en
×

分享给微信好友或者朋友圈

使用微信“扫一扫”功能。
作者简介:

卜亚辉(1985—),男,山东东营人,副研究员,博士,从事机器学习算法在油田开发中的应用、断块油田开发技术的研究工作。E-mail:b.bradley@163.com。

中图分类号:TE319

文献标识码:A

文章编号:1009-9603(2022)04-0135-08

DOI:10.13673/j.cnki.cn37-1359/te.202008033

参考文献 1
贾虎,邓力珲.基于流线聚类人工智能方法的水驱油藏流场识别[J].石油勘探与开发,2018,45(2):312-319.JIA Hu,DENG Lihui.Oil reservoir water flooding flowing area identification based on the method of streamline clustering artifi⁃ cial intelligence[J].Petroleum Exploration and Development,2018,45(2):312-319.
参考文献 2
李道伦,刘旭亮,查文舒,等.基于卷积神经网络的径向复合油藏自动试井解释方法[J].石油勘探与开发,2020,47(3):583-591.LI Daolun,LIU Xuliang,ZHA Wenshu,et al.Automatic well test interpretation based on convolutional neural network for a radial composite reservoir[J].Petroleum Exploration and Development,2020,47(3):583-591.
参考文献 3
ASKARI F R,KHAMEHCHI E.A novel approach to assist history matching using artificial intelligence[J].Chemical Engineering Communications,2015,202(4):513-519.
参考文献 4
LI Yu,ZHANG Tao,SUN Shuyu,et al.Accelerating flash calcula⁃ tion through deep learning methods[J].Journal of Computational Physics,2019,394:153-165.
参考文献 5
ZHANG Liming,WANG Saisai,ZHANG Kai,et al.Cooperative ar⁃ tificial bee colony algorithm with multiple populations for interval multi-objective optimization problems[J].IEEE Transactions on Fuzzy,2018,27(5):2872125.
参考文献 6
张凯,路然然,张黎明,等.基于序列二次规划算法的油藏动态配产配注优化[J].油气地质与采收率,2014,21(1):45-50.ZHANG Kai,LU Ranran,ZHANG Liming,et al.Proration optimi⁃ zation of production and injection rate based on sequential qua⁃ dratic programming[J].Petroleum Geology and Recovery Efficien⁃ cy,2014,21(1):45-50.
参考文献 7
刘巍,刘威,谷建伟,等.利用卡尔曼滤波和人工神经网络相结合的油藏井间连通性研究[J].油气地质与采收率,2020,27(2):118-124.LIU Wei,LIU Wei,GU Jianwei,et al.Research on interwell con⁃ nectivity of oil reservoirs based on Kalman filter and artificial neu⁃ ral network[J].Petroleum Geology and Recovery Efficiency,2020,27(2):118-124.
参考文献 8
贾德利,刘合,张吉群,等.大数据驱动下的老油田精细注水优化方法[J].石油勘探与开发,2020,47(3):629-636.JIA Deli,LIU He,ZHANG Jiqun,et al.Data-driven optimization for fine water injection in a mature oil field[J].Petroleum Explora⁃ tion and Development,2020,47(3):629-636.
参考文献 9
刘巍,刘威,谷建伟.基于机器学习方法的油井日产油量预测 [J].石油钻采工艺,2020,42(1):70-75.LIU Wei,LIU Wei,GU Jianwei.Oil production prediction based on a machine learning method[J].Oil Drilling & Production Tech⁃ nology,2020,42(1):70-75.
参考文献 10
孙致学,姜宝胜,肖康,等.基于新型集成学习算法的基岩潜山油藏储层裂缝开度预测算法[J].油气地质与采收率,2020,27(3):32-38.SUN Zhixue,JIANG Baosheng,XIAO Kang,et al.Prediction of fracture aperture in bedrock buried hill oil reservoir based on nov⁃ el ensemble learning algorithm[J].Petroleum Geology and Recov⁃ ery Efficiency,2020,27(3):32-38.
参考文献 11
侯春华.基于长短期记忆神经网络的油田新井产油量预测方法[J].油气地质与采收率,2019,26(3):105-110.HOU Chunhua.New well oil production forecast method based on long-term and short-term memory neural network[J].Petroleum Geology and Recovery Efficiency,2019,26(3):105-110.
参考文献 12
李兆亮,柳金城,王琳,等.基于神经网络的复杂储层流体分级识别[J].断块油气田,2020,27(4):498-500.LI Zhaoliang,LIU Jincheng,WANG Lin,et al.Fluid hierarchical identification of complex reservoir based on neural network meth⁃ od[J].Fault-Block Oil and Gas Field,2020,27(4):498-500.
参考文献 13
马文礼,李治平,孙玉平,等.基于机器学习的页岩气产能非确定性预测方法研究[J].特种油气藏,2019,26(2):101-105.MA Wenli,LI Zhiping,SUN Yuping,et al.Non-deterministic shale gas productivity forecast based on machine learning[J].Spe⁃ cial Oil & Gas Reservoirs,2019,26(2):101-105.
参考文献 14
王端平.对胜利油区提高原油采收率潜力及转变开发方式的思考[J].油气地质与采收率,2014,21(4):1-4.WANG Duanping.Some thoughts about potential of oil recovery ef⁃ ficiency and development model transition in Shengli district[J].Petroleum Geology and Recovery Efficiency,2014,21(4):1-4.
参考文献 15
王森,冯其红,宋玉龙,等.基于吸水剖面资料的优势通道分类方法——以孤东油田为例[J].油气地质与采收率,2013,20(5):99-102.WANG Sen,FENG Qihong,SONG Yulong,et al.Preferential flow path classification method based on injection profile data-taking Gudong oilfield as an example[J].Petroleum Geology and Recov⁃ ery Efficiency,2013,20(5):99-102.
参考文献 16
朱光普,姚军,张磊,等.特高含水期剩余油分布及形成机理 [J].科学通报,2017,62(22):2 553-2 563.ZHU Guangpu,YAO Jun,ZHANG Lei,et al.Pore-scale investiga⁃ tion of residual oil distributions and formation mechanisms at the extra-high water-cut stage[J].Chinese Science Bulletin,2017,62(22):2 553-2 563.
参考文献 17
李振泉,郭长春,王军,等.特高含水期油藏剩余油分布新认识——以孤岛油田中一区 Ng 3 砂组为例[J].油气地质与采收率,2019,26(6):19-27.LI Zhenquan,GUO Changchun,WANG Jun,et al.New under⁃ standing of remaining oil distribution in oil reservoirs at extrahigh water-cut stage:A case of Upper Ng3 sand group in Zhongyi area,Gudao Oilfield[J].Petroleum Geology and Recovery Effi⁃ ciency,2019,26(6):19-27.
参考文献 18
陶光辉,李洪生,刘斌.特高含水期驱替倍数量化表征及调整对策[J].油气地质与采收率,2019,26(3):129-134.TAO Guanghui,LI Hongsheng,LIU Bin.Quantitative characteriza⁃ tion of displacement multiple and adjustment countermeasures in ultra-high water cut stage[J].Petroleum Geology and Recovery Efficiency,2019,26(3):129-134.
参考文献 19
李阳,杨勇.老油田绿色低成本开发探索与实践[J].油气地质与采收率,2019,26(2):1-6.LI Yang,YANG Yong.Exploration and practice of green low-cost development in old oilfields[J].Petroleum Geology and Recovery Efficiency,2019,26(2):1-6.
参考文献 20
崔传智,李松,杨勇,等.特高含水期油藏平面分区调控方法 [J].石油学报,2018,39(10):1 155-1 161.CUI Chuanzhi,LI Song,YANG Yong,et al.Planar zoning regula⁃ tion and control method of reservoir at ultra-high water cut stage [J].Acta Petrolei Sinica,2018,39(10):1 155-1 161.
参考文献 21
刘丹.井间剩余油富集区识别方法研究[D].西安:西安石油大学,2014.LIU Dan.The recognition method on residual oil of enrichment re⁃ gion in well[D].Xi’an:Xi’an Shiyou University,2014.
参考文献 22
邱锡鹏.神经网络与深度学习[M].北京:机械工业出版社,2020.QIU Xipeng.Neural networks and deep learning[M].Beijing:Chi⁃ na Machine Press,2020.
参考文献 23
杨煜,张炜.TensorFlow平台上基于LSTM神经网络的人体动作分类[J].智能计算机与应用,2017,7(5):41-45.YANG Yu,ZHANG Wei.Human action classification based on LSTM neural network on TensorFlow[J].Intelligent Computer and Applications,2017,7(5):41-45.
参考文献 24
章敏敏,徐和平,王晓洁,等.谷歌TensorFlow机器学习框架及应用[J].微型机与应用,2017,36(10):58-60.ZHANG Minmin,XU Heping,WANG Xiaojie,et al.Application of Google TensorFlow machine learning framework[J].Microcom⁃ puter & Its Application,2017,36(10):58-60.
参考文献 25
李梦洁,董峦.基于PyTorch的机器翻译算法的实现[J].计算机技术与发展,2018,28(10):160-163,167.LI Mengjie,DONG Luan.Implementation of machine translation algorithm based on PyTorch[J].Computer Technology and Devel⁃ opment,2018,28(10):160-163,167.
目录contents

