CN114895359A - DAS (data acquisition system) same-well monitoring real-time microseism effective event denoising method and system - Google Patents
DAS (data acquisition system) same-well monitoring real-time microseism effective event denoising method and system Download PDFInfo
- Publication number
- CN114895359A CN114895359A CN202210817635.5A CN202210817635A CN114895359A CN 114895359 A CN114895359 A CN 114895359A CN 202210817635 A CN202210817635 A CN 202210817635A CN 114895359 A CN114895359 A CN 114895359A
- Authority
- CN
- China
- Prior art keywords
- micro
- pipeline
- seismic
- event
- wave
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 88
- 238000000034 method Methods 0.000 title claims abstract description 81
- 238000012549 training Methods 0.000 claims abstract description 70
- 238000012545 processing Methods 0.000 claims abstract description 18
- 238000005070 sampling Methods 0.000 claims description 18
- 230000002159 abnormal effect Effects 0.000 claims description 16
- 230000007246 mechanism Effects 0.000 claims description 15
- 238000000605 extraction Methods 0.000 claims description 13
- 238000011176 pooling Methods 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 8
- 230000003321 amplification Effects 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 7
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 7
- 238000012935 Averaging Methods 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 5
- 238000011065 in-situ storage Methods 0.000 claims description 3
- 238000007710 freezing Methods 0.000 claims description 2
- 230000008014 freezing Effects 0.000 claims description 2
- 238000005457 optimization Methods 0.000 claims description 2
- 238000005096 rolling process Methods 0.000 claims 8
- 230000008569 process Effects 0.000 description 20
- 239000013307 optical fiber Substances 0.000 description 11
- 239000002184 metal Substances 0.000 description 9
- 230000003287 optical effect Effects 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 6
- 238000010801 machine learning Methods 0.000 description 6
- 239000000835 fiber Substances 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000001629 suppression Effects 0.000 description 4
- 239000004568 cement Substances 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 238000000926 separation method Methods 0.000 description 3
- 239000002002 slurry Substances 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000013434 data augmentation Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 239000012530 fluid Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000009828 non-uniform distribution Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000010924 continuous production Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000001939 inductive effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000005086 pumping Methods 0.000 description 1
- 238000011158 quantitative evaluation Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000011435 rock Substances 0.000 description 1
- 238000011410 subtraction method Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/288—Event detection in seismic signals, e.g. microseismics
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/40—Seismology; Seismic or acoustic prospecting or detecting specially adapted for well-logging
- G01V1/44—Seismology; Seismic or acoustic prospecting or detecting specially adapted for well-logging using generators and receivers in the same well
- G01V1/48—Processing data
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/30—Noise handling
- G01V2210/32—Noise reduction
Landscapes
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Remote Sensing (AREA)
- Environmental & Geological Engineering (AREA)
- Acoustics & Sound (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- Geophysics And Detection Of Objects (AREA)
- Complex Calculations (AREA)
Abstract
The invention relates to a DAS (data acquisition system) same-well monitoring real-time microseism effective event denoising method and system, belonging to the technical field of microseism monitoring. The method comprises the steps of firstly obtaining a training data set, then training a pre-established denoising model by using the training data set to obtain a trained denoising model, finally obtaining a real-time micro-seismic effective event through DAS same-well monitoring, removing pipeline waves in the real-time micro-seismic effective event by using the trained denoising model to obtain a denoised micro-seismic effective event, and accordingly effectively removing the pipeline waves in the DAS same-well monitoring real-time micro-seismic effective event in a mode of establishing the denoising model to avoid the influence of the pipeline waves on the subsequent processing of the micro-seismic effective event. In addition, the method can also acquire new pipeline wave data according to the period setting so as to update the trained denoising model and use the denoising model in the next period, so that the denoising model has good denoising performance all the time along with the hydraulic fracturing.
Description
Technical Field
The invention relates to the technical field of microseism monitoring, in particular to a DAS (data acquisition system) same-well monitoring real-time microseism effective event denoising method and system based on deep learning.
Background
The hydraulic fracturing technology injects high-pressure fluid into a shale reservoir to generate complex artificial fractures, so that the connectivity of the reservoir can be increased, and the single-well yield can be improved. In the hydraulic fracturing process, a fracturing fluid is injected into a fracturing well at a wellhead by using a high-pressure pump, and a subsurface reservoir stratum is fractured due to stress changes to generate micro-earthquakes. The microseism monitoring technology monitors the fracturing process and evaluates the fracturing effect by monitoring microseism signals generated in the hydraulic fracturing process so as to guide and optimize engineering parameters. The microseism monitoring mainly comprises the following steps: effective event pickup, effective event signal enhancement (noise suppression), micro seismic source positioning, micro seismic source analysis, reservoir stress analysis and the like.
DAS is an emerging data acquisition technology which is rapidly developed in recent years, and achieves measurement of strain signals along the axial direction of optical fibers by detecting phase changes of backward Rayleigh scattering optical signals generated by laser pulses on scatterers inside the optical fibers. In the DAS hydraulic fracturing monitoring, optical fibers are distributed along a horizontal well, full-well-section measurement data can be obtained, and 1m space sampling is realized, so that the DAS has the advantages of wide azimuth and high density data.
