摘要
碳酸盐岩储集层的孔隙空间非常复杂,因而进行储集层评价难度很大。近年来.采用神经网络进行储层评价越来越普遍,且取得了较好效果。但由于神经网络存在一些难以克服的缺点.如训练速度慢、网络结构设计缺乏理论指导以及易陷入局部极小点等,使得神经网络的应用受到一定的限制。针对这种情况.提出了利用遗传算法同神经网络相结合的方法来进行储集层评价,在很大程度上弥补了神经网络的这些缺陷。给出了方法的基本理论,并在某地区进行了碳酸盐岩储层评价,评价结果与测井解释和钻井结果吻合得较好。
Carbonate reservoir is more complex It is very difficult to evaluate carbonate reservoir. In recent years, using neural network to evaluate reservoir becomes more and more prevalent, and gets preferable results. However, neural network have some defects, which are difficult to o-vercome, such as low training speed, lack of reasonable guidance on network structure design, getting into local minimum etc. Those defects make application of network confined In this situation, this paper present a method which combines genetic algorithms-neural network to e-valuate reservoir. This method can make up limitations of neural network to maximum extent and have a good effect in some areas reservoir evaluation.
出处
《石油物探》
EI
CSCD
2005年第3期225-228,5,共4页
Geophysical Prospecting For Petroleum
基金
国家自然科学基金(40174039)973项目(G1999043311)资助。
关键词
碳酸盐岩
储集层
孔隙空间
遗传算法
神经网络
储层评价
carbonate rock
reservoir
pore space
genetic algorithms
neural network
reservoir evaluation
