A Novel Permeability Prediction Method for Tight Sandstones Based on Nuclear Magnetic Resonance T2 Spectra: A Case Study From the Sichuan Basin, China

Summary The pore structure of tight sandstone is complex, and its permeability is extremely low. Conventional permeability calculation methods that are widely applied in traditional reservoirs are often not suitable for tight sandstone formations. Accurately evaluating permeability has long been a key challenge in the characterization of tight sandstone reservoirs. Nuclear magnetic resonance (NMR) logging, which reflects pore size distribution and fluid mobility, has become an important technique for pore structure evaluation and permeability prediction. The NMR transverse relaxation time (T2) spectrum contains abundant information on microscopic pore-scale heterogeneity, based on which classical permeability prediction models such as Schlumberger-Doll Research (SDR) and Coates have been developed. However, these conventional models largely rely on a single spectral descriptor. As a result, they tend to be less robust in tight sandstone reservoirs with complex pore structures and are more vulnerable to harsh downhole conditions, such as elevated temperature and pressure, and challenging borehole environments. In this study, based on well-logging and experimental data from the Sichuan Basin in southwest China, a new permeability prediction model is proposed. By deeply mining the multidimensional information contained in the T2 spectrum, the model extracts 19 characteristic parameters from three macroscopic dimensions—rock physical properties, graphical geometric morphology, and mathematical statistical characteristics—further divided into six subcategories, in combination with mercury intrusion porosimetry data. Six permeability-sensitive parameters were selected from the subcategories to construct a systematic parameter fusion framework that preserves as much pore structure information as possible and integrates it into the new model. Comparative validation was conducted using core petrophysical measurements from three stratigraphic intervals in the target blocks exhibiting distinct reservoir qualities. Statistical results indicate that the proposed model significantly improves permeability prediction accuracy relative to the classical SDR and Timur-Coates models, with a more pronounced enhancement in the ultralow-permeability interval (<0.1 md). This study provides a practical and reliable new method for permeability evaluation of ultralow-permeability tight sandstone reservoirs and expands the application potential of NMR technology in oil and gas exploration and development.