CN113130018B - Lithology recognition method based on reservoir meta-target invariant feature description - Google Patents

Lithology recognition method based on reservoir meta-target invariant feature description Download PDF

Info

Publication number
CN113130018B
CN113130018B CN202110426395.1A CN202110426395A CN113130018B CN 113130018 B CN113130018 B CN 113130018B CN 202110426395 A CN202110426395 A CN 202110426395A CN 113130018 B CN113130018 B CN 113130018B
Authority
CN
China
Prior art keywords
information
curve
reservoir
depth
layer
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
Application number
CN202110426395.1A
Other languages
Chinese (zh)
Other versions
CN113130018A (en
Inventor
曹志民
阳璨
吴云
韩建
全星慧
付天舒
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Northeast Petroleum University
Original Assignee
Northeast Petroleum University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Northeast Petroleum University filed Critical Northeast Petroleum University
Priority to CN202110426395.1A priority Critical patent/CN113130018B/en
Publication of CN113130018A publication Critical patent/CN113130018A/en
Application granted granted Critical
Publication of CN113130018B publication Critical patent/CN113130018B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/70—Machine learning, data mining or chemometrics
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/20—Ensemble learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/02—Agriculture; Fishing; Forestry; Mining
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/30—Prediction of properties of chemical compounds, compositions or mixtures
    • 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
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A10/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE at coastal zones; at river basins
    • Y02A10/40—Controlling or monitoring, e.g. of flood or hurricane; Forecasting, e.g. risk assessment or mapping

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • General Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • Crystallography & Structural Chemistry (AREA)
  • Medical Informatics (AREA)
  • Marketing (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Tourism & Hospitality (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Chemical & Material Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Marine Sciences & Fisheries (AREA)
  • Primary Health Care (AREA)
  • Mining & Mineral Resources (AREA)
  • Agronomy & Crop Science (AREA)
  • Game Theory and Decision Science (AREA)
  • Databases & Information Systems (AREA)
  • Animal Husbandry (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Debugging And Monitoring (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a lithology recognition method, in particular to a lithology recognition method based on invariant feature description of a reservoir element target. According to the method, through extraction and description of the invariant features of the logging curve, the automatic layering of the meta-object is realized, meanwhile, the invariant features of the meta-object are embedded into a lithology recognition machine learning model, lithology prediction is finally realized in unknown well machine model application, the generalization capability of the reservoir description is greatly enhanced while the individuality of a local reservoir is maintained, the bottleneck problem that the cross-well popularization capability restricts the machine learning method when the lithology recognition is performed by using the logging curve at present is solved, and the reservoir description precision and reliability are improved.

