摘要
地震信号的特征是由地下岩石、流体的物理特征及其变化直接引起的,有效挖掘隐藏于地震数据中的有关岩性和储层物性信息具有深远而现实的意义。地震属性参数之间多是非线性关系,并且地震属性参数受环境的影响很大。神经网络可以建立属性参数与预测目标之间的高度非线性映射,而遗传算法选择适者生存。改进BP算法,综合BP的快速收敛和GA的全局寻优的特点,具体应用于薄互层储层厚度预测。
Seismic signals are directly affected by the rock and the physical character of liquid. The extraction of information about rock and deposit from the seismic data is of great realistic significance. There exist nonlinear relationships between the seismic attributes, which are affected obviously by the environment. BP algorithm can establish the high nonlinear mapping between the object and the seismic attributes. GA algorithm can select the survival of the fittest. The BP algorithm can quickly converge, and the GA can globally search the fittest result in all data. The improvement of BP algorithm and the application of the GA+BP hybrid algorithm in inter-bed prediction can achieve a satisfactory result.
出处
《物探与化探》
CAS
CSCD
2004年第5期460-462,470,共4页
Geophysical and Geochemical Exploration
关键词
薄互层属性
遗传算法
神经网络
厚度预测
thin inter-bed attribute
GA algorithm
BP Neural network
thickness prediction