DAS monitoring is generally divided into co-well monitoring and adjacent well monitoring. In the same-well monitoring, the monitoring well and the fracturing well are the same horizontal well, and pipeline waves with strong energy can be generated in the working process of the fracturing pump, so that the pipeline waves with strong energy exist in the micro-seismic effective events obtained by the DAS same-well monitoring. The pipe waves have the characteristic of high amplitude, which exceeds the amplitude of microseismic active events; meanwhile, the pipeline wave is transmitted along the monitoring well and has strong periodicity, so that the pipeline wave has large interference on the microseism effective event and cannot be ignored. Meanwhile, the pipeline wave has higher similarity with the microseism effective event, and the frequency band of the pipeline wave is partially overlapped with the frequency band of the microseism effective event in the frequency domain; in the time domain, the waveform of the pipe wave is similar to that of the microseismic active event, so that the pipe wave and the microseismic active event are difficult to separate. In addition, in DAS adjacent well monitoring, a monitoring well and a fracturing well are different horizontal wells, and the pipeline waves which are transmitted along the fracturing well do not exist in data, namely the data only contain microseismic effective events. Because the same monitoring well and the same monitoring system are used for data acquisition in the same-well monitoring and the adjacent-well monitoring, the microseism effective event monitored by the adjacent well has higher similarity with the microseism effective event monitored by the same well.
When a micro-seismic effective event is detected, the micro-seismic effective event generally needs to be subjected to subsequent processing, including seismic source positioning, mechanism inversion and the like, but the pipeline waves generated by DAS same-well monitoring are used as consistent noise and have the characteristics of high amplitude, periodicity and the like, so that the subsequent analysis of the micro-seismic effective event is seriously influenced, and therefore the pipeline waves need to be separated.
Based on the above, a denoising method and a denoising system capable of effectively removing the pipeline waves in the DAS real-time micro-seismic effective event in the same well monitoring are needed.
Disclosure of Invention
The invention aims to provide a denoising method and a denoising system for DAS same-well monitoring real-time micro-seismic effective events, which can effectively remove pipeline waves in the DAS same-well monitoring real-time micro-seismic effective events and avoid influencing the follow-up processing of the micro-seismic effective events.
In order to achieve the purpose, the invention provides the following scheme:
a denoising method for DAS same-well monitoring real-time microseism effective events comprises the following steps:
acquiring a training data set; the training data set comprises a plurality of pipeline wave-containing micro-seismic effective events and pipeline wave-free micro-seismic effective events corresponding to each pipeline wave-containing micro-seismic effective event;
training a pre-established denoising model by using the training data set to obtain a trained denoising model;
and acquiring a real-time micro-seismic effective event through the DAS same-well monitoring, and removing the pipeline waves in the real-time micro-seismic effective event by using the trained denoising model by taking the real-time micro-seismic effective event as input to obtain a denoised micro-seismic effective event.
Optionally, before the real-time micro-seismic effective event is used as an input and the trained denoising model is used to remove the pipeline waves in the real-time micro-seismic effective event, the denoising method further includes: judging whether the monitoring time of the DAS same-well monitoring reaches a preset period or not; if so, updating the training data set by using the pipeline wave data obtained in the DAS same-well monitoring process in the current period to obtain an updated data set, and updating the trained denoising model by using the updated data set to obtain the trained denoising model in the next period.
A DAS in-situ monitoring real-time microseismic active event de-noising system, comprising:
the data set acquisition module is used for acquiring a training data set; the training data set comprises a plurality of pipeline wave-containing micro-seismic effective events and pipeline wave-free micro-seismic effective events corresponding to each pipeline wave-containing micro-seismic effective event;
the training module is used for training a pre-established denoising model by using the training data set to obtain a trained denoising model;
and the denoising module is used for acquiring a real-time micro-seismic effective event through the DAS same-well monitoring, removing the pipeline waves in the real-time micro-seismic effective event by using the trained denoising model by taking the real-time micro-seismic effective event as input, and obtaining the denoised micro-seismic effective event.
Optionally, the denoising system further includes an updating module, configured to determine whether the monitoring time of the DAS monitoring in the same well reaches a preset period; if so, updating the training data set by using the pipeline wave data obtained in the DAS same-well monitoring process in the current period to obtain an updated data set, and updating the trained denoising model by using the updated data set to obtain the trained denoising model in the next period.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides a DAS same-well monitoring real-time micro-seismic effective event denoising method and a system. And then training the pre-established denoising model by using a training data set to obtain the trained denoising model. And finally, acquiring a real-time micro-seismic effective event through DAS (data acquisition system) same-well monitoring, and removing the pipeline waves in the real-time micro-seismic effective event by using a trained denoising model by taking the real-time micro-seismic effective event as input to obtain a denoised micro-seismic effective event, so that the pipeline waves in the DAS same-well monitoring real-time micro-seismic effective event are effectively removed by establishing the denoising model, and the influence of the pipeline waves on the subsequent processing of the micro-seismic effective event is avoided. In addition, the method can also acquire new pipeline wave data according to the period setting so as to update the training data set, further update the trained denoising model, and be used in the next period, so that the denoising model has good denoising performance all the time along with the hydraulic fracturing.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
Fig. 1 is a flowchart of a denoising method according to embodiment 1 of the present invention;
fig. 2 is a schematic block diagram of a denoising method according to embodiment 1 of the present invention;
fig. 3 is a schematic diagram of a mesh structure of the denoising model provided in embodiment 1 of the present invention;
fig. 4 is a system block diagram of a denoising system provided in embodiment 2 of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a denoising method and a denoising system for DAS same-well monitoring real-time micro-seismic effective events, which can effectively remove pipeline waves in the DAS same-well monitoring real-time micro-seismic effective events and avoid influencing the follow-up processing of the micro-seismic effective events.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Example 1:
in the process of monitoring hydraulic fracturing by using the DAS, the DAS has the advantages of wide azimuth and high-density data, and meanwhile, microseism monitoring is used as a continuous and long-time process, so that the obtained monitoring data volume is very large, and the DAS has the characteristic of large data volume. For example, DAS sampling is 1 ms, and for example, according to a single well 2000 well depth, approximately 650GB of data can be obtained in one day according to 1m spatial sampling, and thus it is impossible to process the data by a manual method or a conventional method, and it is necessary to improve the processing efficiency by an artificial intelligence method.
