Lithology identification method and system for shale oil reservoir

Abstract

The invention provides a lithology identification method and system for a shale oil reservoir, and relates to the technical field of geological exploration. According to the method, the logging data from multiple sources, the X-ray diffraction (XRD) data and the microscopic image analysis data are fused, so that the influence of complex and changeable mineral components and other physical parameters on lithology identification is effectively overcome, and the accuracy and reliability of an identification result are remarkably improved. Furthermore, dimension reduction is carried out on the fused features by adopting a principal component analysis (PCA), data redundancy is reduced, the most representative features are extracted, and the efficiency of the model is improved. An attention mechanism is introduced into the lithology identification model, the weight of each logging data feature is adaptively adjusted, the influence of key features on the lithology identification result is enhanced, and the interference of useless features is reduced, so that the accuracy and reliability of lithology identification are further improved, and a more accurate basis is provided for distribution prediction of shale oil.

Description

Technical Field

[0001] The present invention relates to the field of geological exploration technology, and in particular to a lithology identification method and system for shale oil reservoirs. Background Art

[0002] With the continuous deepening of oil and gas exploration and development, unconventional oil and gas reservoirs have been paid more and more attention. Lithology identification is the basis of reservoir evaluation and an important part of reservoir description and logging interpretation. Only by accurately understanding the lithology of the reservoir can we accurately solve its reservoir parameters and judge the oil and gas content of the reservoir. Therefore, the accuracy of lithology identification directly affects the reliability of reservoir evaluation results.

[0003] Since there are many types of mineral components and complex lithology in shale oil reservoirs, some shale formations contain a large amount of interfering substances such as pyrite, which may affect the accuracy and reliability of lithology identification. In the process of lithology identification, it is necessary to comprehensively consider a variety of factors, such as formation pressure, formation temperature, rock mechanical properties, etc. These factors may interfere with and affect lithology identification. At the same time, due to the strong heterogeneity of shale oil reservoirs, there may be significant differences in logging data from different parts, which reduces the reliability of lithology identification results. Summary of the invention

[0004] The present invention provides a shale oil reservoir lithology identification method and system, which can improve the accuracy and reliability of shale oil reservoir lithology identification results and improve the accuracy of shale oil distribution.

[0005] In a first aspect, the present invention provides a method for lithology identification of a shale oil reservoir, the method comprising: obtaining detection data of multiple well logging data of the shale oil reservoir in an area to be identified, the detection data comprising well logging data, X-ray diffraction data and microscopic image analysis data; performing feature fusion based on the detection data of each well logging to obtain fusion features of each well logging; performing dimensionality reduction processing on the fusion features of multiple well loggings using a principal component analysis method to obtain detection features of the area to be identified; obtaining a lithology identification result based on the detection features of the area to be identified and a preset lithology identification model; the lithology identification model adjusts the weight coefficient of the fusion features of each well logging through an attention mechanism during the identification process; and determining the distribution of shale oil in the area to be identified based on the lithology identification result.

[0006] In a possible implementation, principal component analysis is used to perform dimension reduction processing on the fused features of multiple well loggings to obtain the detection features of the area to be identified, including: calculating the covariance matrix between the fused features of multiple well loggings based on the fused features of multiple well loggings; the elements in the covariance matrix are the covariance values ​​between two fused features; constructing a characteristic equation based on the covariance matrix; calculating the eigenvalues ​​and eigenvectors of the characteristic equation; determining the principal component eigenvectors based on the eigenvalues ​​of the characteristic equation; and determining the detection features of the area to be identified based on the principal component eigenvectors.

[0007] In a possible implementation, a lithology identification result is obtained based on the detection features of the area to be identified and a preset lithology identification model, including: calculating the attention score of each principal component eigenvector based on the detection features of the area to be identified and the attention module of the lithology identification model; the attention score is used to characterize the contribution of the principal component eigenvector to the lithology identification result; determining the weight coefficient of each principal component eigenvector based on the attention score of each principal component eigenvector; generating multiple attention eigenvectors based on each principal component eigenvector and the weight coefficient of each principal component eigenvector; determining the lithology identification result based on multiple attention eigenvectors and the neural network module of the lithology identification model.

[0008] In a possible implementation, before obtaining the lithology identification result based on the detection characteristics of the area to be identified and the preset lithology identification model, it also includes: obtaining the detection data of well logging in multiple identified areas, and the lithology identification results of multiple identified areas; based on the detection data of well logging in multiple identified areas, using the principal component analysis method, fusing and reducing the dimension of the detection data of each identified area to obtain the detection characteristics of each identified area; based on the detection characteristics and lithology identification results of each identified area, generating multiple training samples; based on the multiple training samples, performing neural network training to obtain the lithology identification model.

[0009] In a possible implementation, neural network training is performed based on multiple training samples to obtain a lithology recognition model, including: step one, based on multiple training samples and a new neural network module, an initial module is trained; step two, based on each principal component feature vector in the input vector of the multiple training samples and an attention module, a weight coefficient of each principal component vector in the multiple training samples is determined; step three, based on each principal component vector in the multiple training samples and the weight coefficient of each principal component vector in the multiple training samples, multiple attention feature vectors are generated for each training sample; step four, based on the multiple attention feature vectors of each training sample and the initial module, a first recognition result of each training sample is determined; step five, based on the first recognition result of each training sample and the output vector of each training sample, an output vector of the initial module is determined. Error; Step 6, if the difference between the output error of the initial module and the output error in the previous iteration is less than or equal to the set value, execute Step 7; if the difference between the output error of the initial module and the output error in the previous iteration is greater than the set value, then based on the output error of the initial module, adjust the module parameters of the attention module, and repeat Steps 2 to 6 until the difference between the output error of the initial module and the output error in the previous iteration is less than or equal to the set value; Step 7, using multiple attention feature vectors of each training sample as new input vectors, regenerate the training sample; Step 8, based on the regenerated training sample and the initial module, retrain the neural network to obtain the neural network module of the lithology recognition model; Step 9, based on the neural network module and attention module of the lithology recognition model, construct the lithology recognition model.

