖𝑐 --A)/A) x 100)/n in %, with (Lic-A) in absolute values
D = ((M-A)/A) x 100 in %
AC = (D-R-2) in %, with D and R in absolute values. 2% is the maximum uncertainty on GIOS,
prepared in-houseor purchased. When D (in absolute value) is < R (in absolute value) and 2, AC
should be indicated as 0%.
The same calculation should be repeated for each gas ACg in %
ACav = (∑𝑛𝑔=1 𝐴𝐶𝑔/m) x 100 in %, excluding O2 and N2

If more than one laboratory was used for the evaluation of monitor accuracy, take the average of AC av
values of participating laboratories.
Note14: Corrected lab results Lic should be calculated using a sample S with Ls concentrations nearest
to those measured in the oil sample from the transformer (S100 sample in the example below). In this
example (A = 259 ppm), one can also use both the S500 and S100 samples and take the average
accuracy calculated from them.
Example of calculations for one gas:

S = 93 ppm (at 0°C) = 93x(293/273) = 100 ppm (at 20°C).
Ls1 = 91 ppm, Ls2 = 89 ppm (at 20°C)
Ls= 90 ppm (on average)
Lab accuracy = ((90-100)/100)= -10%
Li for 4 duplicate samples: L1=259; L2=230; L3=210; L4=240 ppm

Lic corrected: L1c=288; L2c=256; L3c=233; L4c=267 ppm
A = 261 ppm
Repeatabilty:
((288-261)/261)x100=10.3%; ((256-261)/261)x100=-1.9%
((233-261)/261)x100=-10.7%; ((267-261)/261)x100=2.3%
R = (10.3+1.9+10.7+2.3)/4 =6.3% ~6%
M = 305 ppm (at 25°C) = 305x(293/298) = 300 ppm (at 20°)

D =((300-261)/261)x100=14.9% ~ 15%
ACg = 15-6-2 = 7%
Note: Members of the WG evaluating gas monitors had to strictly follow Procedures F.1 to F.3 to make
sure that the evaluation of accuracy for each monitor tested was comparable. Results that were not
46
obtained following this procedure were not taken into consideration when compiling the results
presented in this TB.
Two examples of accuracy calculations thus performed on the same tested monitor unit are indicated
in App Table G.1 andApp Table G.2of APPENDIX G.
Test program procedure in the field

Prior to following Procedures F.1 to F.3 above for evaluating the accuracy of gas monitors, members of
the WG also had to follow the Test Program Procedure in the Field indicated below 6,17, i.e., identify:
 a utility company accepting to install a monitor unit to be tested on one of its transformers in
service.
 a transformer with “routine” gas concentrations (> 10 ppm of hydrocarbons).
 a transformer owner and substation manager in charge of the transformer.
 a contractor and coordinator for the tests within the utility company.
 someone from the monitor manufacturer providing a unit for testing (e.g., marketing/ field
service manager).
 a local monitor supplier authorized representative for overseeing the supply and testing of the
unit.
 one or two laboratories providing reliable and accurate DGA results.

 agree on an installation and commissioning date.
 agree on a location where to install the monitor unit (bottom or top oil or on radiator pipes).
 install and commission the unit according to the manufacturer specifications and ensure
monitor readings have stabilized (1 to 2 weeks).
 verify its functionality on-site, download database diagnosis files of the unit.
 wait for OK from manufacturer that the unit is operating correctly.
 order and receive sets of GIOS samples in syringes, covering all gas concentrations in the
transformer oil.
 take oil samples from the sampling port of the installed unit.
 send these oil samples to the laboratories for DGA analysis.
 take ppm readings on the unit.

 download data files from the unit.
 uninstall the unit.
 perform calculations of accuracy of laboratory results and monitor readings as indicated in 0,
using the GIOS samples nearest to gas concentrations measured in the oil samples, in order
to properly correct laboratory results.
47
48
APPENDIX G. Examples of monitor accuracy

calculations6,14
App Table G.1 Example 1 of Accuracy Calculations Performed for Test No 1 of Table 6.26.
DGA Lab 1 H2 CH4 C2H4 C2H6 C2H2 CO CO2 O2 N2
Lab measurements of Gas-in-Oil Standards (GIOS)

