Single-phase Earth Fault Detection Using Deep Learning In Distribution Systems

• Main Authors: Xiao-Dan Zeng, 曾曉丹
• Other Authors: Duan-Yu Chen
• Format: Others
• Language: en_US
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/45376206202502557862

碩士 === 元智大學 === 電機工程學系 === 105 === Feature extraction for fault signals is critical and difficult in all kinds of fault detection schemes. A novel simple and effective algorithm of faulty feeder detection in distribution systems using a continuous wavelet transform (CWT) technique and convolutional neural network (CNN) algorithm is presented in this paper. Firstly, the transient signal in the frequency band with fault feature is extracted by CWT and transforms it into a gray scale image, which is used as an image sample to input into the CNN. Then the fault features are extracted adaptively by CNN. Finally, the formation of CNN classifier is conducted by utilizing training samples. As a comparison, using discrete wavelet packet transform extracts the transient signal in the frequency band with fault feature. Based on the statistical data, feature vectors for fault classification are calculated. Then the corresponding Adaboost and support vector machine (SVM) classifiers are trained by the feature vectors. A simulation model is established in PSCAD/EMTDC, and the training of classification system is conducted by utilizing simulating samples. Verification results of the testing cases reveal that it is no need to construct and extract features artificially for the method of faulty feeder detection in distribution system based on CNN. And the proposed algorithm is more excellent than another method in terms of robustness and adaptability.