Informed Inversion of Fiber Optic Strain Data for Real-Time Fracture Geometry in Monitoring Well: A Case Study from the Utah Forge Site

Abstract Fiber optic (FO) strain measurements in offset monitoring wells offer a unique opportunity for real-time fracture diagnostics in Enhanced Geothermal Systems (EGS) projects. However, conventional strain rate inversion approaches face several limitations. They are computationally expensive, suffer from poor temporal coherence, and struggle to accommodate the high dimensionality of multi-cluster fracturing. Moreover, and perhaps more importantly, they often lack compatibility with physical fracture propagation models. This research introduces a novel, computationally tractable inversion framework that couples a simplified 2D fracture mechanics model with FO strain data through a nonlinear particle filter. The objective is to provide fracture geometry parameters such as length, height, width, and asymmetry, in real-time during stimulation. The methodology is new because it embeds the strain data within a forward physics-based simulation loop. A simplified 2D fracture propagation model governs the evolution of fracture attributes over time, capturing dynamic changes in response to stage-level injection parameters and local rock properties. The observed strain response in the monitoring well is assimilated using a particle filter, which is well-suited for nonlinear models and non-Gaussian noise distributions. The particle filter framework allows real-time updating of the model state (fracture geometry) while enforcing smooth temporal evolution and filtering noise in the strain measurements. Key innovations include dimensionality reduction through fracture symmetry assumptions, stress shadow incorporation via asymmetry angle estimation, and efficient localization schemes to manage multi-cluster scenarios. The results are also validated utilizing the FO observations from the Utah FORGE producing well 16B(78)-32 and associated fractional fluid flow logs. The monitoring well 16B(78)-32 is equipped with FO cables cemented behind casing and captured strain evolution during multiple fracturing stages initiated in the injection well, 16B(78)-32. The inversion workflow successfully estimated time-resolved fracture length, height, width, and asymmetry angle for each stage. For Stage 8, the model successfully reflected the non-uniform fracture propagation across clusters with fractures’ half-lengths varying between 300-660 ft, maximum height reached was around 650 ft, and asymmetry angle indicated significant stress shadow interaction. All the values agreed well with the microseismic event cloud evaluations and calibrated complex fracture geometry modeling results. The fracture evolution trends also showed significant agreement with independent diagnostics such as chemical tracer arrival profiles and spinner log data collected post-stimulation. Compared to conventional strain rate inversion, the particle filter method provided a smoother and more physically plausible time series of fracture evolution metrics, with reduced artifacts and greater temporal coherence. To our knowledge, this is the first real-time particle-filter implementation for FO-strain-based multi-cluster monitoring. By integrating physical fracture dynamics with advanced filtering techniques, the methodology described in this paper offers a computationally efficient, temporally smooth, and physically consistent alternative to strain rate inversion. It scales effectively to horizontal well applications with many clusters and can accommodate uncertainty in cluster location, injection timing, and fracture propagation paths. The framework enables near-real-time interpretation of fracture asymmetry, stress interference, and containment behavior; thus, providing critical input for adaptive fracturing strategies. The described workflow is broadly extensible to offset strain monitoring in unconventional plays and EGS applications.