Rock facies identification method and system based on core label and similarity and related equipment
By using a core label and similarity-based method and optimizing feature mapping parameters with a deep learning model, the problem of sparse core data and difficult model training in lithofacies identification is solved. This achieves high-precision, interpretable lithofacies classification with lightweight deployment and high transferability.
Patent Information
- Application Number
- CN202511029223.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-25
- Publication Date
- 2025-10-31
AI Technical Summary
Traditional lithofacies identification methods are limited by the scarcity of core data, making it difficult to meet the needs of cross-well or inter-well comparison and identification. Furthermore, existing models are difficult to train, and the results lack transferability and interpretability.
A core label and similarity-based approach is adopted. Through feature extraction and linear mapping, combined with deep learning models such as bidirectional long short-term memory neural networks, the feature mapping parameters are optimized to maximize intra-class similarity and inter-class differences. A similarity voting mechanism is used for lithofacies identification.
It achieves high-precision, interpretable lithofacies classification under weakly supervised or unsupervised conditions, with lightweight deployment and high portability, making it suitable for engineering applications.
Smart Images
Figure CN120873633A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of lithofacies identification technology, and more specifically, to a lithofacies identification method, system, and related equipment based on core tags and similarity. Background Technology
[0002] In oil and gas exploration and development, lithofacies identification is a fundamental task in geological modeling and reservoir evaluation, especially crucial in sedimentary facies research and lithological modeling. Traditional lithofacies identification typically relies on core observation, thin section analysis, and geochemical testing, which provide accurate lithological information. However, due to cost and operational limitations, core data often only covers a few well sections, resulting in sparse overall data that is insufficient for identifying lithofacies across entire well sections or between wells.
[0003] Meanwhile, well logging curves, as continuously acquireable high-resolution geological response signals, have potential applications in lithofacies identification. Traditional methods attempt to correlate well logging data with lithofacies to construct lithofacies identification models. However, existing methods generally suffer from the following problems: the number of labels corresponding to core samples is limited, making it difficult to train deep learning models using traditional supervised learning methods; secondly, most models are trained in a black-box manner, making it difficult to interpret the classification criteria and gain the approval of engineers; and finally, it is difficult to establish an effective lithological analogy mechanism between different well sections, resulting in non-transferable identification results. Summary of the Invention
[0004] The purpose of this invention is to provide a lithofacies identification method, system, and related equipment based on core labels and similarity, which solves the problem of low accuracy in lithofacies category identification without large-scale core sample training.
[0005] The above-mentioned technical objective of the present invention is achieved through the following technical solution: In a first aspect, the present invention provides a lithofacies identification method based on core tags and similarity, the method comprising: Acquire the logging data to be identified; The first feature vector of the well logging data to be identified is extracted using a pre-trained feature extraction model; The first feature vector is linearly mapped using pre-configured feature mapping parameters to obtain the second feature vector; The features of multiple core samples from the core database are obtained, and the similarity between the second feature vector and the features of each core sample is calculated. The top K core samples with the highest similarity to the second feature vector are selected to form a core sample set; where K is a positive integer. For K core samples in the core sample set, the lithofacies identification result is output by weighted voting based on their similarity to the second feature vector.
[0006] In one implementation, the pre-configured feature mapping parameter determination process includes: Obtain raw well logging data and labels characterizing the lithofacies category of each sampling point at each depth in each well; Sliding sampling is performed along the depth direction of each original logging well according to the preset depth width and step size to obtain the core sample sequence; The core sample sequence and label are combined to obtain a set of core sample pairs; A feature extraction model was obtained by training a pre-constructed bidirectional long short-term memory neural network using a set of core sample pairs. The original feature vector of the core sample sequence was extracted using a feature extraction model; By introducing pre-configured feature mapping parameters, the original feature vector is linearly mapped to obtain the original feature mapping vector; The similarity between core samples is calculated based on the original feature mapping vectors, and the intra-class sample pair set and inter-class sample pair set of core samples are constructed based on the similarity between core samples. Establish a first objective function that maximizes the Wasserstein distance between the set of intra-class sample pairs and the set of inter-class sample pairs; Solve the first objective function to obtain the parameter values of the pre-configured feature mapping parameters.
