Optimizing the prediction of fault-controlled fracture–cave filling using multi-source seismic data: A case study of the Middle Permian Maokou formation, Luzhou area, southern Sichuan Basin, China

Abstract

Accurate prediction of the boundaries and filling degree of fault-controlled fracture–caves is critical for the effective exploration of fault-controlled reservoirs. An optimized method for predicting the filling degree of such reservoir bodies was developed, targeting the Maokou Formation in the Luzhou area of the southern Sichuan Basin. Based on one-dimensional wave equation forward modeling, gradient structure tensor attributes were computed with standard deviations σ1 = 0.6 and σ2 = 1.6, using λ2 as the eigenvalue to delineate boundaries. A threshold-based and normalized fault-controlled fracture–cave probability volume was generated as a constraint for prestack seismic prediction. AVA (Amplitude Versus Angle) forward modeling verified the feasibility of filling degree prediction using prestack data, and original gathers were optimized. Prestack simultaneous inversion yielded elastic parameter volumes. A rock physics interpretation panel was constructed to classify filling degrees, enabling volume estimation for different categories. Results demonstrate that the gradient structure tensor effectively identified boundaries, with a threshold >0.19 indicating fracture–cave development. AVA responses varied by filling degree: fully filled bodies showed no AVA features, partially filled exhibited Class I or II, and unfilled showed Class III characteristics. The Vp/Vs ratio and P-impedance served as effective discriminators. Fracture-caves primarily occurred on both sides of the faults, with unfilled and partially filled ones concentrated near large faults and intersecting smaller ones, comprising 48.81% of the total volume. The prediction results were validated by dynamic verification wells, achieving an 83.3% concordance rate. The integration of multi-source seismic data and a progressive, constraint-driven prediction strategy enables a significant improvement in the accuracy and efficiency of predicting the filling degree in fault-controlled fracture–cave systems.