CN112444856B - Seismic data resolution improvement method based on deep learning - Google Patents
Seismic data resolution improvement method based on deep learning Download PDFInfo
- Publication number
- CN112444856B CN112444856B CN201910818535.2A CN201910818535A CN112444856B CN 112444856 B CN112444856 B CN 112444856B CN 201910818535 A CN201910818535 A CN 201910818535A CN 112444856 B CN112444856 B CN 112444856B
- Authority
- CN
- China
- Prior art keywords
- data
- earthquake
- training
- seismic
- resolution
- 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
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000013135 deep learning Methods 0.000 title claims abstract description 25
- 230000006872 improvement Effects 0.000 title abstract description 5
- 238000012549 training Methods 0.000 claims abstract description 58
- 230000006978 adaptation Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 abstract description 6
- 238000011161 development Methods 0.000 abstract description 5
- 238000004458 analytical method Methods 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 6
- 238000013461 design Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 3
- 239000004215 Carbon black (E152) Substances 0.000 description 2
- 230000003190 augmentative effect Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 229930195733 hydrocarbon Natural products 0.000 description 2
- 150000002430 hydrocarbons Chemical class 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013434 data augmentation Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
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
- G01V1/282—Application of seismic models, synthetic seismograms
-
- 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
- G01V1/30—Analysis
- G01V1/307—Analysis for determining seismic attributes, e.g. amplitude, instantaneous phase or frequency, reflection strength or polarity
-
- 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
-
- 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
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The application provides a seismic data resolution improvement method based on deep learning, which comprises the following steps: s10: acquiring earthquake and well logging data; s20: establishing a depth network training model oriented to seismic data resolution enhancement; s30: and performing deep learning training on the earthquake and well logging data by using the deep network training model to obtain the earthquake and well logging data with high resolution. According to the application, the ground seismic data is automatically enhanced by combining the deep learning method and the well earthquake, the earthquake and well logging data are trained, the resolution of the earthquake and well logging data is improved, and the basis is provided for the analysis of the earthquake and well logging data, so that the data support is provided for the development of the earthquake exploration technology.
Description
Technical Field
The invention relates to the technical field of seismic data processing, in particular to a method for improving seismic data resolution based on deep learning.
Background
The seismic data is limited by acquisition, processing and interpretation techniques, and the resolution of the processed seismic data is low. The high resolution formation data corresponding to the log data is limited by the number of logs and global seismic data cannot be described.
The conventional method for improving the resolution of the seismic data mostly assumes that the seismic data is stable and the noise level does not change with space, but the actual situation does not meet the assumption, so that the effect after the resolution improvement treatment cannot meet the expected requirement.
Deep learning is applied to the fields of images and audios to be mature day by day, and the effect is remarkable. Therefore, on the basis of the existing big data, the well earthquake combination is performed by utilizing deep learning and data driving to improve the resolution of the earthquake data, and the method is particularly urgent and necessary for better oil and gas exploration and production service.
In general, abroad has made preliminary researches on enhancing seismic data based on deep learning and well-seismic combination, and has achieved certain achievements, and these technologies may be different in specific implementation, but the goal is to continuously improve the efficiency and accuracy of seismic interpretation work. At present, simple hydrocarbon reservoirs gradually decrease, and complex geologic bodies such as thin layers, thin interbeds and the like become main targets for exploration and development of the hydrocarbon reservoirs, so that requirements on exploration precision are higher and higher. The automatic enhancement technology of the ground seismic data is realized by combining the deep learning technology with the well earthquake combination, and the corresponding software product is deduced to have important significance for continuously improving the efficiency and the accuracy of the earthquake interpretation work.
Disclosure of Invention
Aiming at the problems in the prior art, the application provides a method for improving the resolution of seismic data based on deep learning, which is used for solving the technical problems.
The application relates to a seismic data resolution improvement method based on deep learning, which comprises the following steps:
s10: collecting earthquake and well logging data;
S20: establishing a depth network training model oriented to seismic data resolution enhancement;
s30: and performing deep learning training on the earthquake and well logging data by using the deep network training model to obtain the earthquake and well logging data with high resolution.
In one embodiment, step S20 includes:
S201: forward modeling is carried out on the actually acquired earthquake and logging data so as to acquire forward modeling data;
S202: obtaining training data according to forward data and well side data;
S203: acquiring a training label according to the actually acquired earthquake and logging data;
s204: and acquiring a deep network training model according to the training data and the training label.
In one embodiment, in step S201, the actually acquired seismic and logging data is forward-developed by means of noise addition and stochastic reconstruction of geologic models.
In one embodiment, step S203 includes:
S2031: performing horizon automatic calibration on the actually acquired earthquake and logging data;
s2032: obtaining a geological model in a random geological model reconstruction mode;
S3033: and obtaining a training label according to the geological model and the uphole label.
