A method of fault location based on graph convolution neural network in optical transport

As the deployment scale of optical transport networks continues to expand, fault location which is an important function of ensuring the healthy operation of optical transport networks, becomes more and more important in optical network operation and maintenance. However, the expansion of the scale of the optical transport network and the deepening of the degree of heterogeneity have resulted in a large amount of fault data during the operation and maintenance of the optical transport networks. The traditional fault location technology lacks effective processing of a large amount of fault data and cannot meet the needs of the intelligent optical transport networks. In recent years, neural network algorithms have continued to develop, among which graph neural networks are particularly brilliant in processing graph-structured data. Using the reasoning ability of graph neural network, this paper proposes a fault location algorithm for optical transport network based on graph convolutional neural network. Taking the network nodes in the optical transport network as the data nodes in the graph convolution, and the fault data of the network nodes as the feature vector, the graph convolutional neural network aggregates the feature information adjacent nodes for each node. Through iteration, each node saves the feature information of each node to different degrees, so as to obtain the local structural features between nodes and the fault features of different nodes. The algorithm proposed in this paper has strong robustness in the case of network topology changes, that is, the algorithm can still adapt to the network when adding or deleting network nodes. The simulation results show that the fault location accuracy rate of the proposed algorithm can reach more than 95%, and the fault can be quickly located, and the location duration is about several milliseconds.