摘要
波形分类方法是现今比较成熟的地震相分析方法,此法已在岩性预测、砂体预测、裂缝性油气藏和隐蔽性油气藏预测中得到广泛的应用。然而,在以往实际应用中采用输入层为二维Kohonen网络进行地震相分类数的选择都是靠经验或试验方法确定的,增加了分类的不确定性和相应的工作量。本文提出了用改进的输出层为一维的自组织神经网络,半自动地估计地震相分类数。理论模型验证和实例应用表明,改进的Kohonen网络能很好地估计地震相分类数,从而实现快速准确地划分地震相。
The waveform classification is the mature method in seismic facies analysis, it has been widely used in lithology prediction, sandbody prediction, fracture reservoir prediction and subtle reservoir prediction. However in the past application of the method the selection of seismic facies cluster classification for 2d Kohonen network which was took as input layer was determined by experience and testing, as a result the uncertainty and work load for the classification were increased, then a improved Kohonen network was introduced in this paper to semi-automatically estimate seismic facies cluster number. The theoretical modeling and field data application show that the improved Kohonen network works very well for the estimation of seismic facies cluster number so that the fast classification of seismic facies was realized.
出处
《石油地球物理勘探》
EI
CSCD
北大核心
2010年第2期265-271,共7页
Oil Geophysical Prospecting
基金
中国石油集团东方地球物理公司中青年创新基金(2009-09-47)
关键词
自组织神经网络
波形分类
地震相
分类数
self-organized neural network, waveform classification, seismic facies, cluster number
