Intelligent identification method for lithology of igneous rock reservoir based on multi-parameter integration
Through the intelligent identification method of multi-parameter comprehensive, the coding and deep learning of heavy magnetoelectric anomalies and seismic attributes are used to solve the problem of lithologic identification of igneous reservoirs, and the efficiency and accuracy of oil and gas exploration are improved.
Patent Information
- Application Number
- CN202310568617.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-18
- Publication Date
- 2024-11-22
AI Technical Summary
In oil and gas exploration, it is difficult for the prior art to effectively identify the lithologic properties of igneous reservoirs in complex geological structures, resulting in high exploration risks and increased costs.
A three-dimensional spatial lithology model is established to improve the recognition accuracy by encoding and deep learning of the abnormalities and seismic properties of heavy magnetoelectric reservoirs.
It has achieved efficient identification of lithologic properties of igneous reservoirs, reduced the risk of oil and gas exploration, and improved the exploration efficiency and accuracy.
Smart Images

Figure CN119001837A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of geophysical exploration, and in particular to an intelligent lithology identification method for igneous reservoirs based on multi-parameter synthesis. Background Art
[0002] Since the discovery of oil and gas in igneous reservoirs, volcanic rocks have become an important growth point for oil and gas reserves. Geophysical exploration is an important means of understanding underground geological targets. Currently, there are four commonly used methods: gravity, magnetic, electric, and seismic. They are the characterization of different physical parameters of the same geological body from different sides. Seismic exploration occupies a dominant position in oil and gas exploration with its high-precision vertical resolution. However, in special lithology bodies such as gravel layers and volcanic rocks, complex surface, complex foreland belts that have experienced multiple tectonic movements, deep and ultra-deep layers, etc., seismic methods are difficult to form effective reflections, and the data quality is poor, which makes exploration face huge challenges. According to the physical properties of volcanic rocks, that is, they usually have the characteristics of high velocity, large seismic wave absorption energy, high density, high magnetic susceptibility, and high resistivity, and different lithology volcanic rocks have different response characteristics in density, magnetic susceptibility, and resistivity, which provides a physical basis for the comprehensive application of multiple geophysical exploration methods. The practice of volcanic rock exploration shows (Li Hui et al., 2011; Yang Hui et al., 2011; Zou Caineng et al., 2011) that the comprehensive geophysical method is a relatively effective method and technology for predicting deep volcanic rocks. It can obtain geophysical information of volcanic rocks from different aspects of gravity, magnetic method and electrical method, divide the lithology and lithofacies of volcanic rocks, identify craters, study the thickness and distribution of volcanic rocks, and assist seismic to enhance the reliability of volcanic rock identification. Gravity and magnetic data have low cost, wide application range, rich low-frequency information, and high lateral resolution. Electromagnetic exploration is a method between gravity and magnetic methods and seismic methods. It has better vertical resolution and stratification ability than gravity and magnetic methods. In view of the complexity of oil and gas exploration and the limitations and multi-solution of single geophysical exploration technology, it is necessary to cooperate, supplement, constrain and feedback multiple data to carry out comprehensive geophysical research, which has become an important direction for the development of petroleum geophysical exploration at home and abroad.
[0003] Many scholars have conducted extensive exploration work in the comprehensive study of volcanic rocks using gravity, magnetism, electricity and seismic methods. Liu Yunxiang et al. (2006) proposed using high-precision gravity, magnetism and time-frequency sounding data to identify volcanic rock properties with differences in density, magnetism and resistivity, and to divide lithofacies. Special processing and interpretation techniques of gravity, magnetism and electricity data are used to separate gravity, magnetism and electricity anomalies, and to extract gravity, magnetism and electricity anomalies related to volcanic rocks for comprehensive geological interpretation, mainly including three technologies: gravity anomaly stripping technology, magnetic anomaly separation and weak signal processing technology, and time-frequency electromagnetic sounding inversion and anomaly extraction technology. When studying the volcanic rock lithology of the Xujiaweizi fault depression in the northern Songliao Basin, Yang Hui et al. (2011) proposed a new method for macroscopic prediction of volcanic rock lithology using gravity seismic stripping enhancement processing method based on the characteristics of gradually increasing density from acidic rock to neutral rock and then to basic rock. The prediction of volcanic rock lithology is achieved through two steps: 1. Establishing a geophysical model based on 3D seismic data, accurately separating gravity anomalies through gravity seismic stripping, that is, removing gravity anomalies caused by density interfaces between strata that are not related to volcanic rocks; 2. Strengthening the weak signal of residual gravity anomalies, removing the influence of large regional fields, and enhancing the weak gravity anomalies caused by deeply buried volcanic rock lithology. When Wang Yuhua et al. (2008) analyzed the exploration of volcanic gas reservoirs in Gulong Fault Depression in Songliao Basin, due to the deep burial of the target volcanic rock, the target gravity and magnetic information obtained was weak. In addition to the conventional gravity and magnetic data processing methods such as terrain correction, magnetic polarization, gravity and magnetic upward and downward extension, residual anomaly extraction, and derivation, they also applied gravity and magnetic weak information extraction, gravity anomaly stripping, gravity, magnetic and seismic joint inversion interpretation and other technologies to separate gravity and magnetic anomalies, extract and enhance gravity and magnetic anomaly information related to volcanic rocks, so as to achieve comprehensive geological interpretation. Xu Ligui et al. (2009) proposed the idea of comprehensively utilizing geophysical data to interpret deep volcanic rocks in superimposed basins: using high-precision gravity and magnetic exploration and electrical data to understand the deep structure of the basin and deep and large faults, preliminarily delineate the distribution of deep volcanic rock bodies, and narrow down and lock in favorable target areas for seismic exploration of deep volcanic rocks; using seismic phase analysis, seismic attribute interpretation, structural mapping, seismic inversion, and comprehensive evaluation as the main technical means, deep volcanic rocks are identified and predicted from qualitative to quantitative aspects, and favorable exploration targets for deep volcanic rocks are implemented and selected. Suo Xiaodong et al. (2009, 2011) and He Zhanxiang et al. (2019) proposed a set of operational technical methods for comprehensive lithology identification using gravity, magnetic, electric and seismic anomalies. According to the differences in rock physical properties, they established gravity, magnetic and electric anomalies and seismic phase characteristic templates for different lithologies. They predicted the planar distribution and lateral changes of volcanic rocks through high-precision gravity and magnetic data. They took advantage of the electrical depth-resistivity section and seismic profile to vertically determine the depth of deep lithologies, and comprehensively determined the spatial position and morphology of volcanic rocks, thereby achieving precise identification of different types of volcanic rock bodies based on multiple parameters such as gravity, magnetism, electricity and seismic.
