Ground fault location research based on multidimensional scaling and density clustering algorithm

Abstract

When a single line-to-ground fault occurs in a current distribution network containing distributed power sources, fault location is primarily performed by current-based directional protection. This implies using the switching information at the nodes; however, this does not sufficiently maximize the fault information. To address these issues, a single-phase grounding fault location methodology is planned for distribution networks supported by a combination of three-dimensional scaling (MDS) and density-based spatial clustering of applications with noise (DBSCAN). First, the network description matrix is made to support the network topology. Second, the zero-sequence and negative-sequence currents collected at the terminals square measure pre-processed to obtain two-dimensional fault data feature vectors. Third, the fault feature vectors square measure is subjected to DBSCAN bunch analysis to obtain the fault zones of the distribution network, thus compensating for the disadvantage that DBSCAN cannot efficiently replicate high-dimensional knowledge. The simulation results reveal that the planned methodology is not tormented by fault location, fault resistance, fault angle, or line sort; therefore, the fault location results in a correct square measure and might effectively utilize the fault data.