CN114063162B - SVM Vp/Vs prediction method based on small sample machine learning - Google Patents
SVM Vp/Vs prediction method based on small sample machine learning Download PDFInfo
- Publication number
- CN114063162B CN114063162B CN202111302883.8A CN202111302883A CN114063162B CN 114063162 B CN114063162 B CN 114063162B CN 202111302883 A CN202111302883 A CN 202111302883A CN 114063162 B CN114063162 B CN 114063162B
- Authority
- CN
- China
- Prior art keywords
- machine learning
- prediction
- data
- small sample
- curve
- 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 25
- 238000010801 machine learning Methods 0.000 title claims abstract description 17
- 238000012549 training Methods 0.000 claims abstract description 23
- 238000012360 testing method Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims 1
- 238000005553 drilling Methods 0.000 abstract 1
- 239000004215 Carbon black (E152) Substances 0.000 description 2
- BVKZGUZCCUSVTD-UHFFFAOYSA-L Carbonate Chemical compound [O-]C([O-])=O BVKZGUZCCUSVTD-UHFFFAOYSA-L 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 229930195733 hydrocarbon Natural products 0.000 description 2
- 150000002430 hydrocarbons Chemical class 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005755 formation reaction Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 239000011148 porous material Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000011435 rock Substances 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 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
- G01V1/30—Analysis
- G01V1/306—Analysis for determining physical properties of the subsurface, e.g. impedance, porosity or attenuation profiles
-
- 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
- G01V1/303—Analysis for determining velocity profiles or travel times
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/62—Physical property of subsurface
- G01V2210/622—Velocity, density or impedance
- G01V2210/6222—Velocity; travel time
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/60—Analysis
- G01V2210/62—Physical property of subsurface
- G01V2210/624—Reservoir parameters
- G01V2210/6242—Elastic parameters, e.g. Young, Lamé or Poisson
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Remote Sensing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Geology (AREA)
- Geophysics (AREA)
- General Engineering & Computer Science (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a SVM Vp/Vs prediction method based on small sample machine learning in the cross field of machine learning and geophysical prediction. The method comprises the steps of firstly, carrying out iterative training on a small sample machine learning SVM model by utilizing wave impedance Ip and Vp curve data of at least 1 well and label data Vp/Vs (dipole acoustic wave data is needed) and achieving the required precision. The trained SVM Vp/Vs prediction model is then applied to other dipole-free acoustic wells Vp/Vs predictions. The method of the invention can be used for carrying out Vp/Vs curve prediction on the dipole-free acoustic wave data well, and has the advantages of rapidness, simplicity and convenience, and the working efficiency is improved by more than 10 times compared with the traditional method. The method is particularly suitable for areas with less drilling.
Description
Technical Field
The invention belongs to the crossing field of machine learning and geophysical prediction.
Background
The longitudinal and transverse wave velocity ratios Vp/Vs are important petrophysical parameters, and are important parameters for calculating the poisson ratio (the poisson ratio is a nonlinear function of Vp/Vs), and the larger the Vp/Vs, the larger the poisson ratio, so Vp/Vs is called the pseudo poisson ratio. Poisson's ratio or Vp/Vs plays an important role in hydrocarbon exploration and development reservoir prediction and hydrocarbon detection. Typically in clastic formations, sandstones exhibit low poisson's ratio or low Vp/Vs characteristics, while mudstones, including coal seams, exhibit high poisson's ratio or high Vp/Vs characteristics, so that poisson's ratio can be used to distinguish between sandstones and mudstones. If oil gas, particularly a gas reservoir, is stored in sandstone, the poisson ratio is lower, so the low poisson ratio is an important index for oil gas detection. In carbonate strata, the Poisson's ratio is also obviously reduced after oil gas is gathered in the reservoir, so searching for low Poisson's ratio or low Vp/Vs abnormality on the basis of predicting the reservoir is also an important technical means for detecting oil gas in the carbonate strata.
