High Impedance Fault Location of Submarine Cable in Seafloor Observation Network Based on Improved LeNet-5

Abstract

The insulation layer of submarine cables in the seafloor observation network can be easily damaged due to the influence of marine environment and human factors, resulting in high impedance faults of submarine cables. However, it has been difficult to distinguish and locate high impedance faults of submarine cables because of the weak transient current signal created by the faults. The existing methods for high impedance fault location fail to effectively utilize the relation between the transient current signal and the fault distance, hence the fault location performance is relatively unsatisfactory. In this paper, we propose a high impedance fault location method based on improved LeNet-5. Firstly, we analyze the relation between the mutation features in the transient current signal and the fault distance. Then we employ the wavelet transform to mine the mutation features in the transient current signal. Finally, improved LeNet-5 is constructed to exploit the mined mutation features to predict the fault location. Experimental results demonstrate that the fault location performance of improved LeNet-5 is better than the existing high impedance fault location methods.