    摘要

    根据高含水油田剩余油分布特点,提出了含油饱和度等值线拟合样本预制作方法及基于人工神经网络的剩余油预测方法。利用数值模拟批量生成不同井距、物性、工作制度等条件下注采井组间剩余油分布场,编写模块自动提取少量含油饱和度等值线,使用多项式函数拟合不同时刻和层位含油饱和度等值线建立拟合参数样本数据集,实现机器学习样本参数量的大幅度降低。使用Tensorflow搭建神经网络模型,学习训练后形成注采井组间含油饱和度等值线预测模型,根据多个井组间等值线图叠加结果重构研究区含油饱和度场。基于高含水油田实际数据,与数值模拟相比,该方法对断层边界、层间剩余油富集区、井间局部零散剩余油均具一定预测能力;同时可将注采井生产动态数据快速转化为含油饱和度场数据,较传统方法的计算速度和定量化程度显著提高。

    Abstract

    According to the distribution characteristics of remaining oil in high water cut oilfields,the pre-making method of fitting samples of oil saturation isolines and the remaining oil prediction method based on the artificial neural network are proposed. This paper applies the numerical simulation method to generate the remaining oil distribution fields between in- jection and production well groups under different well spacings,physical properties,working systems,and other conditions in batches. It programs a module to automatically extract a small amount of oil saturation isolines and constructs fitting pa- rameter sample data set by polynomial functions to fit the oil saturation isolines at different times and horizon. This method can reduce the sample parameters of machine learning. Tensorflow is adopted to construct the neural network model. After the learning and training process,the oil saturation isoline prediction model between injection and production well groups is formed. The oil saturation field is reconstructed according to the superposition results of isoline maps between multiple well groups. Comparison between the actual data of high water cut oilfield and numerical simulation shows that the new method has prediction ability for fault boundary,interlayer residual oil enrichment area,and local scattered remaining oil between wells. This method can quickly convert dynamic data of oil and water wells into saturation field data. Compared with the traditional method,the proposed method significantly improves the calculation speed and quantification.

