Research On Data-Driven Natural Gas Pipeline Network Security Awareness

As the advancement of carbon neutrality and emission peak,natural gas will be the transitional energy of China.Natural gas usage and pipeline mileage in China has increased rapidly nowadays.However,there are serious potential security menace in the long-term operation of gas network.Gas pipeline explosion accident is a frequent occurrence,that posing a great threat to people's properties and lives and causing enormous damage to ecological environment.Therefore,the industry world and academic circles have paid high attention to the study on risk assessment for natural gas pipeline.The main research contents of this thesis include:1.Combined with the temporal and spatial environment of natural gas pipeline network in Zhejiang Province,this thesis studies and constructs a set of security situation awareness index system of natural gas pipeline network.The subjective weight of natural gas pipe network index is calculated by analytic hierarchy process(AHP).On this basis,the entropy weight method is used to fuse the prior knowledge of several experts,and the correction coefficient of each index is determined.The comprehensive weight of each index is obtained by using the correction coefficient and subjective weight integration algorithm.The risk level of the evaluation object is obtained by fuzzy comprehensive evaluation,and the weighted average principle is used to replace the maximum membership principle to establish the comprehensive evaluation matrix,so as to remedy the deficiency of information loss caused by the maximum membership principle.Finally,a case study of a natural gas pipeline in Zhejiang Province is carried out,and the established model can well reflect the static risk situation of natural gas pipeline.2.In this thesis,the unmanned aerial vehicle(UAV)patrols the medium and the high risk areas determined by the static risk situation,and establishes the data set of UAV aerial intrusion images,which mainly includes the third-party intrusion images of excavators,bulldozers and other construction machinery.Aiming at the problems of small target to be detected and unbalanced sample in the dataset,the corresponding data are preprocessed,and small target amplification and Mosaic are used to enhance the data.Aiming at the difficulty of target detection in UAV aerial shooting third-party construction image,YOLOv4 algorithm is adopted as the benchmark network,and two improvements are made.First,k-means clustering is used to optimize the size of the prior frame,so that the size of the prior frame matches the size of the target to be detected.Second,the Recurrent Criss-Cross Attention module is introduced to further strengthen the ability of backbone network feature extraction.Experimental results show that the proposed algorithm can effectively identify third-party construction intrusion events around pipelines.Meanwhile,ablation experiment shows that the improved method can effectively improve the detection effect of third-party intrusion,and the average accuracy of network identification is 4.05% higher than that of original network identification.3.This thesis analyzes the functional requirements and business process of the natural gas pipeline network situation awareness system,adopts the mainstream front and back-end technology stacks such as Node.js,React,G2 Plot,My SQL and so on,designs and implements the data-driven natural gas pipeline network situation awareness system.The above algorithm model is applied to actual production through system development.The system runs smoothly online,has strong compatibility,meets the needs of daily use,and improves the level of pipeline protection management.