﻿ 地震属性优化与网络函数逼近储层砂体厚度预测方法及应用 The Method of Sand Body Thickness Prediction Based on Attribute Optimization and Network Function Approximation and Its Application

Journal of Oil and Gas Technology
Vol.39 No.02(2017), Article ID:20467,6 pages
10.12677/JOGT.2017.392014

The Method of Sand Body Thickness Prediction Based on Attribute Optimization and Network Function Approximation and Its Application

Xueguo Chen

West Branch of the Research Institute of Exploration and Development, Shengli Oilfield Company, SINOPEC, Dongying Shandong

Received: Nov. 16th, 2016; accepted: Feb. 15th, 2017; published: Apr. 15th, 2017

ABSTRACT

The sand body thickness (or content) was an important parameter in oil and gas exploration. Based on seismic data and well-logging interpretation, a method for predicting sand body thickness in reservoirs was proposed in this paper. A sensitive seismic attribute set was established by using attribute optimization and dimension reduction. A neural network was established with the input including sensitive attributes of borehole seismic trace and sand body thickness interpretation of well log data. The training of the network would minimize the error, and on this basis, the sensitive attribute was input for each trace and the corresponding sand body thickness was output through network. In Block H4 of Shengli Oilfield the above method is applied to predict sandstone thickness, and the relative error is less than 20%, and it basically meets the need of oil exploration and production.

Keywords:Sand Body Thickness, Seismic Attribute Optimization, Network Function Approximation, Reservoir Prediction

1. 引言

2. 属性优化与神经网络函数逼近方法

2.1. 地震属性优化

2.2. 神经网络函数逼近方法

Figure 1. The network structure of sand body thickness prediction

(1)

3. 砂体厚度预测流程

1) 地震资料的构造精细解释。在层位标定的基础上，对目的层顶、底界面作地震构造精细解释，并提取目的层顶、底界面之间的地震数据。

2) 地震属性的提取，建立敏感属性集。对目的层段地震数据，作傅里叶变换、小波变换、希尔伯特变换、相关运算、谱比法吸收系数提取等不同的数学运算，得到目的层段的地震属性。

3) 地震属性的优化。对上面提取的地震属性集通过K-L变换优化，确立对砂体厚度敏感的属性集。

4) 神经网络的训练。设计神经网络的结构，根据井旁敏感的地震属性和测井解释砂体厚度或录井测得砂体厚度，建立神经网络训练集，输入设计的网络，按照BP算法原理进行网络训练，确定神经网络内部各个节点之间连接的权系数和节点上的阈值。

5) 储层砂体厚度分布预测。将敏感地震属性集逐道输入训练后的网络，输出即为该道对应的砂体厚度。

4. 应用效果与认识

5. 结论

1) 不同的储层参数，通过地震属性优化分析能提取反映敏感的一组地震属性，优化后的属性集有利于提高储层参数预测精度。

2) 该次研究提出的基于地震属性优化与网络函数逼近储层砂体预测方法是可行的，在胜利油田H4区块的应用中，预测结果与井资料的符合率较高，相对误差均在20%以下。

Figure 2. The prediction of sandstone thickness distribution of in the studied area

Figure 3. The prediction of sandstone thickness distribution of in the studied area

The Method of Sand Body Thickness Prediction Based on Attribute Optimization and Network Function Approximation and Its Application[J]. 石油天然气学报, 2017, 39(02): 30-35. http://dx.doi.org/10.12677/JOGT.2017.392014

1. 1. 贺振华, 黄德济, 胡兴岷, 等. 复杂油气藏地震波场特征方法理论及应用[M]. 成都: 四川科学技术出版社, 1999.

2. 2. 桂志先, 朱广生, 段天友, 等. 不同类型油气藏地震波特征响应初探[C]//中国地球物理学会年刊. 上海: 同济大学出版社, 1997.

3. 3. 刘兰锋, 刘全新, 雍学善, 等. 基于广义S变换的低频瞬时能量谱油气检测技术[J]. 天然气地球科学, 2005, 16(2): 238-241.

4. 4. 赵成, 桂志先. 基于神经网络的储层参数预测方法及应用[J]. 石油天然气学报(江汉石油学院学报), 2005, 27(3): 467-468.

5. 5. 陈遵德. 储层地震属性优化方法[M]. 北京: 石油工业出版社, 1998.

6. 6. 桂志先, 高刚. 油藏地球物理[M]. 北京: 石油工业出版社, 2015.