Enhancing porosity prediction: Integrating seismic inversion utilizing sparse layer reflectivity, and particle swarm optimization with radial basis function neural networks

Seismic inversion, a crucial process in reservoir characterization, gains prominence in overcoming challenges associated with traditional methods, particularly in exploring deeper reservoirs. In this present study, we propose an inversion approach based on modern techniques like sparse layer reflectivity and particle swarm optimization to obtain inverted impedance. The proposed sparse layer reflectivity and particle swarm optimization techniques effectively minimize the error between recorded seismic reflection data and synthetic seismic data. This reduction in error facilitates accurate prediction of subsurface parameters, enabling comprehensive reservoir characterization. The inverted impedance obtained from both methods serves as a foundation for predicting porosity, utilizing a radial basis function neural network across the entire seismic volume. The study identifies a significant porosity zone (>20%) with a lower acoustic impedance of 6000–8500 m/s g cm³, interpreted as a sand channel or reservoir zone. This anomaly, between 1045 and 1065 ms two-way travel time, provides high-resolution insights into the subsurface. The particle swarm optimization algorithm shows higher correlation results, with 0.98 for impedance and 0.73 for porosity, compared to sparse layer reflectivity's 0.81 for impedance and 0.65 for porosity at well locations. Additionally, particle swarm optimization provides high-resolution subsurface insights near well location and across a broader spatial range. This suggests particle swarm optimization's superior potential for delivering higher resolution outcomes compared to sparse layer reflectivity.