Research On Bayesian Stochastic Inversion Constrained By Seismic Data

With the development of the theory of geophysics，the inversion technology hassteped from post-stack inversion into Prestack seismic inversion which can better reflectthe reservoir parameter information. Due to the band limitation of the seismic data,conventional inversion methods are unable to meet the needs of current exploration. Thispaper consider the model parameters as random variables,using stochastic inversionmethod to realize the purpose of obtaining the elastic and reservoir parameters from thepoint of view of probability and statistics.This paper studies the bayesian stochastic inversion constrained by seismic data. Thisis a kind of statistical inversion method which combines geostatistics, geophysicalinversion theory and optimization method. The method based on Markov Chain perturbedsimulation,can effectively explore the model space of priori information by whichgeological statistical is establish, and the model space is searched through the guidance ofprobabilistic transition rules. According to the Bayesian theory,using well-logginginformation as a condition data and seismic data as constraints,which integrate geostatisticsprior into a posterior probability density function of model.The posteriori probabilitydistribution characteristics and reservoir geophysical parameters are obtained through theanalyzing a set of the posterior samples. Because the algorithm adopt the stochasticoptimization to solve the global optimal solution in theory, so the inversion results didn’tdepend on the initial model.This dissertation firstly explain the basic theory of seismic inversion and ofgeostatistics, and extract the stochastic seismic inversion.Then introduce the principle ofMarkov Chain perturbed simulation in detail and combine with bayesian theory to achievethe constrained effectively by seismic data in stochastic inversion.Then discussed theinfluence factors,such as initial model, spatial structure, the signal-to-noise ratio ofseismic data and real seismic data. Finally,the application of real seismic data of A oil fielddemonstrate the reliability of the method.Finally we promote this method to prestackseismic inversion.The testing model show that combines pre-stack inversion with stochastic simulation is available. The method show that the precision of inversion result isavailable, compared with the conventional deterministic seismic inversion,...