神经网络方法及储层动态模型的研究
|
摘要
长期注水开发过程中,地下储层参数在开发流体的作用下都发生了演变,建立能够反映储层参数在空间和时间上变化规律的储层动态模型,对于提高油气采收率、延长油田的寿命有着非常重要的意义。
本文通过综合应用多学科的理论、方法和技术,并借助于计算机的手段,将数学和储层地质紧密结合,提出了一套新的建立储层四维模型的方法。主要取得了以下成果和认识:①分析了神经网络的算法原理和基本结构,并对常用于储层建模中的BP神经网络进行了研究,总结了传统的BP算法的推导。②针对BP算法在储层建模中困难和问题,提出了新的改进措施,即利用遗传算法来优化BP神经网络的权值和阈值,从而能够有效避免网络训练的局部最优化问题,节省网络的训练时间,使得所建模型能更好的的符合精度的要求。③充分利用计算机的手段,通过编程语言来实现遗传算法优化的BP算法,并利用新的算法来进行网络训练,建立起GA-BP神经网络的储层建模系统。④提出了一种新的建立储层四维模型的方法,即首先利用人工神经网络建立起储层参数的预测模型,得到储层参数的四维数据体,然后利用建模软件建立各个开发阶段的三维地质模型,从而实现了储层地质的四维建模。⑤以胜坨油田二区沙二段8砂层组3小层的三角洲前缘相储层为例,建立了储层宏观参数的四维模型,并对所建模型进行验证,能够得到较满意的效果。
Reservoir parameters have been changed by the exploitation fluid. The reservoir dynamic model, which can reflect the reservoir parameter’s variation based on spatial and temporal, is set up. This has a significant meaning to lift the recovery ratio of oilfield and extend the life of oilfield.
Mathematics and reservoir geology have been combined closely by applying synthetically multi-subject theories, methods and technology and making full use of computer. A new method to build the 4D reservoir model is proposed. Main achievements of the study are summarized as following: (1) Traditional BP derivation algorithm is put forward through the study of algorithm principle and basic structure of neural network and the BP neural network which commonly used in reservoir modeling. (2) New improvement measure (optimization weight and threshold of BP neural network based on genetic algorithm) is offered to overcome difficulties of BP algorithm applied in reservoir modeling. Then, the network can effectively avoid local optimization problem of network training and save network training time. The model built in this paper can meet the requirement of high accuracy. (3) Optimization BP algorithm based on genetic algorithm is offered by making full use of computer, especially programming language. And reservoir modeling system of GA-BP neural network is built by the optimization BP algorithm. (4) A new method to built 4D reservoir model is proposed in this paper. Firstly, build a prediction model of reservoir parameters to have its 4Ddata volume. Secondly, build 3D geological models of each development phases by means of modeling software to get the 4D reservoir model. (5) A 4D reservoir model of macro-parameter is built based on change rule of delta reservoir parameters were researched in Layer 83 of Es 2 in the second block of Shengtuo Oilfield exploited by water-flooding for a long time, and have a flavor result by verification.
本文通过综合应用多学科的理论、方法和技术,并借助于计算机的手段,将数学和储层地质紧密结合,提出了一套新的建立储层四维模型的方法。主要取得了以下成果和认识:①分析了神经网络的算法原理和基本结构,并对常用于储层建模中的BP神经网络进行了研究,总结了传统的BP算法的推导。②针对BP算法在储层建模中困难和问题,提出了新的改进措施,即利用遗传算法来优化BP神经网络的权值和阈值,从而能够有效避免网络训练的局部最优化问题,节省网络的训练时间,使得所建模型能更好的的符合精度的要求。③充分利用计算机的手段,通过编程语言来实现遗传算法优化的BP算法,并利用新的算法来进行网络训练,建立起GA-BP神经网络的储层建模系统。④提出了一种新的建立储层四维模型的方法,即首先利用人工神经网络建立起储层参数的预测模型,得到储层参数的四维数据体,然后利用建模软件建立各个开发阶段的三维地质模型,从而实现了储层地质的四维建模。⑤以胜坨油田二区沙二段8砂层组3小层的三角洲前缘相储层为例,建立了储层宏观参数的四维模型,并对所建模型进行验证,能够得到较满意的效果。
Reservoir parameters have been changed by the exploitation fluid. The reservoir dynamic model, which can reflect the reservoir parameter’s variation based on spatial and temporal, is set up. This has a significant meaning to lift the recovery ratio of oilfield and extend the life of oilfield.
