Classification and Localization of Low-Frequency DAS Strain Rate Patterns with Convolutional Neural Networks

2021 ◽  
Author(s):  
Mengyuan Chen ◽  
Jin Tang ◽  
Ding Zhu ◽  
Alfred Daniel Hill
Keyword(s):  
Edge Detection ◽  
Strain Rate ◽  
Real Time ◽  
Event Detection ◽  
Fiber Optic ◽  
Low Frequency ◽  
Propagation Model ◽  
Advanced Technology ◽  
Hydraulic Fractures ◽  
Fiber Optic Sensing

Abstract Distributed acoustic sensing (DAS) has been used in the oil and gas industry as an advanced technology for surveillance and diagnostics. Operators use DAS to monitor hydraulic fracturing activities, to examine well stimulation efficacy, and to estimate complex fracture system geometries. Particularly, low-frequency DAS can detect geomechanical events such as fracture-hits as hydraulic fractures propagate and create strain rate variations. Analysis of DAS data today is mostly done post-job and subject to interpretation methods. However, the continuous and dense data stream generated live by DAS offers the opportunity for more efficient and accurate real-time data-driven analysis. The objective of this study is to develop a machine learning-based workflow that can identify and locate fracture-hit events in simulated strain rate response that is correlated with low-frequency DAS data. In this paper, "fracture-hit" refers to a hydraulic fracture originated from a stimulated well intersecting an offset well. We start with building a single fracture propagation model to produce strain rate patterns observed at a hypothetical monitoring well. This model is then used to generate two sets of strain rate responses with one set containing fracture-hit events. The labeled synthetic data are then used to train a custom convolutional neural network (CNN) model for identifying the presence of fracture-hit events. The same model is trained again for locating the event with the output layer of the model replaced with linear units. We achieved near-perfect predictions for both event classification and localization. These promising results prove the feasibility of using CNN for real-time event detection from fiber optic sensing data. Additionally, we used image analysis techniques, including edge detection, for recognizing fracture-hit event patterns in strain rate images. The accuracy is also plausible, but edge detection is more dependent on image quality, hence less robust compared to CNN models. This comparison further supports the need for CNN applications in image-based real-time fiber optic sensing event detection.

Modeling of fiber-optic strain responses to hydraulic fracturing

Geophysics ◽  
2020 ◽  
Vol 85 (6) ◽  
pp. A45-A50
Author(s):  
Zhishuai Zhang ◽  
Zijun Fang ◽  
Joe Stefani ◽  
James DiSiena ◽  
Dimitri Bevc ◽  
...  
Keyword(s):  
Hydraulic Fracturing ◽  
Strain Rate ◽  
Fiber Optic ◽  
Low Frequency ◽  
Displacement Discontinuity ◽  
Hydraulic Fractures ◽  
Quantitative Interpretation ◽  
Current Application ◽  
Monitoring Well ◽  
And Inversion

We modeled cross-well strain/strain rate responses of fiber optic sensing, including distributed strain sensing (DSS) and low-frequency distributed acoustic sensing (DAS), to hydraulic stimulation. DSS and low-frequency DAS have been used to measure strain or the strain rate to characterize hydraulic fractures. However, the current application of DSS/DAS is limited to acquisition, processing, and qualitative interpretations. The lack of geomechanical models hinders the development of the technology toward quantitative interpretation and inversion. We have developed a strategy to use the displacement discontinuity method to model the strain field around kinematically propagating fractures. For a horizontal monitoring well, modeling results were able to explain the heart-shaped extending pattern before a fracture hit, the polarity flip due to fracture interaction during stimulation, and the V-shaped pattern when a fracture does not intersect with the monitoring well. For a vertical monitoring well, modeling shows the different characters of strain rate responses when a fracture is near and far away from a vertical monitoring well. We also investigated the effects of fractures with various geometries such as elliptic and layered fractures. We compared and verified the modeling with field data from the Hydraulic Fracturing Test Site 2, a research experiment performed in the Permian Basin. Our modeling work can be used to identify patterns in field observations. The results also help to improve acquisition design and lay the groundwork for quantitative interpretation and inversion.