Conventional log data does not contain Vp/Vs curve data nor poisson's ratio curve data, so it is very interesting to predict Vp/Vs with some algorithm using other log data. The prior prediction of Vp/Vs is to predict Vs first and then calculate Vp/Vs (conventional log data contains Vp curves). The traditional method for predicting Vs is a petrophysical modeling method, namely parameters such as bulk modulus, shear modulus, density, pore aspect ratio, bulk modulus, density, porosity, argillaceous content, water saturation and the like of sandstone and mudstone are input by using some theoretical models such as an Xu-White model of a sandy mudstone stratum, and Vp, vs and density curves are obtained through a forward modeling method. The traditional petrophysical modeling method has a plurality of defects:
(1) Traditional petrophysical modeling methods are very cumbersome and require specialized geophysical logging engineers to be able to master the method;
(2) Conventional petrophysical modeling methods require a large number of parameters including rock elastic modulus, reservoir parameters, and reservoir parameters, which are often difficult to obtain;
(3) The traditional petrophysical modeling method has great workload and low efficiency.
The appearance of machine learning will subvert many traditional techniques or processes, and has very strong nonlinear fitting ability, and the work efficiency is greatly improved.
With only a small amount of dipole sonic well data (providing measured Vs data) and conventional well logging data in a research area, vp/Vs can be directly predicted by using a small sample machine learning algorithm (support vector machine SVM). The input characteristic parameter data are typically the wave impedance Ip and the longitudinal wave velocity Vp, which can be obtained from conventional logging data. The tag data is Vp/Vs (Vs is derived from measured dipole sonic logging). As long as 1 well is subjected to dipole sonic logging, the advantage of the small sample machine learning algorithm is achieved. And (3) predicting the machine learning model of the Vp/Vs through iterative training, and then predicting the inspection interval Vp/Vs by using the trained model to analyze the accuracy of the test set data. If this model accuracy is available, it can be used for Vp/Vs prediction of other wells. Compared with the traditional petrophysical modeling method, the method has various advantages, including simple and convenient method flow (common geophysical engineers and geological engineers can master), strong practicability and greatly improved working efficiency (the efficiency can be improved by more than 10 times).
Disclosure of Invention
The invention aims to solve the problem of Vp/Vs prediction of only conventional logging data in the prediction of the seismic reservoir, and can greatly improve the working efficiency.
The invention is realized in particular as follows:
step one, collecting conventional log impedance Ip curves, longitudinal wave velocity Vp curves, and dipole sonic log Vs curves (at least 1 well).
And step two, carrying out normalization processing on the Ip and Vp curves to normalize the Ip and Vp curves to a range of-1 and 1 or a range of 0 and 1.
And step three, determining a training set data layer segment and a test set data layer segment (equivalent to a verification layer segment).
And fourthly, performing iterative training on the small sample machine learning SVM model by using the Ip and Vp curve data of the training set and the label data Vp/Vs. And (3) adjusting key parameters such as Cost penalty coefficients, kernel function Gamma values and epsilon values to ensure that the accuracy meets the requirements, for example, the accuracy of a training set and a testing set is more than 85%, and the accuracy of the testing set is allowed to be slightly lower than that of the training set.
And fifthly, applying the SVM model trained in the step four to other dipole-free acoustic wave well Vp/Vs prediction.
Drawings
FIG. 1 is a graph of the SVM Vp/Vs prediction model prediction Vp/Vs process and results. The first two columns are the input raw wave impedance Ip and longitudinal wave velocity Vp curves, and columns 4 and 5 are Ip and Vp curves normalized to 0 and 1. The column 6 is the superposition (broken line) of the Vp/Vs measured actually and the Vp/Vs predicted by the SVM, most of the two are coincident, and the average accuracy rate of 95% (5% relative error) can be found from the relative error curve of the column 6, so that the high precision can be achieved both in the training set and the test set.
Detailed Description
The following detailed description of the invention refers to the accompanying drawings, which illustrate and explain the invention by way of example only, and are not intended to limit the invention.
The present invention is thus achieved.
Step one, collecting conventional log impedance Ip curves, longitudinal wave velocity Vp curves, and dipole sonic log Vs curves (at least 1 well).