  • 中国东部老油田已普遍进入高含水开发阶段,生产历史复杂,剩余油分布零散,开发效益逐年变差。在低油价及成本上涨的双重压力下,亟待建立更加准确、高效的油藏生产数据分析方法。大数据及人工智能技术的出现可为传统方法的升级提供全新解决途径。目前智能学习算法在油藏工程领域的研究正处于快速发展阶段,研究方向主要包括:流场聚类算法识别水驱油藏流场[1],卷积神经网络自动试井解释[2],智能算法加速数值模拟历史拟合[3] 及闪蒸分离相平衡计算[4],使用代理模型数值模拟历史拟合及方案优化[5-6],通过机器学习量化评价多井分层注水效果、分析注水调整方向[7-8],预测产量递减及裂缝发育[9-11],储层流体分级识别[12],页岩油产能预测[13] 等。

  • 将机器学习应用于剩余油预测仍处于起步研究阶段,面临的挑战包括:第一,影响剩余油分布的参数众多,如何筛选提取特征参数;第二,如何用更少的参数描述剩余油空间分布特征。为此,笔者提出了基于数值模拟正演的样本构建方法,首先从数值模拟结果中提取含油饱和度等值线进行拟合,然后通过人工神经网络(ANN)建立拟合参数随时间和空间的变化关系,最后根据拟合参数重构含油饱和度场。该方法可快速评价剩余油潜力区,为开发调整政策的制定提供参考。

  • 1 方法概述

  • 高含水油田虽然采出程度较高,但仍蕴藏较大开发潜力,取心及矿场动态资料显示,剩余油呈整体分散、局部富集的特点[14-16],有效驱替仅发生在注采井组间局部范围,注采主流线之外依然存在剩余油富集区。矿场实践表明,改变注采方向、注采强度等流场调控手段可实现高含水油田采收率的进一步提高[17-21]。针对高含水期剩余油及其开发特点,设计了基于机器学习的剩余油预测方法(图1),具体步骤包括:①方案设计及模拟计算。影响注采井组间剩余油分布的主要因素包括井距、厚度、渗透率、相对渗透率和地层倾角等,设计多组概念模型调用数值模拟器获得不同时刻及层位的含油饱和度场。②含油饱和度场提取及拟合。对每个时间步及模拟层含油饱和度作等值线,提取等值线对应数据点,选用多种曲线方程拟合,记录每种拟合方法的参数及拟合精度。③样本库构建及学习训练。将各影响因素及对应拟合参数关联建立样本库,通过神经网络模型拟合训练样本,形成含油饱和度等值线预测模型。④注采井组划分及开发指标标定。根据开发历史及动态数据,对各开发阶段每个井组间的流动方向、采液量和含水率进行标定。⑤注采井组间含油饱和度预测。调用训练好的模型,预测各注采井组间含油饱和度等值线拟合参数。⑥含油饱和度场叠加及验证。将多阶段多注采井组含油饱和度场叠加,通过插值获得整个研究区预测结果。将预测结果与矿场动态进行对比,当误差较大时,调整参数重新计算;当误差较小时,将验证后的参数值样本保存入库,随着应用区块的积累逐步建立矿场样本库。

  • 图1 基于机器学习的剩余油预测方法技术路线

  • Fig.1 Roadmap of remaining oil prediction method based on machine learning

  • 2 注采井组间含油饱和度表征

  • 高含水期注采井组间含油饱和度场特征的表征是该方法的关键,需要满足形态保真且数据量小的要求。油藏工程及矿场实践表明高含水期注采井组间含油饱和度分布形态受储层物性、注水井、采油井位置及工作制度等参数影响,通过数值模拟正演可建立各种剩余油模式。针对中-高渗透均质概念模型,设置模型尺寸为1 000 m×1 000 m×10 m,网格数为 50×50×5,渗透率为 300 mD。在一注一采条件下,采油井的含水率分别为 10% 和 80% 时,纵向上选取模型顶部(第 1 层)、中部(第 3 层)、底部 (第5层)3个位置进行数值模拟。由图2可知,含油饱和度形态呈近似的曲线带,随着含水率升高,水驱波及范围不断扩大。其中蓝色代表强驱替区,红色代表未波及区,蓝色至红色之间代表过渡区。过渡区分布范围虽较小,但含油饱和度变化最快。此外受重力作用影响,储层底部较顶部水驱波及范围更大。

  • 图2 含油饱和度数值模拟结果

  • Fig.2 Numerical simulation results of oil saturation

  • 含油饱和度场图虽然能够全面表征其空间特点,但数据量较大,且以整个场数据为样本进行学习训练难度较大。为此,可提取少量含油饱和度等值线,降低数据量及学习难度。编写模块自动读取各时刻开发指标与含油饱和度场数据,绘制并提取指定含油饱和度等值线,获取每条等值线对应的点集坐标。在含油饱和度为 0.5 的条件下,绘制模型第5层的不同时刻含油饱和度等值线(图3),可以看出当初始时刻的含水率为 2% 时,注入水波及范围较小,受模拟网格影响含油饱和度等值线近似四边形;当含水率为5%时,含油饱和度等值线呈水舌形向采油井推进;当含水率为 30%~70% 时,含油饱和度等值线逐渐过渡到椭圆形,随着含水率升高,椭圆长短轴的差异不断缩小;当含水率为 90% 时,含油饱和度等值线接近圆形。