TRUE NORTH 500 2015.09.24
Prepared STP (0ºC) ppm 493 519 500 502 500 515 518 15800 57000
S_500 Prepared IEC (20ºC) ppm 529 557 537 539 537 553 556 16958 61176
GIOS 1 measured IEC (20ºC) ppm 539 546 522 527 491 522 531 17151 60845
Ls_500 Average ppm 541 551 524 528 493 529 532 17278 61739
Lab accuracy_500 % 2.3% -1.2% -2.4% -1.9% -8.1% -4.3% -4.4% 1.9% 0.9%
TRUE NORTH 100 2015.09.24
S_100 Prepared IEC (20ºC) ppm 105.2 106.3 106.3 106.3 105.2 107.3 124.5 16958 62786
GIOS 1 measured IEC (20ºC) ppm 108.1 109.1 105.4 100.1 97.8 105.0 123.0 16326 62423
Ls_100 Average ppm 108.4 109.3 105.2 100.0 97.5 105.5 123.0 16470 62574
Lab accuracy_100 % 3.0% 2.8% -1.0% -5.9% -7.3% -1.7% -1.2% -2.9% -0.3%
TRUE NORTH 10 2015.09.24
GIOS 1 measured IEC (20ºC) ppm 13.6 11.2 10.5 10.6 9.5 43.0 16791 61637
GIOS 2 measured IEC (20ºC) ppm 13.9 11.3 10.3 9.9 9.3 47.0 16717 61814
Ls_10 Average ppm 13.8 11.3 10.4 10.3 9.4 45.0 16754 61726
Lab accuracy_10 % 16.5% 4.8% -3.1% -4.5% -12.4% 31.0% -2.4% -2.5%
Lab measurements of transformer syringe samples 2015.09.23

L1 Sample 1 measured IEC (20º) ppm 722 37.8 8.5 87.0 1699 <500 6329
L1c Sample 1 corrected IEC (20º) ppm 706 36.8 8.9 88.5 1777 6372
A Average of corrected lab samples ppm 699 36.9 8.9 88.8 1777 6831
Sample 1 difference from average % 1.0% -0.3% 0.0% -0.3% 0.0% -6.7%
Sample 2 difference from average % -3.2% -2.1% -1.2% -2.6% -1.1% -4.2%
Sample 2 difference from average % 3.4% 3.4% 5.9% 3.2% 1.3% -0.2%
Sample 4 difference from average % -1.2% -1.1% -4.7% -0.3% -0.2% 11.2%
R Repeatibility of lab samples % 2.2% 1.7% 2.9% 1.6% 0.6% 5.6%
DGA monitor measurements and accuracy evaluation 2015.09.23

Monitor A reading STP (0ºC) ppm 768 39 0.6 7.8 0.2 86 1740 <500 4000
M Monitor A reading IEC (20ºC) ppm 824 41.9 0.6 8.4 0.2 92.3 1867 <500 4293
RRT detection limits_IEC (CIGRÉ Brochure 409) ppm 5 1 1 1 1 25 25 500 2000

Minimum value for evaluation ppm 10 2 2 2 2 50 50 1000 4000
D Difference from average corrected lab samples % 18% 14% -6% 4% 5% -37%
AC Maximum accuracy = (|D|-|R|-2) % 14% 10% 1% 0% 2% 30%
AC av Average max accuracy (without O2, N2) % 5%
49
App Table G.2 Example 2 of Accuracy Calculations Performed for Test No 1 of Table 6.214
DGA Lab 2 H2 CH4 C2H4 C2H6 C2H2 CO CO2 O2 N2
Lab measurements of Gas-in-Oil Standards (GIOS)

TRUE NORTH 500 2015.09.24
S_500 Prepared IEC (20ºC) ppm 532 559 543 547 540 557 573 16528 61068
Ls_500 Average ppm 541 581 552 550 555 582 580 17516 64535
Lab accuracy_500 % 1.7% 4.0% 1.6% 0.5% 2.8% 4.5% 1.1% 6.0% 5.7%
TRUE NORTH 100 2015.09.24
Ls_100 Average ppm 113.5 114.1 110.0 107.9 109.6 113.0 141.6 17664 65767
Lab accuracy_100 % 4.7% 7.3% 1.4% 0.5% 2.1% 7.4% 10.8% 2.2% 3.9%
TRUE NORTH 10 2015.09.24
Ls_10 Average ppm 14.7 11.4 10.8 10.6 10.7 11.7 47.1 17504 65234
Lab accuracy_10 % 24.5% 6.2% 0.3% -1.7% -0.6% 21.1% 37.0% 1.3% 2.5%
Lab measurements of transformer syringe samples 2015.09.23

