04-11-2016, 09:22 AM
1464274528-DRAFTFINAL.pdf (Size: 1.26 MB / Downloads: 5)
ABSTRACT
Conventional practice for transformer dissolved-gas analysis (DGA) is to use
concentrations of several fault gases, with or without total dissolved combustible gas (TDCG),
for evaluating apparent fault severity. We suggest a simpler approach based on the normalized
energy intensity (NEI), a quantity related directly to fault energy dissipated within the
transformer. DGA fault severity scoring based on NEI is shown to be sensitive to all IEC fault
types and to be more responsive to shifts in the relative concentrations of the fault gases than
scoring based on fault gas concentrations. Instead of eight or more gas concentration limits,
NEI scoring requires only two or three limits that can be empirically derived to suit local
requirements for any population of mineral-oil-filled power transformers.
INTRODUCTION
Abnormal energy dissipation inside a power transformer results in partial destruction
of liquid and solid insulating materials. This process produces trace amount of gaseous by
products dissolved in the transformer oil. This fact forms the basis of the paper.
The analysis of these dissolved gases is called dissolved gas analysis or DGA. It
consists of collecting these samples, measuring the samples and interpreting it. The type and
severity of the fault can also be determined by this analysis.
Principle gases found in this analysis are hydrogen, methane, ethane, ethylene, and
acetylene. Oxides of carbon, nitrogen and oxygen are also present in the dissolved oil. The
IEEE and IEC have set certain guidelines for interpreting the observed limits. The IEC has set
two limits in DGA and classifies the gas concentrations accordingly. But the IEEE classifies
values into 4 levels with three limits. IEEE standards are followed here
But this method using dissolved gas concentration analysis was found to have certain
drawbacks and a new method using the concept of NEI (Normalized energy Intensity) was
introduced. This paper mainly deals with the steps involved in calculation of NEI, analysis of
results from NEI, its comparison with old method and the improvements in this new method
Hydrogen and carbon monoxide present in the oil tend to include errors in the analysis
because these gases are caused by non-fault related conditions as well. But a table indicating
CO and H2 concentrations is shown and it is clear that these gases tend to be dominant.
Another problem with TDCG as a fault severity indicator is that it gives methane and
ethane (associated with low and medium-range thermal faults) equal weightage with acetylene
and ethylene (associated with arcing and high range thermal faults). Hence the importance of
the new term NEI or Normalized Energy Intensity
NORMALIZED ENERGY INTENSITY (NEI)
Normalized energy intensity is found out from a set of calculations based on enthalpy
of formation of the respective dissolved gases. Enthalpy of formation of substance B from
substance A is the amount of energy required to produce one mole of B from A, or MB grams
of B, where MB is the molecular mass of substance B expressed in g/mole
Cellulose, the material with which most of the insulations are made of, decomposes at
even ambient temperatures, producing carbon monoxide, carbon dioxide, and water. So, carbon
monoxide and dioxide should be excluded during all calculations. Hydrogen is a fault gas but
can be produced by non-fault related processes like electrolysis of water and reaction of acidic
material with galvanized metal. It is also observed as source of stray gassing in hydro-refined
transformer oils
Excluding these gases from NEI calculation solves the first disadvantage of TDCG
discussed before to an extent. But since these gases form the majority in transformer oil, the
sensitivity of the new method is poorer. An improvement to sensitivity is advised later in the paper though.
CALCULATION OF NEI
Concentration of each HC gas in Μl/L is multiplied by (1L)/ (106Μl) and then by (1 mol)/
(22.4L) to convert the numerator to moles. Then, multiply by (103L/Kl) to convert the
denominator from L to Kl. The resultant value (mol/Kl) is multiplied by the enthalpy of
formation in (Kj/mol) to obtain the Kj/Kl for the gases.
After all these calculations, we arrive at an empirical formula as shown below where the name of each gases indicates their concentrations.
PRELIMINARY STATISTICAL ANALYSIS
DGA databases were contributed by two large US electric utilities, identified here as
Source A and Source B. Incomplete samples (with missing gas data) and known after-failure
samples were removed. One sample per year (the latest) was retained for each transformer, to
avoid bias from closely spaced investigative sampling. NEI was inserted into each sample
record. The composition of the combined database is indicated by Table.
