特种油气藏 ›› 2022, Vol. 29 ›› Issue (4): 128-134.DOI: 10.3969/j.issn.1006-6535.2022.04.018

• 油藏工程 • 上一篇    下一篇

低渗油藏注入水构成岭回归量化评价方法

郑宪宝1,2, 王宏伟1,2, 苗志国1,2, 李美芳1,2   

  1. 1.中国石油大庆油田有限责任公司,黑龙江 大庆 163712;
    2.黑龙江省油层物理与渗流力学重点实验室,黑龙江 大庆 163712
  • 收稿日期:2021-07-13 修回日期:2022-03-30 出版日期:2022-08-25 发布日期:2023-01-09
  • 作者简介:郑宪宝(1978—),男,高级工程师,2003年毕业于大庆石油学院石油工程专业,2006年毕业于该校油气田开发工程专业,获硕士学位,现从事低渗透油田水驱开发及提高采收率方面的研究工作。
  • 基金资助:
    国家科技重大专项“低渗、特低渗复杂油藏规模有效动用关键技术”(2017ZX05013-006)

Quantitative Evaluation Method with Ridge Regression for Injected Water Composition in Low-permeability Reservoirs

Zheng Xianbao1,2, Wang Hongwei1,2, Miao Zhiguo1,2, Li Meifang1,2   

  1. 1. PetroChina Daqing Oilfield Co.,Ltd., Daqing, Heilongjiang 163712, China;
    2. Heilongjiang Key Laboratory of Reservoir Physics and Percolation Mechanics, Daqing, Heilongjiang 163712, China
  • Received:2021-07-13 Revised:2022-03-30 Online:2022-08-25 Published:2023-01-09

摘要: 针对大庆油田朝5区块葡萄花油层存在大量无效注水,常规数值模拟计算效率低下、评价流程复杂的问题,分析了注入水构成,建立了全地层地质模型,首先利用数值模拟方法量化了注入水的构成,然后基于影响注入水的地质开发因素,通过岭回归方法构建了高注采比区块注水构成量化评价模型,并将模型预测结果与数值模拟计算的结果进行了对比验证,证明岭回归构建的模型较为可靠。研究结果表明:目前朝5区块储层吸水占比为55%,未划储层段吸水为28%,泥页岩吸水为8%,外溢水量为12%,建立的注水构成量化评价模型能准确表征朝5区块的注入水构成,岭回归模型相比数值模拟可以在计算精度相差不大的情况下对评价效率有极大的提升。研究成果对于提高油田注水开发水平,保证油田开发具有一定的指导意义。

关键词: 低渗油藏, 高注采比, 注水构成量化评价, 无效注水, 机器学习, 大庆油田

Abstract: Aiming at the problems of ineffective water injection in Putaohua reservoir, Block Chao 5, Daqing Oilfield, low calculation efficiency of conventional numerical simulation and complicated evaluation procedure, the composition of injected water was analyzed and a full stratigraphic geological model was established. Firstly, the composition of injected water was quantified with numerical simulation method, then a quantitative evaluation model for injected water composition in blocks with high injection-to-production ratio was constructed by ridge regression method on the basis of geological development factors affecting the injected water, and the prediction results of the model were compared with the results of numerical simulation, which proved that the model constructed by ridge regression was more reliable. The results of the study showed that, the water absorption ratio was 55% in the reservoir in Block Chao 5, 28% in the non-zoned reservoir interval and 8% in mud shale, and the water overflow was 12%; the established quantitative evaluation model for injected water composition could accurately characterize the injected water composition in Block Chao 5; compared with the numerical simulation, the ridge regression model could greatly improve the evaluation efficiency with little difference in calculation accuracy. There is much for reference of the study results for improving the water-flooding development effect of oilfields and guaranteeing oilfield development.

Key words: low-permeability reservoir, high injection-production ratio, quantitative evaluation of water injection composition, ineffective water injection, machine learning, Daqing Oilfield

中图分类号: