Lake facies shale ancient redox recovery method based on machine learning method

Abstract

The invention relates to the technical field of unconventional oil-gas exploration, in particular to a lacustrine facies shale ancient redox recovery method based on a machine learning method, which comprises the following steps: sample collection and primary treatment; element composition analysis; preprocessing and cleaning data; screening trace elements; pCA (principal component analysis): performing PCA on element concentration screening data of the rock core sample, and extracting main components, which can most explain changes of ancient oxidation-reduction conditions, in all elements of the rock core sample through a dimension reduction technology; and recovery of the ancient redox conditions: based on the main components obtained by PCA, rebuilding the ancient redox conditions of the target lacustrine facies shale in combination with the score change trend of the main components. According to the method, a machine learning method, especially a principal component analysis technology, is introduced to restore the ancient oxidation-reduction condition of the lacustrine facies shale, so that high automation and systematization of an analysis process are realized, and the accuracy and reliability of an analysis result are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of unconventional oil and gas exploration, and in particular to a lacustrine shale paleo-oxidation-reduction recovery method based on a machine learning method. Background Art

[0002] Existing sedimentary paleo-redox restoration methods mainly rely on specific geochemical indicators, such as U, V, Mo, U / Th ratio, V / (V+Ni) ratio, Ce anomaly and Mo isotope, etc. However, the effectiveness of these indicators in different geological backgrounds is limited, especially in areas with significant exogenous debris input and burial diagenesis. Some trace element indicators may fail, resulting in inaccurate restoration results. In addition, when selecting indicators suitable for specific study areas, existing methods are highly complex and easily affected by subjective judgment. How to extract key elements from many indicators remains a challenge. Traditional methods also require a lot of manual operations, such as statistical analysis of strawberry pyrite. This method is not only labor-intensive and costly, but also prone to human errors, resulting in poor consistency and repeatability of results.

[0003] Therefore, current sedimentary paleo-redox restoration methods have the disadvantages of large workload, strong subjectivity, and poor adaptability. Summary of the invention

[0004] The purpose of the present invention is to provide a lacustrine shale paleo-redox restoration method based on a machine learning method to solve the technical problems of large workload, strong subjectivity, and poor adaptability in the prior art.

[0005] In order to solve the above technical problems, the present invention specifically provides the following technical solutions: A method for paleo-redox restoration of lacustrine shale based on a machine learning method comprises the following steps: Step 1, sample collection and preliminary processing: First, core samples are collected from different depths and multiple geological layers in the target lacustrine shale layer, and then all core samples are cleaned and leveled to remove impurities on the surface of the core samples, and the geological layer and depth information of the core samples are recorded; Step 2, elemental composition analysis: using a portable energy dispersive X-ray fluorescence spectrometer pXRF to perform elemental analysis on the core sample obtained after the treatment in step 1, quantitatively determine the concentration of major elements and trace elements in the core sample, and obtain element concentration data of the core sample; Step 3, data preprocessing and cleaning: preprocessing the element concentration data of the core sample obtained after the processing in step 2 to obtain element concentration preprocessing data of the core sample; Step 4, trace element screening: using the correlation analysis method to evaluate the correlation between the elements Ti and Zr indicating terrigenous debris and other elements in the element concentration preprocessing data of the core sample obtained after the processing in step 3, according to the correlation results, the element concentration data that is easily affected by the debris flux is excluded from the element concentration preprocessing data to obtain the element concentration screening data of the core sample; Step 5, principal component analysis (PCA): perform principal component analysis (PCA) on the element concentration screening data of the core samples, and extract the main components that can best explain the changes in ancient redox conditions from all elements of the core samples through dimensionality reduction technology; Step 6, restoration of ancient redox conditions: based on the main components obtained by PCA and combined with the score change trend of the main components, the ancient redox conditions of the target lacustrine shale are reconstructed.

[0006] As a preferred solution of the present invention, the method for preprocessing the element concentration data of the core sample obtained after the processing in step 2 includes: First, for element concentration data below the method detection limit (MDL), its concentration value is set to the method detection limit (MDL) to ensure the integrity of the data; Then, the missing data items in the element concentration data of the core samples were excluded to ensure the data validity of the subsequent analysis; Finally, by calculating the mean and standard deviation of each element concentration data, outliers exceeding the mean plus four times the standard deviation were identified and removed to obtain the element concentration preprocessing data of the core samples.