[0010] In a possible implementation, the logging data includes logging identification, reservoir depth, formation radiation intensity, formation density, resistivity, mineral composition information, porosity, organic carbon content, gas content and permeability; X-ray diffraction data includes mineral composition, crystal structure and brittleness index; microscopic image analysis data includes pore structure information, pore size distribution information, organic matter content, organic matter type, pore connectivity and fluorescence intensity; based on the detection data of each logging, feature fusion is performed to obtain the fusion feature of each logging, including: determining the standard data of each detection data, the standard data is used to characterize the limit value of whether shale oil is exploitable; based on the detection data of each logging and the standard data of each detection data, standardization processing is performed to obtain the standard detection data of each logging; based on the standard detection data of each logging, feature splicing and fusion are performed to obtain the fusion feature of each logging.

[0011] In one possible implementation, the lithology identification result includes the shale oil content of each well logging; based on the lithology identification result, the distribution of shale oil in the area to be identified is determined, including: based on the lithology identification result and the location information of each well logging, the area to be identified is divided to obtain multiple sub-areas; based on the shale oil content of each sub-area, the display color of each sub-area is determined; based on the display color of each sub-area, a shale oil distribution map of the area to be identified is drawn, and the shale oil distribution map represents the distribution of shale oil.

[0012] In a possible implementation, after determining the distribution of shale oil in the area to be identified based on the lithology identification results, it also includes: determining the key exploration area in the area to be identified based on the shale oil distribution map of the area to be identified; obtaining the detection data and surface conditions of multiple well logging in the key exploration area; determining the shale oil distribution map of the key exploration area based on the detection data of multiple well logging in the key exploration area; determining the location of the oil wells in the area to be identified based on the shale oil distribution map of the key exploration area and the surface conditions of the key exploration area.

[0013] In a second aspect, an embodiment of the present invention provides a lithology identification device for a shale oil reservoir, the lithology identification device comprising: a communication module for acquiring multiple well logging detection data of a shale oil reservoir in an area to be identified, the detection data comprising well logging data, X-ray diffraction data and microscopic image analysis data;

[0014] The processing module is used to perform feature fusion based on the detection data of each well logging to obtain the fusion features of each well logging; use the principal component analysis method to perform dimension reduction processing on the fusion features of multiple well logging to obtain the detection features of the area to be identified; based on the detection features of the area to be identified and a preset lithology identification model, a lithology identification result is obtained; the lithology identification model adjusts the weight coefficient of the fusion features of each well logging through the attention mechanism during the identification process; based on the lithology identification result, the distribution of shale oil in the area to be identified is determined.

[0015] In a third aspect, an embodiment of the present invention provides a lithology identification system for a shale oil reservoir, the lithology identification system comprising an electronic device, the electronic device comprising a memory and a processor, the memory storing a computer program, the processor being used to call and run the computer program stored in the memory to execute the steps of the method described in the first aspect and any possible implementation method of the first aspect.

[0016] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and wherein when the computer program is executed by a processor, the steps of the method described in the first aspect and any possible implementation method of the first aspect are implemented.

[0017] The present invention provides a lithology identification method and system for shale oil reservoirs. The present invention fuses well logging data, X-ray diffraction data and microscopic image analysis data of multiple well logging data, and performs lithology identification based on fusion features, thereby reducing the influence of multiple influencing factors in a single well logging data. That is, by fusing multiple well logging data, the influence of various complex and variable physical parameters such as mineral components on the lithology identification results is reduced, thereby improving the accuracy of the lithology identification results. Further, after fusion, the present invention uses principal component analysis to perform dimensionality reduction processing on the fusion features, thereby reducing data redundancy between multiple well logging data, and adaptively adjusting the weights of the fusion features of each well logging data through the attention mechanism in the lithology identification model, thereby enhancing the influence of useful features on the lithology identification results, and reducing the influence of useless features on the lithology identification results, thereby improving the accuracy and reliability of the lithology identification results of the shale oil reservoir, thereby improving the accuracy of the distribution of shale oil. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0019] Figure 1 It is a schematic flow chart of a lithology identification method for a shale oil reservoir provided by an embodiment of the present invention;

[0020] Figure 2 It is a structural schematic diagram of a lithology identification device for a shale oil reservoir provided by an embodiment of the present invention;

[0021] Figure 3 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0023] In the description of the present invention, unless otherwise specified, " / " means "or", for example, A / B can mean A or B. "And / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, "at least one" and "plurality" refer to two or more. The words "first", "second", etc. do not limit the quantity and execution order, and the words "first", "second", etc. do not limit them to be different.

[0024] In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way for easy understanding.

[0025] In addition, the terms "including" and "having" and any variations thereof mentioned in the description of this application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or modules is not limited to the listed steps or modules, but may optionally include other steps or modules that are not listed, or may optionally include other steps or modules that are inherent to these processes, methods, products or devices.

[0026] In order to make the purpose, technical solutions and advantages of the present invention more clear, the following will be described through specific embodiments in conjunction with the accompanying drawings of the present invention.

[0027] like Figure 1 As shown, an embodiment of the present invention provides a lithology identification method for shale oil reservoirs. The method includes steps S101-S105.

[0028] S101, acquiring detection data of multiple well logging of shale oil reservoirs in the area to be identified.

