A Study of Insulation Status Monitoring and Diagnostic System for High Voltage Underground Cable Joints

• Main Authors: Chien-kuo Chang, 張建國
• Other Authors: Ruay-Nan Wu
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/50029303882786288521

博士 === 國立臺灣科技大學 === 電機工程系 === 100 === In recent years, to improve the failure of the insulation of underground cable joints becomes the important issue. The test of partial discharges (PDs) was adopted for high voltage apparatus by many companies. The maintenance and diagnosis of cable joint is getting instant and difficult due to the mass of joints in the underground cable systems. It would be expensive with labor and expense to carry out by the recent partial discharge analyzer. This study proposed a long-term placement of PD monitoring unit at cable joints. It is composed of peak detection circuit and DSP (Digital Signal Processor), for acquiring and converting PD signals of sensors into discharge sequence data, and providing the data to computer for the data storage, analysis and display. Three PD data of underground cable joints were utilized in this paper. In the feature extraction, the discharge region of phase resolved analysis in the reference was interpreted and the improvement was provided. In addition, many discharge quantities were adopted in this paper, such as fractal dimension, discharge phase region, and statistic value. These features still are feasible in the different place. The data process was divided into two parts, off-line and real-tine modes. In the off-time mode, the establishment of the diagnostic rules is performed by the previous data. In the real-time mode, the measurement of partial discharge signal and the diagnosis by the rules are carried out automatically. The insulation status diagnostic rules are the hidden relation between the features which were discovered by the decision tree method. The class of data was divided by feature turning analysis and outlier analysis. The cut condition was calculated by Gini index method. The information gain was utilized to choose the attribute of the node in the developing of decision tree. The pre-pruning method was used to reduce the complexity of the tree. In the result, both case 1 and 2 have higher success ratio than 80%, excepted lower in case 1. The success ratio of in the midterm and final term are follows, case 1 is 88% and 100%, case 2 is 87 and 88%, case 83% and 84%. Hence, the insulation status diagnosis made by several rules provides sufficient result. These decision rules are suitable for partial discharge monitoring unit (PDMU).