Description

Lithology recognition method based on reservoir meta-target invariant feature description
Technical Field
The invention provides a lithology recognition method, in particular to a lithology recognition method based on invariant feature description of a reservoir element target.
Background
Along with the fact that most of old oil fields in China enter a middle and late development stage, exploration and development of unconventional oil and gas resources become the most main way for realizing stable yield and increasing production and prolonging production life of the current old oil fields. Whether the exploration and development of unconventional oil and gas resources of old oil fields or the accurate prediction and development of new oil and gas resources are the unavoidable difficulties faced by the major oil field enterprises in China at present.
Geophysical well logging data is one of the most important sources of information for obtaining reservoir resource description information. Because the use of different types of geophysical log data sets for target reservoir stratification is an important basis for subsequent lithology identification, log phase analysis, reservoir partitioning, and inter-well parameter prediction of oil content, which directly affect the performance of these subsequent applications, it is a primary task to more accurately describe the geological change of the target reservoir in order to effectively solve the above-mentioned problems. The physical nature of layering with well logs is that the target reservoir is divided into multiple small layers with the same geologic features, which in turn can reduce the amount of data that needs to be analyzed for reservoir description and to some extent the effects of non-stratigraphic factors. Obviously, the accurate and reliable lithology identification has very important practical significance for the exploration and development work of oil and gas resources.
At present, related enterprises and academic research institutions at home and abroad have conducted a great deal of researches on description of logging big data reservoirs. In the aspect of theoretical research methods, three main technical methods are adopted at present: a geostatistical based deterministic or stochastic geologic modeling and reservoir characterization method; classical machine learning methods; an integrated learning/deep learning method. However, the development speed of classical geological geophysical technology is slow, and the cost of manpower, time and the like is high, so that the development of communication, storage and computing capabilities cannot be kept pace with; the single machine learning method has large fitting risk and poor popularization capability; the integrated learning method has a certain vitality, but cannot realize the full utilization of the increasingly multi-source heterogeneous well earthquake big data; although the deep learning method shows the capability of superstrong potential value of mining related big data in many other fields, the current well earthquake multi-source heterogeneous big data cannot reach the applicable condition of the deep learning method and cannot fully exert the capability of the deep learning method. However, the storage state of actual oil and gas resources is more and more complex, the change is fast, the modes are scattered and changeable, and the single or simple multi-mode description cannot adapt to the actual situation. Particularly, the method has outstanding effect on the large-scale high-precision logging curve when facing the aspects of the later development and the diving of middle and late old oil fields and the effective utilization of new oil and gas resource reserves. In addition, the statistical characteristics of logging data among different wells often have non-negligible differences, and simple curve standardization processing and the like are difficult to make up for the differences and even bring new information loss.
Therefore, aiming at the layering property of the logging curve thin sand body, the object-oriented multi-geological object sample and the effective identification of the object-oriented inter-well lithology and lithology through the association matching of the object-oriented test data and the sample data are critical technologies for urgent needs of oil fields.
Disclosure of Invention
The invention aims to solve the problem that the original spatial amplitude characteristic of the existing logging curve does not have invariance between wells. The invention provides a lithology recognition method based on invariant feature description of a reservoir element target.
It comprises the following steps:
step S1: obtaining reservoir correlation characteristics by taking correlation measurement of adjacent points in the longitudinal direction from each depth sampling vector and obtaining corresponding correlation difference characteristics by taking difference from measurement distances of the correlation characteristics, thereby realizing correlation invariance characteristic extraction among multiple logging curves;
step S2: extracting tensor features of a multi-curve reservoir structure and extracting local binary pattern LBP texture features of each curve in the longitudinal direction by carrying out transverse singular value decomposition on a neighborhood vector set of each depth sampling vector, and extracting structural invariance features among multi-logging curves;
step S3: obtaining local statistical characteristics through microscopic invariant moment characteristics obtained through the description of statistical information of a logging curve data set and global statistical characteristics by means of macroscopic gray level symbiotic invariance texture characteristics, obtaining interwell invariance characteristics of the logging curve data set, and realizing extraction of the statistical invariance characteristics among multiple logging curves;
step S4: combining and utilizing the correlation difference characteristics obtained in the step S1 and the tensor characteristics obtained in the step S2 to obtain accurate and precise geological edge layering points of the reservoir element target, thereby realizing automatic layering;
step S5: the lithology recognition of each fine small layer is realized by carrying out complete description on available information and unchanged characteristics which can be obtained by a logging curve data set in the small layer;
step S6: carrying out lithology recognition or coding by using the descriptive information obtained in the step five in a multi-channel integrated machine learning mode, and constructing a lithology recognition machine learning model;
step S7: lithology predictions are implemented in unknown well machine model applications.
Preferably, the method for extracting the relevant invariant feature between the multi-log curves in the step S1 is as follows: the correlation features include: pearson correlation coefficient and cosine correlation coefficient, which are calculated by the following equation (1) and equation (2), respectively:
