CN111474585B - Gas layer identification method based on deep learning and rock mechanical parameters - Google Patents
Gas layer identification method based on deep learning and rock mechanical parameters Download PDFInfo
- Publication number
- CN111474585B CN111474585B CN202010183702.3A CN202010183702A CN111474585B CN 111474585 B CN111474585 B CN 111474585B CN 202010183702 A CN202010183702 A CN 202010183702A CN 111474585 B CN111474585 B CN 111474585B
- Authority
- CN
- China
- Prior art keywords
- bulk modulus
- gas layer
- deep learning
- rock
- rock mechanical
- 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
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
- G01V1/40—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
- G01V1/50—Analysing data
Landscapes
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- Remote Sensing (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a gas layer identification method based on deep learning and rock mechanical parameters, which comprises the steps of calculating the rock mechanical parameters and measuring 4 logging curve values, and then respectively carrying out normalization treatment on the calculated result of rock mechanical parameter recombination and the 4 logging curve values; training and learning the normalized data by using a deep learning neural network, and using a learning result for calculating new well data; the gas layer identification method well solves the difficulty of gas layer identification in carbonate strata and sand shale strata, and has high identification accuracy and simple method; the actual gas content of the stratum is divided in stages, and prediction is carried out according to the training result of the neural network, so that the method has higher recognition precision compared with a comprehensive gas content index method, and the gas content can be calculated semi-quantitatively.
Description
Technical Field
The invention belongs to the technical field of petroleum and natural gas exploration and development, and particularly relates to a gas layer identification method based on deep learning and rock mechanical parameters.
Background
Because the carbonate stratum has large pore type change, the pore gap composite reservoir is multiple, the pore heterogeneity is strong, the response error of the three-porosity curve to the rock porosity is large, and the response of the resistivity to the reservoir fluid property is weak, the reservoir fluid property is difficult to accurately judge based on conventional logging data.
The monopole and dipole modes of the MPAL, XAMC and other array acoustic logging instruments can be used for measuring characteristic parameters such as longitudinal and transverse wave speeds of the stratum in both hard stratum and soft stratum, when the stratum contains gas, the longitudinal wave speed is obviously reduced, the longitudinal wave speed is basically unaffected, the longitudinal and transverse wave speed ratio is reduced at the moment, therefore, a gas layer can be judged through the change of the parameters, most common methods are intersection diagrams, difference ratio values and the like, a plate is required to be manufactured, the use is inconvenient, the reaction sensitivity of each parameter to the gas layer is different, and the uncertainty of gas layer identification is increased.
Disclosure of Invention
The invention aims to provide a gas layer identification method based on deep learning and rock mechanical parameters, so as to solve the problems in the prior art.
In order to achieve the above purpose, the invention adopts the following technical scheme:
a gas layer identification method based on deep learning and rock mechanical parameters comprises 4 logging curves and rock mechanical parameters, and then respectively carrying out normalization processing on the results of rock mechanical parameter recombination calculation and 4 logging curve values; training and learning the normalized data by using a deep learning neural network algorithm, and using a learning result for calculating new well data; the rock mechanical parameters comprise bulk modulus XKB, compression coefficient CB, lame constant LAME, rock skeleton bulk modulus XKMX, water bulk modulus XKW, compression coefficient WCB and fluid compression coefficient FCB; the neural network training learning includes data tagging, preprocessing of data, and computation of a sample training dataset and a test dataset.
Further, the calculation formulas of the bulk modulus XKB and the compression coefficient CB are as follows:
YME＝2·SM·(1+POIS) (2)
wherein SM is shear modulus, A is constant, DEN is density, DTS is transverse wave time difference, YME is Young's modulus, POIS is Poisson's ratio;
further, the calculation formula of the LAME constant LAME is as follows:
wherein YME is Young's modulus, POIS is Poisson's ratio;
further, the calculation formula of the rock skeleton bulk modulus XKMX is as follows:
wherein POR is porosity, VSH is clay content, LIME is limestone content, DOLO is dolomite content, XKMA is sandstone skeleton bulk modulus, XKSH is mudstone bulk modulus, XKLM is limestone skeleton bulk modulus, and XKDO is dolomite skeleton bulk modulus;
further, the bulk modulus XKW and the compression coefficient WCB of water are calculated as follows:
wherein A is a constant, DENW is the density of water, and DTCW is the acoustic time difference of water;
further, the fluid compression coefficient FCB is calculated as follows:
wherein CB is a compression coefficient, POR is a porosity, VSH is a shale content, LIME is a limestone content, DOLO is a dolomite content, XKMA is a sandstone skeleton bulk modulus, XKSH is a mudstone bulk modulus, XKLM is a limestone skeleton bulk modulus, and XKDO is a dolomite skeleton bulk modulus.
Further, the data tagging processing method comprises the following steps: the gas layer is classified into 8 types according to the gas yield of the gas layer, labeling treatment is carried out, and a total of 8 different label values of 0-7 are respectively given.
Further, the preprocessing of the data comprises the recombination calculation and normalization of parameters;
the rock mechanical parameters are recombined into 3 attribute values, which are respectively defined as P1, P2 and P3, and the calculation formula is as follows:
the direct assignment of P1, P2 and P3 with a calculation result less than 0 is 0;
uniformly defining the attribute values P1, P2, P3 and 4 log curve values as 7 curve data of C1, C2, C3, C4, C5, C6 and C7, carrying out normalization processing on the C1-C7, wherein a normalization calculation formula is as follows:
wherein C is I,N Normalized to curve I, C I,i For values of different depths of the I-th curve, C I,min Is the minimum value of the I-th curve, C I,max Is the maximum of the I-th curve.
Furthermore, the input layer is provided with three attribute values P1, P2 and P3 and 4 logging curves;
the three attribute values are respectively: rock estimates the sum of the difference between bulk modulus and bulk modulus, the difference between the fluid compression coefficient and the water compression coefficient, and the lame constant and bulk modulus; the 4 log curves are respectively: gamma (GR), deep Resistivity (RD), shallow Resistivity (RS), sonic jet lag (AC).
Further, an LSTM neural network of a cyclic neural network is adopted, the network model is divided into 3 layers, namely an LSTM layer, a Dropout layer and a full-connection layer, an activation function is a sigmoid function, and a loss function is a cross entropy loss function.
The beneficial effects of the invention are as follows:
1. according to the gas layer identification method provided by the invention, the result of the recombination calculation of the logging curve and the rock mechanical parameter is normalized, and is input into the neural network for training, so that the applicable gas layer identification model is obtained. Compared with a linear comprehensive gas-containing index method, the nonlinear neural network has higher identification precision, and can realize semi-quantitative calculation of the gas content;
2. the gas layer identification method provided by the invention is only required to re-extract samples from the data of the new block, label the data and input a neural network to re-calculate a new model when being applied to different blocks, and a calculation method of specific parameters is not required to be considered, and a calculation formula is required to be re-calibrated by a comprehensive gas-containing index method or other methods, so that the method has stronger expandability and is simpler.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention. In the drawings:
FIG. 1 is a flow chart of a gas layer identification method of the present invention;
FIG. 2 is a diagram of the neural network architecture of the gas layer identification method of the present invention.
Detailed Description
The invention will be described in detail below with reference to the drawings in connection with embodiments. It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other.
The following detailed description is exemplary and is intended to provide further details of the invention. Unless defined otherwise, all technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments in accordance with the invention.
As shown in FIG. 1, a gas layer identification method based on deep learning and rock mechanical parameters comprises 4 logging curves and rock mechanical parameters, and then respectively carrying out normalization processing on the results of rock mechanical parameter recombination calculation and 4 logging curve values; training and learning the normalized data by using a deep learning neural network algorithm, and using a learning result for calculating new well data; the rock mechanical parameters comprise bulk modulus XKB, compression coefficient CB, lame constant LAME, rock skeleton bulk modulus XKMX, water bulk modulus XKW, compression coefficient WCB and fluid compression coefficient FCB; the neural network training learning includes data tagging, preprocessing of data, and computation of a sample training dataset and a test dataset.
The rock mechanical parameter calculation method comprises the following steps:
1) The calculation formulas of the bulk modulus XKB and the compression coefficient CB are as follows:
YME＝2·SM·(1+POIS) (2)
where SM is shear modulus, A is constant, DEN is density, DTS is shear wave time difference, YME is Young's modulus, POIS is Poisson's ratio, DTR is shear wave time difference ratio and shear wave time difference ratio of a fully hydrated formation.
2) The calculation formula of the LAME constant LAME is as follows:
wherein YME is Young's modulus and POIS is Poisson's ratio.
3) The calculation formula of the rock skeleton bulk modulus XKMX is as follows:
wherein POR is porosity, VSH is clay content, LIME is limestone content, DOLO is dolomite content, XKMA is sandstone skeleton bulk modulus, XKSH is mudstone bulk modulus, XKLM is limestone skeleton bulk modulus, and XKDO is dolomite skeleton bulk modulus;
the method is a method for calculating the skeleton bulk modulus of the carbonate stratum, and the skeleton bulk modulus of the sand shale stratum is obtained after LIME term and DOLO term in the formula are removed.
4) The volumetric modulus XKW of water and the compression coefficient WCB are calculated as follows:
where A is a constant, DENW is the density of water, and DTCW is the acoustic time difference of water.
5) The fluid compression coefficient FCB is calculated as follows:
wherein CB is a compression coefficient, POR is a porosity, VSH is a shale content, LIME is a limestone content, DOLO is a dolomite content, XKMA is a sandstone skeleton bulk modulus, XKSH is a mudstone bulk modulus, XKLM is a limestone skeleton bulk modulus, and XKDO is a dolomite skeleton bulk modulus.
After the parameters are calculated, in order to judge the gas-containing stratum more intuitively and determine the gas content, a deep learning neural network algorithm is utilized to train and learn the gas production result of the actual gas testing stratum, and then the learning result is used for calculating new well data.
The deep learning neural network algorithm is utilized to carry out data tagging, data preprocessing and sample training data set and test data set arrangement.
1) Data tagging
Although the deep learning algorithm can directly perform regression calculation, the variation range of the gas test result is very large, so that quantitative regression calculation is not directly performed in the method, the gas production is segmented first, different label values are given, the total label values are divided into the following 8 types, and the labels are set as follows:
| sequence number | Label (Label) | Gas production (m) 3 /m) |
| 1 | 0 | <0 (non-air layer) |
| 2 | 1 | 0-100 |
| 3 | 2 | 100-500 |
| 4 | 3 | 500-1000 |
| 5 | 4 | 1000-5000 |
| 6 | 5 | 5000-10000 |
| 7 | 6 | 10000-50000 |
| 8 | 7 | >50000 |
Note that: the gas production in the table is the gas production per meter of formation, and the thickness of the reservoir refers to the reservoir thickness defined by the interpreter.
And labeling the well data with known test gas results in the same block according to the table.
2) Preprocessing of data
Since the results of the above-described calculation of the respective parameters in pairs have a good indication of the gas content, a pre-treatment is required before the results are input to the neural network for training, including a calculation of the recombination of the parameters and normalization.
The above calculated parameters are recombined into 3 attribute values, defined as P1-P3, respectively. Wherein:
for the above 3 attribute values, the direct assignment with the calculation result less than 0 is 0.
Uniformly defining the attribute values P1, P2, P3 and 4 log curve values as 7 curve data of C1, C2, C3, C4, C5, C6 and C7, carrying out normalization processing on the C1-C7, wherein a normalization calculation formula is as follows:
wherein C is I,N Normalized to curve I, C I,i For values of different depths of the I-th curve, C I,min Is the minimum value of the I-th curve, C I,max Is the maximum of the I-th curve.