[0029] In the embodiment of the present application, the detection data includes logging data, X-ray diffraction data and microscopic image analysis data.

[0030] In some embodiments, the well logging data includes well logging identification, reservoir depth, formation radioactivity, formation density, resistivity, mineral composition information, porosity, organic carbon content, gas content and permeability.

[0031] Exemplarily, well logging identification: a number or name that uniquely identifies each well logging. Reservoir depth: the specific location of the reservoir underground. Formation radioactivity intensity: reflects the content of radioactive elements in the formation. Formation density: density information of the formation, which is related to the porosity and mineral composition of the reservoir. Resistivity: a parameter that reflects the conductivity of the formation, which is related to the fluid content and properties in the formation. Mineral composition information: the content and proportion of various minerals in the formation. Porosity: the proportion of pore space in the reservoir, which affects the oil storage capacity of the reservoir. Organic carbon content: the content of organic matter in the reservoir, which is a key indicator for evaluating the potential of shale oil. Gas content: the content of gas in the reservoir, which may affect the extraction of shale oil. Permeability: the ease of fluid flow in the reservoir, which affects the extraction efficiency.

[0032] In some embodiments, the X-ray diffraction data includes mineralogical composition, crystal structure, and brittleness index.

[0033] For example, mineral composition: the types and contents of various minerals in the formation. Crystal structure: the crystal shape and arrangement of minerals. Brittleness index: assesses the brittleness of formation rocks, which has an important impact on drilling and fracturing operations.

[0034] In some embodiments, the microscopic image analysis data is an image analysis of lithology or mineral structure, mainly including pore structure information, pore size distribution information, organic matter content, organic matter type, pore connectivity and fluorescence intensity.

[0035] For example, pore structure information: the shape, size and distribution of pores. Pore size distribution information: the proportion and distribution of pores of different sizes. Organic matter content: the organic matter content obtained by microscopic image analysis is verified with the organic carbon content in the logging data. Organic matter type: different types of organic matter have an important impact on the generation and storage of shale oil. Pore connectivity: the connection between pores, which affects the flow of fluids. Fluorescence intensity: the fluorescence intensity emitted by organic matter under specific light, which can be used to evaluate the type and maturity of organic matter.

[0036] S102: Perform feature fusion based on the detection data of each well logging to obtain the fusion feature of each well logging.

[0037] In some embodiments, the detection data is multiple modalities from different measurement techniques and data types, and each data type provides different information. Therefore, the detection data is also multimodal data.

[0038] As a possible implementation method, the embodiment of the present invention can extract information from data of different modes through multimodal learning, and improve the prediction or recognition effect by combining this information. In the lithology recognition task of the present invention, data of different modes together provide the model with more comprehensive lithology characteristics, thereby improving the recognition accuracy and reliability.

[0039] Exemplarily, the embodiment of the present invention can perform standardization processing on the detection data to eliminate the influence of the dimension on the data comparison and analysis. Step S102 can be specifically implemented as steps S1021-S1023.

[0040] S1021. Determine standard data for each detection data.

[0041] In some embodiments, standard data is used to characterize the boundary value of whether shale oil is recoverable.

[0042] For example, the embodiment of the present invention can determine the standard data of each detection data based on the existing geological and oil and gas reservoir research results. These standard data are used to characterize the limit value of whether shale oil can be mined, such as the lower limit value of porosity, the critical value of organic carbon content, etc.

[0043] S1022. Perform standardization processing based on the detection data of each well logging and the standard data of each detection data to obtain the standard detection data of each well logging.

[0044] For example, the embodiment of the present invention can compare the detection data of each well logging with the corresponding standard data and perform standardization processing. The standardization processing method can be linear transformation, normalization, etc., so that each detection data is converted into a dimensionless and comparable value.

[0045] S1023. Based on the standard detection data of each well logging, feature splicing and fusion are performed to obtain the fusion features of each well logging.

[0046] For example, the embodiments of the present invention can perform feature splicing and fusion of the standardized logging data, X-ray diffraction data and microscopic image analysis data. Feature splicing is to directly combine the feature data from different sources but with correlation to form a larger feature vector. This fusion method retains the integrity of the original data and allows the subsequent analysis model to use all available information.

[0047] Exemplarily, after obtaining standard detection data, the embodiment of the present invention can realize feature concatenation and fusion of various detection data based on a multimodal learning model to obtain fused features.

[0048] The multimodal learning model is obtained by training and learning based on the detection data of known lithology identification results and combining with the attention mechanism. The construction process of the multimodal learning model is detailed in steps A1-A6.

[0049] The embodiment of the present invention can directly splice the standard detection data of each well logging to obtain a standard detection vector, and then input the standard detection vector into a multimodal learning model to adjust the attention weight to obtain a fusion feature.

[0050] S103, using principal component analysis to perform dimension reduction processing on the fusion features of multiple well loggings to obtain detection features of the area to be identified.

[0051] As a possible implementation manner, step S103 may be specifically implemented as steps S1031 - S1035 .

[0052] S1031. Based on the fusion features of the multiple well loggings, calculate the covariance matrix between the fusion features of the multiple well loggings.

[0053] In some embodiments, the elements in the covariance matrix are covariance values ​​between two fused features.

[0054] For example, for the fusion features of n well logs, we can regard them as an n×m matrix, where n is the number of well logs and m is the dimension of each well log fusion feature. Calculate the covariance matrix C of this matrix, which is an m×m matrix, in which the element C(i,j) represents the covariance value between the i-th feature and the j-th feature. The covariance value reflects the linear correlation between two features. The larger the covariance value, the stronger the linear relationship between the two features.

[0055] S1032. Construct a characteristic equation based on the covariance matrix.