In one embodiment, in step S203, training tags are obtained according to high resolution reflection interface data in the seismic and logging data;
In step S30, the low-resolution reflection interface data in the seismic and logging data is used as training data to perform deep learning training.
In one embodiment, between steps S20 and S30 further comprises:
step S200: and optimizing the established deep network training model.
In one embodiment, step S200 includes the selection of super parameters, the variation of learning rate, and the optimization of convergence rate.
In one embodiment, the deep network training model is optimized by establishing a loss function, learning rate adaptation, super parameter search, and network parameter initialization.
In one embodiment, step S10 includes:
s101: collecting earthquake and well logging data;
s102: and filtering the acquired earthquake and well logging data to obtain the filtered earthquake and well logging data.
Compared with the prior art, the application has the following advantages:
According to the application, the ground seismic data is automatically enhanced by combining the deep learning method and the well earthquake, the earthquake and well logging data are trained, the resolution of the earthquake and well logging data is improved, and the basis is provided for the analysis of the earthquake and well logging data, so that the data support is provided for the development of the earthquake exploration technology.
The above-described features may be combined in various suitable ways or replaced by equivalent features as long as the object of the present invention can be achieved.
Drawings
The invention will be described in more detail hereinafter on the basis of embodiments and with reference to the accompanying drawings. Wherein:
FIG. 1 shows a flow chart of a depth learning based seismic data resolution enhancement method according to the present application.
Fig. 2 shows a design of a depth learning based seismic data resolution enhancement method according to the present application.
FIG. 3 shows a schematic representation of the reflectance at the extraction well location.
FIG. 4 shows a schematic drawing of low resolution seismic data at a well location.
Fig. 5 shows a schematic representation of training data augmentation.
FIG. 6 shows a schematic diagram of the augmented training data and labels.
Fig. 7 shows a schematic diagram of a deep network design.
Fig. 8 shows a schematic diagram of the data shown in fig. 4 after processing using the method of the present application.
In the drawings, like parts are designated with like reference numerals. The figures are not to scale.
Detailed Description
The invention will be further described with reference to the accompanying drawings.
Fig. 1 shows a method for improving the resolution of seismic data based on deep learning according to the present application, and fig. 2 shows the design concept of the method. In the design concept of fig. 2, it can be seen that, to perform deep network training, training data and training labels need to be acquired first to obtain a deep network training model, then deep network training is performed on low-resolution seismic data, after training, reflection coefficients of the low-resolution seismic data are obtained, and finally high-resolution seismic data are acquired.
Specifically, the method according to fig. 1 comprises the steps of:
The first step: seismic and well logging data are collected.
In the step, the obtained seismic and well logging data should keep the directional attribute of the seismic data as much as possible, and the seismic data should be cleaned efficiently in the process to reduce the influence of noise.
And a second step of: and establishing a depth network training model oriented to seismic data resolution enhancement.
As shown in fig. 5, the actual acquired seismic and logging data is augmented. Because the deep network training model needs to be built based on a large amount of data, but is limited to development cost and other factors, the actually collected seismic and well logging data is very limited, and therefore, the actually collected seismic and well logging data needs to be expanded.
Specifically, forward modeling may be performed on the actually acquired seismic and logging data to obtain forward modeling data. The method can forward the actually collected seismic and well logging data in a mode of adding noise and randomly reconstructing a geological model so as to obtain enough forward data. And then, obtaining training data according to the obtained forward data and the well side data.
And acquiring a training label according to the actually acquired earthquake and logging data. Specifically, the horizon automatic calibration is performed on the actually collected seismic and logging data, i.e., the depth domain data such as logging data is converted into time domain data. And then the geological model is obtained by means of random reconstruction of the geological model, the initial geological model is arranged according to the geological ages, and the sequence of the geological models in different geological ages is randomly disturbed, so that a new geological model is obtained after being disturbed once. And finally, obtaining a training label according to the geological model and the uphole label.
As shown in fig. 3, a geologic model of the location of the extraction well, i.e., a training tag, is shown. In fig. 4, the data at the corresponding position is training data. As shown in fig. 6, the upper curve is a schematic diagram of the training data after expansion, and the lower curve is a schematic diagram of the training label obtained. With the training labels and training data, the depth network design shown in fig. 7 can be utilized, the time and space attributes of the seismic data are fully utilized, a proper learning algorithm is searched for to obtain the best learning effect, and the depth neural network is trained, so that an effective reflection interface prediction system is formed.
Preferably, the training tags are obtained from high resolution reflection interface data in the seismic and well log data. When the geological model is obtained, the geological model is randomly reconstructed by adopting high-resolution reflection interface data in earthquake and well logging data so as to obtain the geological model.