[0007] As a preferred embodiment of the present invention, the other elements include main elements Si, Al, Fe, Ca, K, Mg, and trace elements Ti, Zr, V, Cr, Ni, and Mo.

[0008] As a preferred solution of the present invention, in step 4, the Pearson correlation analysis method is used to evaluate the correlation between the elements Ti and Zr indicating terrigenous debris and other elements in the pre-processed core sample element concentration data, and the element concentration data is screened according to the correlation results, specifically: When the Pearson correlation coefficient of an element with Ti or Zr is greater than 0.4, the element is considered to be susceptible to the flux of terrigenous debris, and its concentration data is removed from the element concentration preprocessing data of the core sample; When the Pearson correlation coefficient of an element with Ti or Zr is less than or equal to 0.4, it is considered that the element is not easily affected by the terrigenous debris flux, and its concentration data are retained for subsequent analysis.

[0009] As a preferred embodiment of the present invention, the method for extracting the main component that can best explain the change of ancient redox conditions from all elements in the core sample includes: Step 5.1: Construct a standardized data matrix of element concentration screening data, and construct the element concentration screening data of the core samples into a matrix form according to samples and elements; Step 5.2: Perform principal component analysis (PCA) on the standardized data matrix, including calculating the covariance matrix, solving eigenvalues ​​and eigenvectors; Step 5.3: According to the size of the eigenvalue, select the first few principal components whose cumulative variance contribution rate reaches the preset threshold; Step 5.4: Analyze the principal component loadings, determine the explanatory power of each principal component on the original elements, and extract the main components that can best explain the changes in ancient redox conditions.

[0010] Compared with the prior art, the present invention has the following beneficial effects: The present invention restores the ancient redox conditions of lacustrine shales by introducing machine learning methods, especially principal component analysis (PCA) technology, and achieves a high degree of automation and systematization of the analysis process. Compared with the traditional method that relies on a single geochemical indicator, the present invention can comprehensively process a large amount of multivariate data and extract the key components that best reflect the redox conditions, thereby greatly improving the accuracy and reliability of the analysis results.

[0011] Secondly, the present invention significantly improves the quality and consistency of data through systematic data preprocessing, including the exclusion of outliers and the screening of trace elements affected by debris flux, and avoids the interference of human factors on the analysis results. This process ensures that the data analysis of different work areas and well sections can maintain a high adaptability, making the method have good versatility in a wider range of geological environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0013] Figure 1 A flow chart of a method for restoring ancient redox conditions of lacustrine shale based on principal component analysis (PCA) provided in an embodiment of the present invention; Figure 2 A Pearson correlation coefficient variation curve diagram of each trace element and Zr element provided in an embodiment of the present invention; Figure 3The principal component loading diagram of PC-1 and PC-3 in the shale of the upper fourth section of Shahejie Formation in Well Niuye 1 provided in an embodiment of the present invention; Figure 4 The PC-1 and PC-3 scores provided in the embodiment of the present invention are trend diagrams of changes at different depths of the shale in the upper fourth section of the Shahejie Formation in Well Niuye 1. DETAILED DESCRIPTION

[0014] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0015] like Figure 1 As shown, the present invention provides a lacustrine shale paleo-redox restoration method based on a machine learning method, comprising the following steps: Step 1, sample collection and preliminary processing: First, core samples are collected from different depths and multiple geological layers in the target lacustrine shale layer, and then all core samples are cleaned and leveled to remove impurities on the surface of the core samples, and the geological layer and depth information of the core samples are recorded; Step 2, elemental composition analysis: using a portable energy dispersive X-ray fluorescence spectrometer pXRF to perform elemental analysis on the core sample obtained after the treatment in step 1, quantitatively determine the concentration of major elements and trace elements in the core sample, and obtain element concentration data of the core sample; The present invention significantly improves the quality and consistency of data and avoids the interference of human factors on the analysis results through systematic data preprocessing, including the exclusion of outliers and the screening of trace elements affected by debris flux, as follows: Step 3, data preprocessing and cleaning: preprocessing the element concentration data of the core sample obtained after the processing in step 2 to obtain element concentration preprocessing data of the core sample; This process ensures that the data analysis of different work areas and well sections can maintain a high degree of adaptability, making this method widely applicable in a wider range of geological environments.

