【摘要】 针对传统的线性化迭代反演算法对初始模型依赖程度较高、反演过程容易陷入局部最优的问题,本文研究了一种基于差分进化算法的叠前反演方法。该方法利用差分的简单变异操作和一对一的竞争生存策略,对初始模型的依赖程度较低,全局收敛能力较强,且具有操作简单、运算速度快的特点,是解决复杂优化问题的一种有效方法。针对叠前地震反演问题,本文以贝叶斯理论为基础,结合似然函数与先验约束信息,建立反演目标函数,然后利用差分进化算法对初始模型进行优化,直至目标函数取得全局最优值。模型试算验证了该方法是可行的,且对初始模型的依赖程度较弱,具有较好的全局收敛能力;将该方法应用于实际叠前道集数据,得到了分辨率较高的反演结果。更多还原

【Abstract】 As the conventional linear iteration inversion algorithm highly depends on the initial model and may easily lead the solution to be trapped into local optimal solution,this paper studies a pre-stack inversion method on the basis of the differential evolution algorithm.The differential evolution algorithm makes use of simple mutation operations based on difference and one-to-one competitive survival strategies.The algorithm is not strongly dependent on the initial models and has strong global convergence ability,simple operations and fast calculation speed.All these advantages make it an effective technical means to solve complex optimization problems.In view of pre-stacked seismic inversion problems,an objective function is established based on Bayesian theory with the combination of the likelihood function and prior constraint information.Then the initial model is optimized constantly by the differential evolution algorithm until the objective function reaches the global optimum.Model trial indicates that the method is reliable,and also displays weak dependency on the initial model and good global converge ability.Inversion results with high resolution are obtained by applying it to real pre-stacked gathers.更多还原