Figure BDA0003029747890000031
Figure BDA0003029747890000032
wherein ,Si Representing an i-th depth sample vector on the depth axis; cov (S) i ,S i-1 ) Representing covariance of adjacent depth sample vectors; sigma (S) i ) Representing depth sample vector S i Standard deviation of (2);
the distance measure includes: the Euclidean distance measure, the chebyshev distance measure, and the city block distance measure are calculated by formulas (3) (4) (5), respectively:
Figure BDA0003029747890000033
Figure BDA0003029747890000034
Figure BDA0003029747890000035
preferably, the method for extracting the structure invariant feature between the multi-log curves in the step S2 is as follows:
to obtain structural tensor features between log curves, a set of vectors N (S i ) Singular value decomposition is performed on the obtained product:
Figure BDA0003029747890000036
wherein ,λ1 ≥λ 2 For a local depth sample point vector set N (S i ) Then the corresponding depth sample vector S i The structural tensor features of (2) are taken as:
Figure BDA0003029747890000037
for a given curve X, the LBP texture feature is calculated as follows:
Figure BDA0003029747890000038
wherein ,Ni A local neighborhood of an ith depth sample point of the logging curve; f (f) j The binary code value is encoded according to the following rule:
Figure BDA0003029747890000041
preferably, the method for extracting the statistical invariant feature between the multiple logging curves in step S3 specifically includes:
for a local depth sample point vector set N (S i ) The invariant moment is expressed as:
φ 1 ＝η 20 +η 02 (10)
Figure BDA0003029747890000046
φ 3 ＝(η 30 -3η 12 ) 2 +(3η 21 -η 03 ) 2 (12)
φ 4 ＝(η 30 +η 12 ) 2 +(η 21 +η 03 ) 2 (13)
Figure BDA0003029747890000042
φ 6 ＝(η 20 -η 02 )[(η 30 +η 12 ) 2 -(η 21 +η 03 ) 2 ]+4η 11 (η 30 +η 12 )(η 21 +η 03 ) (15)
Figure BDA0003029747890000043
wherein ,
Figure BDA0003029747890000044
m represents the lateral dimension of the depth sample vector set, i.e., the log Qu Xiantiao number; n represents the longitudinal dimension of the depth sampling vector set, namely the local depth sampling point number; n (N) si (i, j) represents the magnitude of the j-th depth sample point of the i-th curve in the depth sample vector set;
providing gray level symbiotic invariance texture feature description in terms of macroscopic interaction statistical information from a logging curve; for differential log pairs (dX, dY), each curve is quantized or grayed separately; gray level symbiotic invariance texture feature expression:
T GLCM (i)＝GLCM(dX(i),dY(i)) (17)
wherein ,
Figure BDA0003029747890000045
at this time, N (dx=i, dy=j) represents the number of dx=i, dy=j in the same depth sample;
tn=l×l is the number of all possible gray scale pairs, where L is the number of gray scale levels.
Preferably, in step S4, the method for obtaining the accurate and precise geological edge layering point of the reservoir element target by combining the relevant difference feature obtained in step S1 and the tensor feature obtained in step S2, thereby realizing automatic layering specifically includes:
the following candidate edge points can be obtained by using the correlation difference feature obtained in step S1:
Figure BDA0003029747890000051
wherein ZCrosss (dCorr) represents the rising zero crossing of the dCorr-related difference feature; n (p) i ) Representing the current point p i Is a local neighborhood of (b); t (T) dCorr Is a threshold constant;
and (3) obtaining the following candidate edge points by using the reservoir structure tensor characteristics obtained in the step S2:
P Ten ＝{p i |(p i ∈Peak(Ten))} (19)
wherein Peak (Ten) represents a Peak point of the Ten feature;
the total edge candidate points are:
P EC ＝∪(P dCorr ,P Ten )。 (20)
preferably, in step S5, the method for identifying lithology of each fine small layer specifically includes:
describing useful information for an in-formation log dataset includes: the thickness of the small layer, the relative height information of the curves in the layer, the absolute height information of the curves in the layer, the shape/form information of each curve in the small layer and the context information of the data in the small layer and the adjacent layers.
Preferably, the thickness information of the small layer: the depth difference of the top and the bottom of the small layer can be directly utilized, namely:
Thick＝Depth bottom -Depth top (21)
information of the relative height of the intra-layer curve:
Figure BDA0003029747890000052
wherein ,
Figure BDA0003029747890000053
a j-th curve representing an i-th small layer;
absolute height information of intra-layer curves:
Figure BDA0003029747890000054
shape/form information of each curve in the small layer:
using the structure tensor and LBP texture features mentioned in step S2;
intra-cell data and context information of adjacent layers:
and describing the small layer context information from a layer thickness comparison relation, an absolute amplitude relation and a layer structure similarity relation, wherein the layer structure similarity relation is calculated by utilizing the correlation of the layer structure tensor, the invariant distance and the LBP texture characteristic information.
According to the lithology recognition method based on the invariant feature description of the reservoir metadata target, a lithology machine learning model is built, the problem that the deviation of the characteristic statistical properties of the conventional inter-well logging curve is large is successfully solved, and the inter-well popularization capacity of the machine learning model is improved. According to the method, through extraction and description of the invariant features of the logging curve, the automatic layering of the meta-object is realized, meanwhile, the invariant features of the meta-object are embedded into a lithology recognition machine learning model, lithology prediction is finally realized in unknown well machine model application, the generalization capability of the reservoir description is greatly enhanced while the individuality of a local reservoir is maintained, the bottleneck problem that the cross-well generalization capability restricts the machine learning method when the lithology recognition is performed by using the logging curve at present is solved, and the reservoir description precision and reliability are improved.
Drawings
FIG. 1 is a diagram of a log interwell domain invariant feature system construction;
FIG. 2 is a flow chart of a lithology recognition method based on reservoir metadata object invariant feature description according to an embodiment of the present invention;
FIG. 3 is an exemplary graph of a characteristic curve associated with logging data of a well according to a lithology recognition method based on reservoir metadata target invariant feature description in accordance with a second embodiment of the present invention;