  • 根据以上分析,含油饱和度随时间变化规律可用一组特征曲线表示,由此设计采用多项式方法进行拟合。在注采井距为 400 m、油藏厚度为 5 m、垂向位置为 0.5 m、采液量为 104 m3 的条件下,获得 3 条含油饱和度等值线(图4a)。以含油饱和度为0.45 的等值线为例,其是由若干等值点组成的(图4b),以注水井为坐标原点(0,0),定义注水井到采油井连线方向为 x 轴正方向,垂直 x 轴作 y 轴,等值点整体呈对称分布,仅使用 y 轴正半轴数据进行拟合。采用三次多项式进行回归计算,其表达式为:

  • f(x)=ax3+bx2+cx+d
    (1)
  • 采用判定系数 R2 评价多项式拟合质量,R2 值越接近1,表示拟合质量越好,其表达式为:

  • R2=i=1N fxi-y-2i=1N yi-y-2
    (2)
  • 由含油饱和度等值线的多项式拟合参数及判定系数(表1)可以看出,R2 值均大于0.9,说明该方法具有较高的拟合精度。每一条含油饱和度等值线的空间分布特征用4个拟合参数(abcd)表示,而原网格数据至少需要 20 个横纵坐标点值,对等值展多项式拟合后只需要保留坐标原点值和拟合参数值,数据量下降至原来的10%以内。

  • 图3 不同时刻含油饱和度等值线(So=0.5,第5层)

  • Fig.3 Oil saturation isolines at different times(So=0.5,Layer5)

  • 图4 含油饱和度等值线多项式拟合结果

  • Fig.4 Polynomial fitting results of oil saturation isolines

  • 表1 含油饱和度等值线的多项式拟合参数及判定系数

  • Table1 Fitting parameters and determination coefficient of oil saturation isolines

  • 3 样本构建及学习训练

  • 参考中-高渗透油藏物性及井网特征设置方案,通过数值模拟获得训练样本数据集。模型尺寸为2 000 m×2 000 m,较大的平面尺寸是为了消除边界对等值线形态的影响,网格数为100×100×5,地层压力为 20 MPa,原油黏度为 10 mPa∙s,体积系数为 1.2,平均渗透率为300 mD。

  • 考虑注采井距(50,200,250,300,350,400 和 450 m)、油藏厚度(2,3,5和9 m)、渗透率(200,400, 600 和 800 mD)3 个参数,共组合 112 个方案。每个方案提取第 1 层、第 3 层、第 5 层含油饱和度等值线 (So=0.3,0.5,0.7)的拟合数据,获得原始样本数据 9 072组,采用随机方式划分为训练集(80%)和测试集(20%)。

  • 学习训练采用TensorFlow搭建人工神经网络模型(图5),其结构为 1个输入层、若干个隐藏层、1个输出层[22-25]。输入层包含 6 个神经元,分别为注采井距、油藏厚度、渗透率、垂向位置、采液量和含油饱和度;隐藏层数及隐藏层包含神经元个数Hij是需要优化的参数,隐藏层数越多,神经元个数越多,其对非线性数据的拟合能力越强;输出层神经元有 4 个参数。

  • 图5 人工神经网络模型示意

  • Fig.5 Schematic diagram of artificial neural network model

  • 为了提高人工神经网络模型的非线性拟合能力,测试多种激活函数的预测效果,本方案备选的激活函数包括 Sigmoid,Tanh 和 Relu 三种非线性函数,表达式分别为:

  • y=11+e-x
    (3)
  • y=ex-e-xex+e-x
    (4)
  • y=max(0,x)
    (5)
  • 模型质量的评价指标采用均方根误差,其表达式为:

  • RMSE=1Ni=1N ( predicted - label )2
    (6)
  • 初始采用 5个隐藏层,每层 10个神经元的网络模型,对比 150 个训练轮次,测试 Sigmoid,Tanh 和 Relu激活函数的均方根误差,其值分别为96,119和 82 m,故选用 Relu 作为所有神经元的激活函数,从而使模型的预测能力达到最强。对隐藏层数和神经元个数开展敏感性分析,对比测试3~8个隐藏层、 5~20个神经元的网络模型,结果显示5个隐藏层,每层 8~16 个神经元模型能获得较好的预测效果。以该模型为基础,继续增加训练轮次,提高模型的训练精度。

  • 由训练集学习曲线(图6)可以看出,500个训练轮次内,前 20 个训练轮次均方根误差快速下降,等值线各点位置的均方根误差由75 m下降至20 m,即该模型预测等值线点的空间距离偏差在20 m以内; 第 20~270 个训练轮次,精度缓慢下降,均方根误差由 20 m 下降至 9.8 m;第 270 个训练轮次后,均方根误差基本稳定在 8.9 m。由测试集预测结果(图7) 可以看出,数据点整体集中于 45°线,离 45°线距离较远的数据点来自中低含水期的样本,预测精度较低;离 45°线距离较近的数据点来自高含水期的样本,整体预测精度较高。

