在测井中用一种组合进化神经网络识别油水层  

Well log interpretation using a combined evolutionary neural network.

作　　者：张学庆[1] 刘燕[1] 肖慈珣[1] 刘争平[1] 杨斌[1] 

机构地区：[1]成都理工大学信息工程与地球物理系,成都610059

出　　处：《石油物探》2001年第4期119-124,140,共7页Geophysical Prospecting For Petroleum

摘　　要：分析了基于进化算法的神经网络 ,指出基于遗传算法的神经网络具有强的全局搜索能力 ,基于进化规划的神经网络具有强的局部寻优能力。在此基础上 ,提出了综合上述两种神经网络优点的组合进化神经网络 ,并应用于油水层测井解释中 ,降低了误判率。After an analysis of evolutionary algorithm-based neural networks, this paper suggests that genetic algorithm-based neural networks have great ability in searching optimal direction globally, while evolutionary programming-based neural networks are powerful in searching the optimal direction locally. By combing the beauty of the kinds of neural networks, we devised a combined evolutionary neural network for discrimination of oil-and-water-bearing formations in the interpretation of log data, thereby the failure rate is decreased.

关 键 词：组合 遗传算法 进化规划 神经网络 测井解释 油水层识别 

分 类 号：P618.13[天文地球—矿床学] P631.84[天文地球—地质学]