The basic theory of DAS demodulation is: the non-uniform distribution of the material density due to the non-uniformity of the temperature during the cooling process during the fabrication of the optical fiber, and the non-uniform distribution of the density of the particles doped for enhancing the scattering effect, causes a difference in the refractive index in the optical fiber, thereby generating rayleigh scattering, which exists in the entire space of the optical fiber. Because the fiber density of each position in the whole fiber is random, the refractive index of each position of the whole fiber is in random distribution microscopically, and the Rayleigh scattering light does not cause frequency drift due to scattering.
Based on the above, before the denoising method of the embodiment is introduced, a DAS layout method and a main hydraulic fracturing flow in the DAS monitoring hydraulic fracturing process are introduced:
the DAS layout mode is as follows: the horizontal well micro-seismic monitoring system based on distributed optical fiber sensing comprises a metal sleeve and a single-mode optical fiber, wherein an armored optical cable is fixed on the outer side of the metal sleeve, a special single-mode optical fiber is arranged in the armored optical cable, a DAS (data acquisition system) modulation and demodulation instrument is placed near a wellhead, and a DAS signal port of the demodulation instrument is connected with the special single-mode optical fiber outside the sleeve.
The main flow of hydraulic fracturing is as follows:
(1) the metal casing and the armored cable are slowly lowered into the drilled borehole simultaneously.
(2) The annular metal clip is arranged at the joint of two metal casing pipes at a wellhead, so that the armored optical cable is fixed and protected from moving and/or being damaged in the casing pipe descending process.
(3) Pumping cement slurry from the well bottom by a high-pressure pump truck, returning the cement slurry to the well head from the well bottom along the annular space between the outer wall of the metal casing and the drill hole, and permanently fixing the metal casing, the armored optical cable and the stratum rock together after the cement slurry is solidified.
(4) And connecting the single mode optical fiber outside the sleeve in the armored optical cable to the DAS signal input end of the DAS modulation and demodulation instrument at the wellhead.
(5) Collecting three-dimensional ground seismic data of the area around the horizontal well and preprocessing the three-dimensional ground seismic data to obtain a three-dimensional seismic longitudinal wave velocity data volume, and then calibrating, adjusting and updating the three-dimensional seismic longitudinal wave velocity data volume by using acoustic logging velocity data to obtain a primary seismic longitudinal wave velocity field of the stratum around the horizontal well.
(6) And (3) sequentially carrying out directional perforation operation on the metal casing at a pre-designed perforation position in the underground, simultaneously recording perforation signals generated during the directional perforation operation by using single mode fibers outside the casing arranged in the underground and DAS modulation and demodulation instruments near a wellhead, and calibrating and updating the primary seismic longitudinal wave velocity field in the step (5) by using the perforation signals to obtain a velocity field finally used for analyzing the effective events of the hydraulic fracturing micro-earthquake.
(7) During hydraulic fracturing operation, the system can be used for carrying out hydraulic fracturing micro-seismic monitoring by using an armored optical cable which is permanently arranged outside a metal sleeve. The method comprises the steps of collecting data by using single mode fibers outside a sleeve arranged underground, transmitting the data to a DAS modulation and demodulation instrument near a wellhead for demodulation, and obtaining a continuously recorded microseism effective event generated when an underground stratum is broken due to adjacent well or same well hydraulic fracturing operation.
(8) And (4) obtaining the relevant information of the micro-seismic effective event by combining the longitudinal wave velocity field of the underground stratum obtained in the step (6), wherein the information comprises the information of the occurrence time, the three-dimensional spatial position, the energy size and the like of the micro-seismic effective event.
(9) After the hydraulic fracturing is finished, performing seismic source mechanism analysis and seismic grade analysis according to the recorded micro-seismic effective events, further obtaining the fracture mechanism of the underground reservoir of the micro-seismic effective events, and improving the distribution characteristics and the distribution rule of the underground reservoir fractures. And the information is integrated, and effective and reliable qualitative and quantitative evaluation is carried out on the reservoir hydraulic fracturing modification effect of the horizontal well.
In the step (7), the microseism effective event monitored in the same well comprises serious pipeline wave interference, which can seriously affect the subsequent analysis performed by using the microseism effective event in the steps (8) and (9), namely in the DAS same well monitoring, the picked microseism effective event comprises the pipeline wave interference, pipeline wave suppression is required, and the microseism effective event without the pipeline wave is extracted. However, at present, there is no denoising technology for real-time microseism effective events monitored in the same well of DAS, and among the technologies similar to this, there are at present: 1) separation methods for conventional microseismic data, e.g., non-ML methods and ML-based methods; 2) the method for separating the pipeline waves of the same well of the conventional detector generally comprises a physical suppression method and a data separation algorithm, wherein the physical suppression method generally comprises the step of adding a damper in a casing, and the data separation algorithm generally comprises a matched subtraction method, a time-space domain filtering method, a vector median filtering method and an SVD (singular value decomposition) method. However, the prior art can not solve the problem of effectively removing the pipeline wave interference in the micro-seismic effective event obtained by the DAS same-well monitoring.
