DWT and SVM Approach based Incipient Fault Detection Methods for Underground Distribution System

This paper presents a supervised machine-learning technique for incipient fault identification in an underground cable. The nature of the incipient fault is aberrant so it is required to classify incipient faults from non incipient faults. The simulation work has been carried out in the PSCAD platform and several cases such as incipient fault, non incipient faults like switching transients are simulated. The wavelet-based technique is used for feature extraction. Both balanced as well as unbalanced datasets have been tested with several Support vector machine techniques, where fine support vector machine outperforms others.