[0004] In the Chinese patent application with application number: CN201810724558.2, a well-controlled self-coding lithology identification method for igneous rock development areas is involved, which belongs to the field of seismic data processing and interpretation for petroleum exploration and development. The well-controlled self-coding lithology identification method for igneous rock development areas uses the lithology data, seismic data and logging data of multiple wells in the selected area to obtain multiple lithology-sensitive seismic attribute data bodies and multiple lithology-sensitive pre-stack and post-stack rock physical parameter data bodies, and performs well-controlled denoising self-coding attribute dimensionality reduction operations on the two data bodies to obtain lithology identification attributes. According to the lithology identification attributes, the lithology of the igneous rock development area is identified, and the lithology spatial distribution of the igneous rock development area is determined, which solves the problem of complex lithology identification in the igneous rock development area, provides a basis for the drilling design of oil and gas fields in the igneous rock development area, and reduces the risks and costs of exploration and development.
[0005] In the Chinese patent application with application number: CN200810007294.5, a geophysical exploration technology method for identifying the lithology of igneous rocks using the combined characteristics of gravity, magnetism and electricity anomalies is involved. The steps are: collect and record the physical characteristics of the core; use k-means clustering to cluster the density, magnetic susceptibility and resistivity of the rock to obtain the core clustering code; perform gravity and magnetic anomaly stripping on the gravity and magnetic exploration results to obtain anomalies, and use the same clustering to perform three-dimensional space anomaly coding; combine the rock physical property clustering code with the gravity, magnetism and electricity anomaly code to reflect the lithology of each igneous rock in the exploration area. This invention develops the method of comprehensive gravity, magnetism and electricity identification of igneous rock lithology from qualitative speculation to quantitative statistics, reduces the subjective factors of human experience, and does not result in different results for different geological interpreters, thereby improving the identification effect of igneous rocks.
[0006] In the Chinese patent application with application number: CN202110210098.3, a method for determining rock lithology, a storage medium and a computer device are involved. The method includes: S100: obtaining the molecular formula of each of a plurality of pure minerals; S200: obtaining the element types in the target igneous rock and the proportional coefficients of each element type; S300: based on the molecular formula of each of a plurality of pure minerals, according to the element types in the target igneous rock and the proportional coefficients of each element type, determining the mineral types contained in the target igneous rock and the mass content of each mineral type by an inversion method; S400: determining the igneous rock types contained in the target igneous rock and the mass content of each igneous rock type contained in the target igneous rock; S500: determining the lithology type of the target igneous rock according to the igneous rock types contained in the target igneous rock and the mass content of each igneous rock type. The application of this method can effectively determine the quantitative relationship between the mineral content of igneous rocks and elements.
[0007] The above existing technologies are significantly different from the present invention and fail to solve the technical problem we want to solve. For this reason, we have invented a new intelligent identification method for igneous reservoir lithology based on multi-parameter integration. Summary of the invention
[0008] The purpose of the present invention is to provide an intelligent igneous reservoir lithology identification method based on multi-parameter integration that can effectively identify the lithology of igneous reservoirs, thereby effectively reducing the risk of oil and gas exploration.
[0009] The purpose of the present invention can be achieved by the following technical measures: an intelligent identification method for lithology of igneous reservoirs based on multi-parameter synthesis, the intelligent identification method for lithology of igneous reservoirs based on multi-parameter synthesis includes:
[0010] Step 1: Statistical analysis of the physical property data of the study area;
[0011] Step 2, establishing a rock sample physical property parameter coding template;
[0012] Step 3, by extracting and encoding gravity, magneto-electric reservoir anomalies, a single corresponding relationship between gravity, magneto-electric anomalies and igneous rock targets is established;
[0013] Step 4, extracting and encoding seismic attributes;
[0014] Step 5, identifying the target lithology of the igneous reservoir and obtaining model A;
[0015] Step 6, using steps 3 and 4 as learning samples to perform neural network prediction of igneous reservoir lithology to obtain model B;
[0016] Step 7: If the error between models A and B is greater than the preset standard, return to step 2; if the error is less than the preset standard, evaluate both models, end and output the results.
[0017] The purpose of the present invention can also be achieved by the following technical measures:
[0018] In step 1, using the physical property data and drilling results in the study area, first complete the sorting and statistical analysis of the rock physical properties of the study target, including density, magnetic susceptibility, resistivity, and velocity, and obtain the statistical average of the physical properties of each lithology formation; and form a bar graph and average value:
[0019]
[0020] In the formula, x and i represent the values and types of samples of different physical properties such as density, magnetic susceptibility, resistivity, and velocity, and m is the number of data tested for the physical property sample.
[0021] The multi-parameter integrated igneous reservoir lithology intelligent identification method further includes, after step 1, performing a de-distortion process on the sample parameters, performing a deviation calculation on the data, and eliminating data with a large deviation and very few samples, so as to achieve a balance and correlation between various types of input data. After step 1, when performing a de-distortion process on the sample parameters, first calculate the standard deviation of the sample:
[0022]
[0023] In the formula, M is the number of samples, x i is the measured value of sample properties, is the average value; then find the coefficient of dispersion:
[0024]
[0025] Subtract the sample property data from the first one and recalculate the dispersion coefficient to obtain the dispersion coefficient dS n-1 If we remove x i After that, dS n-1 If the difference from dS is greater than 5%, then x i Delete it as a distortion point; then repeat the above process until all sample data meet the deviation coefficient requirements. In step 2, according to the needs of geophysical measured data processing and geological-geophysical comprehensive interpretation, a template for the physical property parameters of igneous reservoir rock samples is established and coded.
[0026] In step 2, firstly, the igneous rock is normalized according to its physical property variation range. The processing method used to normalize the physical property three-body parameters is: a characteristic parameter set of the rock sample physical property data is Y = {y1, y2, y3...y i}, define the maximum and minimum values of the feature data as Y max =max{Y i} and Y min =min{Y i}, change Y i The formula for normalization to the [0,1] interval is:
[0027]
[0028] In the formula, Yi is a certain type of physical property data of the i-th core, i is the core number, and Yai is the normalized value;
[0029] YiA, the normalized value representing a certain property of the i-th rock, is projected into the range of (0.0:1.0), and then each type of property is divided into n levels according to the variation law of 1 / n, namely A(0.0:1 / n), B(1 / n:2 / n), C(2 / n:3 / n), D(3 / n:4 / n)···Z(n-1 / n:n / n), and numbered: A\B\C\D\...\Z; level n is the test data, and different values can affect the final result.
