부산대학교 도서관

The ageing process of power transformers is a complex mechanism, which means condition and health assessment of these critical assets is a challenging task. While transformers are in-service, obtaining paper samples for direct analysis cannot easily be achieved, meaning the only practical way of assessing the state of degradation of the paper insulation is through prediction models. In recent years SP Energy Networks have made a significant investment in their ageing transformer fleet by systematically carrying out post-mortems on all power transformers as they are removed and retired from service. These post-mortems allow comparison between the estimated degree of polymerisation (DP), that was calculated from oil sample results and the real DP measured during the Post-mortem. Predicted DP and residual life is one of the main criteria for condition assessment, and ultimately the decision to retire a transformer. However, analysis of the retired transformer paper samples typically indicated that the actual DP results vary and are inconsistent with the estimated DP values, and that in the majority of cases transformers were being retired from service with significant residual life left in the solid insulation. The purpose of this research is to determine the best prediction model for estimating DP and from this to create a new more precise model for SPEN fleet condition assessment. The potential additional benefits that come with the creation of a new more accurate model to predicted DP include the deference of capital expenditure on new transformers, ability to maximize asset life and provide better value to customers and shareholders. For the purposes of this research, 38 transformers and 14 different models were used in total, each tested against the actual DP values and 2-FAL history found within the post-mortem reports provided by SPEN.