大地电磁反演中改进的自适应正则化因子选取研究

通过建立大地电磁(MT)层状地电模型，利用共轭梯度法求解反问题，在给定不同初始模型的条件下对多种正则化因子选取方法进行了计算比较，分析了各种方法的特点和使用条件.结果表明，自适应正则化算法的效果与传统的定值方法如L曲线法相近，但反演过程远比传统方法便捷.为了解决反演依赖于初始模型的局限并增强解的稳定性，基于多种自适应正则化方案的对比分析，提出了改进的自适应正则化方案，选取数据拟合泛函与模型稳定泛函较大的比值为正则化因子的初始值，并提出相应的调整方案自动控制正则化因子衰减.模型试验表明，该方法对初始模型的依赖性低于其他几种自适应的算法，反演结果的稳定性较强，可以进一步提高正则化反演的效率.

An improved adaptive regularized parameter selection method in MT inversion

Based on the conjugate gradient method to solve the inversed problem of layered magnetotelluric (MT) models, a comparative study is made between the inversion results and several other regularization parameter selection methods under different initial models in order to analyze their respective behavior and conditions. The results show that the adaptive regularization algorithm can get a similar inversion result in a faster and easier way. Moreover, with the purpose to reduce the limit due to depending on the initial model and improve the stability of inversion, an improved adaptive regularization method is put forward by selecting a large ratio of the misfit function to the stabilizing function as the initial value of regularization parameter, and an automatic decay coefficient scheme is proposed to reduce the parameter with iteration. Model test shows that the method is less dependent on the initial model with a stable inversion result, which can improve the efficiency of regularization inversion.