The Prestack Inversion Method For Seismic Parameters Based On The GAMH-MCMC Algorithm

The Markov chain Monte Carlo(MCMC)method has gradually attracted the attention of researchers in the field of geophysical inversion.It is a Monte Carlo simulation in heuristic nonlinear inversion algorithms.The core of this algorithm is to construct a Markov chain whose stationary distribution is consistent with the posterior distribution.After a finite number of iterations,the steady state of the Markov chain is obtained,and then the sample of the posterior distribution is obtained,and finally the posterior Samples are sampled for statistical inference.The MCMC algorithm still has many aspects that need improvement,and it is not perfectly applicable to various inversion problems.Including the speed of convergence of the Markov chain,the time cost of running the program and the artificial design,and the time required to adjust the Markov Monte Carlo process are all urgent problems to be solved.Furthermore,the traditional single-chain method obtains a single information and cannot perform sufficient search.It is easy to make the algorithm fall into a local optimal solution,and there is a situation in which the inversion accuracy is not accurate enough.Most of the existing multi-chain Markov methods only consider the parallelism of the chains,ignoring the learning ability between the chain and others,failing to make good use of the diversity of the overall information of the system,and more manual adjustment operations are required in the later stage and so on,these are all issues that need to be studied.In this paper,the improved GAMH-MCMC method is derived by the analysis of stochastic seismic inversion theory and the Monte Carlo method based on overall system.The overall Monte Carlo emphasizes the learning ability between individuals in the Markov chains,and this method takes this idea as a starting point.Using the information interaction ability of Genetic Algorithm(GA),the cross-operation of genetic algorithm is introduced into the Metropolis Hastings(MH)sampling algorithm,which is intended to ensure that the MCMC method can accurately invert different parameters while improving the utilization of system information and minimizing the number of manual tests.This method utilizes the learning ability between Markov chains,saves multi-chain search time and searches for non-unique solutions as much as possible,reduces the fine adjustment required in the Markov convergence process as much as possible,and further improves the algorithm operation efficiency and enhance the stability of the inversion.Ultimately it can provides theoretical guidance and reliable seismic parameters for reservoir prediction and reservoir description.In this paper,it is applied to the post-stack impedance inversion to predict the distribution of oil and gas reservoirs.The pre-stack elastic impedance inversion is used to obtain the sensitive parameters of “sweet-spot” and to find the sweet-spot development area of sandstone reservoirs.In addition,a stochastic inversion of prestack elastic parameters is performed based on the exact zoeppritz equation to obtain the P-wave velocity and S-wave velocity and density information.