摘要
针对阵列侧向测井反演受给定初值影响的问题,将阵列侧向测井反演问题转化为无约束的非线性全局优化问题,提出了一种基于改进差分进化反演的新方法。在30次随机计算中改进差分进化算法与差分进化算法相比,其寻优成功率可以达到93.3%,平均进化代数为差分进化代数的1/8,其最优适应度也仅为差分进化最优适应度的一半。通过对阵列侧向模型进行反演计算,表明该算法满足电测井反演的需求。改进差分进化方法可有效地解决传统优化方法反演结果对初值的依赖性问题。
According to the problem of array laterolog inversion being affected by the given initial value,a new algorithm based on improved difference evolution inversion is proposed.The inversion of array laterolog response is transformed into an unconstrained nonlinear global optimization problem.Compared with improved differential evolution algorithm and differential evolution algorithm in 30 times randomly calculation,the results show that the searching optimization success rate of the improved algorithm reach to 93.3%,while the average evolutional generation decrease to 1/8 of the original,and the optimum fitness is also only half of the difference evolution inversion method.The inversion results for array laterolog model show that the improved difference evolution algorithm can meet the needs of electric logging inversion.This method is more independent of the initial value than the traditional optimal algorithm.
出处
《吉林大学学报（地球科学版）》
EI
CAS
CSCD
北大核心
2010年第5期1199-1204,共6页
Journal of Jilin University：Earth Science Edition
基金
国家科技重大专项专题(2008ZX05035-02)
国家青年自然科学基金项目(40804020)
山东省自然科学基金项目(Y2007F25)
中国石油大学优秀博士学位论文培育基金(B2007-05)
关键词
改进差分进化算法
阵列侧向
反演
鲁棒性
improved difference evolution algorithm
array laterolog
inversion
robustness
