摘要
以濮阳凹陷白庙构造为研究区 ,以沙二下亚段第 1砂层组为目的层 ,根据神经网络原理 ,选取能够反映储层参数及油气分布特征的“S”型函数 ,构造出可以产生任意复杂判断的三层感知器 ,利用误差反传播算法 ,使估价函数最小化。提取 2 0余种地震波反射特征参数 ,将地震道旁的特征参数输入到神经网络中 ,使其学习、训练和记忆 ,给定神经网络各个节点间的连接权值和节点内部的阈值 ,通过神经网络所记忆的知识 ,用钻井、测井、地质、试油等资料进行约束 ,对未知区域的储层特征参数和含油气分布进行预测 ,取得了较好的效果。
Neural network was used to forecast the oil and gas distribution of Baimiao geological structure in Puyang of Shaanxi Province. To get optimal forecast results, a certain layer as the destination layer was taken and the Sigmoid function was used to reflect both the reservoir parameter and the character of oil and gas distribution. Thus, the neural network method was applied to construct the three-layer perception and the cost function was minimized with the error back-propagation algorithm. More than 20 parameters of seismic-wave echo characteristic were selected and the seismic-path side characteristic parameters were considered as the inputs of the neural network. By means of gradient optimal method, the invisible layer parameters and the thresholds of neural networks were determined. The neural network prediction method gave satisfactory forecast results for the Baimiao geological structure.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2003年第5期574-579,共6页
Journal of Northwestern Polytechnical University
基金
国家杰出青年基金 (6 992 5 30 6 )资助
关键词
神经网络
地震波
特征参数
油气预测
Forecasting
Geological surveys
Natural gas
Neural networks
Petroleum prospecting
Seismic waves