The Machine Learning (ML) method is a practical tool complementary to the conventional data analysis method, and is used for Learning data based on a probabilistic model, has the capability of extracting complex patterns and effective features from a data stream, and also has the potential of inducing and deducing new characteristics and new mechanisms, can improve the prediction capability of a system. The complexity of the earth structure, the indirection of geophysical observation and the complexity of digital tools bring great challenges to the development of geophysical research, the combination of data-driven disciplines and model-driven technologies is helpful to deeply explore difficult problems in the field of seismology, accelerate the generation of new knowledge, and the application of machine learning in the field of seismology is continuously expanded. In this context, a machine learning method is an excellent choice.
Based on this, the present embodiment is configured to provide a denoising method for DAS real-time microseism effective events in same well monitoring, as shown in fig. 1 and fig. 2, the denoising method includes:
s1: acquiring a training data set; the training data set comprises a plurality of pipeline wave-containing micro-seismic effective events and pipeline wave-free micro-seismic effective events corresponding to each pipeline wave-containing micro-seismic effective event;
specifically, S1 may include:
(1) and (3) data generation: pipeline wave-free micro-seismic effective events are obtained through DAS adjacent well monitoring, and are obtained from the work area or other work areas in advance, and the work areas with similar reservoir structures are optimized. And acquiring pipeline wave data through DAS same-well monitoring, and adding the pipeline wave data monitored by the same well and the pipeline wave-free micro-seismic effective event monitored by an adjacent well to obtain a pipeline wave-containing micro-seismic effective event corresponding to the pipeline wave-free micro-seismic effective event.
It should be noted that the adjacent well monitoring and the same well monitoring acquire data using the same parameters, but do not necessarily require that the summed pipe wave data and the pipe wave-free microseismic events be acquired simultaneously. A single pipe-wave-free microseismic event contains Nt (which may be 1000) channels, and each channel has Ns (which may be 1000) sampling points. The pipe wave data also contains Nt (which may be 1000) channels, each of which has a total of Ns (which may be 1000) samples.
Preferably, the embodiment can also augment the data in the data generation process to improve the universality of the data set.
Specifically, after the effective micro-seismic event without the pipeline wave is obtained, the denoising method of the embodiment further includes: and carrying out data amplification on the pipeline wave-free micro-seismic effective event to obtain a new pipeline wave-free micro-seismic effective event. Wherein, the data augmentation of the pipeline wave-free micro-seismic significant event may include: the effective micro-seismic events without the pipeline waves are zoomed according to different amplitude scales, wherein the amplitude scales refer to the data size scales of the sampling points of the effective micro-seismic events without the pipeline waves, namely the data sizes of the sampling points of the effective micro-seismic events without the pipeline waves are zoomed according to a certain proportion; moving the positions of the channels which do not contain the pipeline wave micro-seismic effective events, such as translating left and right or translating up and down, so as to change the positions of the channels, for example, the positions of the channel 1 and the channel 3 are exchanged; the method comprises the steps of replacing a normal channel included in the pipeline wave-free micro-seismic effective event with an abnormal channel, namely constructing the abnormal channel in the pipeline wave-free micro-seismic effective event, wherein the abnormal channel can be a bad channel or a missing channel, the bad channel means that the data value of the whole channel is particularly large or small, and the missing channel means that the data in the whole channel is missing.
After the pipeline wave data is obtained, the denoising method of the embodiment further includes: and carrying out data amplification on the pipeline wave data to obtain new pipeline wave data. Wherein, performing data augmentation on the pipeline wave data may include: the method comprises the steps of scaling the pipeline wave data according to different amplitude scales, wherein the amplitude scales refer to data size scales of sampling points of the pipeline wave data, namely scaling the data size of each sampling point of the pipeline wave data according to a certain proportion.
By the method, the acquired pipeline wave data and the pipeline wave-free micro-seismic effective events can be subjected to data amplification, and the acquired new pipeline wave data and the new pipeline wave-free micro-seismic effective events are used for forming pipeline wave-containing micro-seismic effective events.
(2) And repeating the steps until N pipeline wave-free micro-seismic effective events and pipeline wave-containing micro-seismic effective events corresponding to each pipeline wave-free micro-seismic effective event are obtained, and obtaining a training data set.
The outline of the training data set of this example is: the method comprises the steps that 2D pipeline wave-containing micro-seismic effective events and 2D pipeline wave-free micro-seismic effective events are matched and correspond to each other, and the sizes of the two events are consistent. The training dataset of this embodiment may contain Nk (1000) micro-seismic significant events, and thus the data volumes (with and without pipe-wave micro-seismic significant events) are Nk Ns Nt.