And step two, carrying out normalization processing on the Ip and Vp curves to normalize the Ip and Vp curves to be in the range of 0 and 1.
And step three, determining a training set data layer segment and a test set data layer segment (equivalent to a verification layer segment). 1700-1850m in this example are training set intervals, 1850-1950m are test set intervals.
And fourthly, performing iterative training on the small sample machine learning SVM model by using the Ip and Vp curve data of the training set and the label data Vp/Vs. And adjusting key parameters such as Cost penalty coefficient, kernel function Gamma value and epsilon value to ensure that the average accuracy of the training set and the testing set reaches 95 percent.
The above embodiments describe embodiments of the present invention by taking Vp/Vs as an example of log prediction, but do not limit the scope of the present invention. In this embodiment, log data is taken as an example, but it is obvious that the method of the present invention can be easily generalized to three-dimensional pre-stack inversion wave impedance Ip data and Vp data.
Claims (3)
1. A prediction method based on small sample machine learning SVM Vp/Vs is characterized in that at least 1 well Ip and Vp curve data and label data Vp/Vs are used for carrying out iterative training on the small sample machine learning SVM model to reach more than 85% of accuracy; the SVM Vp/Vs prediction model is applied to the prediction of other dipole-free sonic wells Vp/Vs, so that the prediction of the dipole-free sonic well Vp/Vs curve in the exploration or evaluation stage is realized; the implementation steps are as follows:
step one, collecting a conventional logging wave impedance Ip curve, a longitudinal wave velocity curve Vp and a dipole acoustic logging Vs curve (at least 1 well);
step two, carrying out normalization processing on the Ip and Vp curves to normalize the Ip and Vp curves to be in a range of-1 to 1 or 0 to 1;
step three, determining a training set data layer segment and a test set data layer segment (equivalent to a verification layer segment);
step four, performing iterative training on the small sample machine learning SVM model by using the Ip and Vp curve data of the training set and the label data Vp/Vs; the key parameters such as Cost penalty coefficient, kernel function Gamma value and epsilon value are adjusted to enable the accuracy to meet the requirements, for example, the accuracy of a training set and a testing set is more than 85%, and the accuracy of the testing set is allowed to be slightly lower than that of the training set;
and fifthly, applying the SVM model trained in the step four to other dipole-free acoustic wave well Vp/Vs prediction.
2. The method for predicting Vp/Vs by small sample machine learning of claim 1, wherein step four, iteratively training the small sample machine learning SVM model using Ip and Vp curve data of the training set and the label data Vp/Vs; and (3) adjusting key parameters such as Cost penalty coefficients, kernel function Gamma values and epsilon values to ensure that the accuracy meets the requirements, for example, the accuracy of a training set and a testing set is more than 85%, and the accuracy of the testing set is allowed to be slightly lower than that of the training set.
3. The method of claim 1, wherein the fifth step is to apply the SVM model trained in the fourth step to other dipole-free acoustic well Vp/Vs predictions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111302883.8A CN114063162B (en) | 2021-11-05 | 2021-11-05 | SVM Vp/Vs prediction method based on small sample machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111302883.8A CN114063162B (en) | 2021-11-05 | 2021-11-05 | SVM Vp/Vs prediction method based on small sample machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114063162A CN114063162A (en) | 2022-02-18 |
| CN114063162B true CN114063162B (en) | 2024-03-08 |
Family
ID=80273995
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111302883.8A Active CN114063162B (en) | 2021-11-05 | 2021-11-05 | SVM Vp/Vs prediction method based on small sample machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114063162B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115657132B (en) * | 2022-12-29 | 2023-03-10 | 成都捷科思石油天然气技术发展有限公司 | Reservoir prediction method and system based on machine learning |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111596354A (en) * | 2020-05-11 | 2020-08-28 | 同济大学 | Seismic reservoir prediction method considering space constraint under machine learning framework |