[0030] In step 3, a single correspondence between gravity, magnetic and electric anomalies and igneous rock targets is established by extracting and encoding gravity, magnetic and electric reservoir anomalies. The corresponding approach in step 2 is also adopted to normalize the gravity, magnetic and electric anomalies and assign corresponding codes. According to the variation law of 1 / n, each type of anomaly is divided into n levels, namely A (0.0: 1 / n), B (1 / n: 2 / n), C (2 / n: 3 / n), D (3 / n: 4 / n)···Z (n-1 / n: n / n), and assigned numbers: A\B\C\D\...\Z.
[0031] In step 4, by interpreting the attributes of the seismic data, seismic amplitude information, seismic phase characteristics, frequency characteristics, time slice characteristics, and coherent data characteristics reflecting the igneous rock are obtained.
[0032] In step 4, igneous rock is a special rock mass that is obviously different from normal sedimentary rock. On the seismic profile, igneous rock usually has the seismic phase characteristics of plate, arc and hillock. The internal reflection of igneous rock is manifested as weak amplitude on the seismic profile. The reflection characteristics of igneous rock on the time slice are earthworm-like, chaotic reflection, weak or medium amplitude, irregular elliptical or strip-shaped. It appears as black dots or circles on the coherent data volume slice, and is often cut by strip-shaped high coherence anomalies. Therefore, it is difficult to clearly give the properties of basalt and pyroclastic rock respectively for the seismic attributes of igneous rock, and it can only distinguish other rocks such as sedimentary rock and metamorphic rock. Therefore, the seismic attributes of igneous rock are encoded as a large category, that is, the seismic attribute characteristics of igneous rock are encoded as 1, and the seismic attribute characteristics of igneous rock are encoded as 0. The igneous rock target has k characteristics of seismic amplitude, seismic phase, frequency characteristics, time slice characteristics, and coherence characteristics, so the seismic attribute code of igneous rock is: 1-1-1···1 k , the code for those without igneous rock seismic attribute characteristics is 0-0-0···0 k ; The eigenvalue k is the test value.
[0033] In step 5, the lithology codes of the gravity, magnetic and electric anomalies of the igneous rock target are A\B\C\D\...\Z series and the igneous rock codes of the seismic attributes are 1-1-1···1 kSeries, by mapping the three-dimensional model gravity, magnetic and electric data into lithology codes, and mapping the three-dimensional seismic attribute data into lithology codes, the lithology model of the entire three-dimensional space can be obtained through the corresponding codes, which is called Model A.
[0034] In step 6, deep learning is introduced to identify these three-dimensional lithological features. The gravity, magnetic and electric anomalies extracted in step 3 and the seismic attribute data in step 4 are preprocessed by extracting Mel-frequency cepstral coefficients (MFCC) and power spectrum density (PSD). Then, feature splicing is performed and used as the input of the deep belief neural network (DBN). The DBN model is used to extract features layer by layer, train and classify, and finally a three-dimensional igneous reservoir lithological model is output, which is called model B.
[0035] In step 7, feedback analysis is performed to improve recognition accuracy, and model A and model B are compared. Based on the difference between the two, the above method is optimized through feedback, the characteristic values n and k are modified, and the lithology prediction of the encoding mode is re-performed. At the same time, new reservoir gravity, magneto-electric anomalies and seismic attribute parameters, as well as new neural network pre-learning samples, are obtained, and neural network prediction is re-performed; until the difference between model A and model B reaches the preset standard.
[0036] The intelligent identification method of igneous reservoir lithology based on multi-parameter synthesis in the present invention has the following advantages compared with the prior art: first, it fully utilizes all relevant information, while conventional methods do not fully explore and utilize gravity, magnetic, electric and seismic information, and only utilize gravity, magnetic and electric information; second, when applying machine learning methods, the samples have identifiable characteristics, while conventional methods either only use machine learning such as neural networks for identification, or only use fuzzy coding for identification when identifying lithology. The intelligent identification method of igneous reservoir lithology based on multi-parameter synthesis in the present invention can give full play to the advantages and specialties of various geophysical methods, and can evaluate oil and gas targets efficiently and effectively, which is of great significance to accelerating the pace of oil exploration. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] Figure 1 It is a flowchart of a specific embodiment of the method for intelligent identification of igneous reservoir lithology based on multi-parameter integration of the present invention;
[0038] Figure 2 This is a statistical analysis diagram of rock properties in a specific embodiment of the present invention. DETAILED DESCRIPTION
[0039] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.
[0040] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations and / or combinations thereof.
[0041] The present invention relates to a method for identifying igneous reservoir lithology by geophysical prospecting, specifically a multi-parameter comprehensive intelligent igneous reservoir lithology identification method, including the following specific steps: statistical analysis of physical property data in the study area; de-distortion processing of sample parameters; establishment of a rock sample physical property parameter coding template; extraction and coding of gravity, magneto-electric reservoir anomalies to establish a single corresponding relationship between gravity, magneto-electric anomalies and igneous rock targets; seismic attribute extraction and coding; igneous reservoir target lithology identification; neural network prediction of igneous reservoir lithology; feedback analysis to improve identification accuracy. The method can effectively identify igneous reservoir lithology, thereby effectively reducing oil and gas exploration risks.
[0042] like Figure 1 As shown, Figure 1 The flowchart of the igneous reservoir lithology intelligent identification method based on multi-parameter synthesis of the present invention. The igneous reservoir lithology intelligent identification method based on multi-parameter synthesis includes the following steps:
[0043] 1) Statistical analysis of physical property data of the study area
[0044] Using the physical property data and drilling results in the study area, we first complete the sorting and statistical analysis of the rock physical properties (density, magnetic susceptibility, resistivity, velocity, etc.) of the research target, and obtain the statistical average of the physical properties of each lithology and stratum; and form a bar graph and average value
[0045]
[0046] In the formula, P and i represent the values and types of different physical property samples such as density, magnetic susceptibility, resistivity, and velocity, and m is the number of data tested for the physical property sample.
[0047] like Figure 2 This is a statistical chart of the resistivity of various lithologies in the Carboniferous System, among which basalt and andesite have relatively large resistivity, followed by volcanic breccia, and tuff and sandstone have relatively small resistivity.
[0048] 2) De-distortion of sample parameters
[0049] In order to avoid the influence of singular points in data calculation during model construction and identification, it is necessary to calculate the deviation of the data and eliminate data with large deviations and very few samples, so as to achieve balance and correlation between various types of input data. First, calculate the standard deviation of the sample:
[0050]
[0051] In the formula, M is the number of samples, xi is the measured value of sample properties, is the average value.
