摘要
应用地震资料进行井间储层孔隙度预测是当前油气勘探开发中常用的方法,但通常采用地震反演速度进行孔隙度定量预测难以准确反映孔隙度的空间变化。采用逐步回归和神经网络相结合的地震多属性变换法建立孔隙度模型,可实现孔隙度的空间预测,其核心是:首先通过逐步回归和交叉验证技术来确定最优地震属性及其数量,其次应用神经网络建立所选属性与测井解释孔隙度之间的映射关系,最终反演出孔隙度数据体,进而预测孔隙度的空间变化。以冀中坳陷束鹿凹陷西斜坡台家庄区块沙河街组二段(Es_2)为例,对该方法进行了分析验证。以交叉验证法作为质量控制手段,反演出沙河街组二段孔隙度数据体,直观地反映了沙河街组二段高孔隙度砂岩的展布特征。
The utilization of seismic data in crosswell reservoir porosity prediction is a normal method in current hydrocarbon exploration and development.However,the application of seismic inversion velocity for quantitatively predicting porosity cannot accurately reflect the spatial alternation of porosity.Therefore,we adopted seismic multi-attribute transformation method which combined stepwise regression and neural network to build porosity model and realize the spatial prediction of porosity.Firstly,optimal seismic attributes and its quantity are determined by stepwise regression and cross-validation method.Then,the mapping relation between the optimal seismic attributes and logging porosity is established by neural network. Finally,the porosity of data volume was inverted and its spatial distribution was predicted.Taking the second member of Shahejie Formation in Taijiazhuang Block of Taijiazhuang block of West Slope of Shulu Sag for an example,we applied the method on the data with cross-validation method as quality controlling method to invert the porosity of the interval.The data volume can directly reflect the distribution characteristics of high porosity sandstone in the 2nd Member of the Shahejie Formation.
出处
《石油物探》
EI
CSCD
北大核心
2011年第4期393-397,25,共5页
Geophysical Prospecting For Petroleum
关键词
孔隙度预测
地震多属性
逐步回归
交叉验证
神经网络
porosity prediction
seismic multi-attributes
stepwise regression
cross-validation
neural network
