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:

史长林(1970—),男,河北河间人,高级工程师,博士,从事石油地质研究及储层预测工作。E-mail:shchl@cnooc.com.cn。

中图分类号:TE319

文献标识码:A

文章编号:1009-9603(2022)01-0090-08

DOI:10.13673/j.cnki.cn37-1359/te.2022.01.011

参考文献 1
刘飞,赵晓明,冯潇飞,等.基于重力流相的深水水道分类方案研究[J].古地理学报,2021,23(5):951-965.LIU Fei,ZHAO Xiaoming,FENG Xiaofei,et al.Research on clas⁃ sification of deep-water channels based on gravity flow facies[J].Journal of Palaeogeography,2021,23(5):951-965.
参考文献 2
杜威,纪友亮,李其海,等.不同沉积过程尺度下正演数值模拟研究进展及油气地质意义[J].油气地质与采收率,2020,27(2):62-71.DU Wei,JI Youliang,LI Qihai,et al.Sedimentary forward numeri⁃ cal modeling at different sedimentary scales:Progress and hydro⁃ carbon significance[J].Petroleum Geology and Recovery Efficien⁃ cy,2020,27(2):62-71.
参考文献 3
崔龙涛,张倩萍,郭诚,等.坡折带主控下粗粒三角洲沉积演化及储层预测——以KLA油田为例[J].断块油气田,2019,26(4):444-448.CUI Longtao,ZHANG Qianping,GUO Cheng,et al.Sedimentary evolution of coarse-grained delta and prediction of favorable res⁃ ervoir under control of slope zone:taking KLA oilfield for example [J].Fault-Block Oil and Gas Field,2019,26(4):444-448.
参考文献 4
张春生,刘忠保,施冬,等.利用数值模拟方法预测辫状河砂体几何形态[J].石油勘探与开发,2004,31(z1):85-88.ZHANG Chunsheng,LIU Zhongbao,SHI Dong,et al.Braid river sandbody shape forecast by numerical simulation method[J].Pe⁃ troleum Exploration and Development,2004,31(z1):85-88.
参考文献 5
李少华,张昌民,林克湘,等.储层建模中几种原型模型的建立 [J].沉积与特提斯地质,2004,24(3):102-107.LI Shaohua,ZHANG Changmin,LIN Kexiang,et al.The construc⁃ tion of prototype models in reservoir modeling[J].Sedimentary Geology and Tethyan Geology,2004,24(3):102-107.
参考文献 6
岳大力,吴胜和,刘建民.曲流河点坝地下储层构型精细解剖方法[J].石油学报,2007,28(4):99-103.YUE Dali,WU Shenghe,LIU Jianmin.An accurate method for anatomizing architecture of subsurface reservoir in point bar of meandering river[J].Acta Petrolei Sinica,2007,28(4):99-103.
参考文献 7
严科,杨少春,任怀强.基于油藏开发动态的储层四维模型的建立[J].中国石油大学学报:自然科学版,2010,34(1):12-17.YAN Ke,YANG Shaochun,REN Huaiqiang.Building of 4D reser⁃ voir model based on development performance[J].Journal of Chi⁃ na University of Petroleum:Edition of Natural Science,2010,34(1):12-17.
参考文献 8
史长林,张凤红,陈平.水驱实验模拟注水开发对储层的影响 [J].西南石油大学学报:自然科学版,2013,35(5):87-93.SHI Changlin,ZHANG Fenghong,CHEN Ping.Affection of simu⁃ lating water-flooding by water injection tests on reservoir proper⁃ ties[J].Journal of Southwest Petroleum University:Science & Technology Edition,2013,35(5):87-93.
参考文献 9
冷雪梅,杜启振,孟宪军.基于纵横波弹性阻抗联合反演的储层流体检测方法[J].断块油气田,2019,26(3):319-323,363.LENG Xuemei,DU Qizhen,MENG Xianjun.Reservoir fluid detec⁃ tion method based on joint PP-and PS-wave elastic impedance inversion[J].Fault-Block Oil and Gas Field,2019,26(3):319-323,363.
参考文献 10
程丽芳,王晖,尹艳树,等.基于浅层地震资料的曲流河点坝精细解剖——以加拿大某区块Mcmurray组储层为例[J].断块油气田,2019,26(4):464-469.CHENG Lifang,WANG Hui,YIN Yanshu,et al.Fine dissection of meandering river point bars based on shallow seismic data:a case study of Mcmurray reservoir in a block,Canada[J].Fault-Block Oil and Gas Field,2019,26(4):464-469.
参考文献 11
高秋菊,谭明友,张营革,等.陆相页岩油“甜点”井震联合定量评价技术——以济阳坳陷罗家地区沙三段下亚段为例[J].油气地质与采收率,2019,26(1):165-173.GAO Qiuju,TAN Mingyou,ZHANG Yingge,et al.Well-seismic joint technology for quantitative evaluation of“sweet spot”in con⁃ tinental shale oil:A case study of Lower Es3 Member of Luojia ar⁃ ea in Jiyang Depression[J].Petroleum Geology and Recovery Effi⁃ ciency,2019,26(1):165-173.
参考文献 12
岳丽君,钱艳苓.基于地震波形指示反演技术的窄薄储层预测 [J].大庆石油地质与开发,2020,39(5):135-140.YUE Lijun,QIAN Yanling.Prediction of the narrow-thin reservoir based on seismic waveform indication inversion technology[J].Pe⁃ troleum Geology & Oilfield Development in Daqing,2020,39(5):135-140.
参考文献 13
李宏为,李凡异,齐秋红,等.基于地震多属性分析的吉尔嘎朗图凹陷煤层气富集区预测[J].大庆石油地质与开发,2020,39(4):135-142.LI Hongwei,LI Fanyi,QI Qiuhong,et al.Prediction of CBM en⁃ riched area in Jirgalangtu Sag based on the seismic multiple-attri⁃ bute analysis[J].Petroleum Geology & Oilfield Development in Daqing,2020,39(4):135-142.
参考文献 14
张晶玉,范廷恩,王宗俊,等.多属性储层参数反演方法在M油田中的应用[J].油气藏评价与开发,2019,9(1):8-14.ZHANG Jingyu,FAN Ting’en,WANG Zongjun,et al.Application of reservoir parameters inversion method by using multi-attri⁃ butes in M oilfield[J].Reservoir Evaluation and Development,2019,9(1):8-14.
参考文献 15
宋辉,陈伟,李谋杰,等.基于卷积门控循环单元网络的储层参数预测方法[J].油气地质与采收率,2019,26(5):73-78.SONG Hui,CHEN Wei,LI Moujie,et al.A method to predict res⁃ ervoir parameters based on convolutional neural network-gated recurrent unit(CNN-GRU)[J].Petroleum Geology and Recovery Efficiency,2019,26(5):73-78.
参考文献 16
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 17
李政宏,刘永福,张立强,等.数据挖掘方法在测井岩性识别中的应用[J].断块油气田,2019,26(6):713-718.LI Zhenghong,LIU Yongfu,ZHANG Liqiang,et al.Application of data mining method in lithology identification using well log[J].Fault-Block Oil and Gas Field,2019,26(6):713-718.
参考文献 18
李婷婷,王钊,马世忠,等.地震属性融合方法综述[J].地球物理学进展,2015,30(l):378-385.LI Tingting,WANG Zhao,MA Shizhong,et al.Summary of seis⁃ mic attribute fusion method[J].Progress in Geophysics,2015,30(l):378-385.
参考文献 19
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 20
翟明国.大数据定将改变地质——向读者推荐《地球科学大数据挖掘与机器学习》[J].矿物岩石地球化学通报,2018,37(6):1 209..ZHAI Mingguo.Big data will change geology-Big Data Mining and Machine Learning in Geoscience is recommended to the read⁃ er[J].Bulletin of Mineralogy,Petrology and Geochemistry,2018,37(6):1 209.
参考文献 21
林年添,张栋,张凯,等.地震油气储层的小样本卷积神经网络学习与预测[J].地球物理学报,2018,61(10):4 110-4 125.LIN Niantian,ZHANG Dong,ZHANG Kai,et al.Predicting distri⁃ bution of hydrocarbon reservoirs with seismic data based on learn⁃ ing of the small-sample convolution neural network[J].Chinese Journal of Geophysics,2018,61(10):4 110-4 125.
参考文献 22
付超,林年添,张栋,等.多波地震深度学习的油气储层分布预测案例[J].地球物理学报,2018,61(1):293-303.FU Chao,LIN Niantian,ZHANG Dong,et al.Prediction of reser⁃ voirs using multi-component seismic data and the deep learning method[J].Chinese Journal of Geophysics,2018,61(1):293-303.
参考文献 23
林年添,付超,张栋,等.无监督与监督学习下的含油气储层预测[J].石油物探,2018,57(4):601-610.LIN Niantian,FU Chao,ZHANG Dong,et al.Supervised learning and unsupervised learning for hydrocarbon prediction using multi⁃ wave seismic data[J].Geophysical Prospecting for Petroleum,2018,57(4):601-610.
参考文献 24
李国欣,王峰,皮学军,等.非常规油气藏地质工程一体化数据优化应用的思考与建议[J].中国石油勘探,2019,24(2):147-152.LI Guoxin,WANG Feng,PI Xuejun,et al.Optimized application of geology-engineering integration data of unconventional oil and gas reservoirs[J].China Petroleum Exploration,2019,24(2):147-152.
参考文献 25
甘利灯,张昕,王峣钧,等.从勘探领域变化看地震储层预测技术现状和发展趋势[J].石油地球物理勘探,2018,53(1):214-225.GAN Lideng,ZHANG Xin,WANG Yaojun,et al.Current status and development trends of seismic reservoir prediction viewed from the exploration industry[J].Oil Geophysical Prospecting,2018,53(1):214-225.
目录contents