  • 图6 训练集学习曲线

  • Fig.6 Learning curve of training set

  • 图7 测试集预测结果

  • Fig.7 Prediction results of test set

  • 4 实例应用

  • 测试区块为某封闭断块油藏,该油藏被 5 条断层(F1—F5)包围,其中 F1 断层位于油藏的高部位, F5断层位于低部位,砂体厚度为 7.1 m,平均孔隙度为 0.26,平均渗透率为 460 mD,地层原油密度为 0.897 g/cm3,原油黏度为9 mPa∙s,属于中-高孔、中-高渗透、常压、稀油油藏,经过30多年的开发采出程度为 46.9%,含水率为 96%。除天然能量开采阶段外,主要生产阶段采用两注两采方式(注水井为Inj1,Inj2;采油井为 Prod1,Prod2)。开发动态响应显示注采对应关系明确,Inj1—Prod1和 Inj2—Prod2 是主见效方向,采液量分别为 6.78×104,4.68×104 m3,Inj1—Prod2 是次见效方向,采液量为 3.56×104 m3。将注采数据代入训练模型,获得含油饱和度等值线拟合参数,并根据该参数重构含油饱和度场,对于井组间的重叠部分取其最大值,从而获得整个区块的剩余油分布。

  • 对比含油饱和度数值模拟和机器学习预测结果(图8),新方法对该区块整体的剩余油分布特征预测较好。可以看出F1断层附近的高部位及F2和 F3断层附近的断边带剩余油描述最清晰,说明该学习模型对水驱波及的外部边界描述差异较小,其次是 2口注水井之间存在剩余油富集区(图8b),该区域预测能力主要受到注采井距和采液量影响,当注采井距越大且采液量越小时描述越准确,以上剩余油分布位置均符合矿场认识。然而该方法的不足之处主要表现在 Prod1 和 Prod2 井间剩余油边界描述不精准,对井间滞留区的刻画能力较弱,其原因是学习样本来自单个注采井组的含油饱和度,未考虑多注多采条件下流场相互干扰作用。为此,下一步研究将整个注采井网结构作为输入代入神经网络,描述含油饱和度分布从简单的多项式函数变为多种组合函数,不断提高注采井组间含油饱和度分布的预测能力。

  • 5 结论

  • 建立了一种基于机器学习的剩余油预测方法,通过数值模拟建立考虑不同因素影响下的含油饱和度等值线拟合参数样本,采用神经网络算法建立预测模型,并针对高含水油田开展应用测试。预测结果显示新方法在复杂井间联动关系的条件下,依然能较高质量描述高含水油田平面及纵向剩余油分布形态特征,同时计算效率得到大幅度提高。所提出的含油饱和度等值线多项式拟合样本预处理方法,克服了直接采用含油饱和度场作为学习样本时参数多、训练慢的缺点,训练后获得的预测模型具有更灵活、高效的特点。该方法可实现剩余油的快速预测分析,指导新井方案及注采调控方案设计。针对目前方法存在的缺陷提出了改进方向,未来随着考虑因素及样本数据量的扩充,预测精度将进一步提高,具有很大的发展潜力。

  • 图8 含油饱和度数值模拟和机器学习预测结果

  • Fig.8 Results of oil saturation by machine learning and numerical simulation

  • 符号解释

  • abcd——三次多项式系数,无量纲;

  • fx)——拟合函数,无量纲;

  • h——油藏厚度,m;

  • Hij——隐藏层第i层第j个神经元;

  • K——渗透率,mD;

  • label——样本值,无量纲;

  • L——注采井距,m;

  • Ly——垂向位置,m;

  • N——等值点总数;

  • Nl ——采液量,m3

  • i——等值点序号;

  • predicted——预测值,无量纲;

  • R2 ——判定系数,无量纲;

  • RMSE——均方根误差,无量纲;

  • So——含油饱和度,无量纲;

  • xi ——第i个等值点横坐标;

  • yi——第i个等值点纵坐标;

  • y-——等值点纵坐标平均值;

  • XY——空间位置,m。

  • 参考文献

    • [1] 贾虎,邓力珲.基于流线聚类人工智能方法的水驱油藏流场识别[J].石油勘探与开发,2018,45(2):312-319.JIA Hu,DENG Lihui.Oil reservoir water flooding flowing area identification based on the method of streamline clustering artifi⁃ cial intelligence[J].Petroleum Exploration and Development,2018,45(2):312-319.

    • [2] 李道伦,刘旭亮,查文舒,等.基于卷积神经网络的径向复合油藏自动试井解释方法[J].石油勘探与开发,2020,47(3):583-591.LI Daolun,LIU Xuliang,ZHA Wenshu,et al.Automatic well test interpretation based on convolutional neural network for a radial composite reservoir[J].Petroleum Exploration and Development,2020,47(3):583-591.

    • [3] ASKARI F R,KHAMEHCHI E.A novel approach to assist history matching using artificial intelligence[J].Chemical Engineering Communications,2015,202(4):513-519.

    • [4] LI Yu,ZHANG Tao,SUN Shuyu,et al.Accelerating flash calcula⁃ tion through deep learning methods[J].Journal of Computational Physics,2019,394:153-165.

    • [5] ZHANG Liming,WANG Saisai,ZHANG Kai,et al.Cooperative ar⁃ tificial bee colony algorithm with multiple populations for interval multi-objective optimization problems[J].IEEE Transactions on Fuzzy,2018,27(5):2872125.