RELATIVE UNCERTAINTIES IN NEI AND TDCG
Relative uncertainty is determined both directly and indirectly. Direct values are
tabulated and indirect values are found out from the gas concentrations and the formula for
NEI. High value for acetylene is probably due to consideration of low concentration levels.
Mean relative uncertainty in indirect way is 0.18 for NEI and 0.24 for TDCG. Comparing these
results with the values in the below table, we come to know that relative uncertainties of both
NEI and TDCG are not worse than those of the gas concentration measurements upon which they depend
STRATIFICATION BY N2/O2
An exploratory analysis of a large transformer DGA database was conducted for
IEEE/PES Transformers Committee Working Group C57.104 to investigate the effects of
transformer age, MVA rating, kV rating, and oxygen/air ratio on distributions of key gas
concentrations. The largest effect was associated with the proportion of dissolved oxygen vs.
nitrogen. Transformers with more oxygen in the oil tend to have lower concentrations of
dissolved combustible gases. The balance between oxygen and nitrogen in the transformer oil
depends on the oil preservation system of the transformer. Some preservation systems – and
transformers with leaky conservator diaphragms and gaskets – allow atmospheric oxygen to
diffuse into the transformer, while sealed or nitrogen regulated systems generally have very
low oxygen content in the oil relative to nitrogen. Unfortunately, the databases used for
statistical derivation of gas concentration limits for DGA sometimes have incomplete,
unreliable, or entirely missing information about the oil preservation types of the transformers.
In such cases it is useful to classify transformers as low-oxygen or high-oxygen, based on the available DGA data.
DGA LIMITS
IEEE-style gas concentration limits are conventionally based on 90th and 95th
percentile gas concentrations from a large database with post-failure samples removed. If a
four level severity classification is desired, a 98th or 99th percentile limit is added. For this
study, we derived DGA limits for low-oxygen and high-oxygen transformers separately, using
90th, 95th, and 98th percentile gas concentrations.
Tables VI, VII, and VIII show hydrocarbon gas and NEI DGA limits for the combined
database. Similar limits were calculated for the individual data sources A and B. The large
difference between the low-oxygen percentile limits and the high-oxygen ones shows that
separating the transformers into two oxygen level groups for DGA interpretation is worthwhile.
Because of the very low incidence of acetylene in normally operating electric utility
power transformers, the 90th and even 95th percentile acetylene concentrations in many
transformer populations are zero or very close to zero, making those percentiles unsuitable for
use as DGA limits. For the low-oxygen transformers in the combined database, the 90th, 95th,
and 98th percentile acetylene concentrations are 0, 1, and 6 µL/L, respectively. So, for the
purposes of this study, separate acetylene limits are used
DGA SCORING
Each transformer was classified as low-oxygen or high oxygen by comparing its median
N2/O2 with the N2/O2 limit for its respective data source. A set of hydrocarbon gas and NEI
limits was chosen based on the data source and the transformer’s high/low-oxygen
classification.
For each oil sample, each of the hydrocarbon gases (methane, ethane, ethylene, acetylene)
was given a score as follows, based on its reported concentration x and the limits L1,L2,L3 for
the gas according to the transformer’s assigned DGA limits.
ASSIGNMENT OF FAULT TYPE
IEEE-type numeric DGA scores are customarily whole numbers, but for better comparison
of scoring methods we include a fractional part in the score by interpolation. According to
common usage, a score below 2 is considered acceptable for normal operation (although, in
practice, the score is not all that is taken into consideration); a score of 2 or higher motivates
investigation, possible supplementary testing, and possible consideration of corrective action.
A score of at least 2 but less than 3 would usually be considered cautionary, at least 3 but less
than 4 would be reacted to more urgently, and a score of 4 or more would usually be treated as
an emergency or at least as a sign of advanced deterioration.
For samples where at least one of methane, ethylene, or acetylene had a concentration of at
least 10 µL/L, an apparent fault type was assigned, based on the Duval Triangle [11]. The
assignment of an apparent fault type was done for purposes of statistical comparison, not as a
judgment that any particular samples did or did not indicate a transformer fault.
Fault gases are often present in moderate amounts in trouble-free transformers as the
cumulative result of operational stress and incidents such as temporary overloading, hot
weather, and through faults. In those cases, the apparent fault type indicates the general nature
of the dominant process responsible for generating the residual gases found in the transformer.
