Statistical Analysis of Partial Discharge Patterns and Knowledge Extraction in MV Cable Systems

Decisions on maintaining, repairing, refurbishing and replacing cable systems are usually a compromise between what technically is advised and what economically is feasible. To implement an optimal maintenance and/or replacement strategy, asset managers and/or maintenance engineers need appropriate tools to provide them with reliable information on the condition of the cable systems. Partial discharge analysis has a long record as reliable diagnostic tool to assess the integrity and the quality of electrical insulation of power systems. For assessing the condition of MV cable systems, it often is the only source of information. Proper interpretation of the observed discharge patterns, from e.g. on-line PD monitoring, can be used to reveal the discharge source i.e. defect type as well as the physical phenomena behind it. In contrary with a deterministic approach, probabilistic analyses appear to be more appropriate to model the discharge behavior initiated from unknown sources. Analyses of various PD patterns such as discharge height distribution and discharge density distribution as a function of time may provide characteristic statistical parameters to determine the discharge source as well as the degradation process. Research shows that the 2-parameter Weibull distribution provides a potential model to quantify the characteristics of the observed patterns. This paper presents the application of such statistical modeling to the area of on-line power cable diagnostics. Data obtained from laboratory experiments as well as field data have been studied. Acquired knowledge will be implemented to assess the condition of the cable networks.