| 时间延迟神经网络地震油气预测方法 |
| |
| 引用本文： | 刘瑞林,马在田.时间延迟神经网络地震油气预测方法[J].地球物理学报,1997,40(5):710-717. |
| |
| 作者姓名： | 刘瑞林 马在田 |
| |
| 作者单位： | 1. 江汉石油学院物探系, 荆州 434102;
2. 同济大学海洋系地质与地球物理系, 上海 200092 |
| |
| 摘 要： | 本文绘出基于时间延迟神经网络模型的地震油气预测方法及其初步应用结果，不同于通常的孤立模式识别方法．在特征提取阶段，不仅提取地震道中相应目的层单时窗的特征，同时也提取时窗滑动时的特征，这些多时窗的特征信息反映出地层层序的变化．时间延迟神经网络模型通过井旁道特征串的训练，用于表达特征信息与地层含油气情况的复杂关系和特征信息的变化与地层油气聚集的联系．初步应用表明，这种基于时间延迟网络模型的油气预测方法的结果要好于BP网络方法的结果．
|
| 关 键 词： | 时间延迟神经网络 油气预测 特征提取 孔隙介质 |
A METHOD OF TIME-DELAY NEURAL NETWORK FOR RESERVIOR LATERAL PREDICTION |
| |
| LIU RUI-LIN.A METHOD OF TIME-DELAY NEURAL NETWORK FOR RESERVIOR LATERAL PREDICTION[J].Chinese Journal of Geophysics,1997,40(5):710-717. |
| |
| Authors: | LIU RUI-LIN |
| |
| Institution: | 1. Dept of Geophysical Exploration, Jianghan Petroleum Institute, Jingzhou 434102, China;
2. Dept of Marine Geology and Geophysics, Tongji University, Shanghai 200092, China |
| |
| Abstract: | In this paper, We present a method of time-delay neural network (TDNN) for reservior lateral predication. It is different from common isolated pattern recognition method. In the phase of extracting features, we extract not only features in single time-window, but also features in multiple time-window. Those features in multiple time window could represent the variation of features with bine in the bee when the TDNN has been trained with the features of traces near the wells. It can not only represent the complex relationship between features containing oil/gas in the layer,but also the relationship between the variation of features in the layer. Examples show that the TDNN model is more suitable to predict the reservior laterallythan the BP model. |
| |
| Keywords: | Time-delay neural network Reservior lateral prediction Feature extracting Porous media |
| 本文献已被 CNKI 维普 等数据库收录！ |
| 点击此处可从《地球物理学报》浏览原始摘要信息 |
| 点击此处可从《地球物理学报》下载免费的PDF全文 |
|
