Data-Driven Dynamic Control of Post-Fracturing Shut-In and Flowback in Shale Gas Wells and Field Validation

Summary Optimizing post-fracturing shut-in and flowback is critical for maximizing productivity and minimizing sand production in shale gas wells. Current practices, however, often rely on generalized, static guidelines, lacking the adaptability to address well-specific heterogeneity and dynamic downhole conditions. In this paper, we present a data-driven advisory workflow that integrates machine learning (ML) with physics-based control charts to guide flowback operations in real time. The methodology consists of three core components: First, a geological analysis identifies key reservoir characteristics (e.g., fault proximity and clay content) influencing optimal shut-in duration; second, an extreme gradient boosting (XGBoost) model predicts downhole flowing pressure from readily available surface data with an error of less than 1%; third, these predictions enable the calculation of two novel, normalized parameters (in-perforation net pressure and apparent single-perforation velocity), which are plotted against field-calibrated safe operating envelope (SOE) charts. The system advises choke adjustments to maintain the well’s operating trajectory within a “reasonable zone” on these charts, thereby balancing production acceleration with sand control. Field validation in the Dingshan block demonstrated the workflow’s efficacy: In a controlled test, a well following the advisory system’s recommendation successfully increased rate without sand production, while an offset well that deviated from the advice experienced immediate and severe sanding. This work provides a practical and scalable framework for derisking flowback operations through real-time, data-informed guidance, moving beyond rigid, one-size-fits-all protocols.