典型聚类算法在区块抽油机井系统效率分析中的应用  

Application of typical clustering algorithms in analysis of system efficiency of pumping wells in blocks

作　　者：刘合[1] 卢秋羽[2] 朱世佳[1] 蒋薇 王素玲[2] Liu He;Lu Qiuyu;Zhu Shijia;Jiang Wei;Wang Suling(PetroChina Research Institute of Petroleum Ejcploration and Development,Beijing 100083,China;School of Mechanical Science and Engineering,Northeast Petroleum University,Heilongjiang Daqing 163318,China)

机构地区：[1]中国石油勘探开发研究院,北京100083 [2]东北石油大学机械科学与工程学院,黑龙江大庆163318

出　　处：《石油学报》2020年第12期1657-1664,共8页Acta Petrolei Sinica

基　　金：科技部创新方法工作专项项目(2018IM040100)资助。

摘　　要：因采油中—后期抽油机系统效率的影响因素多且数据庞杂无特征,造成系统效率调控效果差。基于大数据挖掘技术,以大庆油田中区的采油区块为研究对象,将地面数据和井下数据相结合,采用典型聚类算法揭示了系统效率的变化规律,针对区块数据,采用了k-means和DBSCAN聚类算法应用于油田数据分析,首先利用组间误差平方和、组内误差平方和与总误差平方和之比确定了最佳k值,用k-means算法对数据集进行聚类。然后通过设定不同M和ε值用DBSCAN算法对数据集进行聚类,通过将聚类结果可视化并对比两种聚类方法的不同之处发现,k-means算法更符合大庆油田中区数据的聚类分析结果,并以k=4时k-means聚类结果给出了区块低效率井的表现特征,为区块井系统效率调控提供方向指导。There are many factors affecting the system efficiency of pumping units in the middle and late stages of oil recovery and the data is complex and featureless,resulting in the poor regulation effect on system efficiency.Based on the big data mining technology,taking an oil production block in the central Daqing oilfield as the research object,the paper reveals the changing law of system efficiency by combining surface data and downhole data,and using typical clustering algorithms for block data,the clustering algorithms of k-means and DBSCAN are used to analyze oilfield data.First,the optimal k value is determined by the ratios of the inter-cluster and intra-cluster error square sum to the total error square sum,and the data set is clustered using the k-means algorithm.Then,the data set is further clustered using the DBSCAN algorithm by setting different Minpts andεvalues.By visualizing the clustering results,this paper compares the differences between the two clustering methods,and finds that the k-means algorithm is more consistent with the cluster analysis results of the central Daqing oilfield,and provides the representation characteristics of low-efficiency wells in the block according to the k-means clustering results when k=4,thus providing guidance for the regulation of system efficiency for wells in the block.

关 键 词：抽油机 系统效率 大数据 聚类分析 影响因素特征 

分 类 号：TE355[石油与天然气工程—油气田开发工程]