[0056] Exemplarily, the characteristic equation is constructed based on the covariance matrix C, and its form is |λE-C|=0, where E is the identity matrix and λ is the eigenvalue. This equation is used to solve the eigenvalue and eigenvector of the covariance matrix C.

[0057] S1033. Calculate the eigenvalue and eigenvector of the characteristic equation.

[0058] For example, the characteristic equation is solved to obtain m eigenvalues ​​λ1,λ2,...,λm and corresponding m eigenvectors v1,v2,...,vm. The size of the eigenvalue reflects the importance of the corresponding eigenvector in the original data set. The larger the eigenvalue, the more information the corresponding eigenvector contains.

[0059] S1034. Determine the principal component eigenvector based on the eigenvalue of the characteristic equation.

[0060] Exemplarily, the embodiment of the present invention can sort the eigenvalues ​​from large to small, and select the eigenvectors corresponding to the first k largest eigenvalues ​​as the principal component eigenvectors. The selection of k is usually based on a certain threshold (such as the cumulative contribution rate of the eigenvalue reaches a certain proportion, such as 80% or 90%) or an empirical value.

[0061] S1035. Determine the detection features of the area to be identified based on the principal component feature vector.

[0062] Exemplarily, the embodiment of the present invention can project the original fusion feature matrix into the subspace formed by the principal component feature vectors to obtain a detection feature matrix after dimensionality reduction. Each row of this matrix corresponds to a detection feature after dimensionality reduction of a well logging, and each column corresponds to a principal component feature.

[0063] The embodiment of the present invention can input the detection features after dimensionality reduction into a preset lithology recognition model, adjust the weight coefficient of the fusion features of each well logging through the attention mechanism, and obtain the lithology recognition result. Based on the lithology recognition result, the distribution of shale oil in the area to be identified is determined.

[0064] S104: Obtain a lithology identification result based on the detection characteristics of the area to be identified and a preset lithology identification model.

[0065] In the embodiment of the present application, the lithology identification model adjusts the weight coefficient of the fusion features of each well logging through the attention mechanism during the identification process.

[0066] In some embodiments, the lithology identification results include the shale oil content of each well log.

[0067] As a possible implementation manner, step S104 may be specifically implemented as steps S1041 - S1044 .

[0068] S1041. Based on each principal component feature vector in the detection features of the area to be identified and the attention module of the lithology identification model, calculate the attention score of each principal component feature vector.

[0069] In some embodiments, the attention score is used to characterize the contribution of the principal component eigenvector to the lithology identification result.

[0070] For example, in the embodiment of the present invention, the attention module receives the detection feature matrix as input, and generates an attention score for each principal component feature vector through a series of calculations (such as multi-layer perceptron, attention mechanism, etc.). This score reflects the importance or contribution of the feature vector in the lithology identification task. An attention score vector with the same number as the principal component feature vector is obtained.

[0071] S1042. Determine a weight coefficient of each principal component eigenvector based on the attention score of each principal component eigenvector.

[0072] Exemplarily, the embodiment of the present invention can normalize the attention score vector through a softmax function or a similar mechanism to ensure that the sum of all weight coefficients is 1. In this way, each principal component feature vector is assigned a weight coefficient between 0 and 1, indicating its relative importance in the final decision. A weight coefficient vector is obtained, whose length is the same as the number of principal component feature vectors.

[0073] S1043. Generate multiple attention feature vectors based on each principal component feature vector and the weight coefficient of each principal component feature vector.

[0074] Exemplarily, the embodiment of the present invention can multiply each principal component eigenvector with its corresponding weight coefficient to obtain a weighted eigenvector. These weighted eigenvectors are attention eigenvectors. A set of attention eigenvectors having the same number as the original principal component eigenvectors is obtained.

[0075] S1044. Determine a lithology identification result based on multiple attention feature vectors and a neural network module of a lithology identification model.

[0076] Exemplarily, the neural network module receives the attention feature vector as input, and further processes and extracts features through a multi-layer neural network (such as a convolutional neural network, a recurrent neural network, a fully connected network, etc.). Finally, the network outputs a probability distribution of one or more categories, indicating the possibility that the area to be identified belongs to different lithologies. According to the probability distribution, the category with the highest probability is selected as the lithology identification result. If the model outputs multiple categories (such as sandstone, mudstone, shale, etc.), the category with the highest probability is selected as the final lithology identification.

[0077] It should be noted that in lithology identification, the complexity and variability of physical parameters such as logging data and mineral components often lead to errors and instability in the identification results, especially in the lithology identification of shale oil reservoirs, where accurate lithology identification is a key step in exploration and development. The present invention dynamically adjusts the weights of the fusion features of each logging data through the attention mechanism in the lithology identification model, so that the influence of useful features on the lithology identification results is enhanced, while reducing the interference of irrelevant or redundant features on the results, further improving the identification accuracy and reliability.

[0078] After the features are fused, the present invention reduces the dimension of the features through principal component analysis (PCA), reduces data redundancy and extracts the most representative features, so as to improve the learning efficiency and recognition accuracy of the model.

[0079] S105. Based on the lithology identification results, determine the distribution of shale oil in the area to be identified.

[0080] As a possible implementation manner, step S105 may be specifically implemented as steps S1051 - S1054 .

[0081] S1051. Based on the lithology identification result and the location information of each well logging, the area to be identified is divided to obtain multiple sub-areas.

[0082] For example, the embodiment of the present invention can divide the area to be identified into multiple sub-areas based on the location information of the well logging. These sub-areas can be grids based on geographic coordinates, layer segments based on well depth, or areas divided according to geological features (such as faults, unconformities, etc.). The division method can be selected according to actual conditions and needs to obtain multiple sub-areas with clear locations and boundaries.

