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

成都理工大学信息工程与地球物理系,成都610059

Dept. of Information Engineering and Geophysics, Chengdu University of Thechnology, Chengdu 610059

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.