    摘要

    机器学习和数据挖掘具有出色的预测、分析、决策和计算能力,在油气勘探开发领域的应用已取得良好的效果。在总结储层预测方法的基础上,分析了不同储层预测方法的适用性及优缺点,应用机器学习方法,挖掘测井和地震数据,预测了储层的岩石类型、空间展布、孔隙度、渗透率和含油饱和度。将该方法与地震反演储层预测对比,结果表明该方法具有明显优势。一是挖掘地震数据蕴含的大量信息并进行多重属性融合,使预测精度提高;二是数据驱动代替经验驱动,使工作流程简化。

    Abstract

    Machine learning and data mining possess excellent abilities of prediction,analysis,decision-making,and calcu- lation and have achieved good results in the field of oil and gas exploration and development. On the basis of summarizing the reservoir prediction methods,this paper analyzes the applicability,advantages and disadvantages of different reservoir prediction methods. It utilizes the machine learning algorithm to predict the rock type,spatial distribution,porosity,permea- bility,and oil saturation of the reservoir by mining logging and seismic data. This method reveals evident advantages com- pared with seismic inversion reservoir prediction:first,mining a large amount of information contained in seismic data and multi-attribute fusion can improve the prediction accuracy;second,data-driven instead of experience-driven can simplify the workflow.

