Research On Geostatistical Inversion Method Based On Information Entropy And Cuckoo Algorithm

Geostatistical stochastic inversion technology combines the advantages of seismic inversion and reservoir stochastic modeling,and has a very important position in the field of reservoir prediction.Geostatistical random inversion can organically combine seismic data and logging data to obtain higher resolution results than conventional inversion,and it is currently a hot research field in seismic exploration.Considering that there are obvious differences in underground reservoir parameters of different lithofacies,some scholars have proposed the use of mixed Gaussian models to fuse lithofacies into random inversion,and proposed a random inversion method based on mixed Gaussian model(GMM-c).This method takes into account the influence of the relative inversion results of different rocks.According to the characteristics of the reservoir parameters to meet the multi-peak statistics,assuming that the reservoir parameters meet the mixed Gaussian model,it can better describe the distribution of the reservoir according to the actual situation.The disadvantage of this method is that it does not consider the effect of the initial rock proportion on the results.In practical applications,there may be misclassifications that may affect the inversion results.To solve this problem,this paper takes the initial rock proportion as the inversion parameter,and proposes a random weight inversion method of variable weights mixed Gaussian model.This method achieves a fine description of reservoir parameter distribution under different lithofacies and simultaneously inverts lithofacies and reservoir parameters.Considering the low resolution of variable weights mixed Gaussian model inversion,the cuckoo algorithm(CS algorithm)is introduced.This method solves the defect of low resolution in the existing inversion results and can obtain high-precision reservoir parameter results.The cuckoo algorithm is a global optimal solution method.In the search process,the root mean square error reduction is used as a constraint,which can effectively control the trend of the error decline during the solution process,so that the result jumps out of the local optimal solution.We use the geostatistic stochastic inversion method(GMM-CSMCMC)combined with GMM-v and the cuckoo algorithm,and use the Markov chain Monte Carlo method(MCMC)to solve,making the inversion results closer to the optimal solution Improve the global optimality of the inversion method.The effectiveness of the method in this paper is verified by the model with noise and the analysis of actual data.Another drawback of random inversion is the lack of stability of the inversion results.Since most conventional MCMC methods use completely random sampling methods,the inversion results are easy to fall into the local optimal solution,or the realization difference of each inversion result is large,which makes the inversion results difficult to evaluate.In response to this problem,this paper introduces an information entropy model that can characterize the stability of the state of matter.The core of the MCMC method lies in solving the posterior probability distribution.In this paper,the information entropy model is introduced into the process of solving the posterior probability.Using information entropy to constrain the change of the posterior probability can improve the stability of the inversion results.For the proposed algorithm,the theoretical data,noise model,and actual data were tested and analyzed to verify the correctness of the algorithm.This paper combines the mixed Gaussian model with variable weight coefficients,the cuckoo algorithm,and the information entropy model to optimize the traditional geostatistical random inversion from three aspects: resolution,stability,and correctness of lithofacies classification.Through various tests,the correctness of geostatistical inversion method based on information entropy and cuckoo algorithm is demonstrated.Therefore,the method proposed in this paper can promote the development of geostatistical inversion process to a certain extent.