协同过滤算法设计抽油机井举升系统  

Design of collaborative filtering algorithm for pumping well lifting system

作　　者：陈林 王相[1] 张雷 张中慧 肖姝 CHEN Lin;WANG Xiang;ZHANG Lei;ZHANG Zhonghui;XIAO Shu(Changzhou University School of Oil and Gas Engineering,Changzhou 213000,Jiangsu,China;SINOPEC Shengli Oilfield Branch Petroleum Engineering Technology Research Institute,Dongying 257015,Shangdong,China)

机构地区：[1]常州大学石油与天然气工程学院 [2]中国石化胜利油田分公司石油工程技术研究院

出　　处：《石油钻采工艺》2023年第3期319-324,共6页Oil Drilling & Production Technology

基　　金：国家自然科学基金“基于大协同过滤深度融合学习的抽油机井复杂工况诊断方法研究”(编号:52204027);中国石化科技攻关项目“大数据技术在油田开发中的应用研究”(编号:P20071);江苏省研究生科研与实践创新计划项目“基于协同过滤的抽油机井举升方案智能推荐方法研究”(编号:KYCX23_3149)。

摘　　要：传统基于采油工程理论的抽油机举升系统设计方法难以有效处理复杂矿场实际情况,设计方案的可靠性有待提升。建立涵盖稠油、低渗、复杂断块等多个油藏类型的数据库,应用协同过滤推荐技术从数据库的抽油机井举升系统设计方案中探索规律,辅助优化设计,提升抽油机井举升系统的效益。通过对搜集的3万余套历史举升方案相关数据规范化处理,得到涵盖油井地质、流体、生产等维度的抽油机井举升系统设计样本库。在此基础上,分析基于用户的协同过滤推荐系统的典型架构,建立了面向抽油机井举升系统设计的推荐算法,能够根据待设计井的地质开发特征,从数据库的历史样本中匹配得到地质开发条件相似度高且运行效果良好的举升设计方案进行推荐。分析15口井的实例,协同过滤举升系统设计的平均泵效提升7.84%,百米吨液耗电下降24%,推荐方案相比于当前方案均有显著效果提升。研究成果为抽油机井举升方案设计提供了新的思路和方法,为油田大数据应用提供了有益借鉴和参考。The traditional design method of pumping unit lifting system based on oil extraction engineering theory is difficult to effectively handle the actual situation of complex mines,and the reliability of the design scheme needs to be improved.Establish a database covering multiple reservoir types such as heavy oil,low permeability,and complex fault blocks,and apply collaborative filtering recommendation technology to explore patterns in the design scheme of pumping well lifting systems in the database,assist in optimizing design,and improve the efficiency of pumping well lifting systems.By standardizing the collected data of over 30000 sets of historical lifting schemes,a sample library for the design of pumping well lifting systems covering dimensions such as oil well geology,fluid,and production was obtained.On this basis,a typical architecture of a user based collaborative filtering recommendation system was analyzed,and a recommendation algorithm for the design of pumping well lifting systems was established.Based on the geological development characteristics of the wells to be designed,a lifting design scheme with high similarity in geological development conditions and good operational performance was matched from historical samples in the database for recommendation.Analyzing the examples of 15 wells,the average pump efficiency of the collaborative filtering lifting system design has increased by 7.84%,and the power consumption per 100 meters of liquid has decreased by 24%.The recommended plan has significantly improved compared to the current plan.The research provides new ideas and methods for the design of pumping well lifting schemes,and provides beneficial references and insights for the application of big data in oil fields.

关 键 词：勘探开发 工程技术 举升系统 协同过滤推荐 模糊综合评价 

分 类 号：TE355[石油与天然气工程—油气田开发工程] TP18[自动化与计算机技术—控制理论与控制工程]