[0007] In one implementation, the pre-configured feature mapping parameter determination process further includes: Obtain raw well logging data and labels characterizing the lithofacies category of each sampling point at each depth in each well; Sliding sampling is performed along the depth direction of each original logging well according to the preset depth width and step size to obtain the core sample sequence; The core sample sequence and label are combined to obtain a set of core sample pairs; A feature extraction model was obtained by training a bidirectional long short-term memory neural network using a core sample set. The original feature vector of the core sample sequence was extracted using a feature extraction model; By introducing pre-configured feature mapping parameters, the original feature vector is linearly mapped to obtain the original feature mapping vector; The similarity between core samples is calculated based on the original feature mapping vectors, and the intra-class sample pair set and inter-class sample pair set of core samples are constructed based on the similarity between core samples. Establish a second objective function that maximizes the Sinkhorn distance between the set of intra-class sample pairs and the set of inter-class sample pairs; Solve the second objective function to obtain the parameter values of the pre-configured feature mapping parameters.
[0008] In one implementation, the similarity is cosine similarity.
[0009] In one implementation scheme, the characteristics of multiple core samples from the core database are obtained, specifically as follows: The third feature vector of each core sample is extracted using a pre-trained feature extraction model; The third feature vector is linearly mapped according to the pre-configured feature mapping parameters to obtain the core sample features.
[0010] A second aspect of the present invention provides a lithofacies identification system based on core tags and similarity, the system comprising: The data acquisition module is used to acquire the logging data to be identified. The feature extraction module is used to extract the first feature vector of the well logging data to be identified using a pre-trained feature extraction model; The feature mapping module is used to linearly map the first feature vector using pre-configured feature mapping parameters to obtain the second feature vector. The similarity calculation module is used to obtain the features of multiple core samples from the core database and calculate the similarity between the second feature vector and each core sample feature. The sample selection module is used to select the top K core samples with the highest similarity to the second feature vector to form a core sample set; where K is a positive integer. The lithofacies identification module is used to output lithofacies identification results by weighting the K core samples in the core sample set according to their similarity to the second feature vector.
[0011] In one implementation, the feature mapping module is used for: Obtain raw well logging data and labels characterizing the lithofacies category of each sampling point at each depth in each well; Sliding sampling is performed along the depth direction of each original logging well according to the preset depth width and step size to obtain the core sample sequence; The core sample sequence and label are combined to obtain a set of core sample pairs; A feature extraction model was obtained by training a pre-constructed bidirectional long short-term memory neural network using a set of core sample pairs. The original feature vector of the core sample sequence was extracted using a feature extraction model; By introducing pre-configured feature mapping parameters, the original feature vector is linearly mapped to obtain the original feature mapping vector; The similarity between core samples is calculated based on the original feature mapping vectors, and the intra-class sample pair set and inter-class sample pair set of core samples are constructed based on the similarity between core samples. Establish a first objective function that maximizes the Wasserstein distance between the set of intra-class sample pairs and the set of inter-class sample pairs; Solve the first objective function to obtain the parameter values of the pre-configured feature mapping parameters.
[0012] In one implementation, the feature mapping module is also used for: Obtain raw well logging data and labels characterizing the lithofacies category of each sampling point at each depth in each well; Sliding sampling is performed along the depth direction of each original logging well according to the preset depth width and step size to obtain the core sample sequence; The core sample sequence and label are combined to obtain a set of core sample pairs; A feature extraction model was obtained by training a bidirectional long short-term memory neural network using a core sample set. The original feature vector of the core sample sequence was extracted using a feature extraction model; By introducing pre-configured feature mapping parameters, the original feature vector is linearly mapped to obtain the original feature mapping vector; The similarity between core samples is calculated based on the original feature mapping vectors, and the intra-class sample pair set and inter-class sample pair set of core samples are constructed based on the similarity between core samples. Establish a second objective function that maximizes the Sinkhorn distance between the set of intra-class sample pairs and the set of inter-class sample pairs; Solve the second objective function to obtain the parameter values of the pre-configured feature mapping parameters.
[0013] A third aspect of the present invention provides an electronic device, including a memory and a processor; A memory for storing computer programs, the computer programs including program instructions; A processor is configured to execute the program instructions to cause the electronic device to perform the steps of a lithofacies identification method based on core tagging and similarity as provided in the first aspect of the invention.
[0014] A fourth aspect of the present invention provides a computer-readable storage medium comprising a computer program that, when executed by one or more processors, implements a lithofacies identification method based on core tagging and similarity as provided in the first aspect of the present invention.