FIG. 4 is an exemplary graph of a difference characteristic curve associated with logging data of a well according to a lithology recognition method based on reservoir metadata target invariant feature description in accordance with a second embodiment of the present invention;
FIG. 5 is an exemplary graph of tensor feature curves of a well log data structure according to a lithology recognition method based on reservoir metadata target invariant feature description according to a third embodiment of the present invention;
FIG. 6 is an exemplary graph of LBP texture characteristics of well logging data according to a lithology recognition method based on reservoir metadata target invariant feature description according to a third embodiment of the present invention;
FIG. 7 is a graph of a constant moment characteristic example of a well log set based on a lithology recognition method described by reservoir metadata target constant characteristics according to a fourth embodiment of the present invention;
FIG. 8 is a graph of a GLCM feature example of a well log set based on a lithology recognition method described by reservoir metadata target invariant feature description according to embodiment five of the present invention;
fig. 9 is a schematic diagram showing a complete description of a small-scale logging Xiang Tezheng based on a lithology recognition method of reservoir metadata target invariant feature description according to a sixth embodiment of the present invention;
fig. 10 is a quantitative description vector structure diagram of intra-layer data and context-related information of a lithology recognition method based on reservoir metadata target invariant feature description according to a sixth embodiment of the present invention;
FIG. 11 is a schematic diagram of a lithology recognition machine learning model of a lithology recognition method based on reservoir metadata target invariant feature description according to a seventh embodiment of the present invention;
FIG. 12 is an experimental result of lithology prediction of a gold 98 well according to the present invention;
FIG. 13 is a detailed view of experimental results of lithology prediction of a gold 98 well according to the present invention;
FIG. 14 is a graph showing the experimental results of lithology prediction of the gold 392 well according to the present invention;
FIG. 15 is a detailed view of experimental results of lithology prediction of the gold 392 well according to the present invention;
FIG. 16 is an experimental result of lithology prediction of a gold 50 well according to the present invention;
FIG. 17 is a detailed view of experimental results of lithology prediction of a gold 50 well according to the present invention.
Detailed Description
Embodiment one:
the lithology recognition method based on the reservoir element target invariant feature description according to the present embodiment is described below with reference to fig. 1, and includes the following steps:
step S1: obtaining reservoir correlation characteristics by taking correlation measurement of adjacent points in the longitudinal direction from each depth sampling vector and obtaining corresponding correlation difference characteristics by taking difference from measurement distances of the correlation characteristics, thereby realizing correlation invariance characteristic extraction among multiple logging curves;
step S2: extracting tensor features of a multi-curve reservoir structure and extracting texture features of a local binary pattern (LBP, local Binary Patterns) in the longitudinal direction of each curve by carrying out transverse singular value decomposition on a neighborhood vector set of each depth sampling vector, and extracting structural invariance features among multi-logging curves;
step S3: obtaining local statistical characteristics through microscopic invariant moment characteristics obtained through the description of statistical information of a logging curve data set and global statistical characteristics by means of macroscopic gray level symbiotic invariance texture characteristics, obtaining interwell invariance characteristics of the logging curve data set, and realizing extraction of the statistical invariance characteristics among multiple logging curves;
step S4: combining and utilizing the correlation difference characteristics obtained in the first step and the tensor characteristics obtained in the step S2 to obtain accurate and precise geological edge layering points of the reservoir element target, thereby realizing automatic layering;
step S5: the lithology recognition of each fine small layer is realized by carrying out complete description on available information and unchanged characteristics which can be obtained by a logging curve data set in the small layer;
step S6: carrying out lithology recognition or coding by using the descriptive information obtained in the step five in a multi-channel integrated machine learning mode, and constructing a lithology recognition machine learning model;
step S7: lithology predictions are implemented in unknown well machine model applications.
Embodiment two:
further limitation of the lithology recognition method based on reservoir element target invariant feature description according to the first embodiment,
the method for extracting the related invariance features in the first step comprises the following steps:
conventional logging features do not adequately take into account the lateral correlation and associated transformation information of multiple logs at the same depth, which is precisely the invariance capability of the well reservoir description.
However, specifically, for a data set formed by multiple curves of a well, since each depth sampling vector is formed by multiple values from different curves, the reservoir correlation characteristics can be obtained by taking correlation metrics of adjacent points in the longitudinal direction of the depth sampling vector on the log. Among the relevant features mentioned in this patent mainly include: pearson correlation coefficient and cosine correlation coefficient. In addition, the calculation of the correlation can be realized by calculating vector uncorrelation, such as Euclidean distance measure, chebyshev distance measure, urban block distance measure and the like.
Meanwhile, after the above-described correlation characteristics are obtained, the corresponding correlation difference characteristics can be obtained by taking the difference. To better illustrate the extraction of log-related invariance features, fig. 1 and 2 show, as examples, the related features and the related difference features of a well target reservoir, respectively.
Embodiment III:
further limitation of the lithology recognition method based on reservoir element target invariant feature description according to the first embodiment,
the method for extracting the structural invariance features in the second step comprises the following steps:
the structural invariant feature refers to a feature which is obtained by detecting or describing a local structure and remains unchanged for geometric transformation and the like, and the basic idea is to extract essential attributes of the local structure for description. Specifically, the structure invariance information related to the patent mainly comprises texture feature descriptions such as structure tensors, local binary patterns and the like.
For a set of well logs, a vector S is sampled for a certain depth i Let N (S) i ) Representing the vector S sampled in depth i A local neighborhood (neighborhood radius is typically set to about 0.5 meters) for the center. Extraction of the structure tensor features can be achieved by analysis of the neighborhood vector set of each depth sample vector. The structure tensor refers to information derived from the log gradient change information that represents local structural invariance. The main direction and the continuous information of the direction of numerical value change in the neighborhood can be found by carrying out singular value decomposition on the local neighborhood.
In addition to the extraction of multi-curve reservoir structure tensor features in the lateral direction using the local depth sample vector set, invariance descriptions can be made for each curve local longitudinal structure feature. The local binary pattern (Local Binary Mode, LBP) is a way of characterizing the invariance encoding of a local structure.
To better illustrate the extraction of log structure invariance features, fig. 3 and fig. 4 respectively show an exemplary graph of tensor feature curves of a certain well log data structure and an exemplary graph of LBP texture feature curves of certain well log data as examples.
Embodiment four:
further limitation of the lithology recognition method based on reservoir element target invariant feature description according to the first embodiment,
the method for extracting the local statistic invariant moment features in the statistic invariant features in the third step comprises the following steps:
the statistical invariant features are purely statistical features obtained by starting with the statistical information of the log data set, have good invariance between wells, and can have better descriptive ability in the aspects of log phase identification and the like. Specifically, the statistical invariant features are divided into local statistical invariant moment features and global statistical features, and the global statistical features are obtained through symbiotic relations of logging data values.
Invariant moment features are features with invariance to translation, rotation, scale, etc. that are composed of first, second and higher order statistical features of the data.
Specifically, for a local depth sample point vector set N (S i ) Its two-dimensional moment representation can be defined as:
Figure BDA0003029747890000091
where M represents the lateral dimension of the depth sample vector set, i.e., the log Qu Xiantiao number; n represents the longitudinal dimension of the depth sample vector set, i.e., the number of local depth sample points;
Figure BDA0003029747890000092
representing the magnitude of the j-th depth sample point of the i-th curve in the depth sample vector set. Correspondingly, the center moment can be expressed as:
Figure BDA0003029747890000101
in order to make the central moment geometrically invariant, formula (2) can be rewritten as the following form of normalized central moment:
Figure BDA0003029747890000102
further, the invariant moment, i.e., hu invariant moment, can be obtained by normalizing the center moment as described above.
To better illustrate the extraction of local statistical invariance features of well logs, an example graph of invariant moment features of a set of well logs is given as an example in FIG. 5.
Fifth embodiment:
further limitation of the lithology recognition method based on reservoir element target invariant feature description according to the first embodiment,
the method for extracting global statistical invariant moment features from the statistical invariant features in the third step comprises the following steps:
the invariant moment information described in embodiment four is presented through the angle of local microscopic statistical information mining. Clearly, reservoir related information mining can be done macroscopically in addition to mining of microscopic statistical information. Therefore, by means of the idea of Gray-Level Co-occurrence Matrix, GLCM in image processing, a macroscopic texture feature description can be given in terms of log versus macroscopic interaction statistics. Specifically, for the differential log curve pair (dX, dY), each curve is quantized (or grayed), for example, 256-level graying processing is performed, even if the amplitude of the differential curve is specified to be between 0 and 255. Next, gray level co-occurrence matrix calculation is performed as follows:
Figure BDA0003029747890000103
where N (dx=i, dy=j) represents the number of dx=i, dy=j in the same depth sample; tn=l×l is the number of all possible gray scale pairs (L is the number of gray scales). After the gray level co-occurrence matrix is obtained, gray level co-occurrence invariance texture features corresponding to the differential curve pair (dX, dY) can be calculated:
T GLCM (i)＝GLCM(dX(i),dY(i)) (5)
to better illustrate the extraction of global statistical invariance features of well logs, fig. 6 shows an example graph of GLCM features for a set of well logs as an example.
Example six:
further limitation of the lithology recognition method based on reservoir element target invariant feature description according to the first embodiment,
in the fifth step, the available information and the unchanged features which can be acquired by the logging curve data set in the small layer are completely described, so that the lithology of each fine small layer is identified by the following method:
describing useful information for an in-formation log dataset includes: basic physical properties such as thickness of the small layer, relative height information of the in-layer curve, absolute height information of the in-layer curve, and the like; shape/morphology information of each curve in the small layer; and context information of data in the small layer and the adjacent layers, as shown in fig. 7.
(1) Layer thickness information: the depth difference of the top and the bottom of the small layer can be directly utilized, namely:
Thick＝Depth bottom -Depth top (6)
(2) Information of the relative height of the intra-layer curve:
Figure BDA0003029747890000111
wherein ,
Figure BDA0003029747890000112
a j-th curve representing an i-th small layer;
(3) Absolute height information of intra-layer curves:
Figure BDA0003029747890000113
(4) Intra-layer curve morphology/shape information:
the description is made by using the structure tensor and LBP texture features of the second method.
(5) Intra-layer data and context information of adjacent layers:
