天然气管网大数据分析方法及发展建议

Approach of big data analysis and suggestions on development of natural gas pipeline network

天然气管网数据具有海量、多源异构、类型多样、价值密度差异大等特点,数据分析与应用难度大。借鉴电网、供应链、互联网等领域大数据研究成果,提出了基于数据处理、数据挖掘、多元数据综合分析的天然气管网大数据分析方法框架。从数据清洗、特征筛选、特征重构3个方面阐述了管网数据处理的方法与功能;结合具体业务和场景,明确了预测预警、模式识别、规则学习与推理是构成管网数据挖掘方法的基础;讨论了运用多元化管网数据的综合方法,指出发展多模态学习和联邦学习是打破管网数据壁垒、形成数据智慧的关键路径。天然气管网大数据分析应该不断完善管网的大数据"生态"、深入研究融合领域知识的机器学习方法,发展跨领域、可解释、可控制的管网大数据分析方法体系,为智慧管网技术提供理论基础。

Due to the characteristics,such as massiveness,multi-source,diversity and great difference of value density,of the data of natural gas pipeline network,it is of great difficulty to do data analysis and application.With reference to the study results of big data in power grid,supply chain and internet,a big data analysis framework of natural gas pipeline network based on data processing,data mining and comprehensive analysis of multivariate data is put forward.Definitely,the method and function of data processing is illustrated in terms of data cleaning,feature selection and reconstruction.Based on the specific business and scenarios,it is defined that prediction and early warning,model identification,rule learning and deduction are the basis to construct the data mining method of pipeline network.In addition,the comprehensive application of diversified data of pipeline network is discussed.It is also pointed out that the development of multimode learning and federated learning is the key to break the data barrier and to form data intelligence of pipeline network.Through big data analysis of natural gas pipeline network,the big data"ecology"of pipeline network shall be improved continuously,the machine learning method of knowledge in data fusion shall be deeply researched,and a cross-border interpretable and controllable big data analysis method system of pipeline network shall be established,so as to provide theoretical support for development of intelligent pipeline network technologies.