scholarly journals Distributed Fiber Optic Sensing for Real-Time Monitoring of Gas in Riser during Offshore Drilling

Sensors ◽  
2020 ◽  
Vol 20 (1) ◽  
pp. 267 ◽  
Author(s):  
Giuseppe Feo ◽  
Jyotsna Sharma ◽  
Dmitry Kortukov ◽  
Wesley Williams ◽  
Toba Ogunsanwo
Keyword(s):  
Real Time ◽  
Fiber Optic ◽  
Full Scale ◽  
Flow Dynamics ◽  
Petroleum Engineering ◽  
Detection Methods ◽  
Marine Riser ◽  
Offshore Drilling ◽  
Well Control ◽  
Fiber Optic Sensing

Effective well control depends on the drilling teams’ knowledge of wellbore flow dynamics and their ability to predict and control influx. Unfortunately, detection of a gas influx in an offshore environment is particularly challenging, and there are no existing datasets that have been verified and validated for gas kick migration at full-scale annular conditions. This study bridges this gap and presents pioneering research in the application of fiber optic sensing for monitoring gas in riser. The proposed sensing paradigm was validated through well-scale experiments conducted at Petroleum Engineering Research & Technology Transfer lab (PERTT) facility at Louisiana State University (LSU), simulating an offshore marine riser environment with its larger than average annular space and mud circulation capability. The experimental setup instrumented with distributed fiber optic sensors and pressure/temperature gauges provides a physical model to study the dynamic gas migration in full-scale annular conditions. Current kick detection methods primarily utilize surface measurements and do not always reliably detect a gas influx. The proposed application of distributed fiber optic sensing overcomes this key limitation of conventional kick detection methods, by providing real-time distributed downhole data for accurate and reliable monitoring. The two-phase flow experiments conducted in this research provide critical insights for understanding the flow dynamics in offshore drilling riser conditions, and the results provide an indication of how quickly gas can migrate in a marine riser scenario, warranting further investigation for the sake of effective well control.

Fracture Hits and Hydraulic-Fracture Geometry Characterization Using Low-Frequency Distributed Acoustic Sensing Strain Data

2021 ◽  
Vol 73 (07) ◽  
pp. 39-42
Author(s):  
Kan Wu ◽  
Yongzan Liu ◽  
Ge Jin ◽  
George Moridis
Keyword(s):  
Strain Rate ◽  
Green’S Function ◽  
Green's Function ◽  
Hydraulic Fracture ◽  
Laser Energy ◽  
Low Frequency ◽  
Forward Model ◽  
Hydraulic Fractures ◽  
Fracture Geometry