[0016] Step 4, trace element screening: the correlation analysis method is used to evaluate the correlation between the elements Ti and Zr indicating terrigenous debris and other elements (including the main elements Si, Al, Fe, Ca, K, Mg, and trace elements V, Cr, Ni, Mo) in the element concentration preprocessing data of the core samples obtained after the processing in step 3. According to the correlation results, the element concentration data that are easily affected by the debris flux are excluded from the element concentration preprocessing data to obtain the element concentration screening data of the core samples; The present invention restores the ancient redox conditions of lacustrine shale by introducing machine learning methods, especially principal component analysis (PCA) technology, and realizes a high degree of automation and systematization of the analysis process, as follows: Step 5, principal component analysis (PCA): perform principal component analysis (PCA) on the element concentration screening data of the core samples, and extract the main components that can best explain the changes in ancient redox conditions from all elements of the core samples through dimensionality reduction technology; Therefore, compared with the traditional method that relies on a single geochemical indicator, the present invention can comprehensively process a large amount of multivariate data and extract the key components that best reflect the redox conditions, thereby greatly improving the accuracy and reliability of the analysis results.

[0017] Step 6, restoration of ancient redox conditions: Based on the main components obtained by PCA and combined with the score change trend of the main components, the ancient redox conditions of the target lacustrine shale are reconstructed.

[0018] In addition, the present invention pays special attention to the sedimentary characteristics of lacustrine shale and designs a highly universal operation process that can be effectively applied in different lacustrine shale research areas, overcoming the problem of insufficient applicability of existing technologies in different sedimentary backgrounds. Through PCA technology, the present invention not only improves the accuracy of lacustrine shale paleo-redox condition restoration, but also significantly reduces the complexity and time cost of data processing.

[0019] The method for preprocessing the element concentration data of the core sample obtained after the processing in step 2 includes: First, for element concentration data below the method detection limit (MDL), its concentration value is set to the method detection limit (MDL) to ensure the integrity of the data; Then, the missing data items in the element concentration data of the core samples were excluded to ensure the data validity of the subsequent analysis; Finally, by calculating the mean and standard deviation of each element concentration data, outliers (or outliers) exceeding the mean plus four times the standard deviation (4σ) are identified and removed to obtain the element concentration preprocessing data of the core samples.

[0020] Other elements include the main elements Si, Al, Fe, Ca, K, Mg, and trace elements Ti, Zr, V, Cr, Ni, and Mo.

[0021] In step 4, the Pearson correlation analysis method is used to evaluate the correlation between the elements Ti and Zr indicating terrigenous debris and other elements in the element concentration data of the pre-processed core samples. According to the correlation results, the element concentration data are screened as follows: When the Pearson correlation coefficient of an element with Ti or Zr is greater than 0.4, the element is considered to be susceptible to the flux of terrigenous debris, and its concentration data is removed from the element concentration preprocessing data of the core sample; When the Pearson correlation coefficient of an element with Ti or Zr is less than or equal to 0.4, it is considered that the element is not easily affected by the terrigenous debris flux, and its concentration data are retained for subsequent analysis.

[0022] Methods to extract the main components that best explain the changes in ancient redox conditions in all elements of the core samples include: Step 5.1: Construct a standardized data matrix of element concentration screening data, and construct the element concentration screening data of the core samples into a matrix form according to samples and elements; Step 5.2: Perform principal component analysis (PCA) on the standardized data matrix, including calculating the covariance matrix, solving eigenvalues ​​and eigenvectors; Step 5.3: According to the size of the eigenvalue, select the first few principal components whose cumulative variance contribution rate reaches the preset threshold (such as 70%); Step 5.4: Analyze the principal component loadings, determine the explanatory power of each principal component on the original elements, and extract the main components that can best explain the changes in ancient redox conditions.

[0023] The present invention provides an embodiment of restoring the ancient redox conditions of lacustrine shale, which is as follows: Sample collection and preparation: Core samples were systematically collected from the lacustrine shale layer of the upper Sha 4th member of the Niuye 1 well in the Dongying Depression, China, to ensure that the samples covered different depths and geological horizons. The collected samples were cleaned and leveled to ensure that the analysis surface was smooth and free of impurities, and the geological horizon and depth information of each sample was recorded in detail. This information will be used for comparison and verification in subsequent analysis.

[0024] Geochemical analysis: The core samples were analyzed using a portable energy dispersive X-ray fluorescence spectrometer (pXRF) to measure the concentrations of major elements (such as Si, Al, Fe, Ca, K, Mg, etc.) and trace elements (such as Ti, Zr, V, Cr, Ni, Mo, etc.). These data provide basic information for subsequent principal component analysis (PCA) to further analyze the ancient redox conditions in the study area.

