基于快速非支配排序遗传算法的VTI介质多分量叠前联合反演

Multicomponent prestack joint inversion for VTI media based on a fast nondominated sorting genetic algorithm

在VTI介质中,由于引入了各向异性参数使得多分量多参数地震反演问题的非线性程度显著增加,因此采用传统的权重加权法构建单目标函数进行反演得到的反演结果往往并不理想.本文以反射率法为基础,结合快速非支配排序遗传算法研究了一种VTI介质的多分量叠前联合反演方法.该方法以反射率法为正演方程,应用互相关原理构建PP波和PSV波的多目标函数,进而采用快速非支配排序遗传算法全局寻优获得VTI介质的厚度、纵横波速度、密度和各向异性参数等多个参数.在正演的过程中,反射率法可以考虑几何扩散、吸收衰减、透射损失、多次波以及纵横波旅行时不匹配等地震波传播效应,更能精确地描述地震波在地下地层中的真实传播情况;在反演的过程中,快速非支配排序遗传算法可以在不引入权重系数的条件下同时优化多个目标函数,获得联合反演问题的Pareto最优解,既不添加权重系数影响又充分利用多分量地震数据.模型测试结果验证了该反演方法的有效性和可行性.

Because introducing anisotropic parameters markedly increases the nonlinearity of the multicomponent or multiparameter prestack joint inversion for VTI media,the conventional weighted stacking inversion method of objective functions which is aimed at constructing a singleobjective function is usually unsuitable.Based on the reflectivity method,a multicomponent prestack joint inversion method for VTI media is studied by using a fast nondominated sorting genetic algorithm(NSGA Ⅱ)in this paper.This approach first uses the reflectivity method as forward equation and establishes a multi-objective function of PP and PSV waves based on the cross-correlation approach,and then uses the NSGA Ⅱto estimate multiple parameters including thickness,P-and S-wave velocity,density and anisotropic parameters by inverting PP and PSV wave seismic records jointly.In the process of forward modeling,the reflectivity method canaccurately describe the propagation characteristics of seismic wave because some wave-propagation effects are considered,such as geometric spreading,absorption attenuation,transmission loss,multiples and travel time mismatching of P-wave and S-wave.In the process of inversion,the NSGA Ⅱ can simultaneously optimize more than one objective function to obtain the Pareto-optimal solutions of the joint inversion problem without introducing weight coefficients,which eliminates the influence of weight coefficients on inversion results and makes full use of multicomponent seismic records.Model tests verify the effectiveness and feasibility of the proposed inversion method.