The propagation process and geometry of hydraulic fractures depend on complex interactions among the induced fractures and the pre-existing rock fabric, the heterogeneous rock properties, and the stress state. Accurate characterization of the resulting complex hydraulic-fracture geometry remains challenging. Fiber-optic-based distributed acoustic sensing (DAS) measurements have been used for monitoring hydraulic fracturing in adjacent treatment wells. DAS requires an optical fiber attached to the wellbore to transmit the laser energy into the reservoir. Each section of the fiber scatters a small portion of the laser energy back to a surface sensing unit, which uses interferometry techniques to determine strain changes along with the fiber. DAS data in offset wells fall in the low-frequency bands, which has been proven to be a powerful attribute for the characterization of the geometry of hydraulic fractures. Numerous recently published field examples demonstrate the potential of low-frequency DAS (LF-DAS) data for the detailed characterization of the hydraulic fracture geometry. Understanding the fracture-induced rock deformation associated with LF-DAS signals would be beneficial for the better interpretation of real-time data. However, interpretation of LF-DAS measurement is challenging due to the complexity of the subsurface conditions, in addition to potential unanticipated completion issues such as perforation failure, stage isolation failure, etc. All current research efforts focus on the qualitative interpretation of field data.In this study, we quantified the hydraulic fracture propagation process and described the fracture geometry by developing a geomechanical forward model and a Green’s function-based inversion model for the LF-DAS data interpretation, substantially enhancing the value of the LF-DAS data in the process. The work has a significant transformative potential, involving a tool package with developed forward and inversion models that can provide crucial insights for the optimization of hydraulic-fracturing treatments and reservoir development. Methodology The tool package can be used directly in the field to interpret LF-DAS data and monitor hydraulic fracture propagation. Raw data from the field measurement can be automatically processed. The geomechanics forward model we developed can quantify and analyze the strain-rate response from the LF-DAS measurements based on the 3D displacement discontinuity method. Fracture hits are detected by calculating three 1D features along the channel (location) axis, i.e., the maximum strain rate, the summation of strain rates, and the summation of strain-rate amplitudes. Channels with fracture hits usually exhibit significant peak values of these three features. We proposed general guide-lines for fracture-hit detection based on the quantitative analysis of strain/strain-rate responses during the multistage fracturing treatment. The details of the forward model can be found in Liu et al. (SPE 202482, 204457, AMRA-2020-1426). Additionally, we developed a novel Green’s function-based inversion model to qualify fracture width and height based on the determined fracture hits. The strain field that is estimated from the integration of the strain rates measured by the LF-DAS data along the offset monitoring well is related to the fracture widths through a geomechanics Green’s function. The resulting linear system of equations is solved by the least-square method. Details can be found in Liu et al. (SPE 204158, 205379, 204225).

Geomechanical Template for Distributed Acoustic Sensing Strain Patterns during Hydraulic Fracturing

SPE Journal ◽  
2021 ◽  
pp. 1-12
Author(s):  
Yunhui Tan ◽  
Shugang Wang ◽  
Margaretha C. M. Rijken ◽  
Kelly Hughes ◽  
Ivan Lim Chen Ning ◽  
...  
Keyword(s):  
Finite Element ◽  
Hydraulic Fracturing ◽  
Strain Rate ◽  
Hydraulic Fracture ◽  
Strain Field ◽  
Low Frequency ◽  
Fracture Propagation ◽  
Fracture Network ◽  
Hydraulic Fractures ◽  
Distributed Acoustic Sensing

Summary Recently more distributed acoustic sensing (DAS) data have been collected during hydraulic fracturing in shale. Low-frequency DAS signals show patterns that are intuitively consistent with the understanding of the strain field around hydraulic fractures. This study uses a fracture simulator combined with a finite element solver to further understand the various patterns of the strain field caused by hydraulic fracturing. The results can serve as a “type-curve” template for the further interpretation of cross-well strain field plots. Incorporating detailed pump schedule and fracturing fluid/proppant properties, we use a hydraulic fracture simulator to generate fracture geometries, which are then passed to a finite element solver as boundary conditions for elastic-static calculation of the strain field. Because the finite element calculated strain is a tensor, it needs to be projected along the monitoring well trajectory to be comparable with the DAS strain, which is uniaxial. Moreover, the calculated strain field is transformed into a time domain using constant fracture propagation velocity. Strain rate is further derived from the simulated strain field using differentiation along the fracture propagation direction. Scenarios including a single planar hydraulic fracture, a single fracture with a discrete fracture network (DFN), and multiple planar hydraulic fractures in both vertical and horizontal directions were studied. The scenarios can be differentiated in the strain patterns on the basis of the finite element simulation results. In general, there is a tensile heart-shaped zone in front of the propagating fracture tip shown along the horizontal strain direction on both strain and strain rate plots. On the sides, there are compressional zones parallel to the fracture. The strain field projects beyond the depth where the hydraulic fracture is present. Patterns from strain rate can be used to distinguish whether the fracture is intersecting the fiber. Along the vertical direction, the transition zone depicts the upper boundary of the fracture. A complex fracture network with DFN shows a much more complex pattern compared with a single planar fracture. Multiple planar fractures show polarity reversals in horizontal fiber because of interactions between fractures. Data from the Hydraulic Fracturing Test Site 2 (HFTS2) experiment were used to validate the simulated results. The application of the study is to provide a template to better interpret hydraulic fracture characteristics using low-frequency DAS strain-monitoring data. To our understanding, there are no comprehensive templates for engineers to understand the strain signals from cross-well fiber monitoring. The results of this study will guide engineers toward better optimization of well spacing and fracturing design to minimize well interference and improve efficiency.