L1 Sample 1 measured IEC (20º) ppm 696 38.4 1.0 9.7 0.1 90.6 1778 1027 8268
L1c Sample 1 corrected IEC (20º) ppm 685 35.8 1.0 9.9 0.1 84.3 1758 996 7950
A Average of corrected lab samples ppm 696 36.0 1.0 9.8 0.1 85.2 1768 1263 8165
Sample 1 difference from average % -1.6% -0.7% 2.6% 0.3% 0.0% -1.0% -0.6% -21.1% -2.6%
Sample 2 difference from average % -3.2% -4.3% -7.7% -13.2% 0.0% -4.1% -1.2% 32.9% -1.2%
Sample 2 difference from average % 1.8% 3.4% 2.6% 16.8% 0.0% 1.2% 0.7% -23.3% -22.0%
Sample 4 difference from average % 3.0% 1.6% 2.6% -3.9% 0.0% 3.9% 1.0% 11.6% 25.8%
R Repeatibility of lab samples % 2.4% 2.5% 3.8% 8.5% 0.0% 2.5% 0.9% 22.2% 12.9%
DGA monitor measurements and accuracy evaluation 2015.09.23

Monitor A reading STP (0ºC) ppm 768 39 0.6 7.8 0.2 86 1740 <500 4000
M Monitor A reading IEC (20ºC) ppm 824 41.9 0.6 8.4 0.2 92.3 1867 <500 4293
RRT detection limits_IEC (CIGRÉ Brochure 409) ppm 5 1 1 1 1 25 25 500 2000

Minimum value for evaluation ppm 10 2 2 2 2 50 50 1000 4000
D Difference from average corrected lab samples % 18% 16% -15% 8% 6% -47%
AC Maximum accuracy = (|D|-|R|-2) % 14% 12% 4% 4% 3% 33%
AC av Average max accuracy (without O2, N2) % 7%
For oxygen and nitrogen, average values for lab accuracy are taken from all True North (TN) standards,
because concentration of these gases in TN10, TN100 and TN500 are similar.
50
APPENDIX H. Diffusion of helium in oil

Two possible reasons for the unexpectedly low readings of hydrogen during Tests No 2 to 5 of Table
6.2 were investigated by the WG17:
1) Hydrogen had been leaking from the commercial gas-in-oil standard (True North, TN) samples used,
so laboratory results were not properly corrected. This was ruled out 6, however, since the TN samples
were received and used for the tests only one week after they had been prepared, well within their
validity period (see section0).
2) Contamination of oil by helium6,17 coming from one or more of the monitors installed on the test loop
during the one-year period, resulting in a false high hydrogen measurement by the DGA laboratory or
laboratories. Consistent with this possibility, the difference between laboratory values and monitor
readings for hydrogen was lower at the beginning of the tests (6 ppm), than at the end one year later
(20 ppm).
If for example10 ppm of helium was present in oil at the end of the tests, and the syringe samples were
analyzed using the common laboratory GC method with Argon carrier gas and TCD detectors, the
measured signal for hydrogen would actually be a combination of hydrogen and helium signals, leading
to reported hydrogen concentration values approximately 10 ppm above actual values.
Monitors, on the other hand, are not affected by helium in oil since they are using it as a carrier gas, or
they incorporate non-GC based hydrogen sensors, and would therefore have reported hydrogen
concentration values lower than those incorrectly reported by labs. This would explain the apparent
inaccuracy of hydrogen monitor readings for Tests No 2 to 5 in Table 6.2(marked with *).
Supporting further this possibility of contamination by helium is the observation that the difference
between lab results and monitor readings is indeed higher after one year of tests, for instance 15 ppm
rather than 6 ppm for unit A.
This possibility was finally confirmed by measurements made by WG members 24, 25, using a special
laboratory set-up allowing to separate and measure separately hydrogen and helium in oil. 10 ppm of
hydrogen was thus found to give the same GC peak as 15 ppm of helium.
In oil samples coming from transformers equipped during 5 years with a GC-based gas monitor using
helium as the carrier gas and provided by13, the contents of helium in oil were thus found to be between
24 and 31 ppm, or 5 ppm per year of helium, or 3 ppm per year equivalent of hydrogen. The helium
contamination rate in ppm per year may be expected to be larger for transformers of smaller volume.
Some GC-based gas monitors are designed to prevent helium from dissolving in the oil.
In the oil loop of Tests No 2 to 5, the helium content was probably higher since it was fully sealed,
possibly 6 to 10 ppm of equivalent hydrogen (hydrogen losses in the main tank of transformers has
been evaluated by WG D1/A2.47 as 7 ppm per year). So, at the end of Tests No 2 to 5, lab values for
hydrogen were more likely to be 39 to 43 ppm instead of 49 ppm, or 20% to 12% too high.
After correcting for the estimated presence of helium in oil at the end of Tests No 2 to 5, the accuracy
of gas monitors for hydrogen thus becomes similar to the accuracy for the other hydrocarbons in Table
6.2 (< 15%). However, since the content of helium in oil was not measured directly but estimated,
these corrected values of accuracy for hydrogen were not included in the average value indicated in
Table 6.2 for these tests.
To put things in perspective17, hydrogen concentrations below 100 ppm are generally not considered to
be an immediate concern, so a few tens of ppm of hydrogen interference are unlikely to cause problems
in most cases. On the other hand, this interference has been observed 4 to complicate DGA analysis on
new transformers with low hydrogen concentrations, by creating discrepancies between laboratory DGA
results and sensitive on-line monitors being used for early fault detection.
51
CIGRE
21, rue d'Artois
75008 Paris - FRANCE
© CIGRE
ISBN : 978-2-85873-485-6

