The Application Study Of Data Mining Method In The State Evaluation Of Power Cable

The cables take up a large proportion in power supply line. The cables used in industry and the business sector in operation for some time will appear aging and hidden trouble for accident. It may cause significant economic losses if not taking timely measures. The cable insulation condition is usually can be divided into good, not good, bad and the failure. Assessing the cable status on the basis of the daily maintenance data, test data and on-line monitoring data has a momentous engineering value and practical significance.Data mining finds out the connotative, unknown, and potentially valuable knowledge and rules. In order to ensure the integrity of systems, easy integration, on the basis of developing high-precision cable status monitoring device, using wired and wireless communications and Internet technologies to collect data and using data mining tools to assess the state of the cable. The main work and conclusions are as follows:①The aging mechanism of power cable insulation system and its method for state detection were summarized. The sorts of data mining technology and its applications in power industry were introduced.②The on-line detection theory and its realizing way of additional low frequency signal was introduced. Because of the leakage current and medium loss tangent value Xanδwere representative, This paper adopted the low frequency signal superposition method, and used the foutier transform technique to calculate the low frequency leakage current and the medium loss tangent value, and the measurement datas were transmited to the data analyzing workstation by the SCADA system.③The data mining theory was used to finds out the isolated point and the center point. To the problem of data inconsistency and missing data, this paper adopted the regression technique of vacancy value filling method to improve the quality of data mining.④The neural network analysis method of on line time sequence was described. Firstly, the liner and sine model were used to forecast the trend of the time sequence data, then adopted the three-layer feed-forward neural network model, the relaxed optimization algorithm was adopted to make the discrete variable continuous based on the judgement of the number of hidden neurons, lastly according to the BP atithmetic the nerve cell data and the weight coefficient were trained together. ⑤The cable insulation state evaluation method based on the decision tree technology was discussed, the decision tree technology was used to judge the cable insulation, and abtained that the relevance between the cable on line and the off line experimental data was not very big, so the decision subtrees could be formed alone. To the subtree compose there must has the correlation attribute, which can translate to a public attribute. The cable on line monitoring datas were affected by the running environment so they had no right standard to judge the cable insulation state. The stae data sets which can be confirmed according to the experience or off line experiments must be selected when forming the classification rule.Through kinds of data of the actual cables, using SPSS software for actual application, the ultimate simulation results indicate that the application of data mining methods can accurately assess the state of the power cable.