  • 储层预测在油气田勘探开发各个阶段都具有重要意义,其预测技术和方法多种多样。根据应用基础资料不同,可概括为基于地质知识库和基于地震资料的两类储层预测方法。第1类是基于地质知识库的储层预测方法,是利用储层的成因、空间特征、边界条件、物理特征和沉积模式等各种实践参数开展的储层预测研究。地质知识库的建立是该方法的关键,主要包括露头知识库、现代沉积知识库、沉积模拟实验知识库以及井资料知识库。基于露头知识库储层预测方法的优点是模型精细准确[1-2],缺点是露头区地质知识库的空间范围有限; 基于现代沉积知识库储层预测方法的优点是模型完整、直观[3],缺点是难以表征地质沉积时期内的沉积特征;基于沉积模拟实验知识库储层预测方法的优点是便于观测,缺点是模型受时空限制[4];基于井资料知识库储层预测方法的优点在于具有针对性、时空关系的一致性,缺点是预测结果不确定性较大[5-8]。第2类是基于地震资料的储层预测方法,是以地震信息为主,综合利用其他资料(地质、测井、油藏等)作为约束,对油气储层的品质参数进行预测的方法,主要包括地震资料反演、地震属性分析、地震属性融合等。地震反演储层预测方法的优势为井震结合表征储层,缺点是地震反演流程比较复杂[9-13];地震属性分析与融合的储层预测方法,优势是用多种属性共同表征储层参数,提高了储层预测的精度[13-17],缺点是属性优选和权重确定都需要复杂计算和反复试验[18]。在实际储层预测中大多是各种资料和方法的综合运用。

  • 地震数据蕴含丰富的油气储层信息,而这些信息与储层特征参数之间一般具有多维、高噪、非线性等关系特征,常规方法难以有效地建立彼此之间的关系,而机器学习方法因其具有模型鲁棒性强、泛化能力高、自动表征复杂多元非线性关系等特点,在储层特征提取、分类、识别及预测等方面具有先天优势[19-24]。本文的研究以珠江口盆地E油田为例。该油田位于中国南海珠江口盆地北部坳陷带,发育岩性-构造油藏。研究区面积为 70 km2,具有 31口井的测井解释数据,地震资料主频为30 Hz,频宽为 0~60 Hz。目的层为一套扇三角洲碎屑沉积,主要岩性为砂岩、粉砂岩和泥岩,储层为中孔中渗透储层,非均质性强。利用机器学习方法,挖掘测井和地震数据之间的非线性关系,预测该油田储层的岩性、物性和含油气性。与传统预测方法相比,该方法不但能提高预测精度,在预测效率方面也大幅提高,值得同类油田的推广和借鉴。

  • 1 预测方法

  • 1.1 样本数据的建立

  • 在储层预测中应用测井解释数据和地震数据,由于测井和地震数据的采集方式和记录方式不同,因此在预测前,需要通过时深标定,确定井震数据的对应关系,同时对测井和地震数据重采样,以及提取井旁道的地震属性。具体步骤包括:①井震关系建立。利用时间域地震层位解释数据和单井时深关系数据建立全工区速度模型,再将各类地震属性数据从时间域转换为深度域,从而建立准确的井震匹配关系,准确的时深关系是影响结果准确性的关键因素。②测井和地震数据重采样。由于测井数据与地震数据采样精度不同,测井数据采样间隔为 0.125 m、地震数据采样间隔约为 5~10 m,通过建立垂向精度为 0.5~1 m 的模型网格,将测井解释成果数据、各类地震属性数据重采样输入网格中,从而获得垂向精度为 0.5~1 m 的机器学习样本数据。如果对所预测储层的极限厚度不确定的情况下,可以在储层厚度统计的基础上,保留厚、中、薄砂层测井数据,建立不同尺度的网格,充分挖掘样本的预测潜能。③样本提取。通过网格数据采样,提取井轨迹处的测井参数数据,提取对应井旁道的地震数据,并提取多种不同的地震属性,将同一深度的测井数据与地震属性数据组合,形成一组样本数据,作为机器学习的输入样本。④样本数据预处理。为确保训练模型的训练效果,还需对提取样本进行质量检查,包括剔除无效值和异常值、数据标准化、归一化等。尤其对于不同地震属性数量级差异的问题,进行标准化处理能够更好的服务于方法计算,如支持向量机,由于其自身的特点,多量纲样本转化为无量纲样本会更适合该方法处理的要求。

  • 通过以上的步骤,就可以获得较好的机器学习样本,并随机选取其中80%的样本作为训练集,20% 的样本作为验证集。

  • 1.2 预测及优化

  • 机器学习方法种类众多,不同工区的地震、测井资料采集情况各有差异,因此没有哪一种方法在所有工区的应用中都表现最佳。本文采用的主要步骤包括:①机器学习方法编制及优化,包括决策树、贝叶斯、支持向量机、深度神经网络、XGBoost, LightGBM等方法。根据预测储层具体情况,对各类机器学习方法进行对比分析,选取最优方法建立模型。②基于机器学习的储层预测,根据预测的研究目的,利用机器学习方法对井间岩性、物性等参数进行预测。③预测结果分析及检验,对预测结果精度进行分析,通过精确率、召回率和分类模型的综合评价指标分析等对预测结果进行评价,根据油藏描述的实践经验完善检验模型。④模型完善及优化,建立模型优化方法,调整参数优化模型,使模型逼近最优化。