3) Neural network construction
As shown in fig. 2, the present invention employs an LSTM neural network of a recurrent neural network. The network model is divided into 3 layers, namely an LSTM layer, a Dropout layer and a full connection layer, the activation function is a sigmoid function, and the loss function is a cross entropy loss function. Wherein the input layer is 3 attribute values of the above-mentioned P1-P3 and 4 other logging curves, and the 3 attribute values are respectively: rock estimation bulk modulus to bulk modulus difference, fluid compression coefficient to water compression coefficient difference, formation compression coefficient to poisson ratio difference, rameow constant to bulk modulus sum, longitudinal and transverse wave velocity ratio to fully hydrated formation longitudinal and transverse wave velocity ratio difference; the 4 log curves are respectively: gamma (GR), deep Resistivity (RD), shallow Resistivity (RS), sonic jet lag (AC).
The invention aims to solve the problem of multiple classification, the output result of the neural network is a vector, the length of the vector is equal to the number 8 of the classifications of the air content, and the classification corresponding to the maximum value is taken as the air content result.
For the purpose, technical solutions and advantages of the present invention to be more apparent, the following technical solutions and flows of the present invention are fully described in conjunction with the processing flow of actual well data of one block:
in the invention, a block in the Suback region of the Changqing oilfield is selected as a research block, and the whole calculation process is as follows:
(1) The feature value extraction function is utilized in the LEAD4.0 software to extract the data of 26 gas test horizons of the 10 gas wells in the block, and the data of 505 depth points are extracted in total. The extracted part of the original data is as follows:
| GR | RD | RS | AC | XKB | XKMX | FCB | LAME | gas yield of rice |
| 21.39 | 345.13 | 401.97 | 166.79 | 8.29 | 9.83 | 4.7 | 5.42 | 1168.6 |
| 24.24 | 677.46 | 779.64 | 168.07 | 8.51 | 9.76 | 3.84 | 5.6 | 1168.6 |
| 18.87 | 172.19 | 204.15 | 165.5 | 8.04 | 9.83 | 6.84 | 5.26 | 1168.6 |
| …… | ||||||||
| 16.28 | 91.95 | 1.6 | 157.93 | 6.58 | 8.32 | 28.51 | 3.04 | 17860.0 |
| 13.64 | 233.48 | 1.64 | 159.14 | 6.64 | 10.03 | 23.18 | 3.26 | 17860.0 |
| 13.19 | 390.11 | 1.69 | 161.34 | 7.21 | 10.15 | 11.05 | 3.93 | 17860.0 |
| 19.18 | 232.04 | 1.68 | 172.02 | 7.38 | 10.27 | 5.68 | 4.15 | 11162.5 |
| 19.18 | 220.43 | 1.68 | 173.7 | 7.38 | 10.27 | 5.68 | 4.15 | 11162.5 |
| 13.23 | 194.14 | 1.68 | 174.96 | 6.92 | 10.28 | 6.24 | 3.78 | 11162.5 |
| 97.42 | 215.45 | 1.7 | 176.54 | 6.02 | 10.06 | 8.3 | 3.5 | 9348.2 |
| 60.61 | 170.21 | 1.79 | 167.1 | 7.48 | 10.18 | 6.42 | 4.83 | 9348.2 |
