CN114048674B - Shale fracturing discrimination method based on shale fracturing experimental data restoration - Google Patents
Shale fracturing discrimination method based on shale fracturing experimental data restoration Download PDFInfo
- Publication number
- CN114048674B CN114048674B CN202111260899.7A CN202111260899A CN114048674B CN 114048674 B CN114048674 B CN 114048674B CN 202111260899 A CN202111260899 A CN 202111260899A CN 114048674 B CN114048674 B CN 114048674B
- Authority
- CN
- China
- Prior art keywords
- shale
- shale fracturing
- fracturing
- data
- experimental data
- 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.)
- Active
Links
- 238000012850 discrimination method Methods 0.000 title claims abstract description 13
- 230000001364 causal effect Effects 0.000 claims abstract description 21
- 238000000034 method Methods 0.000 claims abstract description 19
- 230000004913 activation Effects 0.000 claims abstract description 17
- 238000002474 experimental method Methods 0.000 claims abstract description 14
- 238000003062 neural network model Methods 0.000 claims abstract description 12
- 238000004458 analytical method Methods 0.000 claims abstract description 7
- 238000011176 pooling Methods 0.000 claims abstract description 5
- 238000012549 training Methods 0.000 claims abstract description 5
- 239000013598 vector Substances 0.000 claims description 25
- 230000006870 function Effects 0.000 claims description 18
- 239000011159 matrix material Substances 0.000 claims description 15
- 230000006835 compression Effects 0.000 claims description 10
- 238000007906 compression Methods 0.000 claims description 10
- 230000010339 dilation Effects 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 7
- 238000004422 calculation algorithm Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 230000008901 benefit Effects 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000006837 decompression Effects 0.000 claims description 3
- 210000002364 input neuron Anatomy 0.000 claims description 3
- 239000000203 mixture Substances 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 230000004044 response Effects 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 2
- 239000012530 fluid Substances 0.000 description 10
- 239000007789 gas Substances 0.000 description 10
- 238000012544 monitoring process Methods 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 6
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 6
- 230000015572 biosynthetic process Effects 0.000 description 5
- 238000005755 formation reaction Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 230000035699 permeability Effects 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000000977 initiatory effect Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 239000003345 natural gas Substances 0.000 description 2
- 230000035515 penetration Effects 0.000 description 2
- 239000003208 petroleum Substances 0.000 description 2
- 230000008439 repair process Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 239000011435 rock Substances 0.000 description 2
- 235000015076 Shorea robusta Nutrition 0.000 description 1
- 244000166071 Shorea robusta Species 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000002591 computed tomography Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 125000001475 halogen functional group Chemical group 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000020477 pH reduction Effects 0.000 description 1
- 230000037361 pathway Effects 0.000 description 1
- 239000011148 porous material Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000005086 pumping Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000010008 shearing Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 230000000153 supplemental effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Investigating Strength Of Materials By Application Of Mechanical Stress (AREA)
Abstract
The invention relates to a shale fracturing discrimination method based on shale fracturing experimental data restoration, which comprises the following steps: incomplete waveform data obtained through shale fracturing experiments are obtained, and the obtained data are preprocessed to obtain a supplementary data set; adopting a combination of an expansion causality convolution, a gate activation function and a residual jump connection structure to construct a deep expansion causality convolution network without a pooling layer; training a deep expansion causal convolutional network by using a supplementary data set to obtain a neural network model for predicting shale fracturing missing waveforms; and outputting the obtained predicted missing waveform according to the neural network model to obtain shale fracturing complete waveform data, and judging the shale internal crack type by adopting a moment tensor analysis method. Compared with the prior art, the method and the device can effectively improve the accuracy of shale fracturing discrimination by reliably repairing shale fracturing experimental data.
Description
Technical Field
The invention relates to the technical field of shale fracturing, in particular to a shale fracturing discrimination method based on shale fracturing experimental data restoration.
Background
Shale gas is a clean energy source with methane as a main component, and is particularly important in exploitation and application of the shale gas in a global environment with energy conservation and emission reduction. The shale gas resources in China are quite rich, the resource quantity is in the front of the world, the economic value is huge, and the resource prospect is wide. The method has the advantages that the exploration and development of shale gas are greatly developed, the energy consumption structure taking coal as a main energy source in China is improved, the resource conservation, environment-friendly production mode and consumption mode are realized on the basis of meeting the social and economic development, the method has important significance in improving the living environment of people and improving the ecological civilization level, and meanwhile, chips are also increased for guaranteeing the national energy safety.