describing the context information of the small layer in terms of layer thickness comparison relation, absolute amplitude relation, layer structure similarity relation and the like, wherein the layer thickness comparison information and the absolute amplitude relation are the layer thickness and absolute amplitude values of the record layer and the adjacent layer; the layer structure similarity relationship can be calculated by using the correlation of the information such as layer structure tensor, invariant pitch, LBP texture and the like. For this purpose, a quantitative description vector of the in-layer data and the context-dependent information shown in fig. 8 can be obtained.
Embodiment seven:
further limitation of the lithology recognition method based on reservoir element target invariant feature description according to the first embodiment,
in the sixth step, the descriptive information obtained in the fifth step is applied to lithology recognition or coding by adopting a multi-channel integrated machine learning mode, and the lithology recognition machine learning model is constructed by the following steps:
after the complete description of the information, a plurality of basic learning machines are utilized to form multi-channel integrated machine learning, the meta-characteristics of the logging curve are constructed, and the prediction result is optimized through the meta-machine learning and stratum geological knowledge. A machine learning identification framework schematic diagram constructed by the invention is shown in fig. 9.
In order to verify the effectiveness of the method, 24 wells in the range from 3 to 392 in the zizania work area of the central concave area of the north part of the loose basin are adopted to verify the method. Experimental results show that the method has higher prediction accuracy and is closer to the lithology of the rock core than the conventional logging interpretation accuracy, can accurately reflect the change condition of the sand-mud ratio of the small layer, and has good popularization effect.
Wherein 11 wells such as gold 3, gold 12, gold 23, gold 27, gold 31, gold 37, gold 38, gold 54, gold 58, gold 80, gold 391 and the like are randomly selected as model wells, 10 wells such as gold 20, gold 28G, gold 30, gold 34, gold 40, gold 45, gold 51, gold 55, gold 56, gold 59 and the like are selected as model correction wells, and 3 wells such as gold 50, gold 98, gold 392 and the like with core description fine lithology are selected as test wells.
On the selection of logging curves, GR, DEN, SP, CAL, AC, LLD and LLS7 conventional curves are selected, and the target lithology is as follows: mudstone, siltstone, argillite siltstone, fine sandstone, scale insect layer, and oil shale. Fig. 6-11 show overall and detailed results of the lithology predictions for 3 test wells, respectively.
However, the experimental result also shows that the method has certain defects, because the proportion of fine sandstone in the training sample and the test sample is very small, the prediction effect is poor, and the false detection rate is improved under the condition of ensuring the effective detection rate by emphasizing the detection of the oil shale. For this reason, there is also a need in the subsequent studies to achieve a re-enhancement of the system performance by increasing the processing capacity of small samples of fine sandstone, oil shale, etc.
The invention is characterized in that an interwell domain invariant feature system which can effectively improve the stratum information expression capacity of the logging curve as shown in figure 1 is constructed, and a lithology recognition machine learning model is constructed on the basis of the invariant feature system, so that lithology recognition is realized.
In order to solve the problem that the original spatial amplitude characteristic of the existing logging curve does not have invariance among wells, the invention takes fully mining the invariance characteristic which can effectively represent the reservoir characteristic as a starting point, designs a logging curve invariance characteristic system consisting of related invariance characteristic, structural invariance characteristic and statistical invariance characteristic, solves the problem that the logging curve characteristics among wells are inconsistent to a certain extent, can carry out invariance expression aiming at the intrinsic attribute of a target reservoir, and lays a foundation for subsequent processing and analysis.
Aiming at the problem of large deviation of the characteristic statistics of the conventional inter-well logging curve, the invention constructs an inter-well domain invariant feature system of the logging curve by providing an invariant feature description method facing reservoir meta-targets, simultaneously, in order to better realize the task of lithology recognition, the invention completely describes available information which can be acquired by a logging curve data set in a small layer, and on the basis of the available information and the available information, realizes the lithology recognition of the accurate geological reservoir by utilizing an integrated machine learning model. Therefore, the method describes the invariance characteristics of the logging curve interwell domain from three aspects of related invariance characteristics, structural invariance characteristics and statistical invariance characteristics by constructing the invariance characteristic system of the logging curve interwell domain, thereby effectively solving the problem of invariance expression by utilizing the intrinsic properties of a target reservoir, realizing a method for realizing lithology recognition by effectively utilizing the invariance characteristics of a reservoir element target and multi-channel integrated machine learning, and greatly enhancing the generalization capability of reservoir description while keeping the individuality of a local reservoir.
The invention provides a lithology recognition method based on reservoir element target invariant feature description, which aims to solve the bottleneck problem that the popularization capability of a cross-well is restricted to be introduced by a machine learning method when lithology recognition is carried out by using a logging curve at present. The method adapting to the invention has good self-adapting capability and robustness in the application of lithology recognition by using the logging curves of different wells in different areas. Compared with the conventional method for carrying out reservoir description by utilizing simple curves such as the amplitude characteristic curve of the local spatial data and the like, the method greatly improves the accuracy and the reliability of the target reservoir description. Fig. 2 shows a block diagram of the method of the invention. The key technical content of the invention comprises two parts of establishing a logging curve interwell domain invariant feature system and completely expressing interwell small-layer knowledge.
While the invention has been described with respect to particular embodiments thereof, it will be appreciated that the invention is not limited thereto but may be practiced with modification and alteration within the spirit and scope of the appended claims.

