Root Cause Analysis of Failures for Power Communication Network Based on CNN

Abstract

As the topology of power communication networks becomes more and more complex, the root cause analysis of network failures becomes more and more difficult. Traditionally, it relied on manual analysis of alarm data from the network management system by experts’ experiences to locate the root cause of fault. The accuracy and efficiency in large scale network were hard to guarantee. In this paper, a fault root cause analysis method based on convolutional neural network is proposed. The alarm coding matrix is designed to characterize the network topology and node state, as the labeled input data of convolutional neural network to extract the features of network failures and classify the faults with different features, to achieve the root cause analysis model of network failures. The performance evaluation shows that the proposed fault root cause analysis method has a high accuracy.