Shale has very low permeability and therefore a series of fracture grids naturally occurring inside shale become the most dominant channels for shale gas transport and storage. When natural gas is mined, shale is often subjected to fracturing to form cracks, and shale gas conveying channels are generated, so that the exploitation of the shale natural gas is realized, and therefore shale fracturing experiments have important significance for the exploitation of the shale gas.
Shale fracturing crack propagation is a complex problem related to a plurality of physical processes such as solid fracture, liquid flow, gas diffusion and the like, and provides new challenges and development opportunities for fracture mechanics experiments, theory and numerical simulation research. In order to optimize the shale fracturing process, it is important to accurately detect the location of the natural fracture. The anisotropic behavior of shale creates preferential pathways through shale organization, and in addition, the alignment of natural fractures can also induce anisotropic patterns of fluid flow. In 2001, de Pater et al, university of Delft, netherlands, have autonomously developed active acoustic monitoring techniques, but the techniques can only perform two-dimensional planar positioning, and there are errors in steering or torsional fracture positioning. The most widely applied monitoring technology is a passive acoustic monitoring technology at present, and in a small rock core plate fracturing experiment, a good monitoring effect is obtained, but no report on application to a large-size rock sample is seen yet. The acoustic wave monitoring hydraulic fracture initiation and propagation forms are researched by utilizing large object model experiments such as the middle petroleum exploration court division Wang Yonghui. A series of indoor experiments of shale reservoir crack extension are carried out in China university of Petroleum (Beijing) fracturing acidification laboratory and China petrochemical engineering institute, some important results are obtained, and the form of shale crack extension can be conveniently evaluated through test piece observation or high-energy CT scanning after the experiments. In the hydraulic fracturing operation process, the trend and the form of crack growth need to be monitored, the accuracy of a fracturing area is ensured, the fracturing effect is evaluated, and the fracturing scheme is optimized. The microseism technology can image and monitor a complex fracture network generated by hydraulic fracturing, and the principle is as follows: in the hydraulic fracturing process, formation stress and pore pressure are changed, the stability of weak layers such as natural cracks, layer planes and the like around the cracks is affected, shearing sliding occurs, tiny earthquakes similar to fault initiation are caused, but energy is much smaller, the released earthquake energy can be detected by a seismic wave detector of a monitoring well, and information about a seismic source can be obtained through data processing. When the shale gas well is subjected to hydraulic fracturing operation, a group of detectors are put into the adjacent well of the fracturing well, microseism events generated by fracturing are received and treated, and the distribution of the seismic sources in time and space is determined. The application of the microseism technology is not only embodied in imaging and monitoring, but also integrates information obtained by microseism with information such as geology, geophysics and logging data along with the deep interpretation of microseism information, and can be applied to numerical simulation to simulate a complex fracture network gas reservoir more accurately so as to optimize production prediction.
