High‐resolution reservoir stochastic modelling based on optimized estimation of vertical autocorrelation

Abstract

AbstractThe high‐resolution model of elastic properties is of great significance for fine reservoir characterization and precise oil and gas exploration. However, it is difficult to obtain a satisfactory high‐resolution reservoir model with the existing technologies. In this paper, a novel high‐resolution stochastic modelling strategy based on the fast Fourier transform moving average is proposed. In this strategy, several structural parameters are optimized to improve the rationality of the stochastic model, including vertical autocorrelation length, horizontal autocorrelation length, roughness factor and angle parameter. Among them, the optimization of the vertical autocorrelation length is crucial for vertical high‐resolution modelling. To this end, a nonlinear optimal inversion strategy of the vertical autocorrelation length is designed based on the idea of minimizing the spectral Jensen–Shannon divergence between the modelling result and the logging curve. However, nonlinear inversion is usually unstable, so it is necessary to introduce a regularization operator in the inversion to improve the stability. Considering that the heterogeneity of the subsurface medium is consistent or gradual within the stratums, but discontinuous and abrupt at the interfaces, edge‐preserving regularization is applied to obtain a blocky estimation of the vertical autocorrelation length. The optimal estimation experiment of the vertical autocorrelation length based on the measured logging data shows that the edge‐preserving regularization significantly improves the stability of the nonlinear optimal inversion, and blocky estimation results with sharp edges are obtained. Then, the optimized vertical autocorrelation length and the other structural parameters are applied to fast Fourier transform moving average modelling on an actual reservoir profile. The result shows that the resolution of the model is significantly improved, which realizes the fine reservoir characterization. In addition, the optimized structural parameters effectively constrain the small‐scale heterogeneity and ensure the rationality of the stochastic model.