Prediction of marl reservoir distribution based on facies-constrained reflectivity inversion method

Abstract

Accurate characterization of carbonate reservoirs remains a significant challenge due to complex facies variations and the substantial effects of wave propagation. We propose a facies-constrained reflectivity inversion strategy. The method establishes a relationship between logging data and seismic waveforms, applies clustering analysis using the Self-Organizing Map (SOM) technique, and utilizes the clustering results to constrain the construction of an initial model with realistic lateral variations. Based on this initial model, a Bayesian-based reflectivity inversion is performed, incorporating a modified Cauchy prior distribution to enhance inversion accuracy and stability. The reflectivity method offers a one-dimensional analytical solution to the wave equation, tacking thin layer thicknesses and wave propagation effects into consideration, thereby significantly alleviating inversion problems encountered in marl reservoirs. Compared to traditional inversion methods based on the Zoeppritz equation, the facies-constrained reflectivity inversion delivers higher accuracy and resolution. The application of this technique to identify marl reservoirs in the Lei32 sub-member of the Sichuan Basin has yielded promising results, effectively delineating favorable reservoir areas of approximately 210 km2 and offering strong support for future exploration and development.