See Table IX for the distributions of the apparent fault types for the combined database.
COMPARISON OF SCORING METHODS
For comparison of NEI scoring with HC Gas scoring, our emphasis is on the similarities
and differences of how the two methods rate the severity of generic fault types, given some
reasonable set of DGA limits to work from. We do not assert that the limits used here are
generically useful or that the DGA scores alone are sufficient for DGA interpretation in a
production setting. The following discussion is intended to show that NEI scoring, with an
appropriate choice of limits, is an adequate and effective alternative to HC Gas scoring or any
similar method of fault severity assessment based on gas concentrations. Comparison is
generally done on the basis of two factors-Sensitivity and Rigidity
RIGIDITY
Among samples with any particular HC Gas score, many of the NEI scores will be
higher and many lower (assuming that the alternative NEI limits are used). That is because in
almost all cases the HC Gas score is determined by the concentration of a single “key” gas,
while the NEI score is based on the concentrations of all the hydrocarbon gases in each case.
For example, consider the two hypothetical samples from a low-oxygen transformer shown in
Table XI. The NEI scores are based on the alternative limits in Table
VIII. Based on 100 ppm of ethylene, both samples receive an HC Gas score of 3.04. But Sample
A, with very low levels of hydrocarbon gases other than ethylene, has an NEI score of 2.06,
while Sample B, with much higher levels of non-ethylene hydrocarbon gases, has an NEI score
of 3.78. The NEI scoring method is able to distinguish between two samples that clearly
represent different levels of fault activity, while the HC Gas score, representing only the
maximum of four individual gas concentration scores, and makes no distinction.
Table XII identifies the predominant or “key” hydrocarbon gas or gases, according to the HC
Gas scoring method, for each fault type. The calculations for that table were based on all sample
records from the combined database where the HC Gas score is at least 2.0 and where a fault
type is given. For brevity let those be called “fault samples.” The table shows the number
(“Samples”) and percentage (“%Samples”) of fault samples of the indicated fault type where
the score of the indicated gas is equal to the HC Gas score of the sample. For each fault type,
the gases not listed are dominant in only a small minority of cases.
Thus, for example, for fault type T3, there are 5695 fault samples in which the ethylene score
is equal to the sample’s HC Gas score. Those 5695 records represent 89.7% of all the fault
samples having a fault type of T3.
In summary, for the HC Gas method, almost all of the PD scores are determined by methane;
for T1, almost all scores are determined by methane or ethane; for T2, ethane determines most
of the scores; for T3, ethylene almost always determines the score; and for DT and the arcing
faults, acetylene almost always determines the score.
CONCLUSION
For transformer DGA, the assessment of relative risk or fault severity is conventionally
based on consideration of a combination of combustible gas concentrations. A model of that
approach, based on concentrations in transformer oil of light hydrocarbon gases (the HC Gas
method) was compared with a proposed new approach, based on the fault energy required to
produce the observed hydrocarbon gas concentrations (NEI).
The HC Gas scoring method and similar schemes based on different assortments of gases
require two or three gas concentration limits per gas. The scores, based on gas concentrations
as proxies for fault severity, are not straightforwardly interpretable in terms of fault energy and
do not discriminate between different mixtures of gas produced by more or less severe or
prolonged faults. Because percentile-based gas concentration limits are usually unsuitable for
acetylene, gas concentration scoring methods are forced to use precautionary acetylene limits,
based on engineering judgment, which tend to rate arcing-type faults very differently than
others.
The Normalized Energy Intensity (NEI) DGA scoring method, based on hydrocarbon
fault gas concentrations and enthalpies of formation, provides a numeric fault severity score
that is directly related to the amount of fault energy expended in the oil, even for arcing-type
faults. Because NEI is based on all the hydrocarbon gas concentrations, not just on one at a
time, it responds better to gradual increases in fault severity. It is sensitive to each of the IEC
transformer fault types, and its overall sensitivity can be adjusted with predictable effect by
modifying the limits. For an IEEE-style scale of DGA scores from 1 to 4, only three NEI limits
are needed. The NEI is easily calculated and, with only two or three limits to consider, the
scoring logic is very simple.