scholarly journals OC-0254: MR compatibility of fiber optic sensing for real-time needle tracking

2016 ◽  
Vol 119 ◽  
pp. S116-S117
Author(s):  
M. Borot de Battisti ◽  
B. Denise de Senneville ◽  
M. Maenhout ◽  
G. Hautvast ◽  
D. Binnekamp ◽  
...  
Keyword(s):  
Real Time ◽  
Fiber Optic ◽  
Fiber Optic Sensing

scholarly journals Real-Time Well-Integrity Monitoring Using Fiber-Optic Distributed Acoustic Sensing

SPE Journal ◽  
2019 ◽  
Vol 24 (05) ◽  
pp. 1997-2009 ◽  
Author(s):  
T.. Raab ◽  
T.. Reinsch ◽  
S. R. Aldaz Cifuentes ◽  
J.. Henninges
Keyword(s):  
Life Cycle ◽  
Strain Rate ◽  
Real Time ◽  
Fiber Optic ◽  
Axial Strain ◽  
Fiber Optic Cable ◽  
Integrity Monitoring ◽  
Well Integrity ◽  
Distributed Acoustic Sensing ◽  
Cement Integrity

Summary Proper cemented casing strings are a key requirement for maintaining well integrity, guaranteeing optimal operation and safe provision of hydrocarbon and geothermal resources from the pay zone to surface facilities. Throughout the life cycle of a well, high–temperature/high–pressure changes in addition to shut–in cyclic periods can lead to strong variations in thermal and mechanical load on the well architecture. The current procedures to evaluate cement quality and to measure downhole temperature are mainly dependent on wireline–logging campaigns. In this paper, we investigate the application of the fiber–optic distributed–acoustic–sensing (DAS) technology to acquire dynamic axial–strain changes caused by propagating elastic waves along the wellbore structure. The signals are recorded by a permanently installed fiber–optic cable and are studied for the possibility of real–time well–integrity monitoring. The fiber–optic cable was installed along the 18⅝–in. anchor casing and the 21–in.–hole section of a geothermal well in Iceland. During cementing operations, temperature was continuously measured using distributed–temperature–sensing (DTS) technology to monitor the cement placement. DAS data were acquired continuously for 9 days during drilling and injection testing of the reservoir interval in the 12¼–in. openhole section. The DAS data were used to calculate average–axial–strain–rate profiles during different operations on the drillsite. Signals recorded along the optical fiber result from elastic deformation caused by mechanical energy applied from inside (e.g., pressure fluctuations, drilling activities) or outside (e.g., seismic signals) of the well. The results indicate that the average–axial–strain rate of a fiber–optic cable installed behind a casing string generates trends similar to those of a conventional cement–bond log (CBL). The obtained trends along well depth therefore indicate that DAS data acquired during different drilling and testing operations can be used to monitor the mechanical coupling between cemented casing strings and the surrounding formations, hence the cement integrity. The potential use of DTS and DAS technology in downhole evaluations would extend the portfolio to monitor and evaluate qualitatively in real time cement–integrity changes without the necessity of executing costly well–intervention programs throughout the well's life cycle.

Concurrent Real-Time Distributed Fiber Optic Sensing of Casing Deformation and Cement Integrity Loss

2019 ◽  
Author(s):  
Qian Wu ◽  
Sriramya Nair ◽  
Eric van Oort ◽  
Artur Guzik ◽  
Kinzo Kishida
Keyword(s):  
Real Time ◽  
Fiber Optic ◽  
Fiber Optic Sensing ◽  
Cement Integrity ◽  
Casing Deformation
Sign in / Sign up

Export Citation Format

Share Document