摘要
考虑了一种五层结构的正规化模糊神经网络模型,针对网络结构的优化问题给出了该网络模型的规则层节点的选取方法和相应的反传播学习规则。对于具有较少数输入节点的情况,该网络有较快的训练速度。油藏测井解释中水淹层的识别是石油开发中特别是开发中后期比较突出的一个问题,复杂的地质条件在测井曲线的表现中具有许多模糊性,在各种模糊条件的组合下油藏水淹表现为强水淹、中水淹、弱水淹和无水淹等情形。将正规模糊神经网络用于油藏测井解释中水淹层的识别以提取测井曲线与水淹级别之间的映射关系,从而实现模糊性油藏测井解释中水淹层的识别。实验表明此方法对解决水淹层识别问题具有良好的适应性和实用性。
A normal fuzzy neural network (NFNN) with five layers is proposed. Focusing on the structure optimization of network, a new node selection method and corresponding back propagation learning algorithm are presented. Training is faster with fewer input nodes in this kind of neural network. Water-flooded zone identification in oil well logging is an important problem in oil exploitation, especially in its later period. Complex geology conditions lead to many fuzzy characters in logging curve. In the combination of all kinds of fuzzy conditions, oil water-flooded behaves as strong water-flooded, middle water-flooded, weak water-flooded and none water-flood-ed, etc. NFNN is applied to water-flooded identification in oil well logging to find the relation between well logging and water-flooded level. Experiment results illustrate its practicability.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2001年第1期86-89,共4页
Pattern Recognition and Artificial Intelligence
基金
黑龙江省自然科学基金
关键词
模糊神经网络
学习算法
模式识别
水淹层
油藏测井
Fuzzy Neural Network, Learning Algorithm, Pattern Identification, Normalization, Water-Flooded Zone