Mathematics and reservoir geology have been combined closely by applying synthetically multi-subject theories, methods and technology and making full use of computer. A new method to build the 4D reservoir model is proposed. Main achievements of the study are summarized as following: (1) Traditional BP derivation algorithm is put forward through the study of algorithm principle and basic structure of neural network and the BP neural network which commonly used in reservoir modeling. (2) New improvement measure (optimization weight and threshold of BP neural network based on genetic algorithm) is offered to overcome difficulties of BP algorithm applied in reservoir modeling. Then, the network can effectively avoid local optimization problem of network training and save network training time. The model built in this paper can meet the requirement of high accuracy. (3) Optimization BP algorithm based on genetic algorithm is offered by making full use of computer, especially programming language. And reservoir modeling system of GA-BP neural network is built by the optimization BP algorithm. (4) A new method to built 4D reservoir model is proposed in this paper. Firstly, build a prediction model of reservoir parameters to have its 4Ddata volume. Secondly, build 3D geological models of each development phases by means of modeling software to get the 4D reservoir model. (5) A 4D reservoir model of macro-parameter is built based on change rule of delta reservoir parameters were researched in Layer 83 of Es 2 in the second block of Shengtuo Oilfield exploited by water-flooding for a long time, and have a flavor result by verification.
引文
[1]林博,戴俊生,冀国盛,等.胜坨油田坨7块沙二段91夹层随机建模研究[J] .西南石油大学学报(自然科学版), 2008,30(4) : 11-14
[2]吴胜和,李宇鹏.储层地质建模的现状与展望[J] .海相油气地质,2007,12(3) : 53-59
[3]裘怿楠,贾爱林.储层地质模型10年[J].石油学报,2000,21(4):101-104
[4] Windrow B, et al.Neural Networks Applications in Industry[J].Business and Science.Communication of the ACM,1994,37:93-105
[5]付国强,张国栋,吴义杰,等.河流相砂岩油藏综合地质评价建模[J].石油学报,2000,21(5) : 21-26
[6]李健.三角洲低渗透储层流动单元四维模型及剩余油预测[M].北京:石油工业出版社,2004: 9-81
[7]徐守余,李卫成.储层微观动态模型研究[J].石油实验地质,2004,26(4) : 397-400
[8]张继春,彭仕宓,穆立华,等.流动单元四维动态演化仿真模型研究[J].石油学报,2005,26(1) : 69-73
[9]彭仕宓,尹志军,李海燕.建立储层四维地质模型的新尝试[J] .地质论评,2004,50(6):662-665
[10]高博禹,彭仕宓,陈烨菲.储层动态流动单元及剩余油分布规律[J].吉林大学学报(地球科学版),2005,35(2) : 182-187