More preferably, as shown in fig. 4, the low resolution seismic data at the well location, that is, the low resolution reflection interface data in the seismic and logging data, is extracted as training data, and is subjected to deep learning training.
And a third step of: and performing deep learning training on the earthquake and well logging data by using the deep network training model to obtain the earthquake and well logging data with high resolution.
In a preferred embodiment, the established deep network training model is also optimized. It includes the selection of super parameters, the variation of learning rate, and the optimization of convergence rate. In particular, the deep network training model may be optimized by establishing a loss function, learning rate adaptation, super parameter search, and network parameter initialization.
Fig. 8 is a schematic diagram of high resolution data obtained by processing the low resolution data shown in fig. 4 using the method of the present application. As can be seen from comparing fig. 4 and 8, the resolution of the seismic data is significantly improved after processing.
In summary, the application realizes the automatic enhancement of the ground seismic data by combining the deep learning method and the well earthquake, trains the seismic and well logging data, improves the resolution of the seismic and well logging data, and provides a basis for the analysis of the seismic and well logging data, thereby providing data support for the development of the seismic exploration technology.
Although the invention herein has been described with reference to particular embodiments, it is to be understood that these embodiments are merely illustrative of the principles and applications of the present invention. It is therefore to be understood that numerous modifications may be made to the illustrative embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the different dependent claims and the features described herein may be combined in ways other than as described in the original claims. It is also to be understood that features described in connection with separate embodiments may be used in other described embodiments.
Claims (6)
1. The method for improving the resolution of the seismic data based on deep learning is characterized by comprising the following steps of:
s10: collecting earthquake and well logging data;
S20: establishing a depth network training model oriented to seismic data resolution enhancement;
s30: performing deep learning training on the earthquake and logging data by using the deep network training model to obtain high-resolution earthquake and logging data;
step S20 includes:
S201: forward modeling is carried out on the actually acquired earthquake and logging data so as to acquire forward modeling data;
S202: obtaining training data according to forward data and well side data;
S203: acquiring a training label according to the actually acquired earthquake and logging data;
S204: acquiring a deep network training model according to the training data and the training label;
Step S203 includes:
S2031: performing horizon automatic calibration on the actually acquired earthquake and logging data;
s2032: obtaining a geological model in a random geological model reconstruction mode;
s3033: acquiring a training label according to the geological model and the uphole label, wherein the training label is a reflection coefficient;
The automatic horizon calibration of the actually collected earthquake and logging data comprises the following steps: converting the depth domain data into time domain data; the obtaining the geological model by the way of random reconstruction of the geological model comprises the following steps: randomly disturbing the sequence of geologic models of different geologic ages once to obtain a new geologic model;
in step S203, a training tag is obtained according to high-resolution reflection interface data in the seismic and logging data;
In step S30, the low-resolution reflection interface data in the seismic and logging data is used as training data to perform deep learning training.
2. The method for improving the resolution of seismic data based on deep learning according to claim 1, wherein in step S201, the actually acquired seismic and logging data is forward-developed by means of noise addition and random reconstruction of geologic models.
3. The depth learning based seismic data resolution enhancement method of claim 1, further comprising between steps S20 and S30:
step S200: and optimizing the established deep network training model.
4. A method for improving the resolution of seismic data based on deep learning according to claim 3, wherein step S200 includes the selection of super parameters, the variation of learning rate, and the optimization of convergence rate.
5. The deep learning based seismic data resolution enhancement method of claim 4, wherein the deep network training model is optimized by creating a loss function, learning rate adaptation, super-parametric search, and network parameter initialization.