[0052] Then find the coefficient of dispersion:
[0053]
[0054] Subtract the sample property data from the first one and recalculate the dispersion coefficient to obtain the dispersion coefficient dS n-1 If we remove x i After that, dS n-1 If the difference from dS is greater than 5%, then x i Delete it as a distortion point. Then repeat the above process until all sample data meet the deviation coefficient requirements.
[0055] 3) Establish a rock sample physical property parameter coding template
[0056] According to the needs of geophysical measured data processing and geological-geophysical comprehensive interpretation,
[0057] The physical property parameter template of the igneous reservoir rock sample is encoded. First, it is normalized according to the range of igneous rock physical property changes: the processing method used to normalize the physical property three-body parameters is: a characteristic parameter set of the rock sample physical property data is Y = {y1, y2, y3... y i}, define the maximum and minimum values of the feature data as Y max =max{Y i} and Y min =min{Y i}, change Y i The formula for normalization to the [0,1] interval is:
[0058]
[0059] In the formula, Yi is a certain type of physical property data of the i-th core, i is the core number, and Y i A is the normalized value.
[0060] Y i AThat is, the normalized value representing a certain property of the i-th rock, such as the resistivity of several rocks, is projected into the range of (0.0:1.0). Then, according to the law of 1 / n, each type of property is divided into n levels, namely A(0.0:1 / n), B(1 / n:2 / n), C(2 / n:3 / n), D(3 / n:4 / n)···Z(n-1 / n:n / n), and numbered: A\B\C\D\...\Z. Level n is the test data, and different values can affect the final result.
[0061] For example, the physical properties of basalt in igneous rocks are high resistance, high density, strong magnetism and high speed, four high characteristics, and each type of physical property is normalized and graded. Here, the highest level is coded as Z, and we code it as: ZZZZ. The physical properties of pyroclastic rocks are low resistance, low density, low magnetism and low speed, four low characteristics, and here the lowest level is coded as A, and we code it as: AAAA.
[0062] 4) By extracting and encoding gravity, magneto-electric reservoir anomalies, a single correspondence between gravity, magneto-electric anomalies and igneous rock targets is established. The corresponding approach in (3) is also adopted to normalize the gravity, magneto-electric anomalies and assign corresponding codes. According to the 1 / n variation law, each type of anomaly is divided into n levels, namely A (0.0: 1 / n), B (1 / n: 2 / n), C (2 / n: 3 / n), D (3 / n: 4 / n)...Z (n-1 / n: n / n), and assigned numbers: A\B\C\D\...\Z.
[0063] 5) Seismic attribute extraction and coding
[0064] By interpreting the attributes of seismic data, we can obtain seismic amplitude information, seismic phase characteristics, frequency characteristics, time slice characteristics, coherent data characteristics, etc. reflecting igneous rocks. Igneous rocks are a special rock mass that is significantly different from normal sedimentary rocks. On seismic sections, igneous rocks usually have seismic phase characteristics such as "plate-like", "arc-like", and "hill-like". The internal reflection of igneous rocks is shown as weak amplitude on seismic sections. The reflection characteristics of igneous rocks on time slices are earthworm-like, chaotic reflections, weak or medium amplitudes, and irregular elliptical or strip-like. They appear as black dots or circles on coherent data volume slices, often cut by strip-like high coherence anomalies. Therefore, it is difficult to clearly give the seismic attributes of igneous rocks, such as basalts and pyroclastic rocks, and can only distinguish other rocks such as sedimentary rocks and metamorphic rocks. Therefore, the seismic attributes of igneous rocks are coded as a large category, that is, the seismic attribute characteristics of igneous rocks are coded as 1, and the seismic attribute characteristics of igneous rocks are coded as 0. The igneous rock target has k features such as seismic amplitude, seismic phase, frequency feature, time slice feature, and coherence feature. Therefore, the seismic attribute code given to the igneous rock is: 1-1-1···1 k, the code for those without igneous rock seismic attribute characteristics is 0-0-0···0 k The eigenvalue k is a test value, and some features may not have an obvious correspondence.
[0065] 6) Identification of target lithology of igneous reservoirs
[0066] Lithology codes of gravity, magneto-electric anomalies of igneous rock targets A\B\C\D\...\Z series and igneous rock codes of seismic attributes 1-1-1···1 k Series, by mapping the three-dimensional model gravity, magnetic and electric data into lithology codes, and mapping the three-dimensional seismic attribute data into lithology codes, the lithology model of the entire three-dimensional space (called model A) can be obtained through the corresponding codes.
[0067] 7) Neural network prediction of igneous reservoir lithology
[0068] Deep learning is then introduced to identify these three-dimensional lithological features. The gravity, magnetic and electric anomalies extracted in (4) and the seismic attribute data in (5) are preprocessed by extracting Mel-frequency cepstral coefficients (MFCC) and power spectral density (PSD). The features are then concatenated and used as the input of a deep belief neural network (DBN). The DBN model is used to extract features layer by layer, train and classify them, and finally a three-dimensional igneous reservoir lithological model (referred to as model B) is output.
[0069] 8) Feedback analysis improves recognition accuracy
[0070] Compare model A and model B. According to the difference between the two, feedback optimize the above method, modify the above eigenvalues n and k, and re-predict the lithology of the coding mode. At the same time, obtain new reservoir gravity, magnetoelectric anomaly and seismic attribute parameters, as well as new neural network pre-learning samples, and re-predict the neural network. Until the difference between model A and model B reaches the preset standard.
[0071] Example 1 (Only well logging physical property data)
[0072] 1) Statistical analysis of physical property data of the study area
[0073] Using the drilling and logging results in the study area, we first complete the collation and statistical analysis of the physical properties obtained from the well logging of the research target (density, resistivity, velocity, the number of physical properties obtained from the logging data is generally the same, and the logging generally has no magnetism), obtain the statistical average of the physical properties of each formation, and form a bar chart and average value
[0074]
[0075]
[0076]
[0077] A, B, and C represent density, resistivity, and velocity, respectively.
[0078] 2) De-distortion of sample parameters
[0079] In order to avoid the influence of singular points in data calculation during model construction and identification, it is necessary to calculate the deviation of the data and eliminate data with large deviations and very few samples, so as to achieve balance and correlation between various types of input data. First, calculate the standard deviation of the sample:
[0080]
[0081] In the formula, x i is the measured value of sample properties, is the average value.
