Professional Documents
Culture Documents
Elt 227 3
Elt 227 3
SC A2
STUDY COMMITEE REPORT
A2 Reliability Advisory Group This definition of failure is clearly very wide. The
failure definition used by WG 12.05 was more restrictive,
The SC A2 Reliability Advisory Group has comprised considering only problems requiring the unit to be
the following regular members: removed from service to be repaired.
John Lapworth (UK) Leader A further level of discrimination was provided by the
Sam Hall (UK) definitions of failure with forced outage, following IEEE
Peter Austin (Australia) [2]: ‘Failure of a transformer that requires its immediate
Greg Polovick (Canada) removal from service. This is accomplished either auto-
Luiz Cheim (Brazil) matically or as soon as switching operations can be per-
formed.’ (usually assumed to be within 30 minutes), and
The agreed objectives of the group were to: failure with scheduled outage and defect when the imper-
• survey existing reliability databases and initiatives fection or partial lack of performance can be corrected either
• investigate experiences and difficulties in service or during a normally scheduled outage, i.e. with-
• review applications of reliability data out taking the transformer out of service to correct this spe-
• agree outstanding needs for reliability data cific problem.
• identify how CIGRE SC A2 can contribute
Despite concentrating on the availability or other-
Discussions have been held during CIGRE SC A2 wise of the transformer for service and the seriousness
Committee meetings and at the following special events: of the problem, these definitions could still include
• CIGRE SC A2 Colloquium, Dublin, June 2001: Reli- some problems that might not be recognised as fail-
ability Workshop ures by everyone. For instance, should a transformer
oil leak be considered a failure or would it depend on
• Brazilian Regional CIGRE Conference, Florianop- the consequences ? Certainly if the oil leak resulted in
olis, November 2003: Workspot III a Buchholz alarm, the transformer would normally be
switched out of service as soon as possible for inves-
• Greece Regional CIGRE SC A2 Colloquium, tigations, and this could have a significant impact on
Athens, November 2004: Reliability Tutorial the network.
SC A2
STUDY COMMITEE REPORT
While the network operator is interested primarily considered a fault, but aged insulation resulting from
in the impact on the system, plant specialists tend to rank normal service loading would not. Any discharge activ-
failure in terms of what remedial action is required to ity inside the transformer would also be considered a
restore the equipment functionality. If a suspect trans- fault. A fault would normally only become apparent
former was replaced during a planned outage then this once it had developed to the point that it caused some
would not have an impact on system reliability, and the abnormal change in measured parameters e.g. increases
network operator would not register this as a failure, in dissolved gases.
whereas the asset owner clearly would.
Obviously, the existence of a fault is expected to
increase the probability of a failure, and many failures
are expected to occur as the result of the development of
a fault. However, according to the proposal here, and
contrary to the IEC definition, a fault can also occur with-
out a failure and vice versa (e.g. if a transformer failed as
a result of conditions for which it was not designed).
This 1:1 relationship between failure and fault implies Failure rate = number of failures / (number of trans-
that if there has been no failure then the equipment has formers x number of years)
no fault, which seems unnecessarily restrictive and does
not provide for the description of a deteriorated state However, such a figure can be very misleading, as the
which is expected to lead to a failure. A more intuitive following example illustrates.
definition which better reflects common usage of the
term and fits in with the basic transformer failure model Suppose all transformers in a population operated
proposed by CIGRE Working Group A2.18 (Life Man- perfectly until they failed after about 25 years, obviously
agement) [3] would be along the lines of: any damage or an exaggeration of common experience. The average
deterioration beyond normal wear or ageing likely to have failure rate as derived above would then be 4% p.a. How-
an impact on reliability. ever, this failure rate figure does not represent a reason-
able estimate of the probability of failure of the trans-
A localised hotspot caused by a design deficiency formers during their operating lives, as they were
resulting in excessive local insulation ageing would be perfectly reliable until they failed. If the probabil-
SC A2
STUDY COMMITEE REPORT
ity of failure at every point in their life (hazard rate) had or manufacturing weaknesses. With this more realistic
been assumed to be 4% p.a., then the estimated number failure model the key parameters to be determined are
of failures would have been as shown in the probability the age for onset of unreliability and the mean survival
distributions of Figure 1. time thereafter.
