抽油机井原油产量计量新方法  

A Novel Method of Production Measurement of Sucker-Rod Pumping Wells

作　　者：张战敏 朱丹丹 刘重伯 朱丽萍[2] 闵贵芝 王璇 ZHANG Zhanmin;ZHU Dandan;LIU Zhongbo;ZHU Liping;MIN Guizhi;WANG Xuan(Engineering Technology Research Institute,PetroChina Huabei Oilfield Company,Renqiu,Hebei 062552,China;School of Information Science and Engineering,China University of Petroleum(Beijing),Beijing 102249,China)

机构地区：[1]中国石油华北油田公司工程技术研究院 [2]中国石油大学(北京)信息科学与工程学院

出　　处：《钻采工艺》2023年第2期71-76,共6页Drilling & Production Technology

基　　金：中国石油华北油田公司重大科研项目课题“抽油机井电参数计算产量与工况诊断分析技术研究”(编号:2020-HB-C02)。

摘　　要：产量计量是油田开发生产过程中最重要的基础工作之一,准确的计量结果对于抽油机井工作状态诊断、工作制度调整以及措施方案效果评价都具有重要意义。但传统两相分离式量油和功图量油技术普遍存在投资和维护成本高、自动化程度低、适应性差等问题。为了提高量油效率和精度,利用深度学习技术,以生产大数据样本做支撑,建立电参量油模型,提出了由有功功率数据直接计算产量的新方法。电参量油模型包含特征提取模块和参数融合模块,分别由双层GRU网络和三层全连接网络组成。特征提取模块获取有功功率曲线数据中的生产参数信息,参数融合模块实现了冲程、冲次和泵径特征与有功功率数据挖掘结果的信息融合,提高了电参量油模型的准确性和泛化性。实例井验证结果表明,油井工况越好,实际产量越高,预测产量的精度越高,平均预测精度为94.08%。该研究成果为油井计量开创了新思路,为低成本智能油田建设提供了技术支撑。Production measurement is one of the most important basic works in the process of oilfield development and production.The accurate measurement results of sucker-rod pumping wells are of great significance for the diagnosis of working conditions,the adjustment of working system and the effectiveness evaluation of production measures.While there are many problems in the traditional two-phase separation metering technology and dynamometer diagram metering technology,such as high cost of investment and maintenance,low degree of automation,and poor adaptability under different conditions.In order to improve the efficiency and accuracy of production measurement,this paper uses the deep learning technology and large data samples of production as the support to establish the production prediction model and introduces a new method to directly calculate the production by active power data.The model includes feature extraction module and parameter fusion module,which are composed of double-layer GRU network and three-layer fully connected network respectively.The feature extraction module extracts production parameters from active power curve,and the parameter fusion module realizes the information fusion of stroke,frequency of stroke and pump diameter and active power data mining results,which improves the accuracy of the model based on electric parameters and the generalization ability under different working conditions.The verification results of the example wells show that the better the working conditions,the higher the production and the higher the prediction accuracy.The average prediction accuracy reached 94.08%.The research of this paper gives a new idea for production measurement and provides new insights into low-cost construction of the digital oilfield.

关 键 词：抽油机井 产量计量 有功功率曲线 深度学习 GRU网络 

分 类 号：TE863.1[石油与天然气工程—油气储运工程]