A data-driven strategy for pre-fracturing design of pump rate in deep shale gas fracturing

Abstract

The deep shale formation (>3500 m) reserves rich oil and gas resources but is difficult to exploit due to the high risk and low efficiency of hydraulic fracturing, for instance, the high level of wellhead pressures (∼100 MPa) and difficulties in proppant injection. The sand screen-out occurs frequently in deep shale fracturing, during which the pump schedule is usually designed referring to neighbor wells or relying on human experiences. A scientific criterion for optimizing the pump schedule at field-practical scales is urgent for deep shale fracturing compared with the previous empirical estimation. Therefore, the pump rate for proppant injection (PRPI) is defined as a new strategy for the pre-fracturing design. A data-driven workflow is established to process the field data to fit the PRPI, which integrates numerical models (for feature extractions), machine learning algorithms (for feature optimization) and the STATA software. Field data of sand screen-out cases are collected from deep shale gas fracturing wells. The sand-fluid ratio (the volume ratio of injected proppant and fluids) is optimized as the moderating variable for fitting the PRPI based on the Pearson correlation coefficient and the variable importance measure (VIM, Random Forest algorithm). The pre-designed pump rate is fitted in STATA software under different levels of sand-fluid ratio, based on which a positive linear correlation (between pump rate and sand-fluid ratio) and an exponential correlation (between screen-out probability and sand-fluid ratio) are obtained. A field case of deep shale gas fracturing is also presented to verify and explain the usage of PRPI, in which the optimized pump rate is around 15.7 m3/min. This new strategy provides a scientific and quantitative criterion for the pre-fracturing design of pump schedules compared with empirical estimations based on individual experience, which may be significant for improving proppant injection and securing real-time operations of deep shale fracturing.