The Application of Support Vector Machine Classifier to Commercial Customer Electricity Theft

• Main Authors: Yuen-Tse Cheng, 鄭淵澤
• Other Authors: Dr. Ming-Yuan Cho
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/71035570493855375880

碩士 === 國立高雄應用科技大學 === 電機工程系 === 98 === This thesis proposes support vector machine based pattern recognition technique to classify commercial customer electricity theft by establishing and comparing with the various rational load patterns. First, in this thesis, Taiwan commercial customers’ historical electricity data is collected to derive the summer and non-summer reasonable power consumption model. Moreover, the SVM network model is employed to train the selected commercial customer data set to establish the commercial customer electricity theft classifier and then the electricity theft electricity KWH can be derived by analyzing and recognizing historical data in database. Besides, the man machine interface of server and database design which contains logical schema and physical schema as well as the data transformation service program are developed. In this thesis, data transformation service technique is employed to extract, transfer, and load data from customer information system to proposed SQL server database. Finally, testing data covering the Taipower business district is selected for computer simulation to demonstrate the practicality and effectiveness of the proposed method.