基于支持向量机的致密油藏水平井体积压裂初期产能预测  

Support vector machine-based initial productivity prediction for SRV of horizontal wells in tight oil reservoirs

作　　者：陈浩[1] 张超[1,2] 徐程浩 王智林 李芳芳 尚云志 张苏杰[6] 李旋 CHEN Hao;ZHANG Chao;XU Chenghao;WANG Zhilin;LI Fangfang;SHANG Yunzhi;ZHANG Sujie;LI Xuan(College of Safety and Ocean Engineering, China University of Petroleum, Beijing 102249, China;Research Institute of Exploration and Development, Liaohe Oilfield Company, PetroChina, Panjin, Liaoning 124011, China;Research Institute of Petroleum Exploration and Development, Sinopec Jiangsu Oilfield Company, Yangzhou, Jiangsu 225009,China;CNOOC Energy Development Co., Ltd. Engineering Technology Branch, Tianjin, 300452, China;Research Institute of Exploration and Development of Daqing Oilfield Company Ltd., Daqing, Heilongjiang 163712, China;Bohai Drilling Engineering Co., Ltd., Downhole Technical Service Company, Tianjin 300283, China)

机构地区：[1]中国石油大学(北京)安全与海洋工程学院,北京102249 [2]中国石油辽河油田分公司勘探开发研究院,辽宁盘锦124011 [3]中国石化江苏油田分公司勘探开发研究院,江苏扬州225009 [4]中海油能源发展股份有限公司工程技术分公司,天津300452 [5]大庆油田有限责任公司勘探开发研究院,黑龙江大庆163712 [6]渤海钻探工程有限公司井下技术服务公司,天津300283

出　　处：《中国海上油气》2022年第1期102-109,共8页China Offshore Oil and Gas

基　　金：北京市自然科学基金“致密油藏储层物性及开发动态的超声波声学参数响应机制研究(编号:3173044)”;国家自然科学基金“致密油储层基质渗吸规律及原油动用机理研究(编号:51574257)”;国家973项目“提高致密油储层采收率机理与方法研究(编号:ZX20170138)”部分研究成果。

摘　　要：水平井产能预测和评价是致密油藏合理有效开发的关键环节。由于体积压裂后,缝网分布复杂且相互干扰,传统经验公式法和数学解析法在低渗-致密油藏的产能预测普遍误差较大。本文基于大庆油田M2区块20口水平井体积压裂的钻遇、压裂和试油数据,首先采用皮尔森系数、斯皮尔曼系数和肯德尔系数进行了主控因素分析,筛选和评价了7个主要参数。在此基础上,通过样本数据的标准化、主成分分析预处理和交叉验证方法优选,建立了基于支持向量机的致密油藏水平井体积压裂初期产能预测模型。结果表明,开发初期井数较少时,支持向量机留P交叉验证的预测效果最好,平均相对误差仅为5.4%。随着后续生产开采井数增加,建议优选精度较好、运算速度更快的10折法。根据不同的开发阶段的数据量情况,随时调整预测模型,为目标区块不同开发阶段水平井体积压裂现场施工参数优化和产能评价提供重要参考。The productivity prediction and evaluation of horizontal wells is a key element in the reasonable and effective development of tight oil reservoirs.Due to the complex distribution and mutual interference of fracture network after stimulated reservoir volume(SRV),the traditional empirical formula method and mathematical analysis method generally have large errors in the productivity prediction of low-permeability tight oil reservoirs.Based on the drilling encountering,fracturing and oil testing data of 20 horizontal wells with SRV stimulation in Block M2 of Daqing Oilfield,this paper firstly analyzes the main controlling factors using Pearson coefficient,Spearman coefficient and Kendall coefficient,and selects and evaluates 7 main parameters.On this basis,through the standardization of sample data,analysis preprocessing of principal components and the optimization of cross-validation methods,a support vector machine-based initial productivity prediction model is established for SRV of horizontal wells in tight oil reservoirs.The results show that when the number of wells is small in the early stage of development,the prediction effect of SVM leaving P cross-validation is the best,and the average relative error is only 5.4%.As the number of subsequent production wells increases,it is recommended to adopt the 10-fold method with better accuracy and faster calculation speed.According to the data volume in different development stages,the prediction model can be adjusted at any time to provide important references for the optimization of on-site operation parameters and productivity evaluation of SRV horizontal wells in different development stages of the target block.

关 键 词：致密油藏 压裂水平井产能预测 支持向量机 交叉验证法 皮尔森系数 

分 类 号：TE122.14[石油与天然气工程—油气勘探]