[0082] Then find the coefficient of dispersion:
[0083]
[0084] Subtract the sample property data from the first one and recalculate the dispersion coefficient to obtain the dispersion coefficient dS n-1 If we remove x i After that, dS n-1 If the difference from dS is greater than 5%, then x i Delete it as a distortion point. Then repeat the above process until all sample data meet the deviation coefficient requirements.
[0085] 3) Establish a rock sample physical property parameter coding template
[0086] According to the needs of geophysical measured data processing and geological-geophysical comprehensive interpretation, the physical property parameter template of igneous reservoir rock samples is established and coded. First, the igneous rock physical property is normalized according to the range of its physical property variation: the processing method used to normalize the physical property parameters is: the resistivity characteristic parameter set of the rock sample physical property data is Y = {y1, y2, y3... y 1500}, define the maximum and minimum values of the feature data as Y max =max{Y i} and Y min =min{Y i}, change Y i The formula for normalization to the [0,1] interval is:
[0087]
[0088] In the formula, Yi is a certain type of physical property data of the i-th core, i is the core number, and Y iA is the normalized value.
[0089] Y i A That is, the normalized value representing a certain physical property of the i-th rock, such as the resistivity of several rocks, is projected into the range of (0.0:1.0).
[0090] Then, according to the 1 / 15 change rule, each type of physical property is divided into 15 levels, namely A (0.0: 1 / 15), B (1 / 15: 2 / 15), C (2 / 15: 3 / 15), D (3 / 15: 4 / 15)...J (n-1 / n: n / n), and numbered: A\B\C\D\...\P. Level 15 is test data, and different values can affect the final result.
[0091] For example, the physical properties of basalt in igneous rocks are high resistance, high density, strong magnetism and high speed, three high characteristics, and each type of physical property is normalized and graded. Here, the highest level is coded as 15, and we code it as: PPP. The physical properties of pyroclastic rocks are low resistance, low density and low speed, three low characteristics, and here the lowest level is coded as A, and we code it as: AAA.
[0092] 4) By extracting gravity, magnetic and electric reservoir anomalies, a single corresponding relationship between gravity, magnetic and electric anomalies and igneous rock targets is established. The corresponding approach in 3) is also adopted to normalize the gravity, magnetic and electric anomalies and assign corresponding codes. According to the change law of 1 / 15, each type of anomaly is divided into 15 levels, namely A (0.0: 1 / n), B (1 / n: 2 / n), C (2 / n: 3 / n), D (3 / n: 4 / n)···Z (n-1 / n: n / n), and assigned numbers: A\B\C\D\...\P.
[0093] 5) Seismic attribute extraction and coding
[0094] By interpreting the attributes of seismic data, we can obtain seismic amplitude information, seismic phase characteristics, frequency characteristics, time slice characteristics, coherent data characteristics, etc. reflecting igneous rocks. Igneous rocks are a special rock mass that is significantly different from normal sedimentary rocks. On seismic sections, igneous rocks usually have seismic phase characteristics such as "plate-like", "arc-like", and "hill-like". The internal reflection of igneous rocks is shown as weak amplitude on seismic sections. The reflection characteristics of igneous rocks on time slices are earthworm-like, chaotic reflections, weak or medium amplitudes, and irregular elliptical or strip-like. They appear as black dots or circles on coherent data volume slices, often cut by strip-like high coherence anomalies. Therefore, it is difficult to clearly give the seismic attributes of igneous rocks, such as basalts and pyroclastic rocks, and can only distinguish other rocks such as sedimentary rocks and metamorphic rocks. Therefore, the seismic attributes of igneous rocks are coded as a large category, that is, the seismic attribute characteristics of igneous rocks are coded as 1, and the seismic attribute characteristics of igneous rocks are coded as 0. Igneous rock targets have five characteristics, namely, seismic amplitude, seismic phase, frequency characteristics, time slice characteristics, and coherence characteristics. Therefore, the seismic attribute code given to igneous rock is: 1-1-1···15, and the code without igneous rock seismic attribute characteristics is 0-0-0···05. The eigenvalue number 5 is a test value, and some characteristics may not have an obvious corresponding relationship, which needs to be determined through feedback testing.
[0095] 6) Identification of target lithology of igneous reservoirs
[0096] For the lithology codes A\B\C\D\...\P series of gravity, magnetic and electric anomalies of igneous rock targets and the igneous rock codes 1-1-1···15 series of seismic attributes, the lithology model of the entire three-dimensional space (called model A) can be obtained through the corresponding codes by mapping the three-dimensional model gravity, magnetic and electric data to lithology codes, and mapping the three-dimensional seismic attribute data to lithology codes.
[0097] 7) Neural network prediction of igneous reservoir lithology
[0098] Deep learning is then introduced to identify these three-dimensional lithological characteristics. The gravity, magnetic and electric anomalies and seismic attribute data are preprocessed by extracting Mel-frequency cepstral coefficients (MFCC) and power spectral density (PSD). Then, the features are spliced and used as the input of the deep belief neural network (DBN). The DBN model is used to extract features layer by layer, train and classify, and finally output a three-dimensional igneous reservoir lithological model (called model B).
[0099] 8) Feedback analysis improves recognition accuracy
[0100] Compare model A and model B. According to the difference between the two, feedback optimize the above method, modify the above eigenvalues 15 and 5, and re-predict the lithology. At the same time, add neural network pre-learning samples and re-predict the neural network. Until the difference between model A and model B is relatively small.
[0101] The specific implementation methods described above provide a specific explanation of the purpose and technical solutions of the present invention. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
[0102] Example 2 (Only rock outcrop physical property data)
[0103] 1) Statistical analysis of physical property data of the study area
[0104] Using the physical property data and drilling results in the study area, we first complete the sorting and statistical analysis of the rock physical properties (density, magnetic susceptibility, resistivity, etc.) of the research target, obtain the statistical average of the physical properties of each lithology and stratum, and form a bar chart and average value.
[0105]
[0106]
[0107]
[0108] A, B, and C represent density, magnetic susceptibility, and resistivity, respectively.
[0109] 2) De-distortion of sample parameters
[0110] In order to avoid the influence of singular points in data calculation during model construction and identification, it is necessary to calculate the deviation of the data and eliminate data with large deviations and very few samples, so as to achieve balance and correlation between various types of input data. First, calculate the standard deviation of the sample:
[0111]
[0112] In the formula, x i is the measured value of sample properties, is the average value.
[0113] Then find the coefficient of dispersion:
[0114]
[0115] Subtract the sample property data from the first one and recalculate the dispersion coefficient to obtain the dispersion coefficient dS n-1If we remove x i After that, dS n-1 If the difference from dS is greater than 5%, then x i Delete it as a distortion point. Then repeat the above process until all sample data meet the deviation coefficient requirements.