Claims (7)

1. The lithology recognition method based on the reservoir element target invariant feature description is characterized by comprising the following steps of:
step S1: obtaining reservoir correlation characteristics by taking correlation measurement of adjacent points in the longitudinal direction from each depth sampling vector and obtaining corresponding correlation difference characteristics by taking difference from measurement distances of the correlation characteristics, thereby realizing correlation invariance characteristic extraction among multiple logging curves;
step S2: extracting tensor features of a multi-curve reservoir structure and extracting local binary pattern LBP texture features of each curve in the longitudinal direction by carrying out transverse singular value decomposition on a neighborhood vector set of each depth sampling vector, and extracting structural invariance features among multi-logging curves;
step S3: obtaining local statistical characteristics through microscopic invariant moment characteristics obtained through the description of statistical information of a logging curve data set and global statistical characteristics by means of macroscopic gray level symbiotic invariance texture characteristics, obtaining interwell invariance characteristics of the logging curve data set, and realizing extraction of the statistical invariance characteristics among multiple logging curves;
step S4: combining and utilizing the correlation difference characteristics obtained in the step S1 and the tensor characteristics obtained in the step S2 to obtain accurate and precise geological edge layering points of the reservoir element target, thereby realizing automatic layering;
step S5: the lithology recognition of each fine small layer is realized by carrying out complete description on available information and unchanged characteristics which can be obtained by a logging curve data set in the small layer;
step S6: carrying out lithology recognition or coding by using the descriptive information obtained in the step five in a multi-channel integrated machine learning mode, and constructing a lithology recognition machine learning model;
step S7: lithology predictions are implemented in unknown well machine model applications.
2. The lithology recognition method based on reservoir element target invariant feature description of claim 1, wherein the method for extracting the relevant invariant feature between the multi-log curves in step S1 is as follows: the correlation features include: pearson correlation coefficient and cosine correlation coefficient, which are calculated by the following equation (1) and equation (2), respectively:
Figure FDA0004170051660000011
Figure FDA0004170051660000012
wherein ,Si Representing an i-th depth sample vector on the depth axis; cov (S) i ,S i-1 ) Representing covariance of adjacent depth sample vectors; sigma (S) i ) Representing depth sample vector S i Standard deviation of (2);
the distance measure includes: the Euclidean distance measure, the chebyshev distance measure, and the city block distance measure are calculated by formulas (3) (4) (5), respectively:
Figure FDA0004170051660000021
Figure FDA0004170051660000022
Figure FDA0004170051660000023
3. the lithology recognition method based on reservoir element target invariant feature description of claim 1, wherein the method for extracting the structure invariant feature between the multi-log curves in step S2 is as follows:
to obtain structural tensor features between log curves, a set of vectors N (S i ) Singular value decomposition is performed on the obtained product:
Figure FDA0004170051660000024
wherein ,λ1 ≥λ 2 For a local depth sample point vector set N (S i ) Then the corresponding depth sample vector S i The structural tensor features of (2) are taken as:
Figure FDA0004170051660000025
for a given curve X, the LBP texture feature is calculated as follows:
Figure FDA0004170051660000026
wherein ,Ni For the well logging curveLocal neighborhood of the ith depth sample point of the line; f (f) j The binary code value is encoded according to the following rule:
Figure FDA0004170051660000027
4. the lithology recognition method based on reservoir element target invariant feature description of claim 1, wherein the method for extracting statistical invariant features between multiple logging curves in step S3 specifically comprises:
for a local depth sample point vector set N (S i ) The invariant moment is expressed as:
φ 1 ＝η 20 +η 02 (10)
Figure FDA0004170051660000036
φ 3 ＝(η 30 -3η 12 ) 2 +(3η 21 -η 03 ) 2 (12)
φ 4 ＝(η 30 +η 12 ) 2 +(η 21 +η 03 ) 2 (13)
Figure FDA0004170051660000031
φ 6 ＝(η 20 -η 02 )[(η 30 +η 12 ) 2 -(η 21 +η 03 ) 2 ]+4η 11 (η 30 +η 12 )(η 21 +η 03 ) (15)
Figure FDA0004170051660000032
wherein ,
Figure FDA0004170051660000033
m represents the lateral dimension of the depth sample vector set, i.e., the log Qu Xiantiao number; n represents the longitudinal dimension of the depth sampling vector set, namely the local depth sampling point number;
Figure FDA0004170051660000035
representing the amplitude of the j-th depth sampling point of the i-th curve in the depth sampling vector set;
providing gray level symbiotic invariance texture feature description in terms of macroscopic interaction statistical information from a logging curve; for differential log pairs (dX, dY), each curve is quantized or grayed separately; gray level symbiotic invariance texture feature expression:
T GLCM (i)＝GLCM(dX(i),dY(i)) (17)
wherein ,
Figure FDA0004170051660000034
at this time, N (dx=i, dy=j) represents the number of dx=i, dy=j in the same depth sample; tn=l×l is the number of all possible gray scale pairs, where L is the number of gray scale levels.
5. The lithology recognition method based on the reservoir element target invariant feature description of claim 1, wherein in step S4, the relevant difference features obtained in step S1 and the tensor features obtained in step S2 are combined to obtain the accurate and precise geological edge layering points of the reservoir element target, so that the method for realizing automatic layering specifically comprises the following steps:
the following candidate edge points can be obtained by using the correlation difference feature obtained in step S1:
Figure FDA0004170051660000041
wherein ZCrosss (dCorr) represents the rising zero crossing of the dCorr-related difference feature; n (p) i ) Representing the current point p i Is a local neighborhood of (b); t (T) dCorr Is a threshold constant;
and (3) obtaining the following candidate edge points by using the reservoir structure tensor characteristics obtained in the step S2:
P Ten ＝{p i |(p i ∈Peak(Ten))} (19)
wherein Peak (Ten) represents a Peak point of the Ten feature;
the total edge candidate points are:
P EC ＝U(P dCorr ,P Ten ) (20)。
6. the lithology recognition method based on reservoir element target invariant feature description of claim 1, wherein in step S5, the lithology recognition method for each fine small layer is implemented by performing complete description on available information and invariant features acquired by the intra-small layer logging data set, and specifically includes:
describing useful information for an in-formation log dataset includes: the thickness of the small layer, the relative height information of the curves in the layer, the absolute height information of the curves in the layer, the shape/form information of each curve in the small layer and the context information of the data in the small layer and the adjacent layers.
7. The lithology recognition method based on reservoir element target invariant feature description of claim 1, comprising:
wherein, the thickness information of the minor layer: the depth difference of the top and the bottom of the small layer can be directly utilized, namely:
Thick＝Depth bottom -Depth top (21)
information of the relative height of the intra-layer curve:
Figure FDA0004170051660000042
wherein ,
Figure FDA0004170051660000043
a j-th curve representing an i-th small layer;
absolute height information of intra-layer curves:
Figure FDA0004170051660000044
shape/form information of each curve in the small layer:
using the structure tensor and LBP texture features mentioned in step S2;
intra-cell data and context information of adjacent layers:
and describing the small layer context information from a layer thickness comparison relation, an absolute amplitude relation and a layer structure similarity relation, wherein the layer structure similarity relation is calculated by utilizing the correlation of the layer structure tensor, the invariant distance and the LBP texture characteristic information.
CN202110426395.1A 2021-04-20 2021-04-20 Lithology recognition method based on reservoir meta-target invariant feature description Active CN113130018B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110426395.1A CN113130018B (en) 2021-04-20 2021-04-20 Lithology recognition method based on reservoir meta-target invariant feature description