| 17.1 | 120.88 | 1.79 | 166.64 | 7.63 | 10.24 | 6.53 | 4.91 | 9348.2 |
| 16.77 | 625.24 | 477.74 | 162.24 | 0.33 | 0.09 | 11.03 | 11.03 | 1 |
| 19.42 | 499.5 | 383.25 | 166.11 | 0.32 | 0.09 | 10.96 | 10.96 | 1 |
| 21.72 | 413.39 | 315.09 | 168.12 | 0.31 | 0.1 | 10.41 | 10.41 | 1 |
(2) The attribute value is calculated and labeled for the data, and the result is as follows:
(3) Normalizing the data by using a formula (10);
(4) After data normalization, randomly selecting 95% of data as a training set and 5% of data as a test set;
(5) Inputting training set data and test set data into a neural network for calculation, wherein the result of calculation aiming at the test set is as follows, the input curve part lists normalized results, and the prediction accuracy reaches 92%:
| GR | RD | RS | AC | XKMX-XKB | FCB-WCB | LAME+XKB | actual classification | Prediction classification |
| 0.16 | 0.841 | 0.015 | 0.79 | 0.06 | 0.13 | 0.51 | 0 | 0 |
| 0.05 | 0.010 | 0.016 | 0.82 | 0.15 | 0.27 | 0.41 | 1 | 1 |
| 0.08 | 0.884 | 0.006 | 0.88 | 0.19 | 0.28 | 0.16 | 5 | 5 |
| 0.10 | 0.970 | 0.862 | 0.79 | 0.00 | 0.10 | 0.70 | 0 | 0 |
| 0.07 | 0.854 | 0.007 | 0.83 | 0.24 | 0.47 | 0.22 | 5 | 5 |
| 0.02 | 0.848 | 0.012 | 0.87 | 0.14 | 0.18 | 0.30 | 5 | 5 |
| 0.13 | 0.870 | 0.768 | 0.82 | 0.11 | 0.21 | 0.47 | 3 | 3 |
| 0.07 | 0.882 | 0.770 | 0.80 | 0.00 | 0.09 | 0.67 | 0 | 3 |
| 0.25 | 0.738 | 0.385 | 0.89 | 0.24 | 0.53 | 0.17 | 1 | 3 |
| 0.01 | 0.011 | 0.018 | 0.80 | 0.12 | 0.24 | 0.50 | 1 | 0 |
| 0.03 | 0.009 | 0.019 | 0.82 | 0.40 | 0.75 | 0.39 | 0 | 0 |
| 0.05 | 0.859 | 0.005 | 0.83 | 0.32 | 0.76 | 0.15 | 5 | 5 |
| 0.04 | 0.726 | 0.467 | 0.80 | 0.09 | 0.23 | 0.56 | 0 | 5 |
| 0.06 | 0.010 | 0.016 | 0.80 | 0.61 | 0.86 | 0.35 | 0 | 0 |
| 0.06 | 0.012 | 0.018 | 0.79 | 0.30 | 0.45 | 0.47 | 0 | 0 |
| 0.10 | 0.780 | 0.016 | 0.83 | 0.10 | 0.16 | 0.45 | 1 | 1 |
| 0.30 | 0.737 | 0.380 | 0.99 | 0.28 | 0.79 | 0.00 | 3 | 3 |
| 0.23 | 0.012 | 0.019 | 0.80 | 0.15 | 0.29 | 0.45 | 1 | 1 |
| 0.05 | 0.010 | 0.016 | 0.80 | 0.36 | 0.49 | 0.44 | 0 | 0 |
| 0.04 | 0.008 | 0.009 | 0.81 | 0.57 | 0.87 | 0.21 | 0 | 0 |
| 0.06 | 0.855 | 0.013 | 0.79 | 0.19 | 0.30 | 0.39 | 5 | 4 |
| 0.09 | 0.927 | 0.868 | 0.81 | 0.01 | 0.11 | 0.63 | 3 | 3 |
| 0.05 | 0.010 | 0.010 | 0.84 | 0.41 | 0.84 | 0.23 | 0 | 0 |
| 0.01 | 0.810 | 0.012 | 0.86 | 0.17 | 0.29 | 0.28 | 5 | 5 |
| 0.10 | 0.917 | 0.846 | 0.81 | 0.00 | 0.10 | 0.64 | 3 | 3 |
It will be appreciated by those skilled in the art that the present invention can be carried out in other embodiments without departing from the spirit or essential characteristics thereof. Accordingly, the above disclosed embodiments are illustrative in all respects, and not exclusive. All changes that come within the scope of the invention or equivalents thereto are intended to be embraced therein.
