| Natural gas pipeline network is an important infrastructure for urban development and construction.With the continuous expansion of the coverage of pipeline facilities,the density and complexity of underground pipeline network are increasing,which greatly increases the threat of natural disasters and third-party damage to pipelines.This study systematically analyses the risk of natural disasters and third-party damage to pipelines,establishes the safety level evaluation index system and further constructs the early warning system platform,which is of great significance to improve the safety level of natural gas pipeline network operation.This paper takes the construction of third-party natural gas pipeline damage and natural disaster monitoring and warning system based on large data as the goal,and the main research contents are as follows:（1）Construct early warning indicator system.Through the analysis of literature,the safety level factors of third-party pipeline destruction and natural disasters are analyzed from five aspects:personnel,equipment,management,social environment and natural environment.Quantitative analysis of safety level factors,construction of early warning index system,using ANP and Entropy Weighting method to determine index weight.（2）Put forward comprehensive safety level index of pipeline.Classification of safety level is made from two aspects of third-party damage and natural disasters.Combined with index weight obtained,comprehensive safety level indexof natural gas pipeline is put forward.L calculation method;According tovalue,combined with safety level classification table,can accurately judge third-party damage of natural gas pipelines and safety level of natural disasters.（3）Model training and algorithm comparison.First,preprocess the third-party damage and natural disaster index data;Then the experimental environment is established by using distributed system infrastructure Hadoop and distributed memory iteration calculation framework Spark to filter Spark ML classification algorithm.According to the experimental results,the third-party damage and natural disaster prediction models trained by random forest algorithm and decision tree algorithm have excellent comprehensive performance index,while the calculation efficiency of random forest algorithm is slightly better.Considering the performance of the algorithm and the coordination of the system,the forecasting model trained by random forest algorithm is selected as the model used in the early warning system.（4）Design the overall structure of the system.According to the forecasting model and combining the technologies of Vue,Spark and Web GIS,the system is constructed as a whole according to the system application requirements,function module design and system development.Taking a pipeline in Zhejiang Province as an example,the functions of integrity management,safety level analysis and emergency management are realized to provide data and technical support for safety management and emergency decision-making of pipeline network. |
