一种混合算法在试井参数优化中的应用  

Application of a Mixed Calculation Method in Optimization of Well Test Parameters

作　　者：徐杰[1] 卢德唐[1] 韩伟[1] 

机构地区：[1]中国科学技术大学工程科学软件研究所,安徽合肥230027

出　　处：《油气井测试》2011年第1期14-17,22,共5页Well Testing

基　　金：国家自然科学基金(10672159;10702069);973项目(2006CB705805)联合资助

摘　　要：基于L-M和差分进化的混合算法是利用差分进化算法在一定进化代数后出现的种群聚类特性,将种群识别为不同的聚类区域,以每个聚类的中心为起始点,再利用基于梯度具有局部搜索能力强的L-M算法可以快速找到该聚类区域的最小极值。混合算法兼顾了差分进化全局搜索能力强和L-M局部搜索能力强收敛速度快的优点。将该混合算法应用于试井参数优化中,通过两种不同油藏模型的实例应用,结果表明,该混合算法比单一的算法优化速度更快,收敛精度更高。该混合算法实用性广,能有效的解决存在多局部极值的试井参数优化复杂问题。Based on the group-gathering feature of the mixed methods derived from L-M and difference methods,groups are identified into different gather areas;the least local extreme value in a certain gather area is quickly found out,with the center of each group as the initiate point and by utilizing L-M calculation method which has very strong ability of local searching.This mixed method combines the advantages of difference method which has strong ability of overall searching and L-M method which has strong ability of local searching and great convergence rate.It is successfully used in optimization of well test parameters.Application in two different reservoir models shows that this mixed method makes both of optimization rate and convergence precision higher than any method it contains alone;and so can be applied effectively to many situations,including solve complicated modern well test problems in which many local extreme values exist.

关 键 词：试井 参数优化 油藏参数 混合算法 

分 类 号：O242.23[理学—计算数学]