  • 在地震属性与测井岩性、物性相关性分析的过程中,一些地震属性会表现出与测井解释成果较低的相关性,对于此类情况,笔者不建议直接删除线性相关性低的属性,因为删除此类属性后可能导致模型准确度的下降。因此,在地震属性及方法优选时,应以模型评估得分作为属性删减、方法选取的主要指标,分类问题主要依据分类模型的综合评价指标、回归问题主要依据决定系数进行模型评估。模型评估参数的表达式分别为:

  • F1=2PRP+R
    (1)
  • P=TPTP+FP
    (2)
  • R=TPTP+FN
    (3)
  • R2=1-i=1n yi-y^i2i=1n yi-y-2
    (4)
  • 本文主要提取了均方根振幅、正交振幅、相对波阻抗、叠后甜点、时间衰减、瞬时、振幅包络、背景消除、叠后偏移和基值偏移等 10 种地震属性,选用了 LightGBM,XGBoost,随机森林,LSTM,支持向量机等5种机器学习分类方法进行建模。

  • 2 实例应用

  • 2.1 岩石类型预测

  • 在 E 油田的岩石类型预测中,应用测井岩性数据为 9 015 个,重采样深度域地震属性体 10 种。通过混淆矩阵分析优选机器学习方法,分别用LightG⁃BM,XGBoost,随机森林,LSTM,支持向量机 5 种机器学习分类方法对模型的精确率和召回率进行分析,并计算得到 F 1值,其值越高代表该方法得到的模型准确率越高。5种集成学习方法所得结果的F 1 值均大于 0.78,而 XGBoost 方法在这一研究区较其他方法的结果更优(表1),F 1值达0.89。

  • 表1 不同方法岩性预测结果对比

  • Table1 Performance comparison of lithology prediction by different algorithms

  • 在 E 油田应用 3 口新井来验证预测效果,对比单井岩性解释结果和基于XGBoost方法的岩性预测结果(图1)发现:①泥岩厚度小于等于2 m的识别率很低。②泥岩厚度大于2 m的识别率超过90%。分析认为,基于机器学习的储层预测方法有极限识别厚度,小于极限识别厚度,预测结果会出现严重偏差,这一偏差的出现是由样本自身特点决定。值得提醒的是,由于基础资料的不同,每个区域都会呈现不同的极限厚度,在样本提取和划分时需注意。

  • 图1 3口验证井的岩性预测结果与测井岩性解释结果对比

  • Fig.1 Comparison of lithology prediction and logging lithology interpretation in three verification wells

  • 比较机器学习方法储层预测结果和叠后地震反演预测结果在数据和方法的异同,数据基础都利用了井数据和地震数据,不同之处在于反演方法基于模型迭代,以经验驱动为主,而机器学习方法由数据推理,以数据驱动为主,其效果的精度取决于样本数据和方法。对比 E油田基于 XGBoost方法的储层预测剖面和叠后反演预测剖面(图2)发现:① 两者预测趋势一致,5套主力层均有识别,基于 XGBoost 方法的效果可以达到反演方法预测的效果。 ②对主力砂层内部隔夹层刻画方面,基于 XGBoost 方法的预测结果表现出比反演方法更为细致的结果。③相较反演方法较为平滑的预测结果,基于 XGBoost 方法的预测结果表现出更强的非均质,符合本区域先验地质认识和生产动态认识。因此,在 E油田基于XGBoost方法的预测效果更优。

  • 图2 基于XGBoost方法储层预测剖面与叠后反演预测剖面的对比

  • Fig.2 Comparison between post-stack inversion prediction profile and reservoir prediction profile based on XGBoost

  • 2.2 储层物性预测

  • 在 E 油田的储层物性预测中,使用的样本数据包括孔隙度数据和渗透率数据各 9 015 个,不同于岩性这种分类数据,孔渗数据属于连续数据,所以研究中选用了 LightGBM,XGBoost,随机森林,LSTM 和支持向量机5种机器学习的回归方法对E油田的储层孔隙度、渗透率进行预测,通过 R2 指标对回归模型准确性进行评估,R2 值越接近 1 代表该方法得到的模型准确率越高。模型评估结果(表2)表明, LightGBM 方法在孔隙度预测中得分较高,达 0.84; 而随机森林方法在渗透率预测中表现最佳,达0.73。

  • 选用最佳方法形成孔隙度和渗透率三维预测体,通过截取过新井剖面来验证预测结果和准确性。预测结果(图3)显示,纵向能很好地反映隔夹层发育情况,横向能反映井间物性变化情况。对该区域物性预测的纵横向变化情况符合该油田的地质认识。

  • 表2 不同方法物性预测结果的R2 指标对比

  • Table2 R2 indicator of physical property prediction performance by different algorithms

  • 2.3 储层含油性预测

  • 通过机器学习方法对E油田的含油饱和度进行预测,使用样本数量为 5 420 个,分别分析了 5 种回归方法的预测性能,通过R2 指标对回归模型准确性进行评估。结果表明,相较物性的 R2 指标,含油饱和度指标整体偏低,平均约为 0.5,其中 R2 指标最高为XGBoost方法。预测结果(图4)显示,与实际测井数据对含油层段识别一致,但定量识别仍需改进。

  • 图3 储层物性预测模型

  • Fig.3 Prediction model of reservoir physical property

  • 图4 基于XGBoost方法的含油饱和度预测剖面

  • Fig.4 Oil-saturation prediction profile based on XGBoost algorithm

  • 3 应用效果分析

  • 通过抽取新井的井旁道预测数据与实际测井数据来验证预测结果的可靠性,岩性预测能够很好的识别砂泥岩(图2);物性和含油性参数的预测(图5)显示,预测曲线与实际测井曲线趋势一致,孔隙度和渗透率预测效果优于含油饱和度预测效果,这与模型评估的结果一致。分析认为含油饱和度预测曲线与实际测井曲线差异较大的问题主要是受到井震数据采集时间不同的影响,油藏的含油饱和度数据是一个动态结果,不同时期测得的含油饱和度曲线仅反映当时油藏的情况,与地震数据采集时所反映信息已有较大变化,尤其是注水开发阶段。预测孔渗参数的方法是否适用于预测饱和度参数,这些都值得进一步研究和分析。

