孙彩堂+李玲+王雪+黄维宁+周逢道

摘 要：传统的可控源音频大地电磁法（CSAMT）反演方法属于线性或者局部线性，大都依赖初始模型.而遗传算法因其不依赖初始模型的特点而应用到CSAMT反演中.但是，标准的遗传算法存在早熟、局部收敛等问题.针对这些问题，对标准的遗传算法进行改进，采用排序法和最优保留策略相结合的选择算子，增强其种群多样性并保证其收敛性；采用父子竞争策略和自适应概率法相结合的交叉算子，能够防止好的父代个体被淘汰，又具有适应性.通过理论模型进行算法仿真验证，证明其有效性，说明改进遗传算法较标准遗传算法在CSAMT一维反演中有明显的改善.通过对实测数据进行反演，其结果与地质资料吻合，证明了其适应性.

关键词：可控源音频大地电磁；反演；改进遗传算法

中图分类号：P631.2 文献标志码：A

CSAMT One-dimensional Inversion Based on Modified Genetic Algorithm

SUN Caitang， LI Ling， WANG Xue， HUANG Weining， ZHOU Fengdao

（National Geophysical Exploration Engineering Research Center （Jilin University），

College of Instrumentation and Electrical Engineering， Jilin University， Changchun 130061， China）

Abstract：The traditional method of controlled source audio frequency magnetotelluric （CSAMT） inversion is linear or locally linear， and most of them depend on the initial model. The genetic algorithm is applied to the CSAMT inversion because it does not depend on the initial model. However， the standard genetic algorithm has some problems such as premature convergence and local convergence. The standard genetic algorithm was improved， and the selection operator based on the ranking method and the optimal reservation strategy was used to enhance the diversity of the population and to ensure its convergence； The crossover operator， which is based on the combination of the father and son competition strategy and the adaptive probability method， can prevent the good parent from being eliminated， and has the adaptability. The simulation results show that the improved genetic algorithm is better than the standard genetic algorithm in the one-dimensional inversion of CSAMT. The inversion of the measured data is consistent with the geological data， which proves the adaptability of the improved genetic algorithm.

Key words：controlled source audio-frequency magnetotellurics （CSAMT）； inversion； modified genetic algorithm

在地球物理勘探方法中，可控源音頻大地电磁法[1-3]（CSAMT）是一种重要的方法.反演[4-7]是CSAMT方法的关键环节之一.经典的线性或者局部线性反演方法大都依赖初始模型，或者出现病态矩阵使得反演失败.遗传算法[8-12]（GA）因其不依赖初始模型等优点而被广泛应用.但是，标准的遗传算法可能存在早熟和局部收敛等问题.因此，对标准的遗传算法进行优化是必要的.刘云峰等[13]引进了“小生境”GA法对一维大地电磁测深地电模型进行处理，并对目标函数进行压缩，有利于目标函数相近时选择，计算复杂地电模型十分有效.Elsayed S M等[14]提出了……