Characteristics And Development Stage Recognition Of Air-gap Discharge Within Oil-impregnated Paper Insulation Considering Effect Of Cavity Size

Oil-paper insulation is commonly used in the power transformer. It would be gradually damaged by partial discharge(PD) which is induced by some small defects. The cavity within impregnated insulation pressboard is an extremely common defect type. Cavities could come into being due to some reasons that incomplete impregnation of multiple pressboards during the production of power transformer, thermal expansion and contraction of the impregnant in the pressboards, shaking and small distortion of the winding induced by short-circuit current and so on. The size of cavities generated in different conditions could not be the same, but right now the effect of cavity size on PD activity has not been emphasized yet, the research on characteristics of PD in large cavity is relatively scarce. Besides, the development of air-gap discharge is a long-term dynamic process. The period of insulation degradation varies a lot depending on the field in the cavity. Therefore, carrying out study on air-gap PD characteristics of different size’s cavities and PD severity diagnosis is of great significance to the safe operation of power transformer. This thesis consists of following contents:① Study the effect of cavity height and diameter on PD inception characteristics based on five type cavity configuration. Calculate the mean value and standard deviation of inception voltage, inception field, PD magnitude and inception phase of different cavity configuration. Experiment results show that within the selected cavity size range, comparing to small cavity PD, large cavity PD possessed lower inception field, higher charge magnitude and higher inception phase. Consider that lower surface electron emission rate and lower reverse field in the large cavity are the main reason of higher inception phase.② Carry out accelerated deterioration experiments on large and small cavity model to analyze PD development characteristics. As experiments went on, total PD magnitude of both model presented first increasing, then decreasing, at last dramatically rising variation trend. At first time, the discharge pulses of small cavity model mainly concentrated on 40?~95?and 220?~275? of power frequency cycle, then discharge phase gradually expanded and crossed 0?and 180?。The discharge pulses of large cavity model firstly concentrated on 70?~90?and 250?~270?, the phase also gradually expanded, but when air-gap PD comes to the last stage, due to lower reverse field, positive PD in large cavity can not expand to the negative half cycle and vice versa. 27 dimensional statistical parameters from PRPD pattern are carried out dimension reduction through KPCA. KPCA kept 96.1% of original feature information in the first six principle components. Same PD development clustering results are obtained by Hierarchical clustering and K means clustering method. PD development stage for large and small cavity model are both divided into three stages, initial discharge stage, weak discharge stage and pre-breakdown stage.③ Two ensemble learning algorithm, Random Forest and Adaboost are introduced to PD diagnosis. Established seven classifier model based on WEKA. Take each three development stages of large and small cavity mode as classifier output to detailed Air-gap PD severity diagnosis results. Compared to three conventional classifier, RBFNN, BKSVM and CART, four ensemble learning classifier RF, Ada-RBFNN, Ada-CART and Ada-BKSVM show higher recognition accuracy. After parameter optimization, the accuracy of Ada-BKSVM and RF can reach 96.37% and 94.35% respectively.