Geostatistical inversion method based on seismic waveform similarity

Abstract

Seismic stochastic inversion method has received much attention because of its considerable advantage of having higher vertical resolution than deterministic inversions. However, due to the lack of cross-well data, the inversion results typically exhibit poor lateral continuity. Furthermore, the inversion efficiency is low, and the inversion result is highly random. Therefore, this study proposes a geostatistical seismic inversion method constrained by a seismic waveform. The correlation coefficient of seismic data is used to measure the similarity of the seismic waveforms, replacing the traditional variogram for sequential Gaussian simulation. Under the Bayesian framework, the Monte Carlo-Markov Chain (MCMC) algorithm is combined with the constraints of seismic data to randomly perturb and optimize the simulation results for obtaining the optimized parameter inversion results. The model data tests show that the initial model based on seismic waveform constraints can accurately describe the spatial structure of the subsurface reservoir. In addition, perturbing and optimizing initial model can increase the convergence speed of the Markov chain and effectively improve the accuracy of the inversion results. In this paper, the proposed geostatistical inversion method is applied to the actual seismic data of an oil field. Under the constraints of the stochastic simulation process and objective function, the geological information contained in the seismic waveforms is fully mined, and a theoretical foundation is provided for realizing the multidata joint-constrained seismic inversion.