One of the difficulties in fracture design is the lack of clear insight into the nature of the fracture complexity created during fracturing, microseismic monitoring does not provide sufficient resolution to delineate precise hydraulic fracture faces. Microseismic events are mainly due to shear failure along the natural fracture or faults around the hydraulic fracture, and these event clouds form a "halo" around the hydraulic fracture. In conventional sandstone formations, the width of the observed event cloud is relatively narrow, while in unconventional reservoirs, a wider event cloud is typically observed. Microseismic cloud formation may be due to simple hydraulic fracture created by penetration of deep fluids into shale natural fractures, or the formation of complex hydraulic fracture networks. Although deep fluids are entirely likely to penetrate into the highly permeable, initially well-connected network of natural fractures, the effective permeability of many unconventional reservoirs is very low, and core observations indicate that most of the natural fractures in these shales have been mineralized. Thus, in low permeability shale, the penetration of fluids through the natural fracture network is limited. Murphy et al suggested that fluids may also infiltrate natural fractures due to shear causing natural fracture expansion, but this typically occurs under high stress anisotropy conditions with natural fractures oriented in the principal stress direction 30 to 60, for many shale reservoirs with loose construction environments and small level stress differences, although hydraulic fractures may follow the path of the natural fracture, extensive microseismic clouds strongly indicate that a complex network of fracture hydraulic fractures has been formed; in one field case of Barnett shale, fisher et al provide supporting evidence for complex hydraulic fracturing networks in which the fracturing fluid accidentally connects several adjacent wells that are not within the intended fracture plane, resulting in a yield decline; in analysis of another Barnett case by cispola et al, it was shown that the fracture length predicted by the planar fracture model was much greater than that shown by microseismic data, unless the fluid efficiency was assumed to be very low (less than 10%) in the simulation so that the planar fracture contained a large volume of injected fluid, this inefficiency being in contradiction to the very slow pressure drop typically observed when most shale formations are shut off after pumping. In contrast, a complex hydraulic fracture network may explain that at the same fracture network length, the volume of fluid stored in the fracture network is much greater, but still has higher fluid efficiency due to the small amount of leakage. However, in high-structural stress geological environments, shear fracture may be an important mechanism for the observed broad microseismic clouds and shear-induced potential permeability enhancement. Correctly constructed complex fracture mechanics models help to study fracture mechanisms and provide tools for optimizing fracture design and completion strategies.
The existing research is to obtain crack propagation information of shale samples through shale fracturing experiments, but due to the fact that components of shale are complex, strain gauges attached to the surface of the shale cannot accurately and comprehensively obtain detection signals, so that experimental data are missing, and finally shale fracturing judgment is inaccurate.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a shale fracturing discrimination method based on shale fracturing experimental data restoration, which improves the accuracy of shale fracturing discrimination by restoring the shale fracturing experimental data.
The aim of the invention can be achieved by the following technical scheme: shale fracturing discrimination method based on shale fracturing experimental data restoration comprises the following steps:
s1, acquiring incomplete waveform data obtained by a shale fracturing experiment, and preprocessing the acquired data to obtain a supplementary data set;
s2, constructing a deep expansion causal convolution network without a pooling layer by adopting a combination of expansion causal convolution, a gate activation function and a residual jump connection structure;
s3, training the deep expansion causal convolutional network by using the supplementary data set to obtain a neural network model for predicting shale fracturing missing waveforms;
and S4, outputting the obtained predicted missing waveform according to the neural network model to obtain shale fracturing complete waveform data, and judging the shale internal crack type by adopting a moment tensor analysis method.
Further, the step S1 specifically includes the following steps:
s11, acquiring incomplete waveform data obtained by shale fracturing experiments;
s12, compressing the acquired data by adopting a mu-law compression algorithm, then carrying out one-hot encoding, and then carrying out linear interpolation on the data to obtain a supplementary data set, wherein vectors corresponding to different classifications are obtained after one-hot encoding, the length of the vectors depends on the total quantity of the classifications, the value of the corresponding classification in the vectors is 1, and the rest result vectors are 0.
Further, the compression formula of the μ -law compression algorithm is:
the decompression formula is:
wherein μ takes the value of 100.
Further, the output layer of the deep dilation causal convolutional network constructed in step S2 employs a softmax activation function:
wherein y is i K is the total number of neurons of the output layer for the size before the input neurons are activated, and the output size represents the probability of taking each value;
the output of the softmax activation function corresponds to the vector obtained after the one-hot encoding.
Further, the loss function of the deep dilation causal convolution network constructed in step S2 is a cross entropy loss function:
H(p,q)＝-∑p log(q)
wherein, p is the original correctly classified corresponding one-hot coding vector, and q is the output vector after softmax activation.
Further, the step S3 specifically uses the first 1500 sequences in the supplemental dataset as inputs and the last 1500 sequences as outputs to train the deep dilation causal convolutional network.
Further, the step S4 specifically includes the following steps:
s41, outputting the obtained predicted missing waveform according to the neural network model to obtain shale fracturing complete waveform data;
s42, screening acoustic emission time based on shale fracturing complete waveform data, solving initial motion amplitude, and solving moment tensor matrix by utilizing the initial motion amplitude;
s43, carrying out eigenvalue decomposition on the matrix quantity matrix to obtain three corresponding eigenvalues, and solving the ratio values of the three modes;
s44, judging the crack type of shale fracturing by adopting an advantage judging method of M.Ohtsu and combining the ratio values of the three modes.
