A Fast Physics-Based Data-Driven Surrogate Model for Multibench Multipad Unconventional Reservoirs with Fracture-Driven Interactions (Frac-Hits)

Summary

Fracture-driven interactions (FDIs) between parent and child wells result from the propagation of hydraulic fractures into pressure-depleted regions or existing fracture networks. These interactions can significantly alter pressure and fluid flow pathways in unconventional reservoirs, leading to production interference, increased water production, and reduced recovery. Traditional numerical workflows struggle to fully integrate fracture propagation with multiphase flow in such depleted, fracture-dominated systems, limiting their ability to accurately capture FDI impacts at the field scale.

In this study, we extend a physics-based, data-driven flow network surrogate model (GPSNet) by explicitly incorporating FDI effects through an enhanced multiple interacting continua (MINC) grid design. The proposed formulation effectively captures both rapid stimulated reservoir volume (SRV) decline and potential FDI events among wells. Hydraulic fracture systems and SRV are represented using high-flow channels, while lower-flow channels capture matrix-dominated transient flow. Fracture-to-fracture connections are dynamically activated during hydraulic fracturing events, enabling the model to capture transient pressure depletion and fluid redistribution associated with parent-child well interactions without explicitly simulating fracture propagation.

The approach is validated through a field case involving 16 hydraulically fractured wells across three benches and three pads. GPSNet-FDI demonstrates strong agreement between predicted production metrics—phase rates and bottomhole pressure (BHP)—and observed data after history matching (HM). Moreover, GPSNet uses less than 0.05% of the gridblocks typically required by a full-fidelity model and completes in just 3.4 minutes on a single processor—more than 1,000 times faster than the full-fidelity flow simulation alone, excluding the time-intensive fracture modeling process required to establish the fracture grid in the full-fidelity approach. This computational advantage scales with increasing well count and fracture complexity.

To the best of our knowledge, this study represents the first application of a flow-network model to a field case with FDI impacts and the first implementation of FDI simulation in physics-based, data-driven shale and tight reservoir surrogates without coupling flow and geomechanics. Unlike conventional full-physics models, the proposed approach leverages GPSNet—a fast, reliable surrogate—to enable large-scale simulation studies for unconventional assets. This capability supports rapid, robust decision-making in shale and tight reservoir development and management. Furthermore, the methodology circumvents the computationally prohibitive challenge of explicitly simulating water flow during fracture propagation and demonstrates strong potential for broader application in scenarios where interwell communication is critical.