Quantitative Prediction of Fracture Parameters in Volcanic Reservoirs Using Integrated Rock Physics and BO-BiLSTM Network Approach

Natural fractures significantly influence the production performance of volcanic hydrocarbon reservoirs. However, the complex nature of fracture networks affects the elastic behavior and seismic responses of these reservoirs, creating substantial challenges for seismic-based fracture characterization. This study proposes a method for quantitatively estimating fracture parameters in volcanic formations by integrating a rock physics model with the Bayesian optimized bidirectional long short-term memory (BO-BiLSTM) network architecture. Following the established model for volcanic formations, a model-based inversion approach is introduced to calculate fracture parameters using well log data, incorporating the simulated annealing particle swarm optimization (SA-PSO) algorithm for reliable multiparameter simultaneous estimation. The fracture density is derived from these parameters, enabling a comprehensive assessment of fracture parameters. The validity of the volcanic model and inversion framework is confirmed by the strong correlation between predicted and observed velocities during the inversion process. Subsequently, the BO-BiLSTM architecture is used to build a predictive framework that captures the intricate relationships between fracture parameters and elastic properties. This predictive model is then applied to estimate fracture parameters from seismically derived elastic properties. The seismic-based predictions exhibit strong agreement with results obtained from well log data, demonstrating the applicability of the proposed method. The predicted fracture parameters reflect microscale fracture characteristics and exhibit a structure-related distribution when visualized alongside macroscale fractures, revealing patterns of fracture networks and potential connectivity between larger fractures. These insights enhance the understanding of multiscale fracture systems, providing critical information for volcanic reservoir characterization.