[0116] 3) Establish a rock sample physical property parameter coding template
[0117] According to the needs of geophysical measured data processing and geological-geophysical comprehensive interpretation, the physical property parameter template of igneous reservoir rock samples is established and coded. First, the igneous rock physical property is normalized according to the range of its physical property variation: the processing method used to normalize the physical property parameters is: the resistivity characteristic parameter set of the rock sample physical property data is Y = {y1, y2, y3... y 413}, define the maximum and minimum values of the feature data as Y max =max{Y i} and Y min =min{Y i}, change Y i The formula for normalization to the [0,1] interval is:
[0118]
[0119] In the formula, Yi is a certain type of physical property data of the i-th core, i is the core number, and Y i A is the normalized value.
[0120] Y i A That is, the normalized value representing a certain physical property of the i-th rock, such as the resistivity of several rocks, is projected into the range of (0.0:1.0).
[0121] Then, according to the 1 / 9 change rule, each type of physical property is divided into 9 levels, namely A (0.0: 1 / 9), B (1 / 9: 2 / 9), C (2 / 9: 3 / 9), D (3 / 9: 4 / 9)...J (n-1 / n: n / n), and numbered: A\B\C\D\...\J. Level 9 is the test data, and different values can affect the final result.
[0122] For example, the physical properties of basalt in igneous rocks are high resistivity, high density, strong magnetism, and high speed, four high characteristics, and each type of physical property is normalized and graded. Here, the highest level is coded as 9, and we code it as: JJJ. The physical properties of pyroclastic rocks are low resistivity, low density, and low magnetism, three low characteristics, and here the lowest level is coded as A, and we code it as: AAA.
[0123] 4) By extracting gravity, magnetic and electric reservoir anomalies, a single corresponding relationship between gravity, magnetic and electric anomalies and igneous rock targets is established. The corresponding approach in 3) is also adopted to normalize the gravity, magnetic and electric anomalies and assign corresponding codes. According to the change law of 1 / 9, each type of anomaly is divided into 9 levels, namely A (0.0: 1 / n), B (1 / n: 2 / n), C (2 / n: 3 / n), D (3 / n: 4 / n)···Z (n-1 / n: n / n), and numbered: A\B\C\D\...\J.
[0124] 5) Seismic attribute extraction and coding
[0125] By interpreting the attributes of seismic data, we can obtain seismic amplitude information, seismic phase characteristics, frequency characteristics, time slice characteristics, coherent data characteristics, etc. reflecting igneous rocks. Igneous rocks are a special rock mass that is significantly different from normal sedimentary rocks. On seismic sections, igneous rocks usually have seismic phase characteristics such as "plate-like", "arc-like", and "hill-like". The internal reflection of igneous rocks is shown as weak amplitude on seismic sections. The reflection characteristics of igneous rocks on time slices are earthworm-like, chaotic reflections, weak or medium amplitudes, and irregular elliptical or strip-like. They appear as black dots or circles on coherent data volume slices, often cut by strip-like high coherence anomalies. Therefore, it is difficult to clearly give the seismic attributes of igneous rocks, such as basalts and pyroclastic rocks, and can only distinguish other rocks such as sedimentary rocks and metamorphic rocks. Therefore, the seismic attributes of igneous rocks are coded as a large category, that is, the seismic attribute characteristics of igneous rocks are coded as 1, and the seismic attribute characteristics of igneous rocks are coded as 0. Igneous rock targets have six characteristics, including seismic amplitude, seismic phase, frequency characteristics, time slice characteristics, and coherence characteristics. Therefore, the seismic attribute code given to igneous rock is: 1-1-1···16, and the code without igneous rock seismic attribute characteristics is 0-0-0···06. The eigenvalue number 6 is a test value, and some characteristics may not have an obvious corresponding relationship, which needs to be determined through feedback testing.
[0126] 6) Identification of target lithology of igneous reservoirs
[0127] For the lithology codes A\B\C\D\...\J series of gravity, magnetic and electric anomalies of igneous rock targets and the igneous rock codes 1-1-1···16 series of seismic attributes, the lithology model of the entire three-dimensional space (called model A) can be obtained through the corresponding codes by mapping the three-dimensional model gravity, magnetic and electric data to lithology codes, and mapping the three-dimensional seismic attribute data to lithology codes.
[0128] 7) Neural network prediction of igneous reservoir lithology
[0129] Deep learning is then introduced to identify these three-dimensional lithological characteristics. The gravity, magnetic and electric anomalies and seismic attribute data are preprocessed by extracting Mel-frequency cepstral coefficients (MFCC) and power spectral density (PSD). Then, the features are spliced and used as the input of the deep belief neural network (DBN). The DBN model is used to extract features layer by layer, train and classify, and finally output a three-dimensional igneous reservoir lithological model (called model B).
[0130] 8) Feedback analysis improves recognition accuracy
[0131] Compare model A and model B. According to the difference between the two, feedback optimize the above method, modify the above eigenvalues 9 and 6, and re-predict the lithology. At the same time, add neural network pre-learning samples and re-predict the neural network. Until the difference between model A and model B is relatively small.
[0132] The specific implementation methods described above provide a specific explanation of the purpose and technical solutions of the present invention. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
[0133] Example 3 (with well logging and surface outcrop physical property data)
[0134] 1) Statistical analysis of physical property data of the study area
[0135] Using the physical property data and drilling results in the study area, we first complete the sorting and statistical analysis of the rock physical properties (density, magnetic susceptibility, resistivity, velocity, etc.) of the research target, obtain the statistical average of the physical properties of each lithology and stratum, and form a bar chart and average value.
[0136]
[0137]
[0138]
[0139]
[0140] A, B, C, and D represent density, magnetic susceptibility, resistivity, and velocity, respectively, and the physical property data N are 1515, 65, 1521, and 1515, respectively.
[0141] 2) De-distortion of sample parameters
[0142] In order to avoid the influence of singular points in data calculation during model construction and identification, it is necessary to calculate the deviation of the data and eliminate data with large deviations and very few samples, so as to achieve balance and correlation between various types of input data. First, calculate the standard deviation of the sample:
[0143]
[0144] In the formula, x i is the measured value of sample properties, is the average value.
[0145] Then find the coefficient of dispersion:
[0146]
[0147] Subtract the sample property data from the first one and recalculate the dispersion coefficient to obtain the dispersion coefficient dS n-1 If we remove x i After that, dS n-1 If the difference from dS is greater than 5%, then x i Delete it as a distortion point. Then repeat the above process until all sample data meet the deviation coefficient requirements.