S2: training a pre-established denoising model by using the training data set to obtain a trained denoising model;
before S2, the denoising method of the present embodiment may further include: the training data set is preprocessed to obtain a preprocessed data set, and S2 is performed with the preprocessed data set as a new training data set. Wherein, the preprocessing of the same step is performed on all the data obtained by actual monitoring in the training data set, and then the preprocessing of the training data set may include: and carrying out abnormal channel processing, averaging processing, abnormal value removing processing and data normalization processing on the micro-seismic effective event containing the pipeline waves and the micro-seismic effective event not containing the pipeline waves. The abnormal channel processing means that an abnormal channel is determined, and then interpolation replacement is carried out through an adjacent channel. The averaging process means averaging the current data Ns × Nt. The abnormal value removing process is to determine an abnormal value, and then replace the abnormal value with an interpolation value, wherein the abnormal value is a value outside an interval formed by 95% of the maximum value and 95% of the minimum value of the current data Ns × Nt. The data normalization can adopt any existing normalization mode, such as maximum and minimum normalization, mean normalization and the like. It should be noted that the training data set may be preprocessed using any combination of the above preprocessing methods.
The denoising model used in this embodiment may adopt a deep learning model, and the module architecture thereof may be: the method comprises the steps that features of training data are shrunk and expanded on a spatial scale through a pair of down-sampling and up-sampling operations, a plurality of pairs of the up-sampling and the down-sampling operations can be used for realizing feature extraction on different spatial scales, in order to keep feature details on different scales, the down-sampling and the up-sampling on the same spatial scale are connected through an attention mechanism layer, the attention mechanism layer comprises a convolution layer, a maximum pooling layer, an average pooling layer and a connection layer, the attention mechanism layer has an up-sampling function, and input and output dimensions of the attention mechanism layer are kept consistent. As shown in fig. 3, the denoising model of this embodiment includes an input layer, a feature extraction layer, a full connection layer, and an output layer, which are connected in sequence, where the feature extraction layer includes a first convolution block, a maximum pooling layer, a second convolution block, a maximum pooling layer, a third convolution block, a maximum pooling layer, a fourth convolution block, a fifth convolution block, an upsampling layer, a sixth convolution block, an upsampling layer, a seventh convolution block, an upsampling layer, and an eighth convolution block, which are connected in sequence, the first convolution block is connected with the eighth convolution block through an attention mechanism layer, and has the same spatial scale, the second convolution block is connected with the seventh convolution block through an attention mechanism layer, and has the same spatial scale, the third convolution block is connected with the sixth convolution block through an attention mechanism layer, and has the same spatial scale, and the fourth convolution block is connected with the fifth convolution block through an attention mechanism layer. The spatial dimension of the second volume block is half that of the first volume block, the spatial dimension of the third volume block is half that of the second volume block, and the spatial dimension of the fourth volume block is half that of the third volume block.
More specifically, the first convolution block, the second convolution block, and the third convolution block each include two first convolution layers, the fourth convolution block includes one first convolution layer, the first convolution layer employs a 2D convolution kernel, the convolution size is 3 × 3, each first convolution layer is activated using a correction linear unit, that is, a ReLU activation function is employed, and padding and stride are 1 and 2, respectively. The 1-7 convolutional layer parameters are: 32 × 3 × 3, 32 × 3 × 3, 64 × 3 × 3, 64 × 3 × 3, 128 × 3 × 3, 128 × 3 × 3, 256 × 3 × 3, where the first dimension is the number of convolutions and the last two dimensions are the convolution sizes. The fifth convolution block includes a second convolution layer, and the sixth convolution block, the seventh convolution block, and the eighth convolution block each include two second convolution layers, the second convolution layers employ 2D convolution kernels, the convolution size is 2 × 2, each second convolution layer is activated using a corrected linear cell, that is, employing the ReLU activation function, padding and stride are 1 and 2, respectively.
The input of the denoising model of this embodiment is a pipeline-wave-containing micro-seismic effective event, the output is a predicted denoised micro-seismic effective event, and the label is a pipeline-wave-free micro-seismic effective event corresponding to the pipeline-wave-containing micro-seismic effective event. When the denoising model is trained, the predicted error between the denoised micro-seismic effective event and the pipeline wave-free micro-seismic effective event is used for updating the network parameters of the denoising model, specifically, MSE (mean squared error) can be used as a loss function for calculating the error of the denoising model, and a random gradient descent optimization method is used for training. The dynamic learning rate is set during the training process, the initial value is set to 0.0001, the dynamic learning rate is reduced by half every 50 iterations, the batch processing amount is set to 40, and the number of iterations can be set to 200.
In this embodiment, when the denoising model is trained, the training data set may be further divided into a training set and a test set, where the ratio is 8: 2. and training the denoising model by using a training set, testing the denoising model obtained by training by using a test set, and selecting a model with the highest precision as the trained denoising model.
The model training of the present embodiment is performed on the GPU image processing unit.
S3: and acquiring a real-time micro-seismic effective event through DAS (data acquisition system) same-well monitoring, and removing the pipeline waves in the real-time micro-seismic effective event by using the trained denoising model by taking the real-time micro-seismic effective event as input to obtain a denoised micro-seismic effective event.
When actual monitoring data is analyzed and processed, real-time micro-seismic effective events are collected by DAS same-well monitoring, the real-time micro-seismic effective events comprise pipeline wave noise, the real-time micro-seismic effective events are subjected to preprocessing with the same steps as those of a training data set, the preprocessing can comprise damaged channel processing (interpolation replacement is carried out through adjacent channels), averaging is carried out, abnormal values are removed (the abnormal values are replaced by interpolation values), and data normalization is carried out. And then sending the preprocessed real-time micro-seismic effective events into a trained denoising model to obtain the denoised micro-seismic effective events.