Claims (9)
1. A gas layer identification method based on deep learning and rock mechanical parameters is characterized by comprising the steps of collecting 4 logging curves and rock mechanical parameters, and then respectively carrying out normalization processing on the results of rock mechanical parameter recombination calculation and 4 logging curve values; training and learning the normalized data by using a deep learning neural network algorithm, and using a learning result for gas layer recognition; the rock mechanical parameters comprise bulk modulus XKB, compression coefficient CB, lame constant LAME, rock skeleton bulk modulus XKMX, water bulk modulus XKW, compression coefficient WCB and fluid compression coefficient FCB; the neural network training learning comprises data tagging, preprocessing of data and calculation of a sample training data set and a test data set; the 4 log curves are respectively: gamma GR, deep resistivity RD, shallow resistivity RS, acoustic moveout AC;
the volumetric modulus XKW of water and the compression coefficient WCB are calculated as follows:
where A is a constant, DENW is the density of water, and DTCW is the acoustic time difference of water.
2. The method for identifying the gas layer based on the deep learning and rock mechanical parameters according to claim 1, wherein the calculation formulas of the bulk modulus XKB and the compression coefficient CB are as follows:
YME＝2·SM·(1+POIS) (2)
where SM is shear modulus, A is constant, DEN is density, DTS is transverse wave time difference, YME is Young's modulus, POIS is Poisson's ratio.
3. The method for identifying the gas layer based on deep learning and rock mechanics parameters according to claim 1, wherein the calculation formula of the LAME constant LAME is as follows:
wherein YME is Young's modulus and POIS is Poisson's ratio.
4. The gas layer identification method based on deep learning and rock mechanical parameters according to claim 1, wherein the calculation formula of the rock skeleton bulk modulus XKMX is as follows:
wherein POR is porosity, VSH is clay content, LIME is limestone content, DOLO is dolomite content, XKMA is sandstone skeleton bulk modulus, XKSH is mudstone bulk modulus, XKLM is limestone skeleton bulk modulus, and XKDO is dolomite skeleton bulk modulus.
5. The gas layer identification method based on deep learning and rock mechanics parameters according to claim 1, wherein the calculation formula of the fluid compression coefficient FCB is as follows:
wherein CB is a compression coefficient, POR is a porosity, VSH is a shale content, LIME is a limestone content, DOLO is a dolomite content, XKMA is a sandstone skeleton bulk modulus, XKSH is a mudstone bulk modulus, XKLM is a limestone skeleton bulk modulus, and XKDO is a dolomite skeleton bulk modulus.
6. The gas layer identification method based on deep learning and rock mechanics parameters according to claim 1, wherein the data tagging processing method is as follows: the gas layer is classified into 8 types according to the gas yield of the gas layer, labeling treatment is carried out, and a total of 8 different label values of 0-7 are respectively given.
7. The method for identifying the gas layer based on the deep learning and rock mechanical parameters according to claim 1, wherein the preprocessing of the data comprises the recombination calculation and normalization of the parameters;
the rock mechanical parameters are recombined into 3 attribute values, which are respectively defined as P1, P2 and P3, and the calculation formula is as follows:
the direct assignment of P1, P2 and P3 with a calculation result less than 0 is 0;
uniformly defining the attribute values P1, P2, P3 and 4 log curve values as 7 curve data of C1, C2, C3, C4, C5, C6 and C7, carrying out normalization processing on the C1-C7, wherein a normalization calculation formula is as follows:
wherein C is I,N Normalized to curve I, C I,i For values of different depths of the I-th curve, C I,min Is the minimum value of the I-th curve, C I,max Is the maximum of the I-th curve.
8. The method for identifying the gas layer based on the deep learning and rock mechanical parameters according to claim 7, wherein the input layers are three attribute values P1, P2 and P3 and 4 logging curves;
the three attribute values are respectively: the difference between the bulk modulus and the bulk modulus of the rock framework, the difference between the fluid compression coefficient and the water compression coefficient, and the sum of the lame constant and the bulk modulus.