[0148] 3) Establish a rock sample physical property parameter coding template
[0149] According to the needs of geophysical measured data processing and geological-geophysical comprehensive interpretation, the physical property parameter template of igneous reservoir rock samples is established and coded. First, the igneous rock physical property is normalized according to the range of its physical property variation: the processing method used to normalize the physical property parameters is: the resistivity characteristic parameter set of the rock sample physical property data is Y = {y1, y2, y3... y N}, define the maximum and minimum values of the feature data as Y max =max{Y i} and Y min =min{Y i}, change Y i The formula for normalization to the [0,1] interval is:
[0150]
[0151] In the formula, Yi is a certain type of physical property data of the i-th core, i is the core number, and Y i A is the normalized value.
[0152] Y i A That is, the normalized value representing a certain physical property of the i-th rock, such as the resistivity of several rocks, is projected into the range of (0.0:1.0).
[0153] Then, according to the variation law of 1 / [N / 10], each type of physical property is divided into [N / 10] levels, namely A(0.0:1 / / [N / 10]), B(1 / / [N / 10]:2 / / [N / 10]), C(2 / / [N / 10]:3 / / [N / 10]), D(3 / / [N / 10]:4 / / [N / 10])···J(n-1 / / [N / 10]:n / / [N / 10]), and numbered: A\B\C\D\...\J. Level / [N / 10] is the test data, and different values can affect the final result.
[0154] For example, the physical properties of basalt in igneous rocks are high resistivity, high density, strong magnetism and high speed, four high characteristics, and each type of physical property is normalized and graded. Here, the highest level is coded as / [N / 10], and we code it as: QQQQ. The physical properties of pyroclastic rocks are low resistivity, low density, low magnetism and low speed, four low characteristics, and here the lowest level is coded as A, and we code it as: AAAA.
[0155] 4) By extracting gravity, magnetic and electric reservoir anomalies, a single corresponding relationship between gravity, magnetic and electric anomalies and igneous rock targets is established. The corresponding approach in 3) is also adopted to normalize the gravity, magnetic and electric anomalies and assign corresponding codes. According to the variation law of 1 / / [N / 10], each type of anomaly is divided into / [N / 10] levels, namely A(0.0:1 / n), B(1 / n:2 / n), C(2 / n:3 / n), D(3 / n:4 / n)···Z(n-1 / n:n / n), and assigned numbers: A\B\C\D\...\Q.
[0156] 5) Seismic attribute extraction and coding
[0157] By interpreting the attributes of seismic data, we can obtain seismic amplitude information, seismic phase characteristics, frequency characteristics, time slice characteristics, coherent data characteristics, etc. reflecting igneous rocks. Igneous rocks are a special rock mass that is significantly different from normal sedimentary rocks. On seismic sections, igneous rocks usually have seismic phase characteristics such as "plate-like", "arc-like", and "hill-like". The internal reflection of igneous rocks is shown as weak amplitude on seismic sections. The reflection characteristics of igneous rocks on time slices are earthworm-like, chaotic reflections, weak or medium amplitudes, and irregular elliptical or strip-like. They appear as black dots or circles on coherent data volume slices, often cut by strip-like high coherence anomalies. Therefore, it is difficult to clearly give the seismic attributes of igneous rocks, such as basalts and pyroclastic rocks, and can only distinguish other rocks such as sedimentary rocks and metamorphic rocks. Therefore, the seismic attributes of igneous rocks are coded as a large category, that is, the seismic attribute characteristics of igneous rocks are coded as 1, and the seismic attribute characteristics of igneous rocks are coded as 0. Igneous rock targets have six characteristics, including seismic amplitude, seismic phase, frequency characteristics, time slice characteristics, and coherence characteristics. Therefore, the seismic attribute code given to igneous rock is: 1-1-1···16, and the code without igneous rock seismic attribute characteristics is 0-0-0···06. The eigenvalue number 6 is a test value, and some characteristics may not have an obvious corresponding relationship, which needs to be determined through feedback testing.
[0158] 6) Identification of target lithology of igneous reservoirs
[0159] For the lithology codes A\B\C\D\...\Q series of gravity, magnetic and electric anomalies of igneous rock targets and the igneous rock codes 1-1-1···16 series of seismic attributes, the lithology model of the entire three-dimensional space (called model A) can be obtained through the corresponding codes by mapping the three-dimensional model gravity, magnetic and electric data to lithology codes, and mapping the three-dimensional seismic attribute data to lithology codes.
[0160] 7) Neural network prediction of igneous reservoir lithology
[0161] Deep learning is then introduced to identify these three-dimensional lithological characteristics. The gravity, magnetic and electric anomalies and seismic attribute data are preprocessed by extracting Mel-frequency cepstral coefficients (MFCC) and power spectral density (PSD). Then, the features are spliced and used as the input of the deep belief neural network (DBN). The DBN model is used to extract features layer by layer, train and classify, and finally the three-dimensional igneous reservoir lithological model (called model B) is output.
[0162] 8) Feedback analysis improves recognition accuracy
[0163] Compare model A and model B. According to the difference between the two, feedback optimize the above method, modify the above characteristic values [N / 10] and 6, and re-predict the lithology. At the same time, increase the neural network pre-learning samples and re-predict the neural network. Until the difference between model A and model B is relatively small.
[0164] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions recorded in the aforementioned embodiments or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
[0165] Except for the technical features described in the specification, all other technical features are known technologies to those skilled in the art.
Claims
1. Intelligent identification method of igneous reservoir lithology based on multi-parameter integration, It is characterized in that The intelligent identification method of igneous reservoir lithology based on multi-parameter integration includes: Step 1: Statistical analysis of the physical property data of the study area; Step 2, establishing a rock sample physical property parameter coding template; Step 3, by extracting and encoding gravity, magneto-electric reservoir anomalies, a single corresponding relationship between gravity, magneto-electric anomalies and igneous rock targets is established; Step 4, extracting and encoding seismic attributes; Step 5, identifying the target lithology of the igneous reservoir and obtaining model A; Step 6, using steps 3 and 4 as learning samples to perform neural network prediction of igneous reservoir lithology to obtain model B; Step 7, compare models A and B. If the error is greater than the preset standard, return to step 2. If the error is less than the preset standard, evaluate both and output the results.
2. The method for intelligent identification of igneous reservoir lithology based on multi-parameter synthesis according to claim 1, It is characterized in that In step 1, using the physical property data and drilling results in the study area, first complete the sorting and statistical analysis of the rock physical properties of the study target, including density, magnetic susceptibility, resistivity, and velocity, and obtain the statistical average of the physical properties of each lithology formation; and form a bar graph and average value: In the formula, x and i represent the values and types of samples of different physical properties such as density, magnetic susceptibility, resistivity, and velocity, and m is the number of data tested for the physical property sample.