[0015] Compared with the prior art, the present invention has the following beneficial effects: This invention combines deep feature modeling, feature mapping space optimization, and a similarity voting mechanism to achieve automatic identification of lithofacies categories under weakly supervised or unsupervised conditions. This method fully leverages the distributive separability between existing expert interpretations and a small number of core samples to construct a lithofacies classification process that is both accurate and interpretable, offering advantages such as lightweight deployment, high portability, and strong engineering adaptability. Attached Figure Description
[0016] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a lithofacies identification method based on core tags and similarity provided in an embodiment of the present invention. Figure 2 Core electron microscopy scan images provided in embodiments of the present invention; Figure 3 A schematic diagram illustrating the sliding window sample construction and label extraction provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of feature mapping and similarity matrix construction provided in an embodiment of the present invention; Figure 5 A schematic diagram illustrating the iterative process of minimizing Wasserstein using the SA algorithm, provided in an embodiment of the present invention; Figure 6 This is a schematic diagram comparing the predicted lithofacies of a section of a well with the actual results, provided as an embodiment of the present invention. Detailed Implementation
[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of this invention are only for explaining this invention and are not intended to limit this invention.
[0018] It should be noted that the terms "comprising" or "may include" used in the various embodiments of this application indicate the presence of the claimed function, operation, or element, and do not limit the addition of one or more functions, operations, or elements. Furthermore, as used in the various embodiments of this application, the terms "comprising," "having," and their cognates are intended only to indicate a specific feature, number, step, operation, element, component, or combination of the foregoing, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing, or adding one or more combinations of the foregoing.
[0019] It should be understood that terms such as "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.
[0020] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a lithofacies identification method based on core tags and similarity provided in an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes: S101, acquire the logging data to be identified.
[0021] In this embodiment, the well logging data is obtained by using a fixed-width sliding window (odd number) and sampling in the depth direction of each well with a step size of 1. The well logging data to be identified can be obtained. The well logging data includes gamma rays, resistivity, density, compensated neutron-lithology density, spontaneous potential, etc.
[0022] S102, using a pre-trained feature extraction model to extract the first feature vector of the logging data to be identified.
[0023] In this embodiment, the training process of the feature extraction model is as follows: acquire the original logging data and labels representing the lithofacies category of each sampling point at each depth of each well; perform sliding sampling along the depth direction of each original logging according to the preset depth width and step size to obtain a core sample sequence; combine the core sample sequence and labels to obtain a core sample pair set; use the core sample pair set to train a pre-constructed bidirectional long short-term memory neural network to obtain the feature extraction model.
[0024] Specifically, raw well logging data (such as GR, RT, DEN, CNL, SP, AC, etc.) and the interpretation results that have been discussed and determined (for training the feature extraction model) are collected. Perform data cleaning operations, including: removing null values (NaN) from well sections; deleting duplicate records; and removing outliers (such as jump values and physically unreasonable intervals).
[0025] Normalization and tensor representation: The logging curves of each well are normalized using a standardization method: options include min-max normalization or z-score normalization; all curves are mapped to the standard numerical range to improve model convergence efficiency.
[0026] The processed data is organized into three-dimensional tensors: ,in: This refers to the number of wells logged; C represents the number of depth sampling points for each well; C represents the number of logging items (such as GR, RT, etc.).
[0027] The corresponding expert interpretation label is a two-dimensional tensor: ;in, This represents the total number of lithofacies classifications; Indicates the first Koujingdi Lithofacies classification at each depth point.
[0028] To extract spatial context information, a fixed width is used. A sliding window (odd number) is used to sample in the depth direction with a step size of 1 to construct a training sample sequence. For each well, the first... Extract window segments at each depth location: For each location Construct window data: The label is positioned at its center. .
[0029] This invention constructs well logging sub-sequence samples using a sliding window approach, trains deep neural networks (such as Bi-LSTM, Transformer, TCN, etc.) using known expert interpretation results, extracts feature vectors with geological semantics, and establishes a unified well logging feature representation space as the basis for subsequent similarity retrieval.
[0030] By combining the label and window data, a set of core sample pairs can be obtained. .
[0031] In this embodiment, a bidirectional long short-term memory neural network is used as the training base network, denoted as . Input for training the classification model The output is the lithofacies category; the output of the intermediate feature layer is denoted as... .
[0032] Combined with core samples for assembly Tags in As a supervisory signal, a bidirectional long short-term memory neural network is trained. The intermediate layer extracts features as an "interpreter's perspective" representation. However, those skilled in the art will understand that other neural networks such as GRU, Transformer, and TCN can also be used, but this embodiment will not elaborate further.
