An Innovative Method of Fracability Evaluation for Tight Reservoirs Based on SEL–MECE

Reservoir fracability evaluation is critical for tight reservoir hydraulic fracturing optimization. This study introduces a novel physics-based fracability evaluation framework integrating stacking ensemble learning (SEL) and the marginal effect of the conditional expectation (MECE). First, a multidimensional indicator system was established, covering characteristics such as reservoir geomechanics, rock mechanics, and the development of natural fractures. Second, SEL models were developed to predict open flow capacity, and four performance metrics were compared to select the optimal model from 26 SEL candidates. Finally, to quantify the individual contribution of each fracability indicator while eliminating interference from treatment and petrophysical parameters, the MECE approach was adopted, thereby developing a new fracability model that quantitatively describes the reservoir’s ability to achieve greater stimulated reservoir volume (SRV) under similar hydraulic fracturing parameters. The experimental results indicate that the RF+KNN model demonstrates optimal performance in both prediction accuracy and model stability. Comparing the fracability index with microseismic monitoring data, the linear correlation coefficient between the fracability index and SRV reached 92%, validating the reliability of the fracability evaluation model. This framework provides a transferable interpretable tool for selecting reservoir sweet spots and fracturing parameter optimization.