Life-cycle assessment of high-voltage assets using statistical tool

Nowadays, utilities are confronted with assets reaching or even exceeding their designed life. This in turn, implies the occurrence of upcoming replacements to assure reliable network operation. In this thesis, the application of statistical tools for life-time and residual life assessment of high-voltage assets installed in a power network is presented. More specifically, future failure behaviour is estimated by statistical distribution fitting to the life-data of different assets types. Further focus is put on the obstacles hindering the interpretation of the life-data. The results presented in this thesis are continuation of the cooperation between the Dutch utility Liander and Delft University of Technology, investigating the applicability of statistical tools as a mean to support the replacement and spare parts policies for component populations being in operation. For the latter, assessment of the life-time, as accurate as possible is necessary. This in turn, is based on the limited, incomplete and imperfect data. In order to provide systematic methodology, the thesis works out the limits of the statistical tools applicability, for the condition assessment of populations of high-voltage assets. Particularly, the exceptional failure occurrence, influence of operational condition and possible discontinuities of the failure rate function are of interest. Based on the study-cases, ways to overcome hurdles met during statistical analysis are presented. First of all, based on the real examples of failure data of high-voltage assets, the robustness of the statistical analysis for incomplete life and imperfect data containing singularities is elaborated. Second of all, the fact of varying load of the components population and its influence on the process of statistical analysis is investigated and the results are put forward. In the end, the validity of the statistical analysis is discussed. In particular, the continuous behaviour of the failure rate function is interesting. It has been shown that due to gradual load increase over time, the sudden break in failure rate function and in the failure prediction is not expected to occur. The above mentioned study-cases have been used to determine the extent to which the statistical tools, in particular Weibull curve fitting methodology, can be applied to the life-time assessment of power network components. In particular, the derivation and extrapolation of the time-dependent failure rate function together with its confidence limits is possible and of relevance for the management of the assets and further decision making. Thus, the gap between the statistical tools used to analyse the imperfect data of the populations and decision making processes with respect to these populations, is closed by the study-cases presentation. It has been shown that the statistical analysis is a robust methodology for the asset life-time assessment. More specifically, it can provide answers to the questions with respect to the replacement and spare parts policies of the high-voltage assets installed in power network.Electrical Sustainable EnergyElectrical Engineering, Mathematics and Computer Scienc