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110426395.1A CN113130018B (en) 2021-04-20 2021-04-20 Lithology recognition method based on reservoir meta-target invariant feature description

Publications (2)

Publication Number Publication Date
CN113130018A CN113130018A (en) 2021-07-16
CN113130018B true CN113130018B (en) 2023-05-12

Family

ID=76778460

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110426395.1A Active CN113130018B (en) 2021-04-20 2021-04-20 Lithology recognition method based on reservoir meta-target invariant feature description

Country Status (1)

Country Link
CN (1) CN113130018B (en)

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109655933A (en) * 2017-10-11 2019-04-19 中国石油化工股份有限公司 The Lithology Identification Methods and system on unconventional stratum

Family Cites Families (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2397664B (en) * 2003-01-24 2005-04-20 Schlumberger Holdings System and method for inferring geological classes
US8219319B2 (en) * 2009-12-18 2012-07-10 Chevron U.S.A. Inc. Workflow for petrophysical and geophysical formation evaluation of wireline and LWD log data
CN103149589B (en) * 2013-02-22 2016-01-06 中国石油天然气股份有限公司 Pyrogenic rock Petroleum Exploration Methods and device
US10815758B2 (en) * 2015-01-16 2020-10-27 Schlumberger Technology Corporation Oilfield service selector
CN106951924B (en) * 2017-03-27 2020-01-07 东北石油大学 Seismic coherence body image fault automatic identification method and system based on AdaBoost algorithm
CN109388816A (en) * 2017-08-07 2019-02-26 中国石油化工股份有限公司 A kind of hierarchical identification method of complex lithology
CN109425896B (en) * 2017-08-25 2021-04-30 中国石油天然气股份有限公司 Dolomite oil and gas reservoir distribution prediction method and device
CN110685600B (en) * 2018-06-20 2021-01-19 中国石油化工股份有限公司 Drill bit adjustment prediction method for geosteering
CN109919184A (en) * 2019-01-28 2019-06-21 中国石油大学(北京) A kind of more well complex lithology intelligent identification Methods and system based on log data
CN110097069A (en) * 2019-03-11 2019-08-06 西安科技大学 A kind of support vector machines Lithofacies Identification method and device based on depth Multiple Kernel Learning
CN112528106A (en) * 2019-12-20 2021-03-19 中国石油天然气股份有限公司 Volcanic lithology identification method
CN112016477B (en) * 2020-08-31 2023-06-09 电子科技大学 Logging deposit microphase identification method based on deep learning

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109655933A (en) * 2017-10-11 2019-04-19 中国石油化工股份有限公司 The Lithology Identification Methods and system on unconventional stratum

Also Published As

Publication number Publication date
CN113130018A (en) 2021-07-16

Similar Documents

Publication Publication Date Title
CN109653725A (en) A layer water flooding degree log interpretation method is stored up based on sedimentary micro and the mixed of rock phase
CN103616731B (en) Altered Volcanic Rocks at some Place Effective Reservoirs defining method and device in a kind of oil-gas exploration
US7519476B1 (en) Method of seismic interpretation
CN111596978A (en) Web page display method, module and system for lithofacies classification by artificial intelligence
CN106529667A (en) Logging facies identification and analysis method based on fuzzy depth learning in big data environment
CN111767674B (en) Logging lithology recognition method based on active domain adaptation
Xueqing et al. Bi-LSTM deep neural network reservoir classification model based on the innovative input of logging curve response sequences
CN109613623B (en) Lithology prediction method based on residual error network
CN117576335B (en) Three-dimensional space model data processing method and system for mineral area investigation
CN110412662A (en) Method for prediction of reservoirs of thin interbeded based on seismic multi-attribute deep learning
CN114529110A (en) Lithofacies inversion method and system based on deep neural network model
CN105629304A (en) Sand body superposition mode identification method based on multiple attributes
CN113608265A (en) Reservoir prediction method based on deep hybrid neural network
CN104459790B (en) Oil-gas possibility basin effective reservoir analysis method and device
Wei et al. Characterization of saturation and pressure distribution based on deep learning for a typical carbonate reservoir in the Middle East
Yang et al. Deep-learning missing well-log prediction via long short-term memory network with attention-period mechanism
CN116168224A (en) Machine learning lithology automatic identification method based on imaging gravel content
Peng et al. Predicting rate of penetration in ultra-deep wells based on deep learning method
Guoqiang et al. Construction of well logging knowledge graph and intelligent identification method of hydrocarbon-bearing formation
CN113130018B (en) Lithology recognition method based on reservoir meta-target invariant feature description
WO2024012222A1 (en) Method and apparatus for selecting key geological parameter of block to be surveyed, and device
Liu et al. Identification of tight sandstone reservoir lithofacies based on CNN image recognition technology: A case study of Fuyu reservoir of Sanzhao Sag in Songliao Basin
Dramsch Machine Learning in 4D Seismic Data Analysis
WO2022178507A1 (en) Geologic learning framework
Wang et al. A Lithology Identification Approach Using Well Logs Data and Convolutional Long Short-Term Memory networks

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