Advanced Analysis and Diagnostics for Remote Online PD Monitoring of HV Rotating Machines

Abstract

Partial discharge (PD) testing is considered one of the best techniques for assessing the quality and degradation of insulation systems. On-line Partial Discharge (OLPD) monitoring can detect flaws in insulation at an early stage and trend their development over time. In rotating machinery, OLPD monitoring presents challenges due to the complexity of the stator winding insulation and multiple defects can coexist simultaneously. Thus, PD pulses from various sources superimpose resulting in a complicated Phase Resolved PD (PRPD) pattern. Moreover, irrelevant pulses picked up by OLPD sensors can distort correct diagnosis and prevent accurate trending of PD values. These pulses could originate from on-site noise, cross-coupled PD activity from adjacent phases or high frequency triggers that are necessary for machine functioning. A technique employed that is particularly suited to motors in hazardous gas zones is remote monitoring in which sensors are located at the switchgear end of the machine circuit. While offering many benefits, remote monitoring introduces new challenges for signal de-noising and classification. These can be tackled effectively by novel advanced analysis tools presented in this paper. High-resolution visualization of all captured pulses based on several characteristic features highlight key information regarding the origin of the pulses. A novel set of non-dimensional parameters has been derived that are useful for segregating PD data, along with common features such as amplitude, phase and frequency. All this allows different PD types to be distinguished from each other as clusters. Similarly, PD clusters can be differentiated from noise interference and cross-coupled events. The data system also registers the time for each event which allows the clusters to be trended individually. Features of the PRPD patterns such as shape and polarity predominance are utilized to identify the specific defects of each phase. Expert analysts can then evaluate the overall condition of the stator winding and locate the defected area(s). Maximum utilization of the OLPD monitoring can be achieved by the capability to transfer and analyze the data regularly. Typically, the data obtained in the field is commercially sensitive. Network topology and cloud computing methods have been developed to automatically and securely transfer data from the substation to an analysis site anywhere in the world. In this paper, a variety of rotating machine PRPD patterns that were captured with on-line monitoring are presented via high-resolution visualization and advanced plotting tools. The monitoring technique for the acquisition of the PRPDs was remote monitoring.