砂泥岩地层概率神经网络岩性反演技术应用研究
|
摘要
概率神经网络是一种基于概率密度函数理论的神经网络,能够广泛地应用于模式识别等领域.针对地震岩性反演预测问题,提出了一种具体的概率神经网络方法,包括网络模型的构造和预测识别步骤等.研究区主要目的层为沙溪庙组沙一段湖滩砂及河道砂体,储层单层厚度小,岩性横向变化较大,利用地震资料进行常规储层预测较困难.为此,根据该区储层的测井响应特征、地震属性特征与地质岩性特征的相关性,利用概率神经网络方法对地震属性数据做变换,从而对地层特征进行预测识别.
Probabilistic neural network is based on the theory of probabilistic density function,and it is widely used in many realms,such as patter recognition,etc.A specific method of probabilistic neural network is presented for the lithology inversion of seismic data,in which the construction of the network model and the steps of prediction are included.The target strata in the studied area are the beach sandstone and channel sandstone of Shayi member of Shaximiao formation,and because their reservoir thickness is little and the lateral change of their lithology is great,it is difficult to predict the reservoir lithology based on seismic data by conventional methods.For this reason,According to the correlation among logging responses,seismic attributes and geological lithology characteristics,the probabilistic neural network method is used for transforming the seismic attributes to identify the reservoir lithology information.And a case shows that the technique has a good application result.
Probabilistic neural network is based on the theory of probabilistic density function,and it is widely used in many realms,such as patter recognition,etc.A specific method of probabilistic neural network is presented for the lithology inversion of seismic data,in which the construction of the network model and the steps of prediction are included.The target strata in the studied area are the beach sandstone and channel sandstone of Shayi member of Shaximiao formation,and because their reservoir thickness is little and the lateral change of their lithology is great,it is difficult to predict the reservoir lithology based on seismic data by conventional methods.For this reason,According to the correlation among logging responses,seismic attributes and geological lithology characteristics,the probabilistic neural network method is used for transforming the seismic attributes to identify the reservoir lithology information.And a case shows that the technique has a good application result.
引文
[1]杨文采.地球物理反演的理论与方法[M].北京:地质出版社,1997:242-267.
[2]Zhengping Liu,Jiaqi Liu.Seismic-controlled nonlinear ex-trapolation of well parameters using neural networks[J].Geophysics,1998,63(6):2035-2041.
[3]Paul F M G,Albertus H B.Monte Carlo si mulation ofwells[J].Geophysics,1996,61(3):631-638.
[4]McCormack M D.Neural computing in geophysics[J].The Leading Edge,1991,10(1):11-15.
[5]李宗田,刘伟.非线性约束储层反演技术在卫城油田的应用[J].西安石油大学学报:自然科学版,2005,20(5):17-21.
[6]Masters T.Signal and image processing with neural net-works[M].John Wiley&Sons Incorporation,1994.
[7]Specht D.A general regression neural network:IEEETransaction[C]//Neural Networks.1991:568-576.
[8]韩军,梁全胜,常迈,等.反演技术在低勘探程度区域岩性油气藏勘探中的应用[J].西安石油大学学报:自然科学版,2007,22(1):49-52.
[9]徐旺林,庞雄奇,吕淑英,等.动态概率神经网络及油气概率分布预测[J].石油地球物理勘探,2005,40(1):65-70.
[10]张志国,张福田,高永利.马岭油田中一区储层流动单元研究[J].西安石油大学学报:自然科学版,2006,21(1):35-38.
[2]Zhengping Liu,Jiaqi Liu.Seismic-controlled nonlinear ex-trapolation of well parameters using neural networks[J].Geophysics,1998,63(6):2035-2041.
[3]Paul F M G,Albertus H B.Monte Carlo si mulation ofwells[J].Geophysics,1996,61(3):631-638.
[4]McCormack M D.Neural computing in geophysics[J].The Leading Edge,1991,10(1):11-15.
[5]李宗田,刘伟.非线性约束储层反演技术在卫城油田的应用[J].西安石油大学学报:自然科学版,2005,20(5):17-21.
[6]Masters T.Signal and image processing with neural net-works[M].John Wiley&Sons Incorporation,1994.
[7]Specht D.A general regression neural network:IEEETransaction[C]//Neural Networks.1991:568-576.
[8]韩军,梁全胜,常迈,等.反演技术在低勘探程度区域岩性油气藏勘探中的应用[J].西安石油大学学报:自然科学版,2007,22(1):49-52.
[9]徐旺林,庞雄奇,吕淑英,等.动态概率神经网络及油气概率分布预测[J].石油地球物理勘探,2005,40(1):65-70.
[10]张志国,张福田,高永利.马岭油田中一区储层流动单元研究[J].西安石油大学学报:自然科学版,2006,21(1):35-38.
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心