[11]邓玉珍,徐守余.三角洲储层渗流参数动态模型研究—以胜坨油田二区下第三系沙二段83小层为例[J].石油学报,2003,24(2) : 61-64
[12]孙梦茹,刘文业.河流三角洲储层油藏动态模型和剩余油分布[M].北京:石油工业出版社,2004: 63-84
[13]束青林,张本华.河道砂储层油藏动态模型和剩余油预测[M].北京:石油工业出版社,2004: 102-128
[14]孙国.利用人工神经网络系统建立储层四维地质模型[J].油气地质与采收率,2004,11(3) : 4-6
[15]徐守余,王艳红.利用神经网络建立储层宏观参数动态模型-以胜坨油田二区为例[J].油气地质与采收率,2005,12(6) : 10-12
[16]束青林,张本华,徐守余.孤岛油田河道砂储集层油藏动态模型及剩余油研究[J].石油学报,2005,26(3) : 64-73
[17]李红南,王德军.油藏动态模型和剩余油仿真模型[J].石油学报,2006,27(5) : 83-87
[18]徐守余,李红南.储集层孔喉网络场演化规律和剩余油分布[J].石油学报,2003,24(4) : 48-53
[19]李健,李红南.油藏开发流体动力地质作用对储集层的改造[J].石油勘探与开发,2003,30(5) : 86-89
[20]徐守余.渤海湾盆地胜坨油田二区储层微观渗流场演化研究[J].石油实验地质,2003,25(4) : 381-384
[21] McCulloch.W.S. and W.H.Pitts . A Logical Calculus of the Ideas Immanent in Neuron Activity[J]. Bulletin Mathematical Biophysics,1943,VOL,5: 115-133
[22] Hebb.D.O. The Organization of Behavior.Wiley[C] .New York,1949:17-19
[23]陈翰馥.随机递推估计[M].北京:科学出版社,1984:10-21
[24] Hopfield,J.J.and Tank,D.W.Neural Computation of Decision in Optimization problems[C] .Biolog Cybem,1984:3-68
[25]孙焕泉.油藏动态模型和剩余油分布模式[M].北京:石油工业出版社,2002:5-35
[26]杨少春,潘少伟,等.储层四维建模方法研究,天然气地球科学,2009,20(3) : 420-424
[27]董长虹,等.神经网络及应用[M].北京:国防工业出版社,2005:4-65
[28]张青贵.人工神经网络导论[M].北京:中国水利水电出版社,2004:34-67
[29]宋宜斌,王培进.基于径向基函数神经网络的非线性模型辨识[J].计算机工程,2004,30(5) : 142-146
[30]张秀玲,李海滨.基于RBF神经网络的数字模式方法[J].模式识别与人工智能,2002,15(1) : 93-97
[31]盛仲飙.BP神经网络原理及MATLAB仿真[J].渭南师范学院学报,2008,23(5) : 65-67
[32] Kowalczyk A, Ferra H.Generalization in Feedforword Networks[J],NIPS 7,1995,Cambridge,MA:215-222
[33]周彩兰,刘敏.BP神经网络在石油产量预测中的应用[J].武汉理工大学学报,2009,31(3) : 125-129
[34] Broom head DS,Lowe D. Multivariable. Functional interpolation and adaptive networks[J]. Complex System.1988,2(3):321-355
[35]王永毅,徐健.神经元网络控制[M].北京:机械工业出版社,1998:56-67
[36]袁增任.人工神经网络及其应用.北京:清华大学出版社,1999:33-54
[37]贺清碧,周建丽.BP神经网络收敛性问题的改进措施[J].重庆交通学院学报.2005,24(1): 143-145
[38]谭显胜,周铁军.BP算法改进方法的研究进展[J].怀化学院学报,2006,25(6):126-130
[39]丁承民,张传生,刘辉.遗传算法纵横谈[J].信息与控制,1997,26(1):40-46
[40]雷英杰,张善文,等.遗传算法工具箱及应用[M].西安:西安电子科技大学出版社,2005:45-76
[41]张栋,蔡开元.基于遗传算法的神经网络两阶段学习方案[J].系统仿真学报.2003.15(8):1088-1090
[42]王凤琴,高颖,赵军.基于遗传算法的神经网络优化[J].燕山大学学报,2001.25(3): 236
[43] Irie B,Miyake S.Capabilities of three-layered perceptrons[J].Proceedings of the IEEE Inter Conf on Neural Network,1998(1): 641-648
[44]焦李成.神经网络系统理论[M].西安:西安电子科技大学出版社,1992
[45]高大启.有教师的线性基奉函数前向三层神经网络结构研究[J].计算机学报,1998,21(1):80-86
[46]杨斌,匡立春,等.神经网络及其在石油测井中的应用[M].北京:石油工业出版社,2005
