Study On Foundation Of Fingerprint Database And Pattern Recognition Technology Regarding PDs Of Stator Windings For HV Motors

The content of this paper is an important part of "Mine ventilation and power system security condition monitoring and fault diagnosis of early warning systems"(No:2007BAK29B05). It is based on the current tough situation such as the continuously happened fault in the insulation of the high voltage motor, poor insulation monitoring reliability and the imperfect assessment method.With the development of the modern industry, the number of motors used in the production system is rising with higher single unit capacity. The aspect of large HV motors is more and more extensive, and the function of HV motors is getting much more important. The fault of HV motors would affect the motor themselves as well as the production system and the safety of workers, which could result in economic losses and negative effects. Approximately40%of the faults in HV motors lies in the insulation degradation, and the reliability of the insulation condition depend largely on the stator windings.Different degradation types correspond to different PD signal which has its unique fingerprint. By recognizing the PD signal from online monitoring of the insulation of HV motors, the insulation defects can be detected and the aging degree of insulation could be judged. Therefore, the diagnosis and early warning for potential insulation fault could be performed and the service life of HV motors could be improved, which has very important practical significance.In this thesis, an adjustable width mockup stator core was developed to serve as the grounded electrode used for PD collection. The characteristics of insulation structure of HV motors were concluded and the aspects leading to insulation failure were analyzed. The original, influences and relative fault repairing methods of internal PD, slot PD, corona PD and bar-to-bar PD from HV motors were determined.Based on the PD mechanism of HV motors, the PD model of internal PD, slot PD, corona PD and bar-to-bar PD from HV motors were set up combined with insulation fault investigation from the industrial field.On completion of PD models of internal PD, slot PD, corona PD and bar-to-bar PD from HV motors, the PD fingerprint were collected under relative aging experiment, and the distribution of PD fingerprint was given. The PD fingerprint database was found up.The characteristic value and its relationship of various PD occurred in the ground wall insulation of HV motors was developed. The characteristic values were selected as the basis for distinguishing PD pattern types and insulation fault types. BP neural network classifier was adopted to realize the intelligent recognition of PD types.