Further, in the step S42, the calculated relationship between the initial motion amplitude and the moment tensor matrix is:
wherein A is%x) is the initial amplitude, m is the moment tensor matrix, R is the source-sensor distance, (R) 1 r 2 r 3 ) The seismic source points to the direction cosine of the sensor, cs is the response amplitude of the sensor, and is obtained through lead breaking test, R e (t, r) is the reflectance, k e ＝1.96，d e Is t, r.
Further, the calculation relationship between the three feature values and the three modes in the step S43 is:
γ max /γ min ＝Z 1 +Z 2 +Z 3
γ int /γ max ＝0-0.5Z 2 +Z 3
γ min /γ max ＝-Z 1 -0.5Z 2 +Z 3
wherein, gamma min 、γ max 、γ int Three eigenvalues, Z, of the moment tensor matrix respectively 1 、Z 2 And Z 3 Corresponding to three mode scale values.
Further, the specific process of step S44 is as follows:
when Z is 1 At > 60%, shear cracks were determined;
when Z is 2 +Z 3 If more than 60%, the tensile crack is judged;
when 40% < Z 1 If less than 60%, the mixture was judged to be a mixed crack.
Compared with the prior art, the invention has the following advantages:
the invention combines the expansion causal convolution, the gate activation function and the residual jump connection to construct a deep expansion causal convolution network, and can reliably repair the missing shale fracturing experimental data by predicting the missing data, thereby solving the problems of low reliability of analysis results, difficult utilization of a large amount of data and the like caused by the missing experimental data, improving the utilization rate of the experimental data, and further guaranteeing the accuracy of discriminating the crack types in the subsequent shale test piece.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention;
FIG. 2 is a schematic diagram of a deep dilation causal convolutional network constructed in the examples.
Detailed Description
The invention will now be described in detail with reference to the drawings and specific examples.
Examples
As shown in fig. 1, a shale fracturing discrimination method based on shale fracturing experimental data restoration comprises the following steps:
s1, acquiring incomplete waveform data obtained by a shale fracturing experiment, and preprocessing the acquired data to obtain a supplementary data set;
s2, constructing a deep expansion causal convolution network without a pooling layer by adopting a combination of expansion causal convolution, a gate activation function and a residual jump connection structure;
s3, training the deep expansion causal convolutional network by using the supplementary data set to obtain a neural network model for predicting shale fracturing missing waveforms;
and S4, outputting the obtained predicted missing waveform according to the neural network model to obtain shale fracturing complete waveform data, and judging the shale internal crack type by adopting a moment tensor analysis method.
The embodiment applies the method, and the main process is as follows:
step 1: for incomplete waveform data obtained by shale fracturing experiments, firstly, compressing by adopting a mu-law compression algorithm, then performing one-hot coding on the incomplete waveform data, converting the incomplete waveform data into a classification problem, and then performing linear interpolation on experimental data to obtain a supplementary data set, wherein the compression formula of the mu-law compression algorithm is as follows:
mu has a value of 100;
the decompression formula is:
mu has a value of 100.
Step 2: constructing a deep expansion causal convolution network without a pooling layer (shown in fig. 2) by adopting a combination of expansion causal convolution, a gate activation function and a residual jump connection structure, and training the network by taking the first 1500 sequences and the last 1500 sequences of the supplementary data set in the step 1 as inputs and taking the last 1500 sequences as outputs to obtain a neural network model, wherein the output layer of the neural network is activated by adopting a softmax, and the softmax activation function expression is as follows:
wherein y is i Representing the size of the input neurons before activation, k representing the total number of neurons in the layer, and the output size representing the probability of taking each value;
the loss function is a cross entropy loss function commonly used for general multi-classification problems, and the expression is as follows:
H(p,q)＝-∑p log(q)
wherein, p represents the original correct classification of the corresponding one-hot coding vector, and q is the output vector after softmax activation.