[0033] Then, the feature extraction model is used to analyze the logging data to be identified. By performing feature extraction, the first feature vector can be obtained. .
[0034] S103, the first feature vector is linearly mapped using pre-configured feature mapping parameters to obtain the second feature vector.
[0035] In this embodiment, the process of determining the pre-configured feature mapping parameters includes: extracting the original feature vector of the core sample sequence using a feature extraction model; linearly mapping the original feature vector using pre-configured feature mapping parameters to obtain the original feature mapping vector; calculating the similarity between core samples based on the original feature mapping vector, and constructing the intra-class sample pair set and inter-class sample pair set of core samples based on the similarity between core samples; establishing a first objective function that maximizes the Wasserstein distance between the intra-class sample pair set and the inter-class sample pair set; and solving the first objective function to obtain the parameter values of the pre-configured feature mapping parameters.
[0036] Specifically, the original feature vector of the core sample sequence is extracted using a feature extraction model, i.e. .
[0037] The feature mapping parameters are used to adjust the feature space, making core samples of the same class as close as possible and samples of different classes as far apart as possible. Let the feature mapping vector be... A linear mapping is performed on the original feature vector to obtain the original feature mapping vector. .
[0038] The cosine similarity between core sample pairs in the projected feature space is as follows: ,in, Represents the Euclidean norm; subscript , Indicates core number / index.
[0039] Based on the similarity between core samples, intra-class and inter-class sample pair sets are constructed, as follows: The intra-class sample pair set is... The set of sample pairs between classes is .
[0040] To measure the separability between the "intra-class similarity distribution" and the "between-class similarity distribution" in the mapped feature space, this invention introduces the one-dimensional first-order Wasserstein distance as the optimization objective function. The underlying concept is to transform one distribution into another with minimal "cost." The Wasserstein distance is suitable for small data and has strong statistical stability, but its computational cost is relatively high. However, it can be approximated by the Sinkhorn distance, which has a differentiable form, is suitable for deep network optimization, and is computationally faster.
[0041] remember: It is a set of similarity values calculated from within-class sample pairs; It is a set of similarity values calculated from inter-class sample pairs.
[0042] make: For its empirical distribution, the one-dimensional first-order Wasserstein distance is defined as: ,in: Indicates that all edges are respectively , joint distribution This formula measures the distribution Move to distribution The required EarthMover's Distance (EMD) can be calculated as the mean absolute distance between sample pairs in the discrete case.
[0043] when They were sorted as follows: ,assumed Then we have: .
[0044] In this embodiment, the first objective function is to maximize the Wasserstein distance between the intra-class and inter-class distributions, thereby improving the separability of the feature space. The expression for the first objective function is: Solving this first objective function can be achieved using non-gradient algorithms such as simulated annealing, particle swarm optimization, and genetic algorithms.
[0045] Since the Wasserstein distance is not differentiable, it is not suitable for direct gradient optimization in deep networks. Therefore, this invention introduces the Sinkhorn distance as its regularized approximation to improve optimization efficiency and feasibility.
[0046] If the Sinkhorn distance is used, the principle is similar to that of the Wasserstein distance described above, as follows: set up , Let be a uniformly distributed weight vector, where Indicates and A one-dimensional vector of uniform length, with all elements being 1. The Sinkhorn distance is then defined as: ,in: For joint distribution matrix (transportation plan); For the cost of transportation; This is the entropy regularization term; The entropy regularization strength is... Time approaches the primitive distance.
[0047] The second objective function is: .
[0048] The expression for the second objective function determined using the Sinkhorn distance is: The second objective function can also be implemented using non-gradient algorithms such as simulated annealing, particle swarm optimization, and genetic algorithms, or it can be solved using gradient methods based on automatic differentiability, such as Adam / SGD.
[0049] Based on the solution results of the first objective function or the second objective function, the pre-configured feature mapping parameters can be obtained. The parameter values are obtained, and from this, the second feature vector can be determined. .
[0050] This invention introduces learnable feature mapping vectors By optimizing the vector, the mapped features are statistically characterized by intra-class clustering and inter-class separation. Using Wasserstein distance or Sinkhorn distance as the objective function, the goal is to maximize the separation of inter-class distribution structures, thereby improving the reliability of identification with few samples.
[0051] S104: Obtain the features of multiple core samples from the core database and calculate the similarity between the second feature vector and the features of each core sample.
