Data Assimilation-Based Real-Time Optimization Framework for Hydraulic Fracturing Pumping Operations

Abstract

Real-time regulation of fracturing pumping parameters and risk early warning are critical technologies for improving the development efficiency of shale oil and gas reservoirs while ensuring operational safety. Conventional fracturing operations primarily rely on static design and manual control, which suffer from delayed responses and limited capability for dynamic risk identification and real-time parameter optimization. In this study, a data assimilation-based real-time optimization framework for hydraulic fracturing operations is proposed. A gated recurrent unit (GRU) network is employed to predict fracturing pressure, while an ensemble Kalman filter (EnKF) is introduced as an online parameter updating mechanism to adaptively recalibrate the prediction model, thereby improving prediction accuracy. Based on the predicted fracturing pressure, construction risks are classified into three levels using an inverse-slope indicator and a statistical pressure-threshold criterion, with the pressure-window constraint embedded into the inverse-slope evaluation as a pressure-dependent correction. This unified framework jointly accounts for subsurface formation response and surface pressure-safety constraints, enabling more comprehensive and operationally relevant risk assessment. In combination of pressure prediction with risk early warning, a real-time control strategy incorporating both proactive and reactive regulation is developed. The proposed strategy enables advance pressure prediction and risk assessment during fracturing design, allowing pumping parameters to be optimized under safety constraints. The pressure prediction method is validated using a test data set. Subsequently, four representative field cases are analyzed, including two proactive regulation cases and two reactive regulation cases. Results indicate that, in the two proactive cases, enhanced pumping parameter optimization led to increases in the cumulative proppant volume of 28.4 and 22.8 m3, respectively. In the two reactive cases, early risk warning enabled timely mitigation measures, effectively avoiding operational risks. The proposed real-time control strategy provides a robust solution for improving fracturing performance while reducing operational risks.