[47]王崇骏,于汉滁,陈兆乾,谢俊元.一种基于遗传算法的BP神经网络算法及其应用[J].南京大学学报(自然科学版),2003.39(5):460-464
[48]阮若林.利用遗传算法优化人工神经网络权值[J].咸宁学院学报,2005,25(6):49-51
[49]刘瑞斋.神经网络与遗传算法在优化设计中的结合应用[J].舰船电子工程,2008,169(7) : 127-130
[50]苏金明,黄国明,刘波.MATLAB与外部程序接口[M].北京:电子工业出版社,2003: 167-169
[51]范磊,张运陶,等.基于Matlab的改进BP神经网络及应用[J].西华师范大学学报(自然科学版),2005,26(1) : 70-73
[52]欧阳玉梅,马志强,等.基于MATLAB的遗传神经网络设计与实现[J].信息技术,2008,10(6): 73-80
[53]王玉良.基于MATLAB的遗传算法可视化仿真系统研究[J].计算机系统应用,2004,7(10): 41-43
[54]王佶,李耀雄,等.基于神经网络和遗传算法的实验参数确定[J].武汉理工大学学报,2004,26(5) : 52-54
[55]黄玮,刘卫民,等.基于遗传算法的BP网络最优权值算法[J].昆明冶金高等专科学校学报,2006,22(5) : 37-43
[56]卢洵.基于遗传算法的神经网络优化的研究[J].信息工程学院学报,1999,18(3) : 17-19
[57]孙孟茹,周建林,等.胜坨油田精细地质研究[M].北京:中国石化出版社,2004:2-65
[58]高博禹,彭仕宓,陈烨菲.储层动态流动单元及剩余油分布规律[J].吉林大学学报(地球科学版),2005,35(2) : 182-187
[59] ]严科,杨少春,任怀强.储层宏观非均质性定量表征研究[J].石油学报,2008,29(6) : 870-874
[60]宋万超,孙焕泉,孙国,等.油藏开发流体动力地质作用—以胜坨油田二区为例[J].石油学报,2002,23(3) : 52-55
[2]吴胜和,李宇鹏.储层地质建模的现状与展望[J] .海相油气地质,2007,12(3) : 53-59
[3]裘怿楠,贾爱林.储层地质模型10年[J].石油学报,2000,21(4):101-104
[4] Windrow B, et al.Neural Networks Applications in Industry[J].Business and Science.Communication of the ACM,1994,37:93-105
[5]付国强,张国栋,吴义杰,等.河流相砂岩油藏综合地质评价建模[J].石油学报,2000,21(5) : 21-26
[6]李健.三角洲低渗透储层流动单元四维模型及剩余油预测[M].北京:石油工业出版社,2004: 9-81
[7]徐守余,李卫成.储层微观动态模型研究[J].石油实验地质,2004,26(4) : 397-400
[8]张继春,彭仕宓,穆立华,等.流动单元四维动态演化仿真模型研究[J].石油学报,2005,26(1) : 69-73
[9]彭仕宓,尹志军,李海燕.建立储层四维地质模型的新尝试[J] .地质论评,2004,50(6):662-665
[10]高博禹,彭仕宓,陈烨菲.储层动态流动单元及剩余油分布规律[J].吉林大学学报(地球科学版),2005,35(2) : 182-187
[11]邓玉珍,徐守余.三角洲储层渗流参数动态模型研究—以胜坨油田二区下第三系沙二段83小层为例[J].石油学报,2003,24(2) : 61-64
[12]孙梦茹,刘文业.河流三角洲储层油藏动态模型和剩余油分布[M].北京:石油工业出版社,2004: 63-84
[13]束青林,张本华.河道砂储层油藏动态模型和剩余油预测[M].北京:石油工业出版社,2004: 102-128
[14]孙国.利用人工神经网络系统建立储层四维地质模型[J].油气地质与采收率,2004,11(3) : 4-6
[15]徐守余,王艳红.利用神经网络建立储层宏观参数动态模型-以胜坨油田二区为例[J].油气地质与采收率,2005,12(6) : 10-12
[16]束青林,张本华,徐守余.孤岛油田河道砂储集层油藏动态模型及剩余油研究[J].石油学报,2005,26(3) : 64-73
[17]李红南,王德军.油藏动态模型和剩余油仿真模型[J].石油学报,2006,27(5) : 83-87
[18]徐守余,李红南.储集层孔喉网络场演化规律和剩余油分布[J].石油学报,2003,24(4) : 48-53
[19]李健,李红南.油藏开发流体动力地质作用对储集层的改造[J].石油勘探与开发,2003,30(5) : 86-89
[20]徐守余.渤海湾盆地胜坨油田二区储层微观渗流场演化研究[J].石油实验地质,2003,25(4) : 381-384
[21] McCulloch.W.S. and W.H.Pitts . A Logical Calculus of the Ideas Immanent in Neuron Activity[J]. Bulletin Mathematical Biophysics,1943,VOL,5: 115-133
[22] Hebb.D.O. The Organization of Behavior.Wiley[C] .New York,1949:17-19
[23]陈翰馥.随机递推估计[M].北京:科学出版社,1984:10-21