  • 图5 预测曲线与实际测井曲线的对比分析

  • Fig.5 Comparative analysis of prediction curves and actual logging curves

  • 综上分析机器学习方法储层预测具有以下优势。

  • 简化人工操作流程  常用的地震岩石物理分析、属性分析、正演和反演等储层预测技术都遵循 “找规律、提信息、做解释”的过程。其中“找规律” 就是发现储层和非储层的地震信息差异及其可识别性,“提信息”是把这种可识别的差异信息从地震信号中提取出来,“做解释”是依据差异地震信息解释储层特征[25];“找规律”和“做解释”都需要做大量的相关性分析,如聚类、回归分析等,而基于机器学习方法的储层预测是通过机器学习方法“找规律” 和“做解释”,不需要人工做大量的相关分析。

  • 数据驱动代替经验驱动  传统的地震属性分析、反演过程都需要人工来选择地震属性及地震属性与地质参数间的映射模型,优化映射模型的过程非常复杂,这要求分析人员有丰富的经验;基于机器学习的储层预测依靠数据驱动自动建立地震属性与地质参数之间的映射模型,称之为训练模型,从而降低了分析人员对分析经验的依赖,实现经验驱动向数据驱动的转化。

  • 充分挖掘地震信息并多属性表征储层  基于机器学习方法的储层预测是通过井震结合,选取表征储层参数的多种地震属性,通过数学运算建立训练模型,再通过多训练模型分配各种地震属性的权重融合出一种新的地震属性;融合的地震属性具有多维度表征同一储层参数的特征,并且充分挖掘了地震属性蕴含的大量储层信息(这些储层信息人工解析难度非常大),因此该方法降低储层预测的多解性,提高储层预测精度。

  • 4 结论

  • 机器学习方法为储层预测提供了新的思路和方法。通过恰当样本提取,方法优选,不仅能提高预测精度而且能极大缩短预测时间,可以作为储层预测方法的一种补充。但该方法在储层预测实践中,也有应用条件。首先要有一定数量的样本,也就是油田井网达到一定的密度。在勘探阶段,没有足够的样本量,该方法不适用。其次网格划分要合理,划分过细预测精度提高有限,运算量会大幅增加;网格过粗,则会降低预测精度,所以要根据工区储层厚度分布规律和地震资料品质合理划分网格。最后时深转化要精确,时深错位会造成样本训练集错误,因此该方法对时深关系精确度要求更高。

  • 符号解释

  • F 1——分类模型的综合评价指标;

  • FN ——假的负样本数量,个;

  • FP ——假的正样本数量,个;

  • n ——样本数,个;

  • P ——精确率,%;

  • R ——召回率,%;

  • R2 ——决定系数;

  • TP ——真的正样本数量,个;

  • yi ——样本特征的真实值;

  • y-——样本特征的平均值;

  • y^i ——样本特征的预测值。

  • 参考文献

    • [1] 刘飞,赵晓明,冯潇飞,等.基于重力流相的深水水道分类方案研究[J].古地理学报,2021,23(5):951-965.LIU Fei,ZHAO Xiaoming,FENG Xiaofei,et al.Research on clas⁃ sification of deep-water channels based on gravity flow facies[J].Journal of Palaeogeography,2021,23(5):951-965.

    • [2] 杜威,纪友亮,李其海,等.不同沉积过程尺度下正演数值模拟研究进展及油气地质意义[J].油气地质与采收率,2020,27(2):62-71.DU Wei,JI Youliang,LI Qihai,et al.Sedimentary forward numeri⁃ cal modeling at different sedimentary scales:Progress and hydro⁃ carbon significance[J].Petroleum Geology and Recovery Efficien⁃ cy,2020,27(2):62-71.

    • [3] 崔龙涛,张倩萍,郭诚,等.坡折带主控下粗粒三角洲沉积演化及储层预测——以KLA油田为例[J].断块油气田,2019,26(4):444-448.CUI Longtao,ZHANG Qianping,GUO Cheng,et al.Sedimentary evolution of coarse-grained delta and prediction of favorable res⁃ ervoir under control of slope zone:taking KLA oilfield for example [J].Fault-Block Oil and Gas Field,2019,26(4):444-448.

    • [4] 张春生,刘忠保,施冬,等.利用数值模拟方法预测辫状河砂体几何形态[J].石油勘探与开发,2004,31(z1):85-88.ZHANG Chunsheng,LIU Zhongbao,SHI Dong,et al.Braid river sandbody shape forecast by numerical simulation method[J].Pe⁃ troleum Exploration and Development,2004,31(z1):85-88.

    • [5] 李少华,张昌民,林克湘,等.储层建模中几种原型模型的建立 [J].沉积与特提斯地质,2004,24(3):102-107.LI Shaohua,ZHANG Changmin,LIN Kexiang,et al.The construc⁃ tion of prototype models in reservoir modeling[J].Sedimentary Geology and Tethyan Geology,2004,24(3):102-107.

    • [6] 岳大力,吴胜和,刘建民.曲流河点坝地下储层构型精细解剖方法[J].石油学报,2007,28(4):99-103.YUE Dali,WU Shenghe,LIU Jianmin.An accurate method for anatomizing architecture of subsurface reservoir in point bar of meandering river[J].Acta Petrolei Sinica,2007,28(4):99-103.