As the hydraulic fracturing is a continuous process, the denoising method of the embodiment can be applied to the whole process of the hydraulic fracturing to process the continuously obtained real-time microseism effective events. In order to enable the trained denoising model to be suitable for the whole hydraulic fracturing process and have the best denoising performance on the obtained real-time microseism effective events all the time, the embodiment sets a model updating threshold value in the monitoring process, and when the monitoring time reaches a certain period (for example, 1 day), the model updating process is started to update the trained denoising model.
Specifically, before the real-time micro-seismic effective event is taken as an input and the pipeline waves in the real-time micro-seismic effective event are removed by using the trained denoising model, the denoising method of the embodiment further includes:
(1) judging whether the monitoring time of the DAS same-well monitoring reaches a preset period or not;
the preset period can be set artificially, such as 1 day and 2 days.
(2) If so, updating the training data set by using the pipeline wave data obtained in the DAS same-well monitoring process in the current period to obtain an updated data set, and updating the trained denoising model by using the updated data set to obtain the trained denoising model in the next period.
The updating process of the training data set may include: most of data obtained by the DAS same-well monitoring is pipeline wave data, and the rest data is a micro-seismic effective event containing pipeline waves. In the monitoring process in one period, when a real-time micro-seismic effective event is monitored, denoising is carried out by using a trained denoising model in the current period, pipeline wave data can be collected regularly in the current period, namely, the pipeline wave data with a certain time length is obtained according to a fixed time along with the progress of fracturing so as to update the pipeline wave data, then a training data set is updated by using the pipeline wave data, the pipeline wave-containing micro-seismic effective event is updated by using the updated pipeline wave data, and an updated data set is obtained.
The trained denoising model comprises an input layer, a feature extraction layer, a full connection layer and an output layer which are sequentially connected, and the updating of the trained denoising model by using the updated data set can comprise: freezing the network parameters of the feature extraction layer, updating the mapping relation between the feature extraction layer and the full connection layer by using the updated data set to obtain new network parameters, and updating the trained denoising model for use in the next period.
Specifically, in the updating process, the network structure of the feature extraction layer is frozen, that is, the parameters of the feature extraction layer are kept unchanged, all updated pipeline-wave-containing micro-seismic effective events and part of old pipeline-wave-containing micro-seismic effective events in the training data set are used (the number of the two data is still Nk, and the number of the two data is kept consistent), and the mapping relation between the feature output layer and the full connection layer is trained, that is: extracting the characteristics output by the last layer of the characteristic output layer, carrying out fine-tune (fine-tune) on the model by using new data, and reestablishing the mapping relation from the characteristic output layer to the full connection layer. The transfer training is only carried out on the mapping relation from the characteristic output layer of the model to the full connection layer, and the whole model is not required to be trained, so that the training speed can be improved.
Example 2:
the embodiment is used to provide a system for denoising real-time microseism effective events in DAS co-well monitoring, as shown in fig. 4, the denoising system includes:
a data set obtaining module M1, configured to obtain a training data set; the training data set comprises a plurality of pipeline wave-containing micro-seismic effective events and pipeline wave-free micro-seismic effective events corresponding to each pipeline wave-containing micro-seismic effective event;
the training module M2 is used for training a pre-established denoising model by using the training data set to obtain a trained denoising model;
and the denoising module M3 is used for acquiring a real-time micro-seismic effective event through the DAS same-well monitoring, removing the pipeline waves in the real-time micro-seismic effective event by using the trained denoising model by taking the real-time micro-seismic effective event as input, and acquiring the denoised micro-seismic effective event.
Preferably, the denoising system further comprises an updating module for judging whether the monitoring time of the DAS co-well monitoring reaches a preset period; if so, updating the training data set by using the pipeline wave data obtained in the DAS same-well monitoring process in the current period to obtain an updated data set, and updating the trained denoising model by using the updated data set to obtain the trained denoising model in the next period.
The emphasis of each embodiment in the present specification is on the difference from the other embodiments, and the same and similar parts among the various embodiments may be referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (10)
1. A denoising method for DAS same-well monitoring real-time microseism effective events is characterized by comprising the following steps:
acquiring a training data set; the training data set comprises a plurality of pipeline wave-containing micro-seismic effective events and pipeline wave-free micro-seismic effective events corresponding to each pipeline wave-containing micro-seismic effective event;
training a pre-established denoising model by using the training data set to obtain a trained denoising model;
and acquiring a real-time micro-seismic effective event through the DAS same-well monitoring, and removing the pipeline waves in the real-time micro-seismic effective event by using the trained denoising model by taking the real-time micro-seismic effective event as input to obtain a denoised micro-seismic effective event.
2. The method of claim 1, wherein the acquiring of the training data set specifically comprises:
acquiring a micro-seismic effective event without pipeline waves by DAS adjacent well monitoring, and acquiring pipeline wave data by DAS same-well monitoring; adding the pipeline wave data and the pipeline wave-free micro-seismic effective event to obtain a pipeline wave-containing micro-seismic effective event corresponding to the pipeline wave-free micro-seismic effective event;
repeating the steps of obtaining the micro-earthquake effective event without the pipeline wave through DAS adjacent well monitoring, and obtaining the pipeline wave data through DAS same-well monitoring; adding the pipeline wave data and the pipeline wave-free micro-seismic effective events to obtain pipeline wave-containing micro-seismic effective events corresponding to the pipeline wave-free micro-seismic effective events, until N pipeline wave-free micro-seismic effective events and the pipeline wave-containing micro-seismic effective events corresponding to each pipeline wave-free micro-seismic effective event are obtained, and obtaining a training data set.
