Automated evaluation of prpd patterns for on-line pd monitoring of stator winding

Summary

In this paper, the use of specific tools that enable the condition assessment of stator insulation in rotating electrical machines is described. The signal distinction between noise and PD defects is performed by means of a synchronous multi-channel evaluation technique. An automated PD defect classification system, which identifies the type of stator winding insulation defect and its location, has been developed and implemented in the continuous monitoring system described in this paper. This monitoring system is configured for extended data evaluation. PD source identification is performed either on a regular basis or when triggered by an alarm. In such a case, the automated separation of PD sources is performed by a highly efficient hierarchical density-based clustering algorithm. For each source, the system automatically identifies the phase of signal origin and indicates the phase-resolved PD patterns related to this cluster as noise or PD. The automated PD defect classification uses knowledge-based analysis designed as decision tree. It provides a deterministic decision for clear cases. For unclear cases, where the PD signal can be assigned to more than one PD defect classes, an additional pattern recognition procedure is applied which gives a probabilistic decision. The validation and evaluation of PRPD patterns has been confirmed as successful by different case studies from real PD monitoring system installations.

Additional informations

Authors

A. BELKOV, U. BRONIECKI, L.V. BADICU, B. GORGAN, O. KRAUSE

Keywords

stator winding, partial discharge measurements, PRPD pattern, defect classification, risk assessment