Fault Location of Cabled Seafloor Observation Network Based on Fault Feature and Fault Distance Relation Mining Using Deep Learning

Cabled seafloor observation networks (CSONs) have evolved over the last decade to provide abundant electric energy and data transmission to enable marine science. To ensure reliable operation of the CSONs, the technology for fault location with high speed and accuracy is essential. However, the existing methods for CSON fault location fail to deeply mine the relation between fault feature and fault distance, hence the location performance is relatively unsatisfactory. In this paper, we prove that there is relation between fault distance and transient current signal created by fault. Based on this, we propose a fault location method based on Fault feature and Fault distance Relation Mining, named FaRM. Firstly, we simulate CSON and acquire transient current signal by PSCAD. Then we employ variational mode decomposition (VMD) to mine the fault feature in the transient current signal. Finally, FaRM is proposed to exploit the mined fault feature to predict the fault distance. Experimental results demonstrate that the fault location performance of FaRM is better than the existing CSON fault location methods.