[0025] Data preprocessing: Systematically clean the data measured by the portable energy dispersive X-ray fluorescence spectrometer pXRF.

[0026] First, the element concentration values ​​below the method detection limit (MDL) were set as the MDL concentration values, and then samples with missing data were excluded. By calculating the mean and standard deviation of each element, outliers exceeding the mean plus four times the standard deviation (4σ) were identified and excluded to ensure the integrity and consistency of the data, laying a solid foundation for principal component analysis.

[0027] Trace element feature selection: In order to exclude trace elements that are greatly affected by the debris flux, the Pearson correlation coefficient between each trace element and Zr, an element that can indicate terrigenous debris, is calculated, and the Pearson correlation coefficient change curve is drawn, such as Figure 2 As shown, trace elements with correlation coefficients greater than 0.4 were excluded to ensure that the screened trace elements could accurately reflect the ancient redox conditions.

[0028] Principal component analysis (PCA): After data preprocessing and trace element screening, principal component analysis (PCA) is performed. By calculating the eigenvalue (Eigenvalue) of each principal component, its variance contribution rate (Variations) is determined.

[0029] like Figure 3 As shown in the figure, PCA scores and loadings are shown, such as redox-sensitive elements such as Mo, V, U, and Cr are dominant in PC-1, and sulfur and molybdenum are dominant in PC-3), showing the indicative role of these elements in the sedimentary environment. Specifically, in the analysis of the upper subsection of Sha 4 in Niuye 1 well, the first four principal components (PC-1 to PC-4) cumulatively explained 74.5% of the total variance, of which PC-1 contributed 43.0%, PC-2, PC-3, and PC-4 contributed 18.6%, 7.8%, and 5.0%, respectively. PC-1 is mainly dominated by redox-sensitive elements (such as Mo, V, U, Cr, etc.), which are enriched in oxygen-deficient reducing environments, so changes in PC-1 scores directly reflect the redox conditions in the sedimentary environment. In the samples of Niuye 1 well, the increase in PC-1 scores indicates an enhanced reducing state in the sedimentary environment, while the decrease in scores indicates an enhanced oxidizing condition. Therefore, PC-1 was identified as the main redox condition indicator. PC-3 is dominated by elements such as sulfur (S) and molybdenum (Mo). In particular, sulfur forms pyrite with iron in a highly reducing environment, showing a high loading value. Changes in PC-3 scores can reflect more subtle changes in redox conditions and supplement the analysis results of PC-1. By combining PC-1 and PC-3 scores, the ancient redox environment of the study area can be more comprehensively restored.

[0030] Restoration of ancient redox conditions: By analyzing the changing trends of PC-1 and PC-3 scores at different depths along the upper Sha 4th member of Well Niuye 1, the following results were obtained: Figure 4As shown in the figure, it is found that with the shallowing of the formation depth, the PC-1 and PC-3 score curves show a trend of first suddenly increasing and then gradually decreasing. This change indicates that the sedimentary environment of the upper Sha 4th member of Well Niuye 1 has experienced a process from a sudden increase in reduction to a gradual oxidation.

[0031] Results verification and application: In order to verify the effectiveness of this method, the PC-1 and PC-3 curves were compared with the U element enrichment factor (U) in the laboratory whole rock element analysis of core sample powder by previous researchers. EF ) curves. Figure 4 The results show that the change trends of the two are in good agreement, which further verifies the reliability of this method in restoring ancient redox conditions. This restoration model can not only accurately explain the sedimentary environment evolution of the upper Sha 4th sub-member of Well Niuye 1, but can also be applied to other similar studies, providing technical support for the restoration of ancient redox conditions of lacustrine shales.

[0032] The results of the processing and analysis of the method of the present invention have been verified by geology and geochemistry, proving its rationality and scientificity in practical application. Compared with traditional methods, the present invention not only improves the consistency and objectivity of analysis, but also can provide more reliable technical support for the exploration and development of unconventional oil and gas resources, and has broad promotion and application prospects.

