Hydraulic-Fracture-Width Inversion Using Low-Frequency Distributed-Acoustic-Sensing Strain Data Part II: Extension for Multifracture and Field Application

Summary Low-frequency distributed-acoustic-sensing (LF-DAS) strain data are direct measurements of in-situ rock deformation during hydraulic-fracturing treatments. In addition to monitoring fracture propagation and identifying fracture hits, quantitative strain measurements of LF-DAS provide opportunities to quantify fracture geometries. Recently, we proposed a Green’s function–based algorithm for the inversion of LF-DAS strain data (Liu et al. 2020b) that shows an accurate estimation of fracture width near the monitor well with single-cluster completions. However, multicluster completions with tighter cluster spacings are more commonly adopted in recent completion designs. One main challenge in the inversion of LF-DAS strain data under such circumstances is that strain measurements at fracture-hit locations by LF-DAS are not reliable, which makes the individual contribution of each fracture to the measured strain data indistinguishable. In this study, we first extended the inversion algorithm to handle multiple fractures, investigated the uncertainties of the inversion results, and proposed possible mitigation to the challenges raised by completion designs and field data acquisition through a synthetic case study. Ideally, there are available data on both sides of each fracture so that the inverted width of each fracture can be obtained with a negligible error. In reality, the strain data are usually limited, providing less constraint on the width of individual fracture. Nevertheless, the inversion results provide an accurate estimation of the width summation of all fractures. To evaluate the individual fracture width, a time-dependent constraint is added to the inversion algorithm. We assume that the width at the current timestep is dependent on the width at the previous step and the width variation between the two timesteps. The width variation can be roughly estimated from LF-DAS strain-rate data at the fracture-hit location. This extra constraint helps to improve the inversion performance. Finally, a field example is presented. We show the width summation of all fractures and the width of each individual fracture as a function of treatment time. The time-dependent width profiles show consistent trends with the LF-DAS strain-rate data. The calculated strains from the inverted model match well with the LF-DAS measured strain data. The findings demonstrate the potential of LF-DAS data for quantitative hydraulic-fracture characterization and provide insights on better use of LF-DAS data. The direct information on fracture width helps to calibrate fracturing models and optimize the completion designs.