    • [7] 严科,杨少春,任怀强.基于油藏开发动态的储层四维模型的建立[J].中国石油大学学报:自然科学版,2010,34(1):12-17.YAN Ke,YANG Shaochun,REN Huaiqiang.Building of 4D reser⁃ voir model based on development performance[J].Journal of Chi⁃ na University of Petroleum:Edition of Natural Science,2010,34(1):12-17.

    • [8] 史长林,张凤红,陈平.水驱实验模拟注水开发对储层的影响 [J].西南石油大学学报:自然科学版,2013,35(5):87-93.SHI Changlin,ZHANG Fenghong,CHEN Ping.Affection of simu⁃ lating water-flooding by water injection tests on reservoir proper⁃ ties[J].Journal of Southwest Petroleum University:Science & Technology Edition,2013,35(5):87-93.

    • [9] 冷雪梅,杜启振,孟宪军.基于纵横波弹性阻抗联合反演的储层流体检测方法[J].断块油气田,2019,26(3):319-323,363.LENG Xuemei,DU Qizhen,MENG Xianjun.Reservoir fluid detec⁃ tion method based on joint PP-and PS-wave elastic impedance inversion[J].Fault-Block Oil and Gas Field,2019,26(3):319-323,363.

    • [10] 程丽芳,王晖,尹艳树,等.基于浅层地震资料的曲流河点坝精细解剖——以加拿大某区块Mcmurray组储层为例[J].断块油气田,2019,26(4):464-469.CHENG Lifang,WANG Hui,YIN Yanshu,et al.Fine dissection of meandering river point bars based on shallow seismic data:a case study of Mcmurray reservoir in a block,Canada[J].Fault-Block Oil and Gas Field,2019,26(4):464-469.

    • [11] 高秋菊,谭明友,张营革,等.陆相页岩油“甜点”井震联合定量评价技术——以济阳坳陷罗家地区沙三段下亚段为例[J].油气地质与采收率,2019,26(1):165-173.GAO Qiuju,TAN Mingyou,ZHANG Yingge,et al.Well-seismic joint technology for quantitative evaluation of“sweet spot”in con⁃ tinental shale oil:A case study of Lower Es3 Member of Luojia ar⁃ ea in Jiyang Depression[J].Petroleum Geology and Recovery Effi⁃ ciency,2019,26(1):165-173.

    • [12] 岳丽君,钱艳苓.基于地震波形指示反演技术的窄薄储层预测 [J].大庆石油地质与开发,2020,39(5):135-140.YUE Lijun,QIAN Yanling.Prediction of the narrow-thin reservoir based on seismic waveform indication inversion technology[J].Pe⁃ troleum Geology & Oilfield Development in Daqing,2020,39(5):135-140.

    • [13] 李宏为,李凡异,齐秋红,等.基于地震多属性分析的吉尔嘎朗图凹陷煤层气富集区预测[J].大庆石油地质与开发,2020,39(4):135-142.LI Hongwei,LI Fanyi,QI Qiuhong,et al.Prediction of CBM en⁃ riched area in Jirgalangtu Sag based on the seismic multiple-attri⁃ bute analysis[J].Petroleum Geology & Oilfield Development in Daqing,2020,39(4):135-142.

    • [14] 张晶玉,范廷恩,王宗俊,等.多属性储层参数反演方法在M油田中的应用[J].油气藏评价与开发,2019,9(1):8-14.ZHANG Jingyu,FAN Ting’en,WANG Zongjun,et al.Application of reservoir parameters inversion method by using multi-attri⁃ butes in M oilfield[J].Reservoir Evaluation and Development,2019,9(1):8-14.

    • [15] 宋辉,陈伟,李谋杰,等.基于卷积门控循环单元网络的储层参数预测方法[J].油气地质与采收率,2019,26(5):73-78.SONG Hui,CHEN Wei,LI Moujie,et al.A method to predict res⁃ ervoir parameters based on convolutional neural network-gated recurrent unit(CNN-GRU)[J].Petroleum Geology and Recovery Efficiency,2019,26(5):73-78.

    • [16] 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.

    • [17] 李政宏,刘永福,张立强,等.数据挖掘方法在测井岩性识别中的应用[J].断块油气田,2019,26(6):713-718.LI Zhenghong,LIU Yongfu,ZHANG Liqiang,et al.Application of data mining method in lithology identification using well log[J].Fault-Block Oil and Gas Field,2019,26(6):713-718.

    • [18] 李婷婷,王钊,马世忠,等.地震属性融合方法综述[J].地球物理学进展,2015,30(l):378-385.LI Tingting,WANG Zhao,MA Shizhong,et al.Summary of seis⁃ mic attribute fusion method[J].Progress in Geophysics,2015,30(l):378-385.

    • [19] 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.

    • [20] 翟明国.大数据定将改变地质——向读者推荐《地球科学大数据挖掘与机器学习》[J].矿物岩石地球化学通报,2018,37(6):1 209..ZHAI Mingguo.Big data will change geology-Big Data Mining and Machine Learning in Geoscience is recommended to the read⁃ er[J].Bulletin of Mineralogy,Petrology and Geochemistry,2018,37(6):1 209.

    • [21] 林年添,张栋,张凯,等.地震油气储层的小样本卷积神经网络学习与预测[J].地球物理学报,2018,61(10):4 110-4 125.LIN Niantian,ZHANG Dong,ZHANG Kai,et al.Predicting distri⁃ bution of hydrocarbon reservoirs with seismic data based on learn⁃ ing of the small-sample convolution neural network[J].Chinese Journal of Geophysics,2018,61(10):4 110-4 125.

