研究了用遗传算法来反演各向异性介质的弹性常数。首先概述遗传算法的基本原理，然后以横向各向同性介质的一维模型反演为例，重点讨论了交换概率、更新概率、变异概率和控制温度对反演收敛速度的影响。为改进收敛性能，反演中利用模拟退火法中的控制温度对目标函数作尺度变换，并采用了层剥离技术。对有噪及二维情形也作了考虑。计算表明遗传算法反演是一种良好的非线性反演方法。 In this paper, the genetic algorithm is used to invert the elastic constants of anisotropic media. First of all, the basic principle of genetic algorithm is summarized. Then the one-dimensional inversion of transversely isotropic media is taken as an example. The effects of exchange probability, update probability, mutation probability and control temperature on the convergence rate of inversion are discussed. In order to improve the convergence performance, the target temperature is controlled by the simulated temperature in the simulated annealing method and the delamination technique is used. The noisy and two-dimensional scenarios are also considered. Calculations show that genetic algorithm inversion is a good non-linear inversion method.