Search

This study presents the first demonstration of the transferability of a convolutional neural network (CNN) trained to detect microseismic events in one fiber-optic distributed acoustic sensing (DAS) data set to other data sets. DAS increasingly is being used for microseismic monitoring in industrial settings, and the dense spatial and temporal sampling provided by these systems produces large data volumes (approximately 650 GB/day for a 2 km long cable sampling at 2000 Hz with a spatial sampling of 1 m), requiring new processing techniques for near-real-time microseismic analysis. We have trained the CNN known as YOLOv3, an object detection algorithm, to detect microseismic events using synthetically generated waveforms with real noise superimposed. The performance of the CNN network is compared to the number of events detected using filtering and amplitude threshold (short-term average/long-term average) detection techniques. In the data set from which the real noise is taken, the network is able to detect >80% of the events identified by manual inspection and 14% more than detected by standard frequency-wavenumber filtering techniques. The false detection rate is approximately 2% or one event every 20 s. In other data sets, with monitoring geometries and conditions previously unseen by the network, >50% of events identified by manual inspection are detected by the CNN.

Fiber-optic distributed acoustic sensing (DAS) cables are now used to monitor microseismicity during hydraulic-fracture stimulations of unconventional gas reservoirs. Unlike geophone arrays, DAS systems are sensitive to uniaxial strain or strain rate along the fiber direction and thus provide a 1C recording, which makes identifying the directionality and polarization of incoming waves difficult. Using synthetic examples, we have shown some fundamental characteristics of microseismic recordings on DAS systems for purposes of hydraulic fracture monitoring in a horizontal well in anisotropic (vertical transverse isotropy [VTI]) shales. We determine that SH arrivals dominate the recorded signals because their polarization is aligned along the horizontal cable at the near offset, although SV will typically dominate for events directly above or below the array. The amplitude of coherent shear-wave (S-wave) arrivals along the cable exhibits a characteristic pattern with bimodal peaks, the width of which relates to the distance of the event from the cable. Furthermore, we find that S-wave splitting recorded on DAS systems can be used to infer the inclination of the incoming waves, overcoming a current limitation of event locations that have constrained events to lie in a horizontal plane. Low-amplitude SV arrivals suggest an event depth similar to that of the DAS cable. Conversely, steep arrivals produce higher amplitude SV-waves, with S-wave splitting increasing with offset along the cable. Finally, we determine how polarity reversals observed in the P and SH phases can be used to provide strong constraints on the source mechanisms.

We present a case study demonstrating the use of an “L”-shaped downhole fibre-optic array to monitor microseismicity. We use a relatively simple method to detect events from continuous waveform data, and develop a workflow for manual event location. Locations are defined with a cylindrical coordinate system, with the horizontal axis of the DAS cable being the axis of symmetry. Events are located using three manual “picks”, constraining (1) the zero-offset “broadside” channel to the event (2) the P-S wave arrival time difference at the broadside channel, and (3) the angle, ? of the event from the array. Because the one-component DAS array is unable to record P-wave energy on the broadside channel, the P-wave pick is made indirectly by ensuring that the modeled P- and S-wave moveout curves match the observed data. The ? angle requires that signal is observed on the vertical part of the array, in our case this is possible because an engineered fiber, rather than standard telecommunications fiber, provided a significant reduction in the noise level. Because only three picks need to be made, our manual approach is significantly more efficient than equivalent manual processing of downhole geophone data, where picks for P- and S-waves must be made for each receiver. We find that the located events define a tight cluster around the injection interval, indicating that this approach provides relatively precise and accurate event locations. A surface microseismic array was also used at this site, which detected significantly fewer events, the locations of which had significantly greater scatter than the DAS array locations. We conclude by examining some other aspects of the DAS microseismic data, including the presence of multiple events within very short time windows, and the presence of converted phases that appear to represent scattering of energy from the hydraulic fractures themselves.

