基于遗传-BP神经网络的沉积微相自动识别  

Automatic sedimentary facies identification method based on genetic-BP neural networks

作　　者：许少华[1] 陈可为[2] 梁久祯[1] 郑生民 

机构地区：[1]大庆石油学院计算机科学系,黑龙江安达151400 [2]同济大学海洋地质系,上海200092 [3]大庆油田有限责任公司第六采油厂,黑龙江大庆163113

出　　处：《大庆石油学院学报》2001年第1期51-54,共4页Journal of Daqing Petroleum Institute

基　　金：黑龙江省自然科学基金资助项目 !(F9917)

摘　　要：提出了一种基于神经网络与图象处理技术相结合的沉积微相自动识别方法 .该方法是先将数字化测井曲线和地层参数预处理转化为二值点阵图象模式 ,经过点阵数据编码压缩提取和记忆曲线所表征的地层模式特征 ,然后利用超线性BP算法与遗传算法相结合的方法训练多层前馈神经网络 .所得神经网络稳定、学习收敛速度快 ,同时具有很强的记忆能力和推广能力 。We propose an automatic sedimentary facies identification method based on combination of neural network with image process technology. First, we translate digital well logging curves and stratum parameters into binary image modes. Second, through contracting binary data codes, we distill and store stratum mode characters token by well logging curves. Last, we combine BP algorithm with genetic algorithm to train a multilayers forward neural network. The neural network has the advantages of being stable, fast learning, awfully memorable and generalized ability. This model is suitable to solve problems of sedimentary facies identification.

关 键 词：沉积微相自动识别 测井曲线 神经网络 图象处理 BP计算 遗传算法 

分 类 号：P631.84[天文地球—地质矿产勘探]