3. The method of claim 2, wherein after the pipeline wave-free real-time microseismic active event is obtained, the denoising method further comprises: carrying out data amplification on the pipeline wave-free micro-seismic effective event to obtain a new pipeline wave-free micro-seismic effective event;
wherein, the data amplification of the pipeline wave-free micro-seismic effective event specifically comprises the following steps: scaling the pipeline wave-free micro-seismic effective events according to different amplitude scales; moving the position of each channel included by the pipeline wave-free micro-seismic effective event; and replacing the normal channel included by the pipeline wave-free micro-seismic effective event with an abnormal channel.
4. The method of claim 2, wherein after the pipeline wave data is obtained, the method further comprises: performing data amplification on the pipeline wave data to obtain new pipeline wave data;
wherein, to carry out data amplification to pipeline wave data specifically includes: and scaling the pipeline wave data according to different amplitude scales.
5. The method of claim 1, wherein prior to training a pre-established denoising model with the training data set, the denoising method further comprises: preprocessing the training data set to obtain a preprocessed data set, and taking the preprocessed data set as a new training data set;
wherein preprocessing the training data set specifically comprises: and carrying out abnormal channel processing, averaging processing, abnormal value removing processing and data normalization processing on the pipeline wave-containing micro-seismic effective event and the pipeline wave-free micro-seismic effective event.
6. The DAS same-well monitoring real-time microseism active event denoising method of claim 1, wherein the denoising model comprises an input layer, a feature extraction layer, a full connection layer and an output layer which are connected in sequence; the characteristic extraction layer comprises a first rolling block, a maximum pooling layer, a second rolling block, a maximum pooling layer, a third rolling block, a maximum pooling layer, a fourth rolling block, a fifth rolling block, an upper sampling layer, a sixth rolling block, an upper sampling layer, a seventh rolling block, an upper sampling layer and an eighth rolling block which are connected in sequence; the first convolution block and the eighth convolution block are connected through an attention mechanism layer, and the spatial scales of the first convolution block and the eighth convolution block are the same; the second convolution block and the seventh convolution block are connected through an attention mechanism layer, and the spatial scales of the second convolution block and the seventh convolution block are the same; the third convolution block and the sixth convolution block are connected through an attention mechanism layer, and the spatial scales of the third convolution block and the sixth convolution block are the same; the fourth volume block has the same spatial dimension as the fifth volume block.
7. The DAS same-well monitoring real-time microseism active event denoising method of claim 1, wherein when a pre-established denoising model is trained by the training data set, MSE is used as a loss function, and a stochastic gradient descent optimization method is used for training.
8. The DAS in-situ monitoring real-time micro-seismic significant event denoising method of claim 1, wherein before the real-time micro-seismic significant event is taken as an input and the pipeline waves in the real-time micro-seismic significant event are removed by the trained denoising model, the denoising method further comprises:
judging whether the monitoring time of the DAS same-well monitoring reaches a preset period or not;
if yes, updating the training data set by using the tube wave data obtained in the DAS same-well monitoring process in the current period to obtain an updated data set, and updating the trained denoising model by using the updated data set to obtain the trained denoising model in the next period.
9. The DAS in-situ monitoring real-time microseism active event denoising method of claim 8, wherein the trained denoising model comprises an input layer, a feature extraction layer, a full connection layer and an output layer which are connected in sequence, and the updating the trained denoising model by using the updated data set specifically comprises:
and freezing the network parameters of the feature extraction layer, and updating the mapping relation between the feature extraction layer and the full connection layer by using the updated data set.
10. A DAS same-well monitoring real-time microseismic active event denoising system, comprising:
the data set acquisition module is used for acquiring a training data set; the training data set comprises a plurality of pipeline wave-containing micro-seismic effective events and pipeline wave-free micro-seismic effective events corresponding to each pipeline wave-containing micro-seismic effective event;
the training module is used for training a pre-established denoising model by using the training data set to obtain a trained denoising model;
and the denoising module is used for acquiring a real-time micro-seismic effective event through the DAS same-well monitoring, removing the pipeline waves in the real-time micro-seismic effective event by using the trained denoising model by taking the real-time micro-seismic effective event as input, and obtaining the denoised micro-seismic effective event.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210817635.5A CN114895359B (en) | 2022-07-13 | 2022-07-13 | DAS (data acquisition system) same-well monitoring real-time microseism effective event denoising method and system |
| LU502639A LU502639B1 (en) | 2022-07-13 | 2022-08-09 | A method and system for denoising valid real-time microseismic events by DAS Ko-Well monitoring |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210817635.5A CN114895359B (en) | 2022-07-13 | 2022-07-13 | DAS (data acquisition system) same-well monitoring real-time microseism effective event denoising method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114895359A true CN114895359A (en) | 2022-08-12 |
