一维大地电磁测深几种反演算法的比较研究  

Comparison among several inversion algorithms of 1D MT.

作　　者：冯思臣[1] 王绪本[1] 阮帅[1] 

机构地区：[1]成都理工大学油气藏地质及开发工程国家重点实验室,教育部地球探测与信息技术重点实验室

出　　处：《石油地球物理勘探》2004年第5期594-599,498-628,共6页Oil Geophysical Prospecting

基　　金：国家自然科学基金项目 (4 0 0 740 3 6)资助

摘　　要：大地电磁反演算法主要分为线性与非线性两类。马奎特法、遗传算法以及模拟退火是其中具有代表性的算法。本文将三种方法的反演过程、结果及理论进行了对比研究 ,指出各自的优点与局限性 ,以及使用时应注意的相应条件。研究表明 ,马奎特法寻找目标函数全局最优的能力较差 ,只能局部寻优 ,但运算速度快 ;模拟退火法和遗传算法寻找目标函数全局最优的能力较强 ,但运算速度较慢。当对地下地电参数预先有一定程度了解 ,又可以选取比较合理的初始解时 ,可选择马奎特法 ,否则宜选用其他非线性全局寻优能力强的反演方法。MT inversion algorithms are mainly divided into two categories: linear and non-linear algorithms,among which the Marquit method,genetic algorithm and method of simulated annealing are representative algorithms.The paper compared the inversion procedures,results and theories among these three methods and pointed out each advantages and limitations and corresponding conditions needed to pay attention when these methods are used.The study showed that the Marquit method has poor ability in global optimization of objective function,only has ability in local optimization,but its operation speed is fast;simulated annealing and genetic algorithm have strong ability in global optimization of objective function but slow speed of algorithm.The Marquit method is chosen as inversion method when knowing underground geoelectric parameters in some degree in advance and being able to choose more reasonable preliminary solution,otherwise,the other non-linear inversion algorithms that have strong ability in global optimization are chosen.

关 键 词：大地电磁测深 反演算法 反演方法 模拟退火法 目标函数 代表性 电参数 一维 初始解 全局寻优能力 

分 类 号：TE973.92[石油与天然气工程—石油机械设备] P631[天文地球—地质矿产勘探]