Figure 2: Failure distributions for an onset Clearly, the challenge is to find a simplified and
of unreliability at 20 years acceptable form of survey that still provides meaningful
results.
This is of course the familiar ‘bath tub’ curve. It is
usually assumed that the increasing probability of fail-
ure after a particular age is due to the effects of compo- Uses of reliability data
nent ageing, i.e. oil or paper ageing for transformers, but
for transformers a more likely cause of the onset of unre- It is considered that the main uses to which relia-
liability is probably damage caused by unusual system bility data will be put are:
events, e.g. short circuits, lightning strikes or switching • Check on new manufacturers and designs
transients, particularly when transformers have design • Optimise maintenance
SC A2
STUDY COMMITEE REPORT
• Benchmark against other utilities/service providers quality assurance measures employed in an attempt to
• Manage network reliability ‘build in’ quality and reliability, to identify best practice.
• Justify spares holding
• Estimate end of life This would appear to be a suitable area of activity for
• Manage risk CIGRE SC A2, building on existing initiatives regard-
ing Specification [4] and Design Review [5].
In consideration of this wide range of uses of relia-
bility data a number of general questions arise:
• Are someone else’s experiences relevant? Optimising maintenance
• Can one database serve all uses ?
• How detailed / complete must the data be? The interest here is in determining appropriate main-
• What is the minimum data required ? tenance standards which minimise costs for an accept-
• What about confidentiality ? able reliability, using techniques such as Reliability Cen-
• Which uses really require shared data ? tred Maintenance (RCM). This use will be equipment
• Would a common approach be enough ? and utility specific, possibly seeking to allocate resources
within a population, and will inevitably focus mainly on
the reliability of components.
Benchmarking
The interest here is on the reliability of new trans- This topic is obviously very complicated, involving
formers and is of increasing relevance today as a result business as well as technical issues, and not primarily
of structural changes in the industry: globalisation of concerned with reliability or plant performance, so this
manufacturers and restructuring of utilities, so that there is not considered to be a priority activity for CIGRE
is often no longer a close relationship between a local SC A2.
supplier and an informed customer: today transformer
procurement is increasingly competitive with pressures
to reduce first cost, often involving new foreign suppli- Managing network reliability
ers, and for customers with less experience, with a poten-
tial impact on reliability. The interest here is in obtaining authoritative world-
wide data on the reliability of various plant items so that
The intention would be to collect in-service reliabil- appropriate asset management decisions can be made to
ity data on new transformers, for the first 5 years or so, safeguard an acceptable system performance, e.g. replace-
to identify poor performers and therefore it would be ment of specific equipment or network re-inforcement.
necessary to identify manufacturer. However, the sur- Ideally, this would collect detailed information on all out-
vey should also take into account different practices ages, planned as well as unplanned, on all equipment,
regarding specifications, design reviews, factory test lev- and include minor problems as well as major failures,
els, factory inspections and witnessing and any special but in practice will probably have to make do with
SC A2
STUDY COMMITEE REPORT
The interest here is not so much in the likelihood [3] CIGRE Working Group A2.18,“Life Management
of experiencing a problem with a typical transformer, of Transformers”, CIGRE Technical Brochure No. 227,
but in assessing the likely consequences and risks involved 2003.
in continuing to operate a transformer which is suspected [4] CIGRE Working Group A2.15, “Guide for Cus-
of being faulty. As such, the information required relates tomers Specifications for Transformers 100 MVA and
specifically to faulty transformers, not to the general pop- 123 kV and above”, CIGRE Technical Brochure No. 156,
ulation and should aim to quantify the expected conse- 2000.
quences of particular symptoms, e.g. time to failure, costs
[5] CIGRE Working Group A2.21,“Design Reviews”,
of failure.
CIGRE Technical Brochure No. 204, 2002.
This activity is a natural activity for CIGRE SC A2 as [6] CIGRE Working Group A2.20, “Economics of
a logical follow-on from previous Working Groups, e.g. Transformer Management”, CIGRE Technical Brochure
A2.18 [3] and A2.20 [6]. No. 248, 2004. ■
www.ahlersheinel.de