[0083] S1052. Determine the display color of each sub-region based on the shale oil content of each sub-region.

[0084] For example, for each sub-region, the embodiment of the present invention can calculate the average or weighted average of all well logging shale oil contents within it as the shale oil content of the sub-region. If there is no well logging data in the sub-region, its shale oil content can be estimated by interpolation methods (such as linear interpolation, Kriging interpolation, etc.).

[0085] Each sub-region is then assigned a specific color based on the amount of shale oil in the shale. This is usually accomplished through a color map that maps a continuous range of shale oil content to a set of discrete colors. The design of the color map can be customized as needed to visually represent the amount of shale oil in the shale. Each sub-region has a corresponding display color.

[0086] S1053. Based on the display color of each sub-region, draw a shale oil distribution map of the area to be identified, where the shale oil distribution map represents the distribution of shale oil.

[0087] Exemplarily, the embodiments of the present invention can use geographic information system (GIS) software or drawing tools to draw a map of the area to be identified based on the location information and boundaries of the sub-areas. Then, the corresponding areas are filled according to the display colors of each sub-area to form a shale oil distribution map. During the drawing process, other geological information (such as stratigraphic boundaries, faults, etc.) can also be added as background information to enhance the readability and information content of the map. An intuitive shale oil distribution map can clearly show the distribution of shale oil in the area to be identified.

[0088] The present invention provides a lithology identification method for shale oil reservoirs. By fusing the well logging data, X-ray diffraction data and microscopic image analysis data of multiple well logging data, and based on the fusion features, lithology identification is performed, thereby reducing the influence of multiple influencing factors in a single well logging. That is, by fusing multiple well logging data, the influence of various complex and variable physical parameters such as mineral components on the lithology identification results is reduced, thereby improving the accuracy of the lithology identification results. Further, after fusion, the present invention uses the principal component analysis method to perform dimensionality reduction processing on the fusion features, thereby reducing the data redundancy between multiple well logging data, and through the attention mechanism in the lithology identification model, weights of the fusion features of each well logging are adaptively adjusted, thereby enhancing the influence of useful features on the lithology identification results, and reducing the influence of useless features on the lithology identification results, thereby improving the accuracy and reliability of the lithology identification results of the shale oil reservoir, thereby improving the accuracy of the distribution of shale oil.

[0089] The present invention adopts multimodal data fusion technology, combined with principal component analysis (PCA) and adaptive attention mechanism, which can more effectively reduce data redundancy and noise, and improve the accuracy and stability of lithology identification. Through this innovative method, the accuracy and reliability of shale oil reservoir lithology identification are significantly improved, thereby providing a more accurate prediction of shale oil distribution.

[0090] Optionally, the lithology identification method for shale oil reservoirs provided in an embodiment of the present invention further includes steps S201-S202 before step S104.

[0091] S201, obtaining well logging detection data in multiple identified areas and lithology identification results in multiple identified areas.

[0092] S202, based on the detection data of well logging in multiple identified areas, the detection data of each identified area are integrated and dimensionally reduced by using the principal component analysis method to obtain the detection features of each identified area.

[0093] S203: Generate multiple training samples based on the detection characteristics and lithology identification results of each identified area.

[0094] Exemplarily, the training sample uses the detection feature of each identified area as an input vector and the lithology identification result of each identified area as an output vector.

[0095] S204: Based on multiple training samples, neural network training is performed to obtain a lithology recognition model.

[0096] Exemplarily, step S204 may be specifically implemented as steps one to nine.

[0097] Step 1: Based on multiple training samples and a new neural network module, an initial module is trained.

[0098] Exemplarily, the embodiment of the present invention can use the input vector of the training sample to train the new neural network module to obtain an initial module. During the training process, the network output is calculated by forward propagation, and the network parameters are adjusted by back propagation to minimize the error between the output vector and the true output vector.

[0099] Step 2: Based on the principal component feature vectors in the input vectors of the multiple training samples and the attention module, the weight coefficient of each principal component vector in the multiple training samples is determined.

[0100] Exemplarily, the attention module receives an input vector as input, and generates a weight coefficient for each principal component feature vector through a series of calculations (such as multi-layer perceptron, attention mechanism, etc.). This weight coefficient reflects the importance or contribution of the feature vector in the recognition task.

[0101] Step three, based on the principal component vectors in the multiple training samples and the weight coefficients of the principal component vectors in the multiple training samples, generate multiple attention feature vectors for each training sample.

[0102] Exemplarily, the embodiment of the present invention can multiply each principal component vector with its corresponding weight coefficient to obtain a weighted feature vector. These weighted feature vectors are attention feature vectors.

[0103] Step 4: Based on the multiple attention feature vectors of each training sample and the initial module, determine the first recognition result of each training sample.

[0104] Exemplarily, an embodiment of the present invention may use the attention feature vector as the input of the initial module, and obtain the first recognition result of each training sample through forward propagation calculation.

[0105] Step 5: Based on the first recognition result of each training sample and the output vector of each training sample, determine the output error of the initial module.

[0106] Exemplarily, the embodiment of the present invention can calculate the error between the first recognition result of each training sample and the true output vector to obtain the output error of the initial module.

[0107] Step six, if the difference between the output error of the initial module and the output error in the previous iteration process is less than or equal to the set value, execute step seven; if the difference between the output error of the initial module and the output error in the previous iteration process is greater than the set value, then based on the output error of the initial module, adjust the module parameters of the attention module, and repeat steps two to six until the difference between the output error of the initial module and the output error in the previous iteration process is less than or equal to the set value.

[0108] Exemplarily, the module parameters of the attention module include weights, biases, etc. of the attention mechanism.

[0109] Step 7: Use multiple attention feature vectors of each training sample as new input vectors to regenerate the training sample.

