混合优化波阻抗反演方法研究
|
摘要
地震反演常用的线性算法具有较快的收敛速度,但是易陷入局部最优解。因此需要引进一些非线性优化算法求解全局最优解。近年来相继出现了模拟退火、遗传算法、禁忌搜索算法和混沌搜索算法等,虽然这些算法具有较强的全局优化性能,但是其计算速度慢,远远不能满足实际生产的要求。如何将上述两类算法结合起来实现优势互补成为了反演中的一个重要课题之一。文章提出的混合优化波阻抗反演方法综合了共轭梯度算法和模拟退火算法的优点,在模拟退火反演框架内加入共轭梯度迭代算法,即在模拟退火反演过程中,当目标函数值满足给定的条件时,进行一定次数的共轭梯度迭代反演,最终以模拟退火反演结果来判断其收敛性。实际计算表明,该方法不仅收敛速度快,而且抗干扰能力强,计算得到的波阻抗剖面能较好的反映地层地质特征。
The linear algorithms commonly used in seismic inversion are characterized by rapid convergence rate but often local optimal solutions. Some nonlinear algorithms are needed to achieve global optimal solutions. In recent years, several nonlinear algorithms have been developed successively, including simulated annealing, genetic algorithm, taboo searching and chaos searching. Although these algorithms are highly capable of global optimization, their computation speeds are too slow to meet the requirements of practical application. How to integrate the linear algorithms with nonlinear ones to make full use of their advantages is an important task in seismic inversion. The hybrid optimum impedance inversion method presented in this paper absorbs the advantages of conjugate gradient and simulated annealing algorithms. The conjugate gradient algorithm is integrated with simulated annealing inversion. That is to say, several times of conjugate gradient iterative inversion will be performed when the value of the object function satisfy the given conditions during simulated annealing inversion, and the convergence will be finally judged according to the simulated annealing inversion results. Real computation shows that this method not only has rapid convergence rate, but also has strong anti-interference ability, and the achieved impedance profile can well represent the geologic features of the formations.
The linear algorithms commonly used in seismic inversion are characterized by rapid convergence rate but often local optimal solutions. Some nonlinear algorithms are needed to achieve global optimal solutions. In recent years, several nonlinear algorithms have been developed successively, including simulated annealing, genetic algorithm, taboo searching and chaos searching. Although these algorithms are highly capable of global optimization, their computation speeds are too slow to meet the requirements of practical application. How to integrate the linear algorithms with nonlinear ones to make full use of their advantages is an important task in seismic inversion. The hybrid optimum impedance inversion method presented in this paper absorbs the advantages of conjugate gradient and simulated annealing algorithms. The conjugate gradient algorithm is integrated with simulated annealing inversion. That is to say, several times of conjugate gradient iterative inversion will be performed when the value of the object function satisfy the given conditions during simulated annealing inversion, and the convergence will be finally judged according to the simulated annealing inversion results. Real computation shows that this method not only has rapid convergence rate, but also has strong anti-interference ability, and the achieved impedance profile can well represent the geologic features of the formations.
引文
[1]LAVIELLEM.2DBayesiandeconvolution[J].Geophys ics,1991,56(12):2008.
[2]COOKEDA,SCHNEIDERWA.Generalizedlinearin versionofreflectionseismicdata[J].GeoPhysics,1983,48(6).
[3]周竹生,等.广义共轭梯度算法[J].物探与化探,1996,25(5).
[4]姚姚.地球物理非线性反演模拟退火算法的改进[J].地球物理学报,1995,38(5).
[5]TSALLISC.Possiblegeneralizationofboltzmann gibbs statistics[J].JStatPhys,1988,52.
[6]INGBERL.Veryfastsimulatedreannealing[J].Math ComputModeling,1989,12.
[7]朱仕军,杜志敏,等.储层定量建模[M].北京:石油工业出版社,2005.
[2]COOKEDA,SCHNEIDERWA.Generalizedlinearin versionofreflectionseismicdata[J].GeoPhysics,1983,48(6).
[3]周竹生,等.广义共轭梯度算法[J].物探与化探,1996,25(5).
[4]姚姚.地球物理非线性反演模拟退火算法的改进[J].地球物理学报,1995,38(5).
[5]TSALLISC.Possiblegeneralizationofboltzmann gibbs statistics[J].JStatPhys,1988,52.
[6]INGBERL.Veryfastsimulatedreannealing[J].Math ComputModeling,1989,12.
[7]朱仕军,杜志敏,等.储层定量建模[M].北京:石油工业出版社,2005.
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心