Optimizing petrophysical property prediction in fluvial-deltaic reservoirs: a multi-seismic attribute transformation and probabilistic neural network approach

Conventional techniques, which depend on geostatistical modeling, frequently fail to capture reservoir variability, especially when well data are sparse. To overcome this limitation, we develop a combined approach that integrates Multi-Seismic Attribute Transformation (MSAT) and Probabilistic Neural Network (PNN) techniques. By leveraging data from eight wells and post-stack seismic data from the Poseidon 3D area, gas-saturated deltaic-fluvial settings, we analyzed twelve seismic attributes, among which acoustic impedance low-frequency attribute, relative geological time, amplitude envelope, and amplitude-weighted frequency emerge as the most significant. Using the MSAT approach, we established strong connections between these attributes and porosity. Six attributes provide a correlation coefficient of 0.65 with the target log. To enhance precision, PNN and fine-tuning the sigma factor using well drop-out cross-validation analysis was utilized. This leads to a significant enhancement in model accuracy, reaching 76%. In addition, when the model training is concentrated within a 10-millisecond range around the Plover reservoir zone, the accuracy increases significantly to 89%. Our approach has proven to be highly effective, with a success rate of 73% as evidenced by validation through well drop-out analysis. This demonstrates that our method surpasses traditional methods. This innovative integration of seismic attribute-driven methodologies has indicated a major leap forward in identifying and characterizing reservoir heterogeneity. The results demonstrate that it enables more efficient future well-planning strategies and deepens our comprehension of deltaic-fluvial reservoir dynamics.