[0033] The present invention restores the ancient redox conditions of lacustrine shales by introducing machine learning methods, especially principal component analysis (PCA) technology, and realizes a high degree of automation and systematization of the analysis process. Compared with the traditional method that relies on a single geochemical indicator, the present invention can comprehensively process a large amount of multivariate data and extract the key components that best reflect the redox conditions, thereby greatly improving the accuracy and reliability of the analysis results. Secondly, the present invention significantly improves the quality and consistency of the data through systematic data preprocessing, including the exclusion of outliers and the screening of trace elements affected by debris flux, and avoids the interference of human factors on the analysis results. This process ensures that the data analysis of different work areas and well sections can maintain a high adaptability, making the method have good versatility in a wider range of geological environments.

[0034] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A method for paleo-redox restoration of lacustrine shale based on machine learning method, characterized in that: The following steps are involved: Step 1, sample collection and preliminary processing: First, core samples are collected from different depths and multiple geological layers in the target lacustrine shale layer, and then all core samples are cleaned and leveled to remove impurities on the surface of the core samples, and the geological layer and depth information of the core samples are recorded; Step 2, elemental composition analysis: using a portable energy dispersive X-ray fluorescence spectrometer pXRF to perform elemental analysis on the core sample obtained after the treatment in step 1, quantitatively determine the concentration of major elements and trace elements in the core sample, and obtain element concentration data of the core sample; Step 3, data preprocessing and cleaning: preprocessing the element concentration data of the core sample obtained after the processing in step 2 to obtain element concentration preprocessing data of the core sample; Step 4, trace element screening: using the correlation analysis method to evaluate the correlation between the elements Ti and Zr indicating terrigenous debris and other elements in the element concentration preprocessing data of the core sample obtained after the processing in step 3, according to the correlation results, the element concentration data that is easily affected by the debris flux is excluded from the element concentration preprocessing data to obtain the element concentration screening data of the core sample; Step 5, principal component analysis (PCA): perform principal component analysis (PCA) on the element concentration screening data of the core samples, and extract the main components that can best explain the changes in ancient redox conditions from all elements of the core samples through dimensionality reduction technology; Step 6, restoration of ancient redox conditions: based on the main components obtained by PCA and combined with the score change trend of the main components, the ancient redox conditions of the target lacustrine shale are reconstructed.

2. The method for lacustrine shale paleo-redox restoration based on a machine learning method according to claim 1, characterized in that: The method for preprocessing the element concentration data of the core sample obtained after the processing in step 2 includes: First, for element concentration data below the method detection limit (MDL), its concentration value is set to the method detection limit (MDL) to ensure the integrity of the data; Then, the missing data items in the element concentration data of the core samples were excluded to ensure the data validity of the subsequent analysis; Finally, by calculating the mean and standard deviation of each element concentration data, outliers exceeding the mean plus four times the standard deviation were identified and removed to obtain the element concentration preprocessing data of the core samples.

3. The method for lacustrine shale paleo-redox restoration based on machine learning method according to claim 1, characterized in that: The other elements include main elements Si, Al, Fe, Ca, K, Mg, and trace elements Ti, Zr, V, Cr, Ni, Mo.

4. The method for lacustrine shale paleo-redox restoration based on machine learning method according to claim 1, characterized in that: In step 4, the Pearson correlation analysis method is used to evaluate the correlation between the elements Ti and Zr indicating terrigenous debris and other elements in the element concentration data of the pre-processed core samples. According to the correlation results, the element concentration data are screened as follows: When the Pearson correlation coefficient of an element with Ti or Zr is greater than 0.4, the element is considered to be susceptible to the flux of terrigenous debris, and its concentration data is removed from the element concentration preprocessing data of the core sample; When the Pearson correlation coefficient of an element with Ti or Zr is less than or equal to 0.4, it is considered that the element is not easily affected by the terrigenous debris flux, and its concentration data are retained for subsequent analysis.

5. The method for lacustrine shale paleo-redox restoration based on machine learning method according to claim 1, characterized in that: Methods to extract the main components that best explain the changes in ancient redox conditions in all elements of the core samples include: Step 5.1: Construct a standardized data matrix of element concentration screening data, and construct the element concentration screening data of the core samples into a matrix form according to samples and elements; Step 5.2: Perform principal component analysis (PCA) on the standardized data matrix, including calculating the covariance matrix, solving eigenvalues ​​and eigenvectors; Step 5.3: According to the size of the eigenvalue, select the first few principal components whose cumulative variance contribution rate reaches the preset threshold; Step 5.4: Analyze the principal component loadings, determine the explanatory power of each principal component on the original elements, and extract the main components that can best explain the changes in ancient redox conditions.

Images