6. The depth learning based seismic data resolution enhancement method of claim 1, wherein step S10 comprises:
s101: collecting earthquake and well logging data;
s102: and filtering the acquired earthquake and well logging data to obtain the filtered earthquake and well logging data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910818535.2A CN112444856B (en) | 2019-08-30 | 2019-08-30 | Seismic data resolution improvement method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910818535.2A CN112444856B (en) | 2019-08-30 | 2019-08-30 | Seismic data resolution improvement method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112444856A CN112444856A (en) | 2021-03-05 |
| CN112444856B true CN112444856B (en) | 2024-05-17 |
Family
ID=74735307
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910818535.2A Active CN112444856B (en) | 2019-08-30 | 2019-08-30 | Seismic data resolution improvement method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112444856B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113945974B (en) * | 2021-06-10 | 2022-05-24 | 中国矿业大学(北京) | Seismic data reconstruction method, device, equipment and medium |
| CN113608258B (en) * | 2021-07-16 | 2024-01-19 | 长江大学 | Self-consistent deep learning method for constructing high-resolution wave impedance inversion label |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101206264A (en) * | 2007-11-08 | 2008-06-25 | 符力耘 | Method for inversion of high resolution non-linear earthquake wave impedance |
| CN103389513A (en) * | 2013-08-02 | 2013-11-13 | 中国石油集团川庆钻探工程有限公司 | Method for improving seismic data resolution by using acoustic logging data constraint inversion |
| WO2018201114A1 (en) * | 2017-04-28 | 2018-11-01 | Pioneer Natural Resources Usa, Inc. | High resolution seismic data derived from pre-stack inversion and machine learning |
| CN108802812A (en) * | 2017-04-28 | 2018-11-13 | 中国石油天然气股份有限公司 | Well-seismic fusion stratum lithology inversion method |
| WO2018229469A1 (en) * | 2017-06-12 | 2018-12-20 | Foster Findlay Associates Limited | A method for validating geological model data over corresponding original seismic data |
-
2019
- 2019-08-30 CN CN201910818535.2A patent/CN112444856B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101206264A (en) * | 2007-11-08 | 2008-06-25 | 符力耘 | Method for inversion of high resolution non-linear earthquake wave impedance |
| CN103389513A (en) * | 2013-08-02 | 2013-11-13 | 中国石油集团川庆钻探工程有限公司 | Method for improving seismic data resolution by using acoustic logging data constraint inversion |
| WO2018201114A1 (en) * | 2017-04-28 | 2018-11-01 | Pioneer Natural Resources Usa, Inc. | High resolution seismic data derived from pre-stack inversion and machine learning |
| CN108802812A (en) * | 2017-04-28 | 2018-11-13 | 中国石油天然气股份有限公司 | Well-seismic fusion stratum lithology inversion method |
| WO2018229469A1 (en) * | 2017-06-12 | 2018-12-20 | Foster Findlay Associates Limited | A method for validating geological model data over corresponding original seismic data |
Non-Patent Citations (3)
| Title |
|---|
| Vertical resolution enhancement of seismic data with convolutional U-net;Yonggyu Choi et al.;SEG International Exposition and 89th Annual Meeting;第2页 * |
| 吴大奎等.地震、测井资料联合反演的神经网络算法研究.天然气工业.2004,第24卷(第3期),第55-57页. * |
| 地震、测井资料联合反演的神经网络算法研究;吴大奎等;天然气工业;第24卷(第3期);第55-57页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112444856A (en) | 2021-03-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8271244B2 (en) | Hybrid stratigraphic layering using pseudo-wheeler space | |
| RU2013112257A (en) | IDENTIFICATION OF SEDIMENTATION SYSTEMS | |
| CN112444856B (en) | Seismic data resolution improvement method based on deep learning | |
| CN107765308B (en) | Reconstruct low-frequency data frequency domain full waveform inversion method based on convolution thought Yu accurate focus | |
| WO2013176799A1 (en) | System and method for predicting rock strength | |
| FR3057963A1 (en) | MULTIVARIABLE ANALYSIS OF SEISMIC DATA, MICROSISMIC DATA, AND PETROPHYSICAL PROPERTIES IN FRACTURE MODELING | |
| CN107831542A (en) | DDW high accuracy depths domain well shake matching process | |
| WO2011103297A2 (en) | Estimating internal multiples in seismic data | |
| CN113015926A (en) | Passive seismic imaging | |
| CN104155687A (en) | Phase control post-stack acoustic wave impedance inversion method | |
| CN106226841A (en) | A kind of fluvial facies three-dimensional sedimentary facies model Decided modelling method | |
| Visser et al. | Introduction to the Groningen static reservoir model | |
| Cheng et al. | Meta-Processing: A robust framework for multi-tasks seismic processing | |
| CN107515425A (en) | Suitable for the earthquake prediction method of actic region glutenite deposition connected component | |
| CN103543478A (en) | Geologic morphological interpolation KM (Kriging and Multiple-point geostatistics) method | |
| CN111796326A (en) | Method and system for constructing sequence stratum framework based on Unet network | |
| Saad et al. | SiameseFWI: A deep learning network for enhanced full waveform inversion | |
| CN112649893B (en) | Multi-data and multi-parameter fusion modeling method and system for thin reservoir | |
| US20230084403A1 (en) | Machine learning driven dispersion curve picking | |
| CN105277981A (en) | Non-consistent time-lapse seismic bin matching method based on wave field continuation compensation | |
| Peace et al. | Imaging new opportunities and play concepts in the Adriatic Sea and Levantine Basin | |
| CN115880455A (en) | Three-dimensional intelligent interpolation method based on deep learning | |
| CN111751886A (en) | Shale gas reservoir fracture modeling method based on microseism monitoring data | |
| Baillie et al. | The Gippsland Basin: reprocessing reveals new opportunities in a ‘mature’basin | |
| CN111983689B (en) | Near-source electromagnetic seismoelectric joint GR inversion 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 |