9. The method for identifying the gas layer based on the deep learning and rock mechanical parameters according to claim 1, wherein an LSTM neural network of a cyclic neural network is adopted, the network model is divided into 3 layers, namely an LSTM layer, a Dropout layer and a full-connection layer, an activation function is a sigmoid function, and a loss function is a cross entropy loss function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010183702.3A CN111474585B (en) | 2020-03-16 | 2020-03-16 | Gas layer identification method based on deep learning and rock mechanical parameters |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010183702.3A CN111474585B (en) | 2020-03-16 | 2020-03-16 | Gas layer identification method based on deep learning and rock mechanical parameters |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111474585A CN111474585A (en) | 2020-07-31 |
| CN111474585B true CN111474585B (en) | 2023-04-25 |
Family
ID=71747474
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010183702.3A Active CN111474585B (en) | 2020-03-16 | 2020-03-16 | Gas layer identification method based on deep learning and rock mechanical parameters |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111474585B (en) |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5251286A (en) * | 1992-03-16 | 1993-10-05 | Texaco, Inc. | Method for estimating formation permeability from wireline logs using neural networks |
| CN108596780B (en) * | 2018-04-28 | 2019-05-21 | 中国石油大学(华东) | A kind of Reservoir type division methods and system based on multiple information |
| CN110133724B (en) * | 2019-05-06 | 2021-01-29 | 中国石油天然气集团有限公司 | Method for identifying gas layer by applying array acoustic logging data |
-
2020
- 2020-03-16 CN CN202010183702.3A patent/CN111474585B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111474585A (en) | 2020-07-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103590827B (en) | Based on the compact clastic rock natural gas well PRODUCTION FORECASTING METHODS of Reservoir Classification | |
| CN111767674B (en) | Logging lithology recognition method based on active domain adaptation | |
| CN109214026A (en) | A kind of shale gas horizontal well initial productivity prediction technique | |
| CN107526106A (en) | Quick seismic waveform sorting technique based on semi-supervised algorithm | |
| CN112213797A (en) | Method for predicting reservoir productivity by using logging-phase combined post-stack seismic attributes | |
| CN104047598A (en) | Method for predicating productivity of nonhomogeneity ancient karst carbonate reservoir | |
| Zhu et al. | Rapid identification of high-quality marine shale gas reservoirs based on the oversampling method and random forest algorithm | |
| CN103630939A (en) | Air layer identification and evaluation method | |
| CN104570109B (en) | A kind of method of reservoir hydrocarbons prediction | |
| CN112796738A (en) | Stratum permeability calculation method combining array acoustic logging and conventional logging | |
| CN113033648A (en) | Method for realizing logging interpretation by using machine learning algorithm | |
| Guoqiang et al. | Construction of well logging knowledge graph and intelligent identification method of hydrocarbon-bearing formation | |
| CN111474585B (en) | Gas layer identification method based on deep learning and rock mechanical parameters | |
| CN104880737A (en) | Multivariate Logistic method using logging information to identify type of underground fluid | |
| CN111155980A (en) | Water flow dominant channel identification method and device | |
| CN116413831A (en) | Multi-well synthetic seismic record automatic calibration and construction interpretation and reservoir prediction method | |
| Feng et al. | Interpretable Lost Circulation Analysis: Labeled, Identified, and Analyzed Lost Circulation in Drilling Operations | |
| Zhong et al. | Intelligent interpretation method for lithology recognition using real-time LWD data | |
| Deng et al. | An Automated Data-Driven Workflow for Identifying Fractured Horizontal Well Sweet Spots in Shale Reservoirs | |
| CN114021663B (en) | Industrial process off-line data segmentation method based on sequence local discrimination information mining network | |
| CN112983406B (en) | Natural gas hydrate reservoir parameter index evaluation method | |
| Damasceno et al. | Source rock evaluation from rock to seismic: Integrated machine learning based workflow | |
| Zhang | Quantitative prediction of median particle size of conventional logged sandstone based on machine learning algorithm | |
| CN115961952A (en) | Reservoir fluid comprehensive discrimination method based on combination parameters in oil and gas reservoir | |
| Li et al. | Application of neural network technique for logging fluid identification in low resistance reservoir |
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 |