    • [22] 付超,林年添,张栋,等.多波地震深度学习的油气储层分布预测案例[J].地球物理学报,2018,61(1):293-303.FU Chao,LIN Niantian,ZHANG Dong,et al.Prediction of reser⁃ voirs using multi-component seismic data and the deep learning method[J].Chinese Journal of Geophysics,2018,61(1):293-303.

    • [23] 林年添,付超,张栋,等.无监督与监督学习下的含油气储层预测[J].石油物探,2018,57(4):601-610.LIN Niantian,FU Chao,ZHANG Dong,et al.Supervised learning and unsupervised learning for hydrocarbon prediction using multi⁃ wave seismic data[J].Geophysical Prospecting for Petroleum,2018,57(4):601-610.

    • [24] 李国欣,王峰,皮学军,等.非常规油气藏地质工程一体化数据优化应用的思考与建议[J].中国石油勘探,2019,24(2):147-152.LI Guoxin,WANG Feng,PI Xuejun,et al.Optimized application of geology-engineering integration data of unconventional oil and gas reservoirs[J].China Petroleum Exploration,2019,24(2):147-152.

    • [25] 甘利灯,张昕,王峣钧,等.从勘探领域变化看地震储层预测技术现状和发展趋势[J].石油地球物理勘探,2018,53(1):214-225.GAN Lideng,ZHANG Xin,WANG Yaojun,et al.Current status and development trends of seismic reservoir prediction viewed from the exploration industry[J].Oil Geophysical Prospecting,2018,53(1):214-225.

  • 参考文献

    • [1] 刘飞,赵晓明,冯潇飞,等.基于重力流相的深水水道分类方案研究[J].古地理学报,2021,23(5):951-965.LIU Fei,ZHAO Xiaoming,FENG Xiaofei,et al.Research on clas⁃ sification of deep-water channels based on gravity flow facies[J].Journal of Palaeogeography,2021,23(5):951-965.

    • [2] 杜威,纪友亮,李其海,等.不同沉积过程尺度下正演数值模拟研究进展及油气地质意义[J].油气地质与采收率,2020,27(2):62-71.DU Wei,JI Youliang,LI Qihai,et al.Sedimentary forward numeri⁃ cal modeling at different sedimentary scales:Progress and hydro⁃ carbon significance[J].Petroleum Geology and Recovery Efficien⁃ cy,2020,27(2):62-71.

    • [3] 崔龙涛,张倩萍,郭诚,等.坡折带主控下粗粒三角洲沉积演化及储层预测——以KLA油田为例[J].断块油气田,2019,26(4):444-448.CUI Longtao,ZHANG Qianping,GUO Cheng,et al.Sedimentary evolution of coarse-grained delta and prediction of favorable res⁃ ervoir under control of slope zone:taking KLA oilfield for example [J].Fault-Block Oil and Gas Field,2019,26(4):444-448.

    • [4] 张春生,刘忠保,施冬,等.利用数值模拟方法预测辫状河砂体几何形态[J].石油勘探与开发,2004,31(z1):85-88.ZHANG Chunsheng,LIU Zhongbao,SHI Dong,et al.Braid river sandbody shape forecast by numerical simulation method[J].Pe⁃ troleum Exploration and Development,2004,31(z1):85-88.

    • [5] 李少华,张昌民,林克湘,等.储层建模中几种原型模型的建立 [J].沉积与特提斯地质,2004,24(3):102-107.LI Shaohua,ZHANG Changmin,LIN Kexiang,et al.The construc⁃ tion of prototype models in reservoir modeling[J].Sedimentary Geology and Tethyan Geology,2004,24(3):102-107.

    • [6] 岳大力,吴胜和,刘建民.曲流河点坝地下储层构型精细解剖方法[J].石油学报,2007,28(4):99-103.YUE Dali,WU Shenghe,LIU Jianmin.An accurate method for anatomizing architecture of subsurface reservoir in point bar of meandering river[J].Acta Petrolei Sinica,2007,28(4):99-103.

    • [7] 严科,杨少春,任怀强.基于油藏开发动态的储层四维模型的建立[J].中国石油大学学报:自然科学版,2010,34(1):12-17.YAN Ke,YANG Shaochun,REN Huaiqiang.Building of 4D reser⁃ voir model based on development performance[J].Journal of Chi⁃ na University of Petroleum:Edition of Natural Science,2010,34(1):12-17.

    • [8] 史长林,张凤红,陈平.水驱实验模拟注水开发对储层的影响 [J].西南石油大学学报:自然科学版,2013,35(5):87-93.SHI Changlin,ZHANG Fenghong,CHEN Ping.Affection of simu⁃ lating water-flooding by water injection tests on reservoir proper⁃ ties[J].Journal of Southwest Petroleum University:Science & Technology Edition,2013,35(5):87-93.

    • [9] 冷雪梅,杜启振,孟宪军.基于纵横波弹性阻抗联合反演的储层流体检测方法[J].断块油气田,2019,26(3):319-323,363.LENG Xuemei,DU Qizhen,MENG Xianjun.Reservoir fluid detec⁃ tion method based on joint PP-and PS-wave elastic impedance inversion[J].Fault-Block Oil and Gas Field,2019,26(3):319-323,363.

    • [10] 程丽芳,王晖,尹艳树,等.基于浅层地震资料的曲流河点坝精细解剖——以加拿大某区块Mcmurray组储层为例[J].断块油气田,2019,26(4):464-469.CHENG Lifang,WANG Hui,YIN Yanshu,et al.Fine dissection of meandering river point bars based on shallow seismic data:a case study of Mcmurray reservoir in a block,Canada[J].Fault-Block Oil and Gas Field,2019,26(4):464-469.

