Effect of slickwater-alternate-slurry injection on proppant transport at field scales: A hybrid approach combining experiments and deep learning

Abstract

Proppant transport in underground fractures plays a key role in mitigating sand screen-out and enhancing the stimulated production for hydraulic fracturing. The effects of field pumping schedules, however, are not fully studied. We investigate the effect of slickwater-alternate-slurry injection on proppant transport at field-practical scales. A new hybrid approach is proposed to directly connect experimental studies with field operations, which consists of observation experiments, calculations, and deep learning (DL) workflow. The experiments reveal that the alternate injection induces the unexpected proppant ridge. The modified calculations (considering the ridge height) are proposed to extract features for training the DL algorithm. The workflow predicts the downhole pressure (mainly governed by proppant injection) for error analyses. Approximately 20.2% of the error is eliminated by considering the proppant ridge, thus demonstrating its effect on proppant injection. The predictions are significantly improved in early and late periods of fracturing operations when the fracture is initially created and highly filled. The sensitivity analysis suggests that the pump rate may dominate the ridge height compared with other hydraulic parameters. The study of proppant ridge complements the mechanisms of proppant transport, which is essential for controlling fracturing pressure and boosting the proppant injection.