混合遗传算法在大地电磁一维反演中的应用

遗传算法是近些年来产生和发展的一种模拟生物进化过程的自适应启发式全局优化的搜索算法。它不完全依赖于初始猜测，且具有全局收敛的特点，可以被用来解决各种复杂的实际问题，如工程优化设计，人工智能和决策系统，以及地球物理反演等。尽管遗传算法是一种效率很高的全局优化算法，但许多仿真结果表明，它具有计算时间长，局部搜索能力弱的缺点。而共轭梯度法属于非启发式全局优化搜索方法，收敛速度快，但容易陷入局部极值，且严重依赖初始猜测。根据遗传算法和共轭梯度法的特点，这里提出了一种混合遗传算法，用来进行地球物理反演。该算法既具有遗传算法的全局收敛性，又有共轭梯度法的快速收敛性，经实际应用，取得了良好的效果。

Application of hybrid genetic algorithm in the magnetotelluric 1-D inversion method

The genetic algorithm is a self-adaptive heuristic global optimization search method based on the mechanism of biologic evolution which is adopted and developed recently. Because of the characteristics of partly depending the primary estimation and having the global convergence, it can be used to solve different complex problems, such as the optimal design of projects, artificial intelligence, strategic system, geophysical inversion and so on. Although the genetic algorithm is an efficient global-optimization method, the simulation results shows the disadvantages of time-consuming and vain local-researching ability of algorithm. The conjugate gradient algorithm belongs to a non-heuristic global-optimization search method with the characteristics of swift convergence, easily dumping into local extreme value and severely depending on the primary estimation. This essay adopts a hybrid genetic algorithm for geophysical inversion based on the properties of the genetic algorithm and the conjugate gradient algorithm. The method has the attributes of the global-convergence of the genetic algorithm and the swift convergence of the conjugate gradient and achieved good results in practical application