A Machine Learning-Based Faulty Line Identification for Smart Distribution Network

This paper presents a machine learning-based faulty-line identification method in smart distribution networks. The proposed method utilizes post-fault root-mean-square (rms) values of voltages measured at the main substation and at selected nodes as well as fault information obtained by fault current identifiers (FCIs) and intelligent electronic re-closers (IE-CRs). The information from FCIs and IE-RCs are first used to identify the faulty region in the network. The normalized rms values of voltages are then utilized as the input to the support vector machine (SVM) classifiers to identify the faulty-line according to the pre-determined fault type. The IEEE 123-node distribution test system is simulated in ATP software. MATLAB is used to process the simulated transients and to apply the proposed method. The performance of the method is tested for different fault inception angles (FIA) and different fault resistances with satisfactory results.