[0110] Step eight, based on the regenerated training samples and the initial module, the neural network is retrained to obtain the neural network module of the lithology recognition model.

[0111] For example, the embodiment of the present invention can use the regenerated training samples to retrain the initial module (or the iterated module) to obtain the neural network module of the lithology recognition model. The forward propagation and back propagation algorithms are also used in the training process to minimize the output error.

[0112] Step nine: construct a lithology recognition model based on the neural network module and attention module of the lithology recognition model.

[0113] For example, the embodiment of the present invention can combine the neural network module and the attention module to construct a complete lithology recognition model. The attention module is responsible for extracting important features in the input vector and passing it to the neural network module for further classification or regression tasks.

[0114] In this way, the embodiment of the present invention can train a more accurate lithology identification model by combining the attention mechanism and neural network training before lithology identification, thereby improving the accuracy and reliability of the lithology identification results of the shale oil reservoir.

[0115] Optionally, the lithology identification method for shale oil reservoirs provided in an embodiment of the present invention further includes steps S301-S304 after step S105.

[0116] S301. Determine a key exploration area in the area to be identified based on a shale oil distribution map of the area to be identified.

[0117] For example, the embodiments of the present invention can analyze the shale oil distribution map and identify shale oil-rich areas, which usually have thicker shale oil layers, higher reserves and good recoverability. Considering factors such as geological structure, faults, caprocks and source rocks, further screen and determine key exploration areas. Evaluate the economic and feasibility of key exploration areas, including development costs, technical difficulties and market potential. Obtain the boundary and location information of key exploration areas, as well as related geological and geophysical parameters.

[0118] S302: Acquire detection data and surface conditions of multiple wells in the key exploration area.

[0119] Exemplarily, embodiments of the present invention can arrange logging plans in key exploration areas, determine the location and number of logging, and ensure that key areas can be fully covered. Perform logging operations to obtain logging detection data, including logging curves (such as resistivity, acoustic wave time difference, natural gamma, etc.), core analysis data, geochemical data, etc. Collect surface condition data of key exploration areas, including topography, soil type, vegetation cover, climate conditions, traffic conditions, etc. Then obtain logging detection data and surface condition data, which will be used for subsequent analysis and modeling.

[0120] S303: Determine a shale oil distribution map of the key exploration area based on the detection data of multiple wells in the key exploration area.

[0121] Exemplarily, the embodiments of the present invention can process and interpret the logging data, including data cleaning, quality control, data correction, etc. The underground structure and shale oil reservoir characteristics of the key exploration area are reconstructed based on the logging data by applying technologies such as geostatistics, geophysical inversion and reservoir modeling. Based on the reconstructed underground structure and shale oil reservoir characteristics, a shale oil distribution map of the key exploration area is generated, showing the distribution, thickness, reserves and potential recoverability of the shale oil layer. Then, a shale oil distribution map of the key exploration area is obtained, which provides more detailed and accurate shale oil reservoir information.

[0122] S304: Based on the shale oil distribution map of the key exploration area and the surface conditions of the key exploration area, determine the location of the oil wells in the area to be identified.

[0123] For example, the embodiments of the present invention can analyze the shale oil distribution map to determine the shale oil enrichment area and the potential high-yield well location. Consider the surface conditions, including topography, soil type, vegetation cover, climate conditions, traffic conditions, etc., to evaluate the feasibility and safety of the oil well location. Comprehensively consider factors such as economic efficiency, technical difficulty, environmental protection and regulatory requirements to determine the final oil well location. Then, the oil well location information of the area to be identified is obtained, including the coordinates, quantity, well type and drilling plan of the oil well.

[0124] In this way, the embodiment of the present invention can identify the lithology of the shale oil reservoir, narrow down the key exploration area according to the identification results, re-analyze the key exploration area, and comprehensively determine the location of the oil well suitable for exploitation in combination with the surface conditions, thereby facilitating shale oil exploitation.

[0125] Optionally, the lithology identification method for shale oil reservoirs provided in an embodiment of the present invention, before step S102, performs multimodal learning training through steps A1-A6 to obtain a multimodal learning model.

[0126] A1. Obtain standard detection features and lithology identification models for multiple identified areas.

[0127] A2. Build the initial model based on the attention module and neural network module.

[0128] Among them, the model parameters of the initial model include attention parameters and neural network parameters.

[0129] A3. Based on multiple standard detection features, feature splicing is performed to obtain multiple fusion features.

[0130] A4. Take each standard detection feature as input and each fusion feature as output to construct multiple training samples.

[0131] A5. Based on multiple training samples, the initial model is trained to obtain a trained initial model.

[0132] A6. Based on multiple standard detection features and lithology recognition models, the trained initial model is optimized to obtain a multimodal learning model.

[0133] Exemplarily, step A6 can be implemented through steps A61-A64.

[0134] A61, the number of iterations is set to 1, the trained initial model is used as the current fusion model, and any standard detection feature is selected as the current standard detection feature.

[0135] A62. For the current standard detection features and lithology identification model, obtain the current output result and set the current output result as the optimal solution.

[0136] A63. Based on the current standard detection features and the current fusion model, the current fusion features are obtained.

[0137] A64. Based on the current fusion features and lithology identification model, the current output result is obtained.

[0138] A65. Compare the current output result with the optimal solution. If the current output result is better than the optimal solution, the current output result is determined as the optimal solution.

[0139] A66: If the number of iterations is greater than the maximum number of iterations, then exit the iteration process and execute step A68. If the number of iterations is less than or equal to the maximum number of iterations, then execute step A67.

[0140] A67. Update the model parameters of the current fusion model, and repeat steps A63-A66 with the updated current fusion model.