Summary Fibre-optic Distributed Acoustic Sensing (DAS) cables are now used to monitor microseismicity during hydraulic fracture stimulations of unconventional gas reservoirs. Unlike geophone arrays, DAS systems are sensitive to uniaxial strain along the fibre direction and thus provide a single-component recording, which makes identifying the directionality of incoming waves difficult to infer. Using synthetic examples, we show some fundamental characteristics of microseismic recordings on DAS systems for purposes of hydraulic fracture monitoring in a horizontal well in anisotropic (VTI) shales. We demonstrate that SH arrivals dominate the recorded signals since their polarization is aligned along the horizontal cable at near offset. The amplitude of the SH phase along the cable exhibits a characteristic pattern with bimodal peaks, the width of which relates to the distance of the event from the cable. Furthermore, we find that shear-wave splitting recorded on DAS systems can be used to infer the inclination of the incoming waves, overcoming a current limitation of event locations which have constrained events to lie in a horizontal plane. Low amplitude qSV arrivals suggest an event depth similar to that of the DAS cable. Conversely, steep arrivals produce higher amplitude qSV waves, with shear-wave splitting increasing with offset along the cable.

Summary Emerging acquisition systems based on fiber optic technologies such as Distributed Acoustic Sensing are enabling dense spatial and temporal sampling of strain fields. This has resulted in a large increase in the volume in the volume of data and the rate at which data is generated. These increases can be interpreted as satisfying two of the 3 ‘V's of ‘Big Data’ i.e. Volume and Velocity (the third V refers to data Variety). In this presentation we show how we have used big data technologies such as Apache Hadoop and Apache Spark to tackle these data issues for the specific problem of microseismic event detection. Furthermore, traditional approaches were thought to be unlikely to efficiently scale to these new data so we turned to machine learning approaches based on computer vision. Rather than trying state of the art technologies such as Convolutional Neural Nets we decided to try a mature technology used for face detection known as a Haar Cascade. We have tested this approach on field data and found that this approach can work well and are motivated to try newer machine learning techniques with the expectation of moving beyond microseismic event detection.

Dipole sonic logs acquired in near-vertical pilot wells and over the build section of nearby horizontal production wells are inverted to determine the five elastic constants characterizing a transversely isotropic formation, under the assumption of lateral homogeneity. Slowness values from a single depth in the vertical well are combined with data from the corresponding depth in the deviated well; these data are then inverted using nonlinear optimization to derive the five elastic constants. The technique is demonstrated on data from the Haynesville Shale in Texas. Estimates of the anisotropy are in line with a priori expectations; the Thomsen e and γ parameters are well correlated and generally possess positive anellipticity, with Thomsen’s e greater than Thomsen’s δ.

Mudrocks, defined to be fine‐grained siliclastic sedimentary rocks such as siltstones, claystones, mudstones and shales, are often anisotropic due to lamination and microscopic alignments of clay platelets. The resulting elastic anisotropy is often non‐negligible for many applications in the earth sciences such as wellbore stability, well stimulation and seismic imaging. Anisotropic elastic properties reported in the open literature have been compiled and statistically analysed. Correlations between elastic parameters are observed, which will be useful in the typical case that limited information on a rock's elastic properties is known. For example, it is observed that the highest degree of correlation is between the horizontal elastic stiffnesses C11 and C66. The results of statistical analysis are generally consistent with prior observations. In particular, it is observed that Thomsen's ɛ and γ parameters are almost always positive, Thomsen's ɛ and γ parameters are well correlated, Thomsen's δ is most frequently small and Thomsen's ɛ is generally larger than Thomsen's δ. These observations suggest that the typical range for the elastic properties of mudrocks span a sub‐space less than the five elastic constants required to fully define a Vertical Transversel Isotropic medium. Principal component analysis confirms this and that four principal components can be used to span the space of observed elastic parameters.

Dipole sonic log data recorded in a vertical pilot well and the associated production well are analyzed over a [Formula: see text]-ft section of a North American gas shale formation. The combination of these two wells enables angular sampling in the vertical direction and over a range of inclination angles from 54° to 90°. Dipole sonic logs from these wells show that the formation’s average properties are, to a very good approximation, explained by a transversely isotropic medium with a vertical symmetry axis and with elastic parameters satisfying [Formula: see text], but inconsistent with the additional ANNIE relation ([Formula: see text]). More importantly, these data clearly show that, at least for fast anisotropic formations such as this gas shale, sonic logs measure group slownesses for propagation with the group angle equal to the borehole inclination angle. Conversely, the data are inconsistent with an interpretation that they measure phase slownesses for propagation with the phase angle equal to the borehole inclination angle.