DGA Monitoring Systems

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

DGA Monitoring Systems

Uploaded by

Copyright:

Available Formats

D1/A2

DGA monitoring systems

M.DUVAL, Convenor CA I.BOCSI, Secretary HU

This Technical Brochure describes:

2. Types of on-line gas monitors on the market ........................................................ 11

3. On-line gas monitors for on-load tap-changers .................................................... 19

4. Installation of on-line gas monitors........................................................................ 21

6. Evaluation of accuracy of on-line monitors ........................................................... 25

APPENDIX A. definitions, abreviations and symbols ..................................................... 31

APPENDIX B. links and references .................................................................................. 31

APPENDIX D. DGA monitors using Gas Chromatography (GC) measurement methods .

APPENDIX E. Gas-in-oil standards .................................................................................. 41

APPENDIX F. Procedure of WG D1/A2.47 for evaluating the accuracy of gas monitors

APPENDIX G. Examples of monitor accuracy calculations6,14 ....................................... 49

APPENDIX H. Diffusion of helium in oil ........................................................................... 51

Figures and Illustrations

Work on the first two items is reported in another TB of the WG [B1].

2. Types of on-line gas monitors on the market

Extraction Gas measurement Advantages Limitations

2.2 Types of on-line gas monitors versus faults possible to identify in

Table 2.2 Definition of the Types and Sub-Types of Faults or Stresses

Fault or Stress Definition

Application Type of Gases Faults Possible to Identify Faults not Possible to

2.3 Detection of hydrogen with types of monitors M1, M1* and M2 in

Table 2.4 Monitors Suitable for Different Types of Transformers[B2], [B4]

Type of monitor Gases Transformers

2.4 Detection of carbon monoxide with monitors of type M2 in

2.5 Fault identification in transformers with on-line gas monitors

2.6 Detection of fast-occurring faults in transformers with on-line gas

Figure 2.1 Sudden Change in Gas Formation and Type of Fault

2.7 Costs and benefits of various DGA monitoring techniques for

Table 2.5 Costs/Benefits of Various DGA Monitoring Techniques

DGA Monitoring Number of Avoided Avoided Costs

3. On-line gas monitors for on-load tap-changers

In general, the following faults can occur on on-load tap-changers:

Number of gases Faults to Identify

H2, CH4, C2H6, C2H4, - all 5 faults, also at early stage

4. Installation of on-line gas monitors

4.1 Installation on transformers

4.2 Installation on on-load tap-changers9

oil sampling loop

Figure 4.1 Oil Sampling Loop for On-Line Gas Monitors

5. Accuracy of DGA results coming from on-line

5.2 Accuracy of DGA results coming from transformers

5.3 effect of poor accuracy on DGA diagnosis

Figure 5.1 Effect of Poor Accuracy on Uncertainty of Diagnosis of Faults in Transformers[B14],[B15]

5.4 effect of poor accuracy of monitor readings on gassing trends

This is illustrated schematically in

6. Evaluation of accuracy of on-line monitors

Table 6.1 List of On-Line Gas Monitors Evaluated by WG D1/A2.47

Manufacturer Model Number Evaluated for

Table 6.2 Accuracy of On-Line Gas Monitors Tested by Members of WG D1/A2.47

APPENDIX A. definitions, abreviations and symbols

APPENDIX B. links and references

Fieldsend-Roxborough, 17 M. Duval, 18 M. Grisaru, 19 H.J. Knab, 20

APPENDIX C. DGA on-line gas monitors using

IR absorption of different gases

Main IR-based technologies used in online DGA monitors

Factors affecting gas analysis with IR

1. Contamination of optical components.

How to evaluate an IR-based online monitor in operation

Validation of IR-based DGA monitors in laboratory

APPENDIX D. DGA monitors using Gas

Advantages of GC methods for DGA