According to the technical scheme, the compressed signal amplitude is uniformly divided into mu equal parts according to the value of mu-law after compression, and then one-hot coding is carried out on the mu equal parts, wherein the one-hot coding is one of the codes aiming at the classification problem in the current machine learning, and a vector is mainly used for representing the final classification result. The vector length depends on the total number of classes, the value of the corresponding class in the vector being 1, the remaining resulting vectors being 0, typically corresponding to the output of the softmax activation function.
In the present invention, it is first assumed that the magnitude of the voltage value at each time point does not exceed the known maximum voltage V after μ -law in the entire channel max And a minimum voltage value V min And will be interval V min ,V max ]Mu parts are divided on average, and one-hot encoding is carried out on each voltage value according to the mu parts. Each voltage value corresponds to a vector with the length mu, if the voltage value falls in the 5 th interval, the fifth element of the vector is 1, the other elements are 0, and the other conditions are the same. In this way, the regression problem is converted into a multiple classification problem. The output of the model is an integer multiple of mu, and for each mu-length vector, the vector is activated by a softmax probability activation function, so that the position with the highest probability in the obtained vector is the section where the voltage value corresponding to the output is located. The method can effectively enhance the stability of the model and achieve the purpose of accelerating convergence.
Step 3: predicting the missing waveform by using the trained neural network, so as to repair and obtain the complete waveforms of six channels;
step 4: the method comprises the steps of utilizing a complete waveform, judging the type of cracks in a shale test piece through moment tensor analysis, specifically, screening acoustic emission events, then solving a primary motion amplitude A (x), and then solving a moment tensor m through the primary motion amplitude A (x), wherein the formula is as follows:
wherein R is the distance between the seismic source and the sensor, (R) 1 r 2 r 3 ) The seismic source points to the direction cosine of the sensor, cs is the response amplitude of the sensor, and is obtained through lead breaking test, R e (t, r) is a reflection coefficient, and the solving formula is as follows:
wherein k is e ＝1.96，d e Is t, r.
After the m matrix is obtained, the m matrix is decomposed into three eigenvalues gamma of the m matrix min 、γ max 、γ int Finally, solving the proportion of the three modes according to the following formulaThe formula is as follows:
γ max /γ min ＝Z 1 +Z 2 +Z 3
γ int /γ max ＝0-0.5Z 2 +Z 3
γ min /γ max ＝-Z 1 -0.5Z 2 +Z 3
finally, judging the crack type by adopting an advantage judging method of M.Ohtsu:
(1) When Z is 1 At > 60%, shear cracks were determined;
(2) When Z is 2 +Z 3 If more than 60%, the tensile crack is judged;
(3) When 40% < Z 1 If less than 60%, the mixture was judged to be a mixed crack.
In summary, when the data is preprocessed, the invention forms the supplementary data set through data interpolation, thereby increasing the data density, effectively extracting the data characteristics, and obtaining a waveform prediction result by using the supplementary data which is closer to the original waveform with higher accuracy;
in the neural network model constructed by the invention, the adopted dilation-causal convolution is a convolution neural network which uses a filter on a region larger than the length of the convolutional neural network by skipping a specific step of input values, and increases the interval of elements required by kernel correspondence on the premise of keeping the kernel size unchanged, which is equivalent to the convolution neural network which uses a zero amplification original filter to obtain a larger filter, but the efficiency is obviously improved, and compared with a normal convolution neural network, the dilation convolution can effectively lead the network to run on a larger scale;
according to the invention, the constructed neural network is utilized to predict the missing waveform of the shale experimental data, so that the utilization rate of the experimental data can be effectively improved, the reliability of the judgment result of the shale fracturing experimental crack type is improved, and the problem of low available samples caused by missing of the experimental data is solved.
