Gas Reservoir Identification Based on Nonstationary and Anisotropic FFT-MA Stochastic Modeling

Summary High-resolution reservoir modeling is a crucial technique for the precise identification of gas reservoirs, holding significant importance in guiding natural gas development. However, the nonstationarity and statistical anisotropy of subsurface media present immense challenges to the reliable implementation of high-resolution reservoir modeling. In response to the nonstationarity and anisotropy of complex reservoirs, we propose a novel stochastic modeling method based on fast Fourier transform moving average (FFT-MA). In this method, variational mode decomposition (VMD) is introduced to decompose logging curves into a series of sparse components with specific center frequencies and narrow bandwidths. Subsequently, the autocorrelation functions of each component are computed and synthesized, thereby inferring the nonstationary vertical autocorrelation functions of logging curves. In addition, to characterize the anisotropic and lateral nonstationary features of the reservoir, angle parameters and lateral autocorrelation functions are extracted from seismic records. Lastly, considering the high sensitivity of Lamé constants (λ and μ) and their density-combined counterparts (λρ and μρ) to gas-bearing reservoirs, FFT-MA stochastic modeling is applied to λρ, μρ, and ρ. Gas identification is then performed based on the joint probability distribution extracted from logging data. The proposed method is tested in the sand-shale reservoirs of the Yinggehai Basin, China. The results indicate that the stochastic models for λρ, μρ, and ρ effectively characterize the nonstationary and anisotropic features of complex reservoirs, significantly enhancing the resolution and accuracy of gas identification.