    • [6] 张凯,路然然,张黎明,等.基于序列二次规划算法的油藏动态配产配注优化[J].油气地质与采收率,2014,21(1):45-50.ZHANG Kai,LU Ranran,ZHANG Liming,et al.Proration optimi⁃ zation of production and injection rate based on sequential qua⁃ dratic programming[J].Petroleum Geology and Recovery Efficien⁃ cy,2014,21(1):45-50.

    • [7] 刘巍,刘威,谷建伟,等.利用卡尔曼滤波和人工神经网络相结合的油藏井间连通性研究[J].油气地质与采收率,2020,27(2):118-124.LIU Wei,LIU Wei,GU Jianwei,et al.Research on interwell con⁃ nectivity of oil reservoirs based on Kalman filter and artificial neu⁃ ral network[J].Petroleum Geology and Recovery Efficiency,2020,27(2):118-124.

    • [8] 贾德利,刘合,张吉群,等.大数据驱动下的老油田精细注水优化方法[J].石油勘探与开发,2020,47(3):629-636.JIA Deli,LIU He,ZHANG Jiqun,et al.Data-driven optimization for fine water injection in a mature oil field[J].Petroleum Explora⁃ tion and Development,2020,47(3):629-636.

    • [9] 刘巍,刘威,谷建伟.基于机器学习方法的油井日产油量预测 [J].石油钻采工艺,2020,42(1):70-75.LIU Wei,LIU Wei,GU Jianwei.Oil production prediction based on a machine learning method[J].Oil Drilling & Production Tech⁃ nology,2020,42(1):70-75.

    • [10] 孙致学,姜宝胜,肖康,等.基于新型集成学习算法的基岩潜山油藏储层裂缝开度预测算法[J].油气地质与采收率,2020,27(3):32-38.SUN Zhixue,JIANG Baosheng,XIAO Kang,et al.Prediction of fracture aperture in bedrock buried hill oil reservoir based on nov⁃ el ensemble learning algorithm[J].Petroleum Geology and Recov⁃ ery Efficiency,2020,27(3):32-38.

    • [11] 侯春华.基于长短期记忆神经网络的油田新井产油量预测方法[J].油气地质与采收率,2019,26(3):105-110.HOU Chunhua.New well oil production forecast method based on long-term and short-term memory neural network[J].Petroleum Geology and Recovery Efficiency,2019,26(3):105-110.

    • [12] 李兆亮,柳金城,王琳,等.基于神经网络的复杂储层流体分级识别[J].断块油气田,2020,27(4):498-500.LI Zhaoliang,LIU Jincheng,WANG Lin,et al.Fluid hierarchical identification of complex reservoir based on neural network meth⁃ od[J].Fault-Block Oil and Gas Field,2020,27(4):498-500.

    • [13] 马文礼,李治平,孙玉平,等.基于机器学习的页岩气产能非确定性预测方法研究[J].特种油气藏,2019,26(2):101-105.MA Wenli,LI Zhiping,SUN Yuping,et al.Non-deterministic shale gas productivity forecast based on machine learning[J].Spe⁃ cial Oil & Gas Reservoirs,2019,26(2):101-105.

    • [14] 王端平.对胜利油区提高原油采收率潜力及转变开发方式的思考[J].油气地质与采收率,2014,21(4):1-4.WANG Duanping.Some thoughts about potential of oil recovery ef⁃ ficiency and development model transition in Shengli district[J].Petroleum Geology and Recovery Efficiency,2014,21(4):1-4.

    • [15] 王森,冯其红,宋玉龙,等.基于吸水剖面资料的优势通道分类方法——以孤东油田为例[J].油气地质与采收率,2013,20(5):99-102.WANG Sen,FENG Qihong,SONG Yulong,et al.Preferential flow path classification method based on injection profile data-taking Gudong oilfield as an example[J].Petroleum Geology and Recov⁃ ery Efficiency,2013,20(5):99-102.

    • [16] 朱光普,姚军,张磊,等.特高含水期剩余油分布及形成机理 [J].科学通报,2017,62(22):2 553-2 563.ZHU Guangpu,YAO Jun,ZHANG Lei,et al.Pore-scale investiga⁃ tion of residual oil distributions and formation mechanisms at the extra-high water-cut stage[J].Chinese Science Bulletin,2017,62(22):2 553-2 563.

    • [17] 李振泉,郭长春,王军,等.特高含水期油藏剩余油分布新认识——以孤岛油田中一区 Ng 3 砂组为例[J].油气地质与采收率,2019,26(6):19-27.LI Zhenquan,GUO Changchun,WANG Jun,et al.New under⁃ standing of remaining oil distribution in oil reservoirs at extrahigh water-cut stage:A case of Upper Ng3 sand group in Zhongyi area,Gudao Oilfield[J].Petroleum Geology and Recovery Effi⁃ ciency,2019,26(6):19-27.

    • [18] 陶光辉,李洪生,刘斌.特高含水期驱替倍数量化表征及调整对策[J].油气地质与采收率,2019,26(3):129-134.TAO Guanghui,LI Hongsheng,LIU Bin.Quantitative characteriza⁃ tion of displacement multiple and adjustment countermeasures in ultra-high water cut stage[J].Petroleum Geology and Recovery Efficiency,2019,26(3):129-134.

