Study On Stochastic Inversion Method Of The Pre-Stack Seismic Waveform Constraints

Seismic inversion method plays an important role in oil and gas reservoir prediction.It is an effective way to obtain various elastic parameters of underground media and reveal the distribution,physical properties and oil and gas properties of underground reservoirs.With the continuous advancement of oilfield exploration and development,the accuracy requirements for reservoir prediction are getting higher and higher,and high-precision reservoir stochastic inversion technology has emerged as the times require.Seismic stochastic inversion is a highresolution inversion method based on statistics and global optimization algorithms.It has certain advantages in identifying thin inter-reservoirs within a tuning scale.The Markov Chain Monte Carlo(MCMC)method is a global optimized nonlinear stochastic inversion method,the method avoids the complicated marginal integral calculation problem in the Bayesian formula for its unique algorithm structure.Based on the generated stable Markov chain consistent with the Bayesian posterior probability density function,the statistical analysis of this method can obtain the statistical characteristics of the complex multi-dimensional posterior density function,and obtain the optimal solution of the inverse problem and its related uncertainties.However,the stochastic inversion method still has problems such as poor lateral continuity and low computational efficiency.Aiming at the problems of the above conventional stochastic inversion methods,this paper has carried out the research on the stochastic inversion method constrained by pre-stack seismic waveforms based on the theory of MCMC stochastic inversion method.In this paper,the correlation coefficient is used to measure the waveform similarity between seismic data,and the correlation coefficient between seismic data is fully used to replace the correlation coefficient between the model parameters to be predicted,and a new ordinary pseudo-kriging constrained by seismic waveform is derived.The interpolation formula further established the initial model with seismic waveform indication.On this basis,a posterior probability density distribution with cooperative constraints of observed seismic data and logging data is constructed under the Bayesian framework,and the parameters of the model to be inverted are simulated randomly for many times by combining Metropolis-Hastings algorithm.The posterior mean is used as the optimal solution of the model parameters,and the posterior variance is used to evaluate the reliability of inversion results,aiming at improving the stability and lateral continuity of inversion.In this paper,the stochastic inversion method based on the seismic waveform constraint has been successively performed post-stack inversion and prestack inversion.After adding the initial model of seismic waveform indication,the convergence speed of the Markov chain is effectively accelerated,which makes the Markov chain converge to the global optimal solution more accurately and quickly.The model tests have verified the effectiveness of this method.The actual data processing shows that this method has greater application potential in improving the inversion resolution and lateral continuity compared with conventional sparse pulse inversion.