Claims (6)
1. A shale fracturing discrimination method based on shale fracturing experimental data restoration is characterized by comprising the following steps:
s1, acquiring incomplete waveform data obtained by a shale fracturing experiment, and preprocessing the acquired data to obtain a supplementary data set;
s2, constructing a deep expansion causal convolution network without a pooling layer by adopting a combination of expansion causal convolution, a gate activation function and a residual jump connection structure;
s3, training the deep expansion causal convolutional network by using the supplementary data set to obtain a neural network model for predicting shale fracturing missing waveforms;
s4, outputting the obtained predicted missing waveform according to the neural network model to obtain shale fracturing complete waveform data, and judging the shale internal crack type by adopting a moment tensor analysis method;
the step S1 specifically comprises the following steps:
s11, acquiring incomplete waveform data obtained by shale fracturing experiments;
s12, compressing the acquired data by adopting a mu-law compression algorithm, then carrying out one-hot encoding, and then carrying out linear interpolation on the data to obtain a supplementary data set, wherein vectors corresponding to different classifications are obtained after one-hot encoding, the length of the vectors depends on the total quantity of the classifications, the value of the corresponding classification in the vectors is 1, and the rest result vectors are 0;
the compression formula of the mu-law compression algorithm is as follows:
the decompression formula is:
wherein μ has a value of 100;
the output layer of the deep dilation causal convolution network constructed in the step S2 adopts a softmax activation function:
wherein y is i K is the total number of neurons of the output layer for the size before the input neurons are activated, and the output size represents the probability of taking each value;
the output of the softmax activation function corresponds to a vector obtained after one-hot encoding;
the loss function of the deep dilation causal convolution network constructed in the step S2 is a cross entropy loss function:
H(p,q)＝-∑plog(q)
wherein, p is the original correctly classified corresponding one-hot coding vector, and q is the output vector after softmax activation.
2. The shale fracturing discrimination method based on shale fracturing experimental data restoration according to claim 1, wherein the step S3 specifically uses the first 1500 sequences in the supplementary data set as input and the last 1500 sequences as output to train the deep expansion causal convolutional network.
3. The shale fracturing discrimination method based on shale fracturing experimental data restoration according to claim 1, wherein the step S4 specifically comprises the following steps:
s41, outputting the obtained predicted missing waveform according to the neural network model to obtain shale fracturing complete waveform data;
s42, screening acoustic emission time based on shale fracturing complete waveform data, solving initial motion amplitude, and solving moment tensor matrix by utilizing the initial motion amplitude;
s43, carrying out eigenvalue decomposition on the matrix quantity matrix to obtain three corresponding eigenvalues, and solving the ratio values of the three modes;
s44, judging the crack type of shale fracturing by adopting an advantage judging method of M.Ohtsu and combining the ratio values of the three modes.
4. The shale fracturing discrimination method based on shale fracturing experimental data restoration according to claim 3, wherein the calculation relation between the initial motion amplitude and the moment tensor matrix in the step S42 is:
wherein A (x) is the initial motion amplitude, m is the moment tensor matrix, R is the distance between the source and the sensor, (R) 1 r 2 r 3 ) The seismic source points to the direction cosine of the sensor, cs is the response amplitude of the sensor, and is obtained through lead breaking test, R e (t, r) is the reflectance, k e ＝1.96，d e Is t, r.
5. The shale fracturing discrimination method based on shale fracturing experimental data restoration according to claim 4, wherein the calculation relation between the three characteristic values and the three modes in the step S43 is as follows:
γ max /γ min ＝Z 1 +Z 2 +Z 3
γ int /γ max ＝0-0.5Z 2 +Z 3
γ min /γ max ＝-Z 1 -0.5Z 2 +Z 3
wherein, gamma min 、γ max 、γ int Three eigenvalues, Z, of the moment tensor matrix respectively 1 、Z 2 And Z 3 Corresponding to three mode scale values.