    • [19] 李阳,杨勇.老油田绿色低成本开发探索与实践[J].油气地质与采收率,2019,26(2):1-6.LI Yang,YANG Yong.Exploration and practice of green low-cost development in old oilfields[J].Petroleum Geology and Recovery Efficiency,2019,26(2):1-6.

    • [20] 崔传智,李松,杨勇,等.特高含水期油藏平面分区调控方法 [J].石油学报,2018,39(10):1 155-1 161.CUI Chuanzhi,LI Song,YANG Yong,et al.Planar zoning regula⁃ tion and control method of reservoir at ultra-high water cut stage [J].Acta Petrolei Sinica,2018,39(10):1 155-1 161.

    • [21] 刘丹.井间剩余油富集区识别方法研究[D].西安:西安石油大学,2014.LIU Dan.The recognition method on residual oil of enrichment re⁃ gion in well[D].Xi’an:Xi’an Shiyou University,2014.

    • [22] 邱锡鹏.神经网络与深度学习[M].北京:机械工业出版社,2020.QIU Xipeng.Neural networks and deep learning[M].Beijing:Chi⁃ na Machine Press,2020.

    • [23] 杨煜,张炜.TensorFlow平台上基于LSTM神经网络的人体动作分类[J].智能计算机与应用,2017,7(5):41-45.YANG Yu,ZHANG Wei.Human action classification based on LSTM neural network on TensorFlow[J].Intelligent Computer and Applications,2017,7(5):41-45.

    • [24] 章敏敏,徐和平,王晓洁,等.谷歌TensorFlow机器学习框架及应用[J].微型机与应用,2017,36(10):58-60.ZHANG Minmin,XU Heping,WANG Xiaojie,et al.Application of Google TensorFlow machine learning framework[J].Microcom⁃ puter & Its Application,2017,36(10):58-60.

    • [25] 李梦洁,董峦.基于PyTorch的机器翻译算法的实现[J].计算机技术与发展,2018,28(10):160-163,167.LI Mengjie,DONG Luan.Implementation of machine translation algorithm based on PyTorch[J].Computer Technology and Devel⁃ opment,2018,28(10):160-163,167.

  • 参考文献

    • [1] 贾虎,邓力珲.基于流线聚类人工智能方法的水驱油藏流场识别[J].石油勘探与开发,2018,45(2):312-319.JIA Hu,DENG Lihui.Oil reservoir water flooding flowing area identification based on the method of streamline clustering artifi⁃ cial intelligence[J].Petroleum Exploration and Development,2018,45(2):312-319.

    • [2] 李道伦,刘旭亮,查文舒,等.基于卷积神经网络的径向复合油藏自动试井解释方法[J].石油勘探与开发,2020,47(3):583-591.LI Daolun,LIU Xuliang,ZHA Wenshu,et al.Automatic well test interpretation based on convolutional neural network for a radial composite reservoir[J].Petroleum Exploration and Development,2020,47(3):583-591.

    • [3] ASKARI F R,KHAMEHCHI E.A novel approach to assist history matching using artificial intelligence[J].Chemical Engineering Communications,2015,202(4):513-519.

    • [4] LI Yu,ZHANG Tao,SUN Shuyu,et al.Accelerating flash calcula⁃ tion through deep learning methods[J].Journal of Computational Physics,2019,394:153-165.

    • [5] ZHANG Liming,WANG Saisai,ZHANG Kai,et al.Cooperative ar⁃ tificial bee colony algorithm with multiple populations for interval multi-objective optimization problems[J].IEEE Transactions on Fuzzy,2018,27(5):2872125.

    • [6] 张凯,路然然,张黎明,等.基于序列二次规划算法的油藏动态配产配注优化[J].油气地质与采收率,2014,21(1):45-50.ZHANG Kai,LU Ranran,ZHANG Liming,et al.Proration optimi⁃ zation of production and injection rate based on sequential qua⁃ dratic programming[J].Petroleum Geology and Recovery Efficien⁃ cy,2014,21(1):45-50.

    • [7] 刘巍,刘威,谷建伟,等.利用卡尔曼滤波和人工神经网络相结合的油藏井间连通性研究[J].油气地质与采收率,2020,27(2):118-124.LIU Wei,LIU Wei,GU Jianwei,et al.Research on interwell con⁃ nectivity of oil reservoirs based on Kalman filter and artificial neu⁃ ral network[J].Petroleum Geology and Recovery Efficiency,2020,27(2):118-124.

    • [8] 贾德利,刘合,张吉群,等.大数据驱动下的老油田精细注水优化方法[J].石油勘探与开发,2020,47(3):629-636.JIA Deli,LIU He,ZHANG Jiqun,et al.Data-driven optimization for fine water injection in a mature oil field[J].Petroleum Explora⁃ tion and Development,2020,47(3):629-636.

    • [9] 刘巍,刘威,谷建伟.基于机器学习方法的油井日产油量预测 [J].石油钻采工艺,2020,42(1):70-75.LIU Wei,LIU Wei,GU Jianwei.Oil production prediction based on a machine learning method[J].Oil Drilling & Production Tech⁃ nology,2020,42(1):70-75.

    • [10] 孙致学,姜宝胜,肖康,等.基于新型集成学习算法的基岩潜山油藏储层裂缝开度预测算法[J].油气地质与采收率,2020,27(3):32-38.SUN Zhixue,JIANG Baosheng,XIAO Kang,et al.Prediction of fracture aperture in bedrock buried hill oil reservoir based on nov⁃ el ensemble learning algorithm[J].Petroleum Geology and Recov⁃ ery Efficiency,2020,27(3):32-38.