| CN112230283A (en) * | 2020-10-12 | 2021-01-15 | 北京中恒利华石油技术研究所 | Seismic porosity prediction method based on logging curve support vector machine modeling |
| WO2021130512A1 (en) * | 2019-12-23 | 2021-07-01 | Total Se | Device and method for predicting values of porosity lithofacies and permeability in a studied carbonate reservoir based on seismic data |
| CN113267816A (en) * | 2021-05-27 | 2021-08-17 | 北京珠玛阳光科技有限公司 | Ultrahigh-resolution data fusion implementation method for seismic logging based on small sample machine learning |
-
2021
- 2021-11-05 CN CN202111302883.8A patent/CN114063162B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021130512A1 (en) * | 2019-12-23 | 2021-07-01 | Total Se | Device and method for predicting values of porosity lithofacies and permeability in a studied carbonate reservoir based on seismic data |
| CN111596354A (en) * | 2020-05-11 | 2020-08-28 | 同济大学 | Seismic reservoir prediction method considering space constraint under machine learning framework |
| CN112230283A (en) * | 2020-10-12 | 2021-01-15 | 北京中恒利华石油技术研究所 | Seismic porosity prediction method based on logging curve support vector machine modeling |
| CN113267816A (en) * | 2021-05-27 | 2021-08-17 | 北京珠玛阳光科技有限公司 | Ultrahigh-resolution data fusion implementation method for seismic logging based on small sample machine learning |
Non-Patent Citations (1)
| Title |
|---|
| "二步法"测井曲线校正及其应用效果分析;钟翔;牛华伟;安勤华;刘婷婷;;海洋石油;20200922(第03期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114063162A (en) | 2022-02-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Islam et al. | Advanced petroleum reservoir simulation: Towards developing reservoir emulators | |
| CN105445791A (en) | Stratum aperture pressure prediction method based on variety earthquake attributes | |
| US8255165B2 (en) | Method for predicting differences in subsurface conditions | |
| CN104181585A (en) | Shear wave estimation method and system in geophysical exploration | |
| CN104769458A (en) | Cauchy distribution-based poststack wave impedance retrieval method | |
| CN105182424A (en) | Method and device of reservoir porosity quantitative forecast based on patchy saturation model | |
| CN110954948A (en) | Physical property parameter inversion method and system for rock physical constraint reservoir | |
| CN111399044A (en) | Reservoir permeability prediction method and device and storage medium | |
| CN105697002A (en) | Method for recognizing coal measure strata lithology | |
| US11630228B1 (en) | Physical embedded deep learning formation pressure prediction method, device, medium and equipment | |
| CN106154343A (en) | The method calculating the oil saturation of fine and close oil reservoir | |
| CN111767674A (en) | Well logging lithology identification method based on active domain adaptation | |
| Elahi* et al. | Characterization of fracture length and conductivity from tracer test and production data with Ensemble Kalman filter | |
| CN114063162B (en) | SVM Vp/Vs prediction method based on small sample machine learning | |
| Gamal et al. | Sonic logs prediction in real time by using random forest technique | |
| Acheampong et al. | Quantitative interpretation of time‐lapse seismic data at Farnsworth field unit: Rock physics modeling, and calibration of simulated time‐lapse velocity responses | |
| CN110118994B (en) | Continental facies hydrocarbon source rock quantitative prediction method based on seismic inversion and machine learning | |
| CN106568918B (en) | Shale organic carbon content TOC prediction method | |
| Zhang et al. | Controls of interlayers on the development and distribution of natural fractures in lacustrine shale reservoirs: A case study of the Da'anzhai member in the Fuling area in the eastern Sichuan Basin | |
| CN115586572B (en) | Seismic rock physical analysis inversion method for pore parameters and reservoir parameters | |
| CN104280767A (en) | Sparse-spike inversion method based on Cauchy distribution | |
| Ge et al. | Determination of total organic carbon (TOC) in tight reservoir using empirical mode decomposition-support vector regression (EMD-SVR): a case study from XX-1 Basin, Western China | |
| CN115877464A (en) | Lithology identification method and device, computer equipment and storage medium | |
| CN113608258B (en) | Self-consistent deep learning method for constructing high-resolution wave impedance inversion label | |
| CN110716235B (en) | Sand shale well logging pore structure 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 |