基于大数据和神经网络的油气管道打孔盗油(气)事件预测方法研究  

Research on Prediction Method of Oil and Gas Stolen by Drilling Holes on Oil and Gas Pipeline Based on Big Data and Neural Network

作　　者：韩永波[1] 张鹏宇 田娜[1] Han Yongbo;Zhang Pengyu;Tian Na(Petro-CyberWorks Information Technology Co.,Ltd.,Beijing 100007;Capital University of Economics and Business,Beijing 100026)

机构地区：[1]石化盈科信息技术有限责任公司,北京100007 [2]首都经济贸易大学,北京100026

出　　处：《中外能源》2020年第8期41-44,共4页Sino-Global Energy

摘　　要：近年来,油气管道上存在打孔盗油(气)事件给长输管道企业带来巨大困扰。管道一旦遭受破坏,将造成巨大的财产损失和环境污染。目前国内一些专家学者已经开展管线打孔盗油(气)的预测及预防研究,利用简化、假设以及支持向量机技术建立预测模型,但尚没有将事件所有相关数据有效利用起来。基于相关性分析、BP神经网络的油气管道打孔盗油(气)事件预测模型,充分考虑打孔盗油(气)事件的不确定性、影响因素多、分布广、难于控制、破坏性大等特点,通过引入相关性分析原理,分析预测模型中涵盖事件相关因素,减少神经网络中不必要的输入变量,提高预测模型训练效率。在信息时代的当下,随着数据采集手段的不断进步及网络共享领域范围的不断扩大,管道打孔盗油(气)方面积累的数据量将越来越巨大,充分利用大数据进行全面学习,可以避免数据简化造成信息的丢失,模型预测计算的准确性、密集性也将相应提高,因此基于大数据的打孔盗油(气)预测模型的应用将具有广阔前景。In recent years,oil and gas theft incident through drilling holes on oil and gas pipelines has brought great trouble to long-distance pipeline enterprises.Once the pipeline is damaged,it will cause huge loss of property and environmental pollution.At present some experts in China has studied the prediction of oil and gas stolen by drilling holes on the oil and gas pipelines.Simplifying,hypothesis and technology of support vector machine are applied to the establishment of prediction model,but all related data are not used effectively during the whole incident.Oil(gas)theft incident through drilling holes on oil and gas pipelines has the characteristics of uncertainty,many influencing factors,wide distribution,being difficult to control and being destructive.Taking full advantage of large data and introducing correlation analysis and BP neural net-work,the prediction model is established.By analyzing event-related factors,the unnecessary input variables in neural network are reduced,and the training efficiency of prediction model is improved.At current information age,the data of oil and gas theft incident will more and more along with the continuous progress of data collection methods and continuous expansion of network sharing.Full usage of big data for comprehensive learning can avoid data loss resulted from data simplifying,and accuracy and denseness of prediction calculation of the model may be improved.The prediction model of oil and gas theft incident based on big data will have wide application prospect.

关 键 词：油气管道 打孔盗油(气) 相关性分析 BP神经网络 预测模型 

分 类 号：TE89[石油与天然气工程—油气储运工程]