| CN114895359B CN114895359B (en) | 2022-09-13 |
Family
ID=82730181
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210817635.5A Active CN114895359B (en) | 2022-07-13 | 2022-07-13 | DAS (data acquisition system) same-well monitoring real-time microseism effective event denoising method and system |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN114895359B (en) |
| LU (1) | LU502639B1 (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102472825A (en) * | 2009-08-07 | 2012-05-23 | 埃克森美孚上游研究公司 | Drilling advisory systems and methods utilizing objective functions |
| CN102893183A (en) * | 2010-03-05 | 2013-01-23 | 瓦勒吉有限责任公司 | Active noise injection computations for improved predictability in oil and gas reservoir discovery and characterization |
| US20200088897A1 (en) * | 2018-09-14 | 2020-03-19 | Bp Corporation North America Inc. | Machine Learning-Based Analysis of Seismic Attributes |
| US20210072416A1 (en) * | 2019-09-10 | 2021-03-11 | Halliburton Energy Services, Inc. | Microseismic velocity models derived from historical model classification |
| CN113296148A (en) * | 2021-05-25 | 2021-08-24 | 电子科技大学 | Microseismic identification method based on time domain and wavelet domain dual-channel convolutional neural network |
| CN113330185A (en) * | 2018-11-29 | 2021-08-31 | Bp探索操作有限公司 | Event detection using DAS features using machine learning |
| CN113962244A (en) * | 2020-07-03 | 2022-01-21 | 中国石油化工股份有限公司 | Rayleigh wave seismic data noise removal method, storage medium and electronic device |
-
2022
- 2022-07-13 CN CN202210817635.5A patent/CN114895359B/en active Active
- 2022-08-09 LU LU502639A patent/LU502639B1/en active IP Right Grant
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102472825A (en) * | 2009-08-07 | 2012-05-23 | 埃克森美孚上游研究公司 | Drilling advisory systems and methods utilizing objective functions |
| CN102893183A (en) * | 2010-03-05 | 2013-01-23 | 瓦勒吉有限责任公司 | Active noise injection computations for improved predictability in oil and gas reservoir discovery and characterization |
| US20200088897A1 (en) * | 2018-09-14 | 2020-03-19 | Bp Corporation North America Inc. | Machine Learning-Based Analysis of Seismic Attributes |
| CN113330185A (en) * | 2018-11-29 | 2021-08-31 | Bp探索操作有限公司 | Event detection using DAS features using machine learning |
| US20210072416A1 (en) * | 2019-09-10 | 2021-03-11 | Halliburton Energy Services, Inc. | Microseismic velocity models derived from historical model classification |
| CN113962244A (en) * | 2020-07-03 | 2022-01-21 | 中国石油化工股份有限公司 | Rayleigh wave seismic data noise removal method, storage medium and electronic device |
| CN113296148A (en) * | 2021-05-25 | 2021-08-24 | 电子科技大学 | Microseismic identification method based on time domain and wavelet domain dual-channel convolutional neural network |
Non-Patent Citations (2)
| Title |
|---|
| SARAH DONNE 等: "Wave height quantification using land based seismic data with grammatical evolution", 《2014 IEEE CONGRESS ON EVOLUTIONARY COMPUTATION (CEC)》 * |
| 马孜卓 等: "基于非监督学习的水力压裂裂缝平面识别", 《中国地震》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114895359B (en) | 2022-09-13 |
| LU502639B1 (en) | 2023-02-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114879252B (en) | DAS (data acquisition system) same-well monitoring real-time microseism effective event identification method based on deep learning | |
| CN115291281B (en) | Real-time micro-earthquake magnitude calculation method and device based on deep learning | |
| RU2321064C2 (en) | Method for building a reversible three-dimensional hydrodynamic model of earth, calibrated in real time during drilling | |
| US20230213669A1 (en) | Intelligent geophysical data acquisition system and acquisition method for shale oil and gas optical fiber | |
| CA3032777C (en) | Multivariate analysis of seismic data, microseismic data, and petrophysical properties in fracture modeling | |
| CN114035228B (en) | Tunnel seismic wave velocity inversion method and system based on deep learning | |
| US9939541B2 (en) | Layered linear inversion techniques for locating microseismic activity | |
| CN116378648B (en) | Near-bit stratum detection method and device based on while-drilling acoustic wave forward looking | |
| CN112180429B (en) | Unfavorable geological structure detection system and method by utilizing tunnel blasting vibration inversion | |
| CN103913768A (en) | Method and device for modeling superficial layer in earth surface based on seismic wave data | |
| WO2014018704A1 (en) | Methods for interpretation of time-lapse borehole seismic data for reservoir monitoring | |
| CN114895359B (en) | DAS (data acquisition system) same-well monitoring real-time microseism effective event denoising method and system | |
| CN115373029B (en) | Real-time micro-seismic source mechanism calculation method and system based on deep learning | |
| CN116227309B (en) | Construction method and system of shield tunneling digital twin stratum fused with multi-source data | |
| WO2022125107A1 (en) | Deep learning methods for wellbore leak detection | |
| CN115453624A (en) | Mine following excavation seismic signal imaging method | |
| CN107679614B (en) | Particle swarm optimization-based real-time sound wave time difference extraction method | |
| CN112031755B (en) | Fracturing monitoring device, method and system, electronic equipment and storage medium | |
| CN112991531B (en) | Dynamic construction method of refined three-dimensional hydrogeologic model | |
| CN109061737B (en) | Reservoir prediction method and device based on synthetic seismic record | |
| US20240230942A9 (en) | Method and System for Real-Time Calculating a Microseismic Focal Mechanism Based on Deep Learning | |
| Li et al. | Deep learning for quantitative hydraulic fracture profiling from fiber optic measurements | |
| CN114063169B (en) | Wave impedance inversion method, system, equipment and storage medium | |
| US11762112B2 (en) | Method of obtaining seismic while drilling signal | |
| CN114397701A (en) | Artificial intelligence four-dimensional seismic inversion method based on CNN method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |