摘要
一些油藏系统的偏微分方程模型,经过变换能化为非线性函数项级数,级数的每一项均为地层参数θ的复杂非线性函数.地层参数是试井解释的依据,因而要求其估值应具有全局最优性,又因上述函数为多峰函数,在极值点处关于θ的变化很敏感,使问题更为困难,现有迭代法均未奏效.为了解决了这个问题,提出了一种新型混合遗传算法.实际应用表明,用上述方法建模具有很高的精确度,模糊的平均相对误差在1%以内,并且能求出地层参数的全局最优估值.
Partial differential equation models of some oil reservoir systems can be transformed, such as well test interpretation, into series composed of nonlinear function terms. Every term is complex nonlinear function of stratigraphic parameters θ. Stratigraphic parameters are the foundation of well test interpretation, so their global optimal estimate values must be sought. And the problem beoames more difficult because aforesaid functions are multimodal functions and very sensitive in extreme point about the change of θ. No iteration methods exist in literatures effect. A new pattern of hybrid genetic algorithms is developed to solve above problem. It has high precision to build models of above systems, the average relative errors being within 1%. Moreover it can obtain global optimal estimate values of stratigraphic parameters.
出处
《哈尔滨理工大学学报》
CAS
北大核心
2009年第3期29-32,共4页
Journal of Harbin University of Science and Technology
基金
黑龙江省自然科学基金(TF2005-26)
关键词
系统辨识
函数型连接神经网络
二阶学习算法
遗传算法
收敛性
system identification
functional link nets
second order learning algorithms
genetic algorithms
convergence