6. The shale fracturing discrimination method based on shale fracturing experimental data restoration according to claim 5, wherein the specific process of step S44 is as follows:
when Z is 1 At > 60%, shear cracks were determined;
when Z is 2 +Z 3 If more than 60%, the tensile crack is judged;
when 40% < Z 1 If less than 60%, the mixture was judged to be a mixed crack.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111260899.7A CN114048674B (en) | 2021-10-28 | 2021-10-28 | Shale fracturing discrimination method based on shale fracturing experimental data restoration |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111260899.7A CN114048674B (en) | 2021-10-28 | 2021-10-28 | Shale fracturing discrimination method based on shale fracturing experimental data restoration |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114048674A CN114048674A (en) | 2022-02-15 |
| CN114048674B true CN114048674B (en) | 2024-03-01 |
Family
ID=80206742
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111260899.7A Active CN114048674B (en) | 2021-10-28 | 2021-10-28 | Shale fracturing discrimination method based on shale fracturing experimental data restoration |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114048674B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106988739A (en) * | 2017-05-19 | 2017-07-28 | 中国石油集团川庆钻探工程有限公司 | Shale reservoir fracturing fracture is recognized and explanation evaluating method |
| CN110175723A (en) * | 2019-05-31 | 2019-08-27 | 西南石油大学 | A kind of shale gas well refracturing fracture initiation direction and extension path prediction technique |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2013230933B2 (en) * | 2012-03-06 | 2015-12-10 | Ion Geophysical Corporation | Model predicting fracturing of shale |
| GB2503903B (en) * | 2012-07-11 | 2015-08-26 | Schlumberger Holdings | Fracture monitoring and characterisation |
-
2021
- 2021-10-28 CN CN202111260899.7A patent/CN114048674B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106988739A (en) * | 2017-05-19 | 2017-07-28 | 中国石油集团川庆钻探工程有限公司 | Shale reservoir fracturing fracture is recognized and explanation evaluating method |
| CN110175723A (en) * | 2019-05-31 | 2019-08-27 | 西南石油大学 | A kind of shale gas well refracturing fracture initiation direction and extension path prediction technique |
Non-Patent Citations (2)
| Title |
|---|
| Moment tensor analysis of transversely isotropic shale based on the discrete element method;Chong Zhaohui;Li Xuehua;Hou Peng;Chen Xiangyu;Wu Yuechao;;International Journal of Mining Science and Technology(第03期);507-514 * |
| 各向异性HTI介质中的剪张源矩张量特征研究;唐杰;张文征;温雷;李聪;;石油物探(第05期);652-657 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114048674A (en) | 2022-02-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Keranen et al. | Induced seismicity | |
| CN108009705B (en) | Shale reservoir compressibility evaluation method based on support vector machine technology | |
| Wang et al. | Quantitative prediction of oil and gas prospects of the Sinian-Lower Paleozoic in the Sichuan Basin in central China | |
| Geng et al. | Predicting seismic-based risk of lost circulation using machine learning | |
| Tariq et al. | A rigorous data-driven approach to predict Poisson's ratio of carbonate rocks using a functional network | |
| Li et al. | Prediction of CO2 leakage risk for wells in carbon sequestration fields with an optimal artificial neural network | |
| Tian et al. | Quantitative evaluation of key geological controls on regional Eagle Ford shale production using spatial statistics | |
| CN105697002A (en) | Method for recognizing coal measure strata lithology | |
| Rahman et al. | Probabilistic analysis of Vette fault stability in potential CO2 storage site Smeaheia, offshore Norway | |
| Zhai et al. | Analysis of acoustic emission events induced during stress unloading of a hydraulic fractured Longmaxi shale sample | |
| Baró et al. | Fracking and labquakes | |
| Porter et al. | Addressing the risks of induced seismicity in subsurface energy operations | |
| Zhai et al. | Effects of bedding planes on progressive failure of shales under uniaxial compression: Insights from acoustic emission characteristics | |
| Hou et al. | Data-driven optimization of brittleness index for hydraulic fracturing | |
| Yu et al. | Analysis of factors of productivity of tight conglomerate reservoirs based on random forest algorithm | |
| Igonin et al. | Seismic anisotropy reveals stress changes around a fault as it is activated by hydraulic fracturing | |
| Du et al. | Prediction equations for the effective number of cycles of ground motions for shallow crustal earthquakes | |
| WO2022011015A1 (en) | Optimization of discrete fracture network (dfn) using streamlines and machine learning | |
| CN114048674B (en) | Shale fracturing discrimination method based on shale fracturing experimental data restoration | |
| Omidvar Eshkalak | Synthetic geomechanical logs and distributions for Marcellus Shale | |
| Grasso et al. | Did wastewater disposal drive the longest seismic swarm triggered by fluid manipulations? Lacq, France, 1969–2016 | |
| Healy et al. | De-risking the energy transition by quantifying the uncertainties in fault stability | |
| Chao et al. | A prediction of borehole stability while drilling preliminary prospecting wells based on seismic impedance | |
| Tiancheng et al. | Source mechanism studies of acoustic emission in large-scale hydraulic fracturing experiment | |
| Liu et al. | Machine learning assisted recovery of subsurface energy: a review |
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 |