    • [11] 侯春华.基于长短期记忆神经网络的油田新井产油量预测方法[J].油气地质与采收率,2019,26(3):105-110.HOU Chunhua.New well oil production forecast method based on long-term and short-term memory neural network[J].Petroleum Geology and Recovery Efficiency,2019,26(3):105-110.

    • [12] 李兆亮,柳金城,王琳,等.基于神经网络的复杂储层流体分级识别[J].断块油气田,2020,27(4):498-500.LI Zhaoliang,LIU Jincheng,WANG Lin,et al.Fluid hierarchical identification of complex reservoir based on neural network meth⁃ od[J].Fault-Block Oil and Gas Field,2020,27(4):498-500.

    • [13] 马文礼,李治平,孙玉平,等.基于机器学习的页岩气产能非确定性预测方法研究[J].特种油气藏,2019,26(2):101-105.MA Wenli,LI Zhiping,SUN Yuping,et al.Non-deterministic shale gas productivity forecast based on machine learning[J].Spe⁃ cial Oil & Gas Reservoirs,2019,26(2):101-105.

    • [14] 王端平.对胜利油区提高原油采收率潜力及转变开发方式的思考[J].油气地质与采收率,2014,21(4):1-4.WANG Duanping.Some thoughts about potential of oil recovery ef⁃ ficiency and development model transition in Shengli district[J].Petroleum Geology and Recovery Efficiency,2014,21(4):1-4.

    • [15] 王森,冯其红,宋玉龙,等.基于吸水剖面资料的优势通道分类方法——以孤东油田为例[J].油气地质与采收率,2013,20(5):99-102.WANG Sen,FENG Qihong,SONG Yulong,et al.Preferential flow path classification method based on injection profile data-taking Gudong oilfield as an example[J].Petroleum Geology and Recov⁃ ery Efficiency,2013,20(5):99-102.

    • [16] 朱光普,姚军,张磊,等.特高含水期剩余油分布及形成机理 [J].科学通报,2017,62(22):2 553-2 563.ZHU Guangpu,YAO Jun,ZHANG Lei,et al.Pore-scale investiga⁃ tion of residual oil distributions and formation mechanisms at the extra-high water-cut stage[J].Chinese Science Bulletin,2017,62(22):2 553-2 563.

    • [17] 李振泉,郭长春,王军,等.特高含水期油藏剩余油分布新认识——以孤岛油田中一区 Ng 3 砂组为例[J].油气地质与采收率,2019,26(6):19-27.LI Zhenquan,GUO Changchun,WANG Jun,et al.New under⁃ standing of remaining oil distribution in oil reservoirs at extrahigh water-cut stage:A case of Upper Ng3 sand group in Zhongyi area,Gudao Oilfield[J].Petroleum Geology and Recovery Effi⁃ ciency,2019,26(6):19-27.

    • [18] 陶光辉,李洪生,刘斌.特高含水期驱替倍数量化表征及调整对策[J].油气地质与采收率,2019,26(3):129-134.TAO Guanghui,LI Hongsheng,LIU Bin.Quantitative characteriza⁃ tion of displacement multiple and adjustment countermeasures in ultra-high water cut stage[J].Petroleum Geology and Recovery Efficiency,2019,26(3):129-134.

    • [19] 李阳,杨勇.老油田绿色低成本开发探索与实践[J].油气地质与采收率,2019,26(2):1-6.LI Yang,YANG Yong.Exploration and practice of green low-cost development in old oilfields[J].Petroleum Geology and Recovery Efficiency,2019,26(2):1-6.

    • [20] 崔传智,李松,杨勇,等.特高含水期油藏平面分区调控方法 [J].石油学报,2018,39(10):1 155-1 161.CUI Chuanzhi,LI Song,YANG Yong,et al.Planar zoning regula⁃ tion and control method of reservoir at ultra-high water cut stage [J].Acta Petrolei Sinica,2018,39(10):1 155-1 161.

    • [21] 刘丹.井间剩余油富集区识别方法研究[D].西安:西安石油大学,2014.LIU Dan.The recognition method on residual oil of enrichment re⁃ gion in well[D].Xi’an:Xi’an Shiyou University,2014.

    • [22] 邱锡鹏.神经网络与深度学习[M].北京:机械工业出版社,2020.QIU Xipeng.Neural networks and deep learning[M].Beijing:Chi⁃ na Machine Press,2020.

    • [23] 杨煜,张炜.TensorFlow平台上基于LSTM神经网络的人体动作分类[J].智能计算机与应用,2017,7(5):41-45.YANG Yu,ZHANG Wei.Human action classification based on LSTM neural network on TensorFlow[J].Intelligent Computer and Applications,2017,7(5):41-45.

    • [24] 章敏敏,徐和平,王晓洁,等.谷歌TensorFlow机器学习框架及应用[J].微型机与应用,2017,36(10):58-60.ZHANG Minmin,XU Heping,WANG Xiaojie,et al.Application of Google TensorFlow machine learning framework[J].Microcom⁃ puter & Its Application,2017,36(10):58-60.

    • [25] 李梦洁,董峦.基于PyTorch的机器翻译算法的实现[J].计算机技术与发展,2018,28(10):160-163,167.LI Mengjie,DONG Luan.Implementation of machine translation algorithm based on PyTorch[J].Computer Technology and Devel⁃ opment,2018,28(10):160-163,167.