电阻率测深数据的遗传算法和最小二乘法反演

遗传算法作为一种模拟生物进化过程的全局优化算法，由于对初始模型的依赖性较小，且不易收敛于局部极值中，在地球物理反演中发挥着越来越重要的作用，但这种方法只能得到次优解而不能得到最优解．最小二乘法是地球物理反演中最成熟的反演方法之一，反演中可以得到最优解，但是受初始模型的影响，容易收敛于局部极值甚至发散．为在电阻率测深曲线反演中得到全局最优解，采用先使用遗传算法再用最小二乘法反演方法，对较为复杂的3层和3层以上大地模型进行了计算，其模型参数误差只有6％。

Inversion of GA and least square for resistivity sounding data

As a global optimization method in simulated bios evolution progress, GA plays a more and more important role in the inversion of geophysics with many disadvantages , such as not being able to reach optimal result. The least square is one of the most ripen inverting methods which can get the optimal values. However, it is easily affected by the initial model. To get the global and optimal values of resistivity sounding data inversion, a joint inversion method of GA and least square is presented. It is shown from the computations that the joint method inherits both the advantages of GA and least square. The model errors of inverting 3 or more strata layered earth are 6% less.