3. The method for intelligent identification of igneous reservoir lithology based on multi-parameter synthesis according to claim 1, It is characterized in that The method for intelligent identification of igneous reservoir lithology based on multi-parameter integration also includes, after step 1, performing de-distortion processing on sample parameters, performing deviation calculation on data, and eliminating data with large deviations and very few samples, so as to achieve balance and correlation between various types of input data.
4. The method for intelligent identification of igneous reservoir lithology based on multi-parameter synthesis according to claim 3, It is characterized in that After step 1, when performing the sample parameter dedistortion processing, first calculate the standard deviation of the sample: In the formula, M is the number of samples, x i is the measured value of sample properties, is the average value; Then find the coefficient of dispersion: Subtract the sample property data from the first one and recalculate the dispersion coefficient to obtain the dispersion coefficient dS n-1 If we remove x i After that, dS n-1 If the difference from dS is greater than 5%, then x i Delete it as a distortion point; then repeat the above process until all sample data meet the deviation coefficient requirements.
5. The method for intelligent identification of igneous reservoir lithology based on multi-parameter synthesis according to claim 1, It is characterized in that In step 2, according to the needs of geophysical measured data processing and geological-geophysical integrated interpretation, a template for the physical property parameters of igneous reservoir rock samples is established and coded.
6. The method for intelligent identification of igneous reservoir lithology based on multi-parameter synthesis according to claim 5, It is characterized in that In step 2, firstly, the igneous rock is normalized according to its physical property variation range. The processing method used to normalize the physical property three-body parameters is: a characteristic parameter set of the rock sample physical property data is Y = {y 1 ,y 2 ,y 3 …y i }, define the maximum and minimum values of the feature data as Y max =max{Y i } and Y min =min{Y i }, change Y i The formula for normalization to the [0,1] interval is: In the formula, Yi is a certain type of physical property data of the i-th core, i is the core number, is the normalized value; That is, the normalized value characterizing a certain physical property of the i-th rock is projected into the range of (0.0:1.0), and then each type of physical property is divided into n levels according to the variation law of 1 / n, namely A(0.0:1 / n), B(1 / n:2 / n), C(2 / n:3 / n), D(3 / n:4 / n)···Z(n-1 / n:n / n), and numbered: A\B\C\D\...\Z; level n is the test data, and different values can affect the final result.
7. The method for intelligent identification of igneous reservoir lithology based on multi-parameter synthesis according to claim 6, It is characterized in that In step 3, a single correspondence between gravity, magnetic and electric anomalies and igneous rock targets is established by extracting and encoding gravity, magnetic and electric reservoir anomalies. The corresponding approach in step 2 is also adopted to normalize the gravity, magnetic and electric anomalies and assign corresponding codes. According to the variation law of 1 / n, each type of anomaly is divided into n levels, namely A (0.0: 1 / n), B (1 / n: 2 / n), C (2 / n: 3 / n), D (3 / n: 4 / n)···Z (n-1 / n: n / n), and assigned numbers: A\B\C\D\...\Z.
8. The method for intelligent identification of igneous reservoir lithology based on multi-parameter synthesis according to claim 1, It is characterized in that In step 4, by interpreting the attributes of the seismic data, seismic amplitude information, seismic phase characteristics, frequency characteristics, time slice characteristics, and coherent data characteristics reflecting the igneous rock are obtained.
9. The method for intelligent identification of igneous reservoir lithology based on multi-parameter synthesis according to claim 8, It is characterized in that In step 4, igneous rock is a special rock mass that is obviously different from normal sedimentary rock. On the seismic profile, igneous rock usually has the seismic phase characteristics of plate, arc and hillock. The internal reflection of igneous rock is manifested as weak amplitude on the seismic profile. The reflection characteristics of igneous rock on the time slice are earthworm-like, chaotic reflection, weak or medium amplitude, irregular elliptical or strip-shaped. It appears as black dots or circles on the coherent data volume slice, and is often cut by strip-shaped high coherence anomalies. Therefore, it is difficult to clearly give the properties of basalt and pyroclastic rock respectively for the seismic attributes of igneous rock, and it can only distinguish other rocks such as sedimentary rock and metamorphic rock. Therefore, the seismic attributes of igneous rock are encoded as a large category, that is, the seismic attribute characteristics of igneous rock are encoded as 1, and the seismic attribute characteristics of igneous rock are encoded as 0. The igneous rock target has k characteristics of seismic amplitude, seismic phase, frequency characteristics, time slice characteristics, and coherence characteristics, so the seismic attribute code of igneous rock is: 1-1-1···1 k , the code for those without igneous rock seismic attribute characteristics is 0-0-0···0 k ; The eigenvalue k is the test value.
10. The method for intelligent identification of igneous reservoir lithology based on multi-parameter synthesis according to claim 9, It is characterized in that In step 5, the lithology codes of the gravity, magnetic and electric anomalies of the igneous rock target are A\B\C\D\...\Z series and the igneous rock codes of the seismic attributes are 1-1-1···1 k Series, by mapping the three-dimensional model gravity, magnetic and electric data into lithology codes, and mapping the three-dimensional seismic attribute data into lithology codes, the lithology model of the entire three-dimensional space can be obtained through the corresponding codes, which is called Model A.
11. The method for intelligently identifying lithology of igneous reservoirs based on multi-parameter synthesis according to claim 10, It is characterized in that In step 6, deep learning is introduced to identify these three-dimensional lithological features. The gravity, magnetic and electric anomalies extracted in step 3 and the seismic attribute data in step 4 are preprocessed by extracting Mel-frequency cepstral coefficients (MFCC) and power spectrum density (PSD). Then, feature splicing is performed and used as the input of the deep belief neural network (DBN). The DBN model is used to extract features layer by layer, train and classify, and finally a three-dimensional igneous reservoir lithological model is output, which is called model B.
12. The method for intelligent identification of igneous reservoir lithology based on multi-parameter synthesis according to claim 11, It is characterized in that In step 7, feedback analysis is performed to improve recognition accuracy, and model A and model B are compared. Based on the difference between the two, the above method is optimized through feedback, the characteristic values n and k are modified, and the lithology prediction of the encoding mode is re-performed. At the same time, new reservoir gravity, magneto-electric anomalies and seismic attribute parameters, as well as new neural network pre-learning samples, are obtained, and neural network prediction is re-performed; until the difference between model A and model B reaches the preset standard.
