摘要
地球物理反演的局部线性方法易使解陷入局部极值,并严重依赖初始模型,而传统的遗传算法在优化应用中存在局部搜索能力弱、早熟收敛等问题,为此,提出了一种解决地球物理反演问题的并行实数编码混合遗传算法(MRCGA)。该方法采用拟网格法初始种群、综合交叉策略和线性算子,实现了并行实码混合遗传算法。理论模型试算证明了该算法反演地震波阻抗的有效性。
Locally linearized methods in geophysical inverse problem are apt to trap themselves in local minima and their solutions are heavily dependent on the selection of initial solutions,while standard genetic algorithms have problems of poor local search ability and premature convergence.This paper presented a real-coded hybrid genetic parallel algorithm(MRCGA)which assigns the initial population with virtual mesh method,and uses crossover strategy and linear operator to achieve coarsely-grained paralleling.The effectiveness of the method in wave impedance inversion has been tested on theoretical model.
出处
《勘探地球物理进展》
2009年第4期265-269,231-232,共5页
Progress in Exploration Geophysics
关键词
MPI
遗传算法
实数编码
初始种群
综合交叉策略
线性算子
波阻抗反演
MPI
genetic algorithm
real-coded
initial population
integrated cross-strategy
linear operator
wave impedance inversion
