Cave reservoir characterization method driven by GA-KPCA and geological knowledge.

Abstract

This paper presents a novel method for cave reservoir characterization based on the Genetic Algorithm (GA) and Kernel Principal Component Analysis (KPCA), aimed at improving the precision of reservoir characterization through adaptive multi-attribute fusion. Sensitive seismic attributes are first extracted using geophysical algorithms and their correlations are analyzed based on geological interpretation. Initial attribute weights are then determined scientifically, ensuring reliable geological input for the fusion process. KPCA, with its strong nonlinear analysis capabilities, is used for efficient clustering and feature extraction of complex cave data, while GA optimizes KPCA's key bandwidth parameter to enhance search efficiency. The GA-KPCA method was validated using both synthetic cave model data and real carbonate rock field data in Tarim Basin, demonstrating significant advantages over traditional methods. The results indicate that the proposed approach effectively addresses the limitations of existing techniques, improving the reservoir identification success rate by approximately 33%, and offering an innovative and efficient solution for cave reservoir exploration and development. This method not only contributes to the advancement of cave reservoir characterization but also provides valuable theoretical and practical insights for future research in the field.