[0141] A68: Determine the current fusion model corresponding to the optimal solution as the multimodal learning model corresponding to the current standard detection feature, and update the current standard detection feature, and repeat steps A62-A68 to obtain the multimodal learning model corresponding to each standard detection feature.

[0142] A69. Based on the multimodal learning model corresponding to each standard detection feature, the model parameters are optimized to obtain the globally optimal multimodal learning model.

[0143] Exemplarily, the globally optimal multimodal learning model can be obtained as follows.

[0144] B1. Determine a plurality of first fusion features based on each standard detection feature and a multimodal learning model corresponding to each standard detection feature.

[0145] B2. Based on the multiple first fusion features and the lithology identification model, a first lithology identification result is obtained.

[0146] B3. Select any set of model parameters and determine the target multimodal learning model.

[0147] B4. Based on each standard detection feature and the target multimodal learning model, multiple second fusion features are obtained.

[0148] B5. Based on the multiple second fusion features and the lithology identification model, a second lithology identification result is obtained.

[0149] B6. Compare the difference between the first lithology identification result and the second lithology identification result to obtain the comprehensive error.

[0150] B7. Update the model parameters to obtain an updated target multimodal learning model, and repeat steps B4-B7 with the updated target multimodal learning model to obtain a target multimodal learning model with the minimum comprehensive error.

[0151] B8. Determine the target multimodal learning model with the smallest comprehensive error as the globally optimal multimodal learning model.

[0152] In this way, the embodiment of the present invention can obtain a multimodal learning model through multimodal data fusion and attention mechanism, combined with the lithology recognition model obtained through training, through repeated iterations, to facilitate multi-source data fusion. That is, the present invention fuses logging data, X-ray diffraction (XRD) data, and microscopic image analysis data from different sources, which provide key information for lithology recognition from three aspects: physics, mineral composition, and microstructure. By fusing these multimodal data, more comprehensive and representative lithology characteristics can be obtained, thereby improving the accuracy of recognition.

[0153] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0154] The following is an embodiment of the device of the present invention. For details not described in detail therein, reference may be made to the corresponding method embodiment described above.

[0155] Figure 2 The schematic diagram of the structure of a lithology identification device for shale oil reservoir provided by an embodiment of the present invention is shown. The lithology identification device 400 includes a communication module 401 and a processing module 402 .

[0156] Communication module 401, used to obtain multiple well logging detection data of shale oil reservoirs in the area to be identified, the detection data including well logging data, X-ray diffraction data and microscopic image analysis data;

[0157] Processing module 402 is used to perform feature fusion based on the detection data of each well logging to obtain the fusion features of each well logging; use the principal component analysis method to perform dimensionality reduction processing on the fusion features of multiple well logging to obtain the detection features of the area to be identified; based on the detection features of the area to be identified and a preset lithology identification model, obtain the lithology identification result; the lithology identification model adjusts the weight coefficient of the fusion features of each well logging through the attention mechanism during the identification process; based on the lithology identification result, determine the distribution of shale oil in the area to be identified.

[0158] Figure 3 Schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. Figure 3 As shown, the electronic device 500 includes: a processor 501, a memory 502, and a computer program 503 stored in the memory 502 and executable on the processor 501. When the processor 501 executes the computer program 503, the steps in the above-mentioned method embodiments are implemented, for example Figure 1 Alternatively, when the processor 501 executes the computer program 503, the functions of each module / unit in the above-mentioned device embodiments are realized, for example, Figure 2 The functions of the communication module 401 and the processing module 402 are shown.

[0159] Exemplarily, the computer program 503 may be divided into one or more modules / units, which are stored in the memory 502 and executed by the processor 501 to implement the present invention. The one or more modules / units may be a series of computer program instruction segments that can implement specific functions, and the instruction segments are used to describe the execution process of the computer program 503 in the electronic device 500. For example, the computer program 503 may be divided into Figure 2 A communication module 401 and a processing module 402 are shown.

[0160] The processor 501 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0161] The memory 502 may be an internal storage unit of the electronic device 500, such as a hard disk or memory of the electronic device 500. The memory 502 may also be an external storage device of the electronic device 500, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 500. Further, the memory 502 may also include both an internal storage unit of the electronic device 500 and an external storage device. The memory 502 is used to store the computer program and other programs and data required by the terminal. The memory 502 may also be used to temporarily store data that has been output or is to be output.

[0162] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A lithology identification method for shale oil reservoirs, characterized in that: include: Acquire multiple well logging detection data of shale oil reservoirs in the area to be identified, wherein the detection data includes well logging data, X-ray diffraction data and microscopic image analysis data; Based on the detection data of each well logging, feature fusion is performed to obtain the fusion features of each well logging; The principal component analysis method is used to reduce the dimension of the fusion features of multiple well logging data to obtain the detection features of the area to be identified; Based on the detection characteristics of the area to be identified and a preset lithology identification model, a lithology identification result is obtained; the lithology identification model adjusts the weight coefficient of the fusion characteristics of each well logging through an attention mechanism during the identification process; Based on the lithology identification result, the distribution of shale oil in the area to be identified is determined.

2. The lithology identification method of shale oil reservoir according to claim 1, characterized in that: The principal component analysis method is used to perform dimension reduction processing on the fusion features of multiple well logging to obtain the detection features of the area to be identified, including: Based on the fusion features of multiple well logging, a covariance matrix between the fusion features of multiple well logging is calculated; the elements in the covariance matrix are the covariance values ​​between two fusion features; Based on the covariance matrix, construct a characteristic equation; Compute eigenvalues ​​and eigenvectors of the characteristic equation; Determining a principal component eigenvector based on the eigenvalues ​​of the characteristic equation; Based on the principal component feature vector, the detection feature of the area to be identified is determined.

