Study On Modified Simulated Annealing Method For Multi-wave Pre-stack AVA Inversion

The pre-stack seismic data contains much more information of the amplitudecharacteristics. Inversion to extract stratigraphic and lithologic parameters based onthe relationship between changes in the amplitude characteristics of information has agreat significance for the study of lithology identification and hydrocarbon detection.In this thesis, multi-wave pre-stack AVA inversion is one of the inversion mode,that isbased on pre-stack seismic AVA angular trace gather records inversion P-wave velocity,S-wave velocity and density of lithological parameters. As the non-linear relationshipbetween the changes information of the amplitude characteristics in the pre-stackseismic data and the lithologic parameters, the Nonlinear method of inverse is themost suitable method of inverse used in the study of multi-wave pre-stack AVAinversion.On the background of seismic wave propagation theory,this thesis focuses on thesimulated annealing algorithm to study the multi-wave pre-stack AVA inversion. Thedata used in the multi-wave pre-stack AVA inversion is the AVA angular trace gatherrecords of amplitude varies with the angle.First,we forward simulate P-P and P-SV wave’s AVA curve by Zoeppritz exactformula based on the three-layer theoretical model.Then we forward simulate P-P andP-SV wave’s AVA angular trace gather records.The basic simulated annealing algorithm is a nonlinear algorithm for globaloptimization.It has a lot of advantages,and has been very widely used in variousfields. However, the basic simulated annealing algorithm there are many shortcomings,this obstruct its access to a wider range of applications.So, many scholars at domesticand international have to improve and optimize it. In this thesis, we will continue to optimize the one widely recognized very fast simulated annealing algorithm. Withchanged the formula of annealing probability acceptance to the one which moresimilar with the physical annealing process, added capability of memory andoptimized search mechanism under the guidance of memory,it becomes the very fastsimulated annealing with memory guide. Then test their computing performance withfour test functions. And compare the computing performance of this algorithm withthe very fast simulated annealing algorithm.The memory guided very fast simulated annealing algorithm is applied toinversion simple two-layers model. It is shown that joint inversion can improve theaccuracy of inversion to a certain extent. Then apply the multi-wave inversion inmultilevel models. It is shown that Simultaneous inversion parameter increases withthe number of layers increased. So it increass the difficulty of the simultaneousinversion parameters and the accuracy of the inversion parameters decreased. We addsome noise to theoretical model’s AVA angular trace gather records, and inversionresults indicate that memory guided very fast simulated annealing algorithm has has acertain extent ability of anti-noise. Then apply the joint inversion to the horizontalgradient model, and to the real seismic data finally.