    • [11] 高秋菊,谭明友,张营革,等.陆相页岩油“甜点”井震联合定量评价技术——以济阳坳陷罗家地区沙三段下亚段为例[J].油气地质与采收率,2019,26(1):165-173.GAO Qiuju,TAN Mingyou,ZHANG Yingge,et al.Well-seismic joint technology for quantitative evaluation of“sweet spot”in con⁃ tinental shale oil:A case study of Lower Es3 Member of Luojia ar⁃ ea in Jiyang Depression[J].Petroleum Geology and Recovery Effi⁃ ciency,2019,26(1):165-173.

    • [12] 岳丽君,钱艳苓.基于地震波形指示反演技术的窄薄储层预测 [J].大庆石油地质与开发,2020,39(5):135-140.YUE Lijun,QIAN Yanling.Prediction of the narrow-thin reservoir based on seismic waveform indication inversion technology[J].Pe⁃ troleum Geology & Oilfield Development in Daqing,2020,39(5):135-140.

    • [13] 李宏为,李凡异,齐秋红,等.基于地震多属性分析的吉尔嘎朗图凹陷煤层气富集区预测[J].大庆石油地质与开发,2020,39(4):135-142.LI Hongwei,LI Fanyi,QI Qiuhong,et al.Prediction of CBM en⁃ riched area in Jirgalangtu Sag based on the seismic multiple-attri⁃ bute analysis[J].Petroleum Geology & Oilfield Development in Daqing,2020,39(4):135-142.

    • [14] 张晶玉,范廷恩,王宗俊,等.多属性储层参数反演方法在M油田中的应用[J].油气藏评价与开发,2019,9(1):8-14.ZHANG Jingyu,FAN Ting’en,WANG Zongjun,et al.Application of reservoir parameters inversion method by using multi-attri⁃ butes in M oilfield[J].Reservoir Evaluation and Development,2019,9(1):8-14.

    • [15] 宋辉,陈伟,李谋杰,等.基于卷积门控循环单元网络的储层参数预测方法[J].油气地质与采收率,2019,26(5):73-78.SONG Hui,CHEN Wei,LI Moujie,et al.A method to predict res⁃ ervoir parameters based on convolutional neural network-gated recurrent unit(CNN-GRU)[J].Petroleum Geology and Recovery Efficiency,2019,26(5):73-78.

    • [16] 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.

    • [17] 李政宏,刘永福,张立强,等.数据挖掘方法在测井岩性识别中的应用[J].断块油气田,2019,26(6):713-718.LI Zhenghong,LIU Yongfu,ZHANG Liqiang,et al.Application of data mining method in lithology identification using well log[J].Fault-Block Oil and Gas Field,2019,26(6):713-718.

    • [18] 李婷婷,王钊,马世忠,等.地震属性融合方法综述[J].地球物理学进展,2015,30(l):378-385.LI Tingting,WANG Zhao,MA Shizhong,et al.Summary of seis⁃ mic attribute fusion method[J].Progress in Geophysics,2015,30(l):378-385.

    • [19] 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.

    • [20] 翟明国.大数据定将改变地质——向读者推荐《地球科学大数据挖掘与机器学习》[J].矿物岩石地球化学通报,2018,37(6):1 209..ZHAI Mingguo.Big data will change geology-Big Data Mining and Machine Learning in Geoscience is recommended to the read⁃ er[J].Bulletin of Mineralogy,Petrology and Geochemistry,2018,37(6):1 209.

    • [21] 林年添,张栋,张凯,等.地震油气储层的小样本卷积神经网络学习与预测[J].地球物理学报,2018,61(10):4 110-4 125.LIN Niantian,ZHANG Dong,ZHANG Kai,et al.Predicting distri⁃ bution of hydrocarbon reservoirs with seismic data based on learn⁃ ing of the small-sample convolution neural network[J].Chinese Journal of Geophysics,2018,61(10):4 110-4 125.

    • [22] 付超,林年添,张栋,等.多波地震深度学习的油气储层分布预测案例[J].地球物理学报,2018,61(1):293-303.FU Chao,LIN Niantian,ZHANG Dong,et al.Prediction of reser⁃ voirs using multi-component seismic data and the deep learning method[J].Chinese Journal of Geophysics,2018,61(1):293-303.

    • [23] 林年添,付超,张栋,等.无监督与监督学习下的含油气储层预测[J].石油物探,2018,57(4):601-610.LIN Niantian,FU Chao,ZHANG Dong,et al.Supervised learning and unsupervised learning for hydrocarbon prediction using multi⁃ wave seismic data[J].Geophysical Prospecting for Petroleum,2018,57(4):601-610.

    • [24] 李国欣,王峰,皮学军,等.非常规油气藏地质工程一体化数据优化应用的思考与建议[J].中国石油勘探,2019,24(2):147-152.LI Guoxin,WANG Feng,PI Xuejun,et al.Optimized application of geology-engineering integration data of unconventional oil and gas reservoirs[J].China Petroleum Exploration,2019,24(2):147-152.

    • [25] 甘利灯,张昕,王峣钧,等.从勘探领域变化看地震储层预测技术现状和发展趋势[J].石油地球物理勘探,2018,53(1):214-225.GAN Lideng,ZHANG Xin,WANG Yaojun,et al.Current status and development trends of seismic reservoir prediction viewed from the exploration industry[J].Oil Geophysical Prospecting,2018,53(1):214-225.