Transformer life prediction using data from units removed from service and thermal modelling

Understanding how and when transformers are likely to fail is of critical importance to the asset management of large networks. Ideally a reliable end-of-life model would be established based on years of previous experience over several lifecycles of equipment of a type representative of that still...

Full description

Saved in:
Bibliographic Details
Main Authors: Paul, Jarman, Ruth, Hooton, Leon, Walker, Qi, Zhong, Ishak, Mohd Taufiq, Zhongdong, Wang
Format: Conference or Workshop Item
Published: 2010
Subjects:
Online Access:http://eprints.uthm.edu.my/1713/
http://eprints.uthm.edu.my/1713/1/A2_212_2010_%2D_Final.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Understanding how and when transformers are likely to fail is of critical importance to the asset management of large networks. Ideally a reliable end-of-life model would be established based on years of previous experience over several lifecycles of equipment of a type representative of that still in service. For large power transformers however establishing such a model is not easy because we are still in the first asset lifecycle and many of the transformers installed when the National Grid system was first developed are still in service well beyond their original design life. Maintaining the reliability of the network is clearly a priority, but this must be done efficiently and economically . Replacement plans must therefore be properly targeted both tactically, to replace those transformers in the worst condition and at the most critical sites, and strategically to make sure the replacement numbers are sustainable in terms of available resources such as outage capacity, supplier capability and investment funding. Given the expected strongly non-linear relationship between system reliability and plant reliability and the possibility of hidden failures in the system where plant in service has aged to a condition where it cannot withstand the power flows resulting from a system failure elsewhere, the best possible understanding of ageing and failure is required. This paper gives the historical failure rate of the transformers used on the National Grid system in the UK and shows that failures to date are random in nature and not statistically age related. This means that traditional approaches to building a statistical end-of-life model cannot be used. Analysis of the insulation of transformers removed from service for any reason indicates a very wide range of condition, some samples show severe thermal ageing and it is clear that age-related failures can be expected if replacement is not carried out, other samples show little ageing and for these transformers it appears that very long lifetimes might be expected if other ageing mechanisms do not become apparent. Transformer condition assessment techniques are now very good and provide a sound basis for replacing transformers before failure. A condition assessment scoring system based on anticipated remaining useful lifetime has been implemented and condition based replacement has made a major contribution to maintaining system reliability. The success of the condition based replacement process however has a profound biasing effect on the failure statistics as transformers are not left in service to fail, so that these statistics cannot be used directly for lifetime modelling.