3. The lithology identification method of shale oil reservoir according to claim 2, characterized in that: The lithology identification result is obtained based on the detection characteristics of the area to be identified and the preset lithology identification model, including: Based on each principal component feature vector in the detection features of the area to be identified and the attention module of the lithology identification model, an attention score of each principal component feature vector is calculated; the attention score is used to characterize the contribution of the principal component feature vector to the lithology identification result; Based on the attention score of each principal component eigenvector, determining the weight coefficient of each principal component eigenvector; Based on each principal component eigenvector and a weight coefficient of each principal component eigenvector, a plurality of attention eigenvectors are generated; Based on multiple attention feature vectors and a neural network module of a lithology recognition model, a lithology recognition result is determined.

4. The lithology identification method of shale oil reservoir according to claim 1, characterized in that: Before obtaining the lithology identification result based on the detection features of the area to be identified and the preset lithology identification model, the method further includes: Acquire well logging detection data in multiple identified areas and lithology identification results in multiple identified areas; Based on the well logging detection data in multiple identified areas, the principal component analysis method is used to fuse and reduce the dimension of the detection data of each identified area to obtain the detection characteristics of each identified area; generating a plurality of training samples based on the detection characteristics and lithology identification results of each identified area; Based on the multiple training samples, neural network training is performed to obtain the lithology identification model.

5. The lithology identification method of shale oil reservoir according to claim 1, characterized in that: The neural network training is performed based on the multiple training samples to obtain the lithology identification model, including: Step 1: Based on the multiple training samples and the new neural network module, an initial module is trained; Step 2: determining a weight coefficient of each principal component vector in the plurality of training samples based on each principal component feature vector in the input vector of the plurality of training samples and an attention module; Step 3: Based on the principal component vectors in the multiple training samples and the weight coefficients of the principal component vectors in the multiple training samples, generate multiple attention feature vectors for each training sample; Step 4: determining a first recognition result of each training sample based on a plurality of attention feature vectors of each training sample and the initial module; Step 5, determining the output error of the initial module based on the first recognition result of each training sample and the output vector of each training sample; Step 6: If the difference between the output error of the initial module and the output error in the previous iteration is less than or equal to the set value, execute step 7; if the difference between the output error of the initial module and the output error in the previous iteration is greater than the set value, adjust the module parameters of the attention module based on the output error of the initial module, and repeat steps 2 to 6 until the difference between the output error of the initial module and the output error in the previous iteration is less than or equal to the set value; Step 7: Regenerate the training sample using multiple attention feature vectors of each training sample as new input vectors; Step 8, based on the regenerated training samples and the initial module, retrain the neural network to obtain the neural network module of the lithology recognition model; Step nine: construct the lithology recognition model based on the neural network module and attention module of the lithology recognition model.

6. The lithology identification method of shale oil reservoir according to claim 1, characterized in that: The logging data include logging identification, reservoir depth, formation radiation intensity, formation density, resistivity, mineral composition information, porosity, organic carbon content, gas content and permeability; the X-ray diffraction data include mineral composition, crystal structure and brittleness index; the microscopic image analysis data include pore structure information, pore size distribution information, organic matter content, organic matter type, pore connectivity and fluorescence intensity; The feature fusion is performed based on the detection data of each well logging to obtain the fusion feature of each well logging, including: Determine standard data for each detection data, wherein the standard data is used to characterize a limit value of whether shale oil is exploitable; Based on the detection data of each well logging and the standard data of each detection data, standardization processing is performed to obtain standard detection data of each well logging; Based on the standard detection data of each well logging, feature splicing and fusion are performed to obtain the fusion features of each well logging.

7. The lithology identification method of shale oil reservoir according to claim 1, characterized in that: The lithology identification results include the shale oil content of each well log; Determining the distribution of shale oil in the area to be identified based on the lithology identification result includes: Based on the lithology identification result and the location information of each well logging, the area to be identified is divided to obtain a plurality of sub-areas; Based on the shale oil content of each sub-region, determine the display color of each sub-region; Based on the display color of each sub-region, a shale oil distribution map of the area to be identified is drawn, and the shale oil distribution map represents the distribution of the shale oil.

8. The lithology identification method of shale oil reservoir according to claim 1, characterized in that: After determining the distribution of shale oil in the area to be identified based on the lithology identification result, the method further includes: Determining a key exploration area in the area to be identified based on a shale oil distribution map of the area to be identified; Obtain detection data and surface conditions from multiple wells in key exploration areas; Determine a shale oil distribution map of the key exploration area based on the detection data of multiple wells in the key exploration area; Based on the shale oil distribution map of the key exploration area and the surface conditions of the key exploration area, the location of the oil wells in the area to be identified is determined.

9. A lithology identification device for shale oil reservoirs, characterized in that: include: A communication module, used to obtain multiple well logging detection data of shale oil reservoirs in the area to be identified, wherein the detection data includes well logging data, X-ray diffraction data and microscopic image analysis data; The processing module is used to perform feature fusion based on the detection data of each well logging to obtain the fusion features of each well logging; the principal component analysis method is used to perform dimension reduction processing on the fusion features of multiple well logging to obtain the detection features of the area to be identified; Based on the detection characteristics of the area to be identified and a preset lithology identification model, a lithology identification result is obtained; the lithology identification model adjusts the weight coefficient of the fusion characteristics of each well logging through an attention mechanism during the identification process; Based on the lithology identification result, the distribution of shale oil in the area to be identified is determined.

10. A lithology identification system for shale oil reservoirs, characterized in that: The lithology identification system includes an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to execute the steps of the method as described in any one of claims 1 to 8.

Images