摘要
探讨了应用神经网络技术预测储层孔隙度的方法 .选取储层的深度、厚度、岩性及砂地体积比 4个因素确立了预测孔隙度的神经网络结构 ,经网络的学习训练过程 ,确定了网络各层的连接权值 ,从而得到了稳定的孔隙度预测网络 .应用该网络对松辽盆地北部深层登娄库组孔隙度预测表明 ,所确定的孔隙度分布与实际状况符合 .应用该方法 ,不需要大量的数据 ,也不需要各参数之间的关系 ,即可得到不同区域内各因素对孔隙度的影响程度 .该方法适用于早期勘探缺乏足够资料条件下的孔隙度预测 .
This paper discusses the methods for predicting reservoir porosity from relatively few data during the exploration period. The depth, thickness, lithology and ratio of sandstone to reservor are selected to build a neural network. After network training, a stable network structure is established. Using this neural network to calculate the porosity of reservoirs sets up the relationship between different influential factors and porosity, and can also determine the degree to which porosity is affected. This method is used in Denglouku group of Songliao basin and the prediction results show a higher precision than normal mathematic fitting method and this method proves to be easier and more efficient.
出处
《大庆石油学院学报》
CAS
北大核心
2003年第2期5-7,共3页
Journal of Daqing Petroleum Institute
基金
国家重点基础研究发展规划项目 ( 2 0 0 1CB2 0 910 4-0 2 )
关键词
油气勘探
神经网络
储层
孔隙度预测
petroleum and gas exploration
neural network
reservoir
porosity prediction