Seismic technology has been used successfully to detect geomechanically induced signals in repeated seismic experiments from more than a dozen fields. To explain geomechanically induced time-lapse (4D) seismic signals, we use results from coupled reservoir and geomechanical modeling. The coupled simulation yields the 3D distribution, over time, of subsurface deformation and triaxial stress state in the reservoir and the surrounding rock. Predicted changes in triaxial stress state are then used to compute changes in anisotropic P- and S-wave velocities employing a stress sensitive rock-physics transform. We predict increasing vertical P-wave velocities inside the reservoir, accompanied by a negative change in P-wave anisotropy [Formula: see text]. Conversely, in the overburden and underburden, we have predicted a slowdown in vertical P-wave velocity and an increase in horizontal velocities. This corresponds to positive change in P-wave anisotropy [Formula: see text]. A stress sensitive rock-physics transform that predicts anisotropic velocity change from triaxial stress change offers an explanation for the apparent difference in stress sensitivity of P-wave velocity between the overburden and the reservoir. In a modeled example, the vertical velocity speedup per unit increase in vertical stress [Formula: see text] is more than twice as large in the overburden as in the reservoir. The difference is caused by the influence of the stress path [Formula: see text] (i.e., the ratio [Formula: see text] between change in minimum horizontal effective stress [Formula: see text] and change in vertical effective stress [Formula: see text]) on vertical velocity. The modeling suggests that time-lapse seismic technology has the potential to become a monitoring tool for stress path, a critical parameter in failure geomechanics.

C011 Linking geomechanics and seismics: Stress effects on time-lapse multicomponent seismic data Abstract 1 Time-lapse stress effects on seismic data can have a similar magnitude as time-lapse effects caused by changes in saturation and pore pressure. Production-induced stress changes occur not only in the reservoir interval itself but are also manifested in the overburden. Furthermore because stress fields caused by reservoir production are directionally dependent the resulting changes in the elastic properties yield an anisotropic seismic response. To study effects in seismic data we have developed a workflow that allows the prediction/estimation of changes in (anisotropic) seismic observations from stress

Summary Successful production from the Clair field requires an understanding of the reservoir fracture system. BP, as operator of the Clair field, is currently investigating the use of surface seismic attributes, such as anisotropy, for fracture characterisation. A multi-azimuthal VSP was acquired to measure the seismic anisotropy and relate it to fracture information from core and FMI logs. Two analysis techniques applied to these data are described in this paper: (a) Polarisation/slowness inversion across the geophone array. This gives an estimate of the -wave anisotropy local to the tool setting depth in the borehole. (b) A travel-time analysis from sea level to the individual geophone settings which measures the total anisotropy in the overburden combined with any fracture-related effects. The results are comparable with available core and FMI data acquired in this well providing encouragement for further studies on fracture characterisation from surface seismic.

Seismic anisotropy is a complicated non-linear phenomenon in which quasi compressional(qP) and two quasi shear waves(qS1 and qS2) propagate in most directions isolated from singularity effects (Crampin 1991). Problems arise in inverting shear wave data due to the presente of singularities and the deviation of group from phase velocities so that linearization approaches are invalid (Chapman and Pratt 1992). Genetic Algorithms (GA) are non-linear optimization methods which are receiving increasing attention in the geophysics community (Sen & Stoffa 1991, Sambridge & Drijkoningen 1992) due to their robustness, efficiency and wide range of applicability. Just as simulated annealing is analogous to crystal annealing so are GA's analogous to the evolutionary processes of reproduction, crossover and mutation in nature. GA's use a population of models which are combined in such a way that more successful parameter combinations receive an increasing sampling rate. Previous attempts to invert shear wave data by MacBeth (1991) used a systematic search of a database of anisotropic materials. This approach is computationally intensive and is restricted by the precalculated database size. GA's appear to offeramore flexible and efficient anisotropic roversion scheme.