[24] Hopfield,J.J.and Tank,D.W.Neural Computation of Decision in Optimization problems[C] .Biolog Cybem,1984:3-68
[25]孙焕泉.油藏动态模型和剩余油分布模式[M].北京:石油工业出版社,2002:5-35
[26]杨少春,潘少伟,等.储层四维建模方法研究,天然气地球科学,2009,20(3) : 420-424
[27]董长虹,等.神经网络及应用[M].北京:国防工业出版社,2005:4-65
[28]张青贵.人工神经网络导论[M].北京:中国水利水电出版社,2004:34-67
[29]宋宜斌,王培进.基于径向基函数神经网络的非线性模型辨识[J].计算机工程,2004,30(5) : 142-146
[30]张秀玲,李海滨.基于RBF神经网络的数字模式方法[J].模式识别与人工智能,2002,15(1) : 93-97
[31]盛仲飙.BP神经网络原理及MATLAB仿真[J].渭南师范学院学报,2008,23(5) : 65-67
[32] Kowalczyk A, Ferra H.Generalization in Feedforword Networks[J],NIPS 7,1995,Cambridge,MA:215-222
[33]周彩兰,刘敏.BP神经网络在石油产量预测中的应用[J].武汉理工大学学报,2009,31(3) : 125-129
[34] Broom head DS,Lowe D. Multivariable. Functional interpolation and adaptive networks[J]. Complex System.1988,2(3):321-355
[35]王永毅,徐健.神经元网络控制[M].北京:机械工业出版社,1998:56-67
[36]袁增任.人工神经网络及其应用.北京:清华大学出版社,1999:33-54
[37]贺清碧,周建丽.BP神经网络收敛性问题的改进措施[J].重庆交通学院学报.2005,24(1): 143-145
[38]谭显胜,周铁军.BP算法改进方法的研究进展[J].怀化学院学报,2006,25(6):126-130
[39]丁承民,张传生,刘辉.遗传算法纵横谈[J].信息与控制,1997,26(1):40-46
[40]雷英杰,张善文,等.遗传算法工具箱及应用[M].西安:西安电子科技大学出版社,2005:45-76
[41]张栋,蔡开元.基于遗传算法的神经网络两阶段学习方案[J].系统仿真学报.2003.15(8):1088-1090
[42]王凤琴,高颖,赵军.基于遗传算法的神经网络优化[J].燕山大学学报,2001.25(3): 236
[43] Irie B,Miyake S.Capabilities of three-layered perceptrons[J].Proceedings of the IEEE Inter Conf on Neural Network,1998(1): 641-648
[44]焦李成.神经网络系统理论[M].西安:西安电子科技大学出版社,1992
[45]高大启.有教师的线性基奉函数前向三层神经网络结构研究[J].计算机学报,1998,21(1):80-86
[46]杨斌,匡立春,等.神经网络及其在石油测井中的应用[M].北京:石油工业出版社,2005
[47]王崇骏,于汉滁,陈兆乾,谢俊元.一种基于遗传算法的BP神经网络算法及其应用[J].南京大学学报(自然科学版),2003.39(5):460-464
[48]阮若林.利用遗传算法优化人工神经网络权值[J].咸宁学院学报,2005,25(6):49-51
[49]刘瑞斋.神经网络与遗传算法在优化设计中的结合应用[J].舰船电子工程,2008,169(7) : 127-130
[50]苏金明,黄国明,刘波.MATLAB与外部程序接口[M].北京:电子工业出版社,2003: 167-169
[51]范磊,张运陶,等.基于Matlab的改进BP神经网络及应用[J].西华师范大学学报(自然科学版),2005,26(1) : 70-73
[52]欧阳玉梅,马志强,等.基于MATLAB的遗传神经网络设计与实现[J].信息技术,2008,10(6): 73-80
[53]王玉良.基于MATLAB的遗传算法可视化仿真系统研究[J].计算机系统应用,2004,7(10): 41-43
[54]王佶,李耀雄,等.基于神经网络和遗传算法的实验参数确定[J].武汉理工大学学报,2004,26(5) : 52-54
[55]黄玮,刘卫民,等.基于遗传算法的BP网络最优权值算法[J].昆明冶金高等专科学校学报,2006,22(5) : 37-43
[56]卢洵.基于遗传算法的神经网络优化的研究[J].信息工程学院学报,1999,18(3) : 17-19
[57]孙孟茹,周建林,等.胜坨油田精细地质研究[M].北京:中国石化出版社,2004:2-65
[58]高博禹,彭仕宓,陈烨菲.储层动态流动单元及剩余油分布规律[J].吉林大学学报(地球科学版),2005,35(2) : 182-187
[59] ]严科,杨少春,任怀强.储层宏观非均质性定量表征研究[J].石油学报,2008,29(6) : 870-874
[60]宋万超,孙焕泉,孙国,等.油藏开发流体动力地质作用—以胜坨油田二区为例[J].石油学报,2002,23(3) : 52-55