Abstract: Porosity is an important parameter in reservoir description.The paper discusses the principle of porosity prediction with seismic data based on the propagation theory of seismic wave in diphase medium.A method of Complete Utilization of Sample Information(CUSI) to predict porosity is presented considering the fact that both the complexity of seismic porosity prediction and the function approximation of BP network need complete sample information and the low efficiency of learning(not suitable for seismic feature optimization).This method uses the capability of local approximation of CUSI network and establishes a function relation between seismic feature and porosity with borehole datd and the seismic data around the borehole to predict the porosity.The paper also presents the optimization principle of seismic features and the seismic feature optimization method based on genetic algorithm in porosity prediction with CUSI network.The result of the application showed that the method improved the porosityprediction precision with seismic datd.

Key words: reservoir prediction, reservoir description, genetic algorithm, feature optimization, seismic interpretation

摘要： 孔隙度是油气藏描述的一个重要参数.基于双相介质中地震波传播理论,论述了地震孔隙度预测原理.考虑到地震孔隙度预测的复杂性与BP网络函数逼近需要利用全体样本的信息、学习效率低(不适于用来优选地震特征)等不足,提出采用完全利用样本信息(CUSI)的网络做孔隙度预测.该方法利用CUSI网络的局部逼近功能,依据井孔数据与井旁地震数据建立地震特征与孔隙度的函数关系来预测孔隙度.在此基础上还提出了CUSI网络孔隙度预测中的地震特征优化原理和基于遗传算法的地震特征优化方法.实际应用结果表明:此方法明显改善了地震孔隙度的预测精度,具有实用价值.

关键词: 储层预测, 油藏描述, 遗传算法, 特征优化, 地震解释