Microseismic analysis over a single horizontal distributed acoustic sensing fiber using guided waves

Abstract

A single horizontal distributed acoustic sensing (DAS) fiber is notoriously challenging for microseismic analysis even when it is close to recorded events. Due to its uniaxial measurement, locations suffer from circular ambiguity. Nonetheless, in unconventional plays, the presence of dispersive guided waves in the DAS records can partially resolve such ambiguity. If the reservoir has lower seismic velocities than its surrounding medium, it can act as a waveguide. In this case, guided waves are generated only by microseismic events occurring inside or close to the reservoir, and their propagation is confined to the reservoir. We first train a machine learning model for microseismic event detection using a unique data set of almost 7000 manually picked events and an equal number of noise windows. Applying the trained model to 10 stimulation stages from two offset wells yields more than 100,000 event detections, with a higher sensitivity than manual labeling. Detected events undergo a localization procedure based on the dispersion properties of guided waves, estimated in-situ from known perforation shots. Location results allow us to reconstruct the spatio-temporal pattern of fracture development. We observe a dominant fracture propagation direction for all stages, which indicates the effect of the regional stress in the reservoir. We qualitatively validate the direction and extent of the fracture growth by perforation shot analysis. We have found the first application of microseismic event location with a single straight fiber, which is considered impossible without a waveguide structure.