摘要
本文基于前馈多层网络结构发展了一种稳健的油气识别技术。全局优化方法的引入和学习样本输入模式的改进克服了BP算法的一些缺陷,该技术适用于检测由于地下岩性和孔隙流体性质变化而引起的波形特征的细微变化和进行储集层横向预测。本文给出的合成数据和实际数据算例证实了该识别技术的稳健性和有效性。
A robust recognition technique for detectiong hydrocarbon is developed, which is based on a feed forward multilayer neural network structure. Along with using generalized simulated annealing method and changing input pattern of learning samples, some limitations of BP algorithm are resolved. This technique is suitable for the detection of subtle changes in wave shape and nature ca,used by variations of subsurface litholog and pore-fluid properties and the lateral prediction of reservoir. The synthetic and real examples prove that the recognition technique is robust and effective.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1994年第1期47-52,共6页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金会
中国科学院
中国石油管理局联合资助
关键词
人工神经网络
油气识别
Artificial Neural Network, Generalized Simulated Annealing, Reservoir, Hydrocarbon Recognition.