[0052] In this embodiment, a pre-trained feature extraction model is used to extract the third feature vector of each core sample; the third feature vector is then linearly mapped according to pre-configured feature mapping parameters to obtain the core sample features. Let the first core sample in the core database be... i The characteristics of each core sample are: Then the similarity between the second feature vector and the features of each core sample is: ;in: Denotes the Euclidean norm; The total number of core samples represents the number of samples. The greater the similarity, the closer the two samples are in the optimized feature space.
[0053] S105, select the top K core samples with the highest similarity to the second feature vector to form a core sample set; where K is a positive integer.
[0054] In this embodiment, the following is set The number of similar samples used for weighted voting is typically much smaller than the total number of core samples. This indicates the number of "nearest neighbors" used in the voting. The sample with the highest similarity to the sample to be identified is selected. Each core sample constitutes a core sample set: , among which, each pair Labels representing core samples and their similarity; The function selects the top K values in reverse order of their values.
[0055] By calculating the similarity between unknown samples and known core samples in an optimized feature space, and selecting Top-K nearest neighbor samples for weighted voting, the lithofacies category is determined. This method achieves high-precision classification without end-to-end supervised training, conforming to the "analogy-recognition" process of geological interpretation.
[0056] S106: For the K core samples in the core sample set, the lithofacies identification result is output by weighted voting based on their similarity to the second feature vector.
[0057] In this embodiment, based on The lithofacies category is determined by weighted voting based on the K most similar samples in the sample. Where R represents the number of all possible lithofacies categories. This involves calculating the weighted sum of similarities belonging to each category in the Top-K samples, and selecting the category corresponding to the maximum value as the lithofacies identification result. This invention can output the Top-K similarity distribution and confidence score for each sample to be identified, enhancing the engineering interpretability of the model. It is suitable for scenarios involving collaborative annotation with geological experts, model result quality control, and dynamic risk assessment.
[0058] This embodiment uses combined logging and core data from a Middle Eastern carbonate oilfield for verification research. A total of 43 vertical wells (i.e., core data) were collected from the study block. The well depth range covers the main reservoir section, with a logging data sampling interval of 0.1 m, and an average of approximately [number] logging points per well depth. Six logging curves (i.e., C=6C=6C=6) were collected at each depth point, including conventional logging terms such as gamma ray (GR), resistivity (RT), density (DEN), neutron porosity (CNL), acoustic transit time (AC), and spontaneous potential (SP). Among them, the lithofacies classification data that has been interpreted by experts were classified into lithofacies type R=6.
[0059] Real core data only provides lithofacies category labels for a small number of well sections, serving as a small-sample monitoring basis for this method. A total of [number missing] known core labels are available. Each core sample underwent thorough geochemical and electron microscopy analysis (as shown in the attached image). Figure 2 Experimental methods such as thin section analysis (as shown) are used to accurately determine its lithofacies classification, which has a very high degree of reliability.
[0060] The scanned electron microscope analysis table is as follows:
[0061] For each well's data, a sliding window was used to extract subsequences. The window width was set to a maximum of 1.2m above and below the data point for each sample. Training sample pairs were constructed using the expert label at the center point of the window as the supervision signal. Total extracted One training sample is used to train the feature extractor.
[0062] A feature extractor is constructed using a bidirectional LSTM network structure. The output feature dimension is 64, that is: .
[0063] After training the model using expert labels as supervision signals, the model was then used for feature encoding extraction on all samples (including core samples), such as... Figure 3 As shown.
[0064] like Figure 4 As shown, for all core samples Extract its feature vector A core feature-label set is constructed. A linear feature mapping parameter is introduced. Remap the feature vectors to all core sample pairs (i,j) and calculate their similarity distance in the projection space. If Add in-class sample pairs set ;like Add inter-class sample pairs .
[0065] The one-dimensional Wasserstein distance metric is used as the separation between the two distributions as the optimization objective. Using simulated annealing algorithm to Optimization is performed to obtain the optimal feature mapping parameters. This maximizes the separability of sample distributions of different lithofacies in the projected space. The Wasserstein value continuously decreases as the simulated annealing algorithm progresses, eventually converging to a stable value around 4000 iterations. The converged parameters can then be used as a vector. As attached Figure 5 .
[0066] Logging samples to be identified First extract features For each core sample , calculation and Cosine similarity in the mapping space.
[0067] The top K=5 core samples with the highest similarity were selected to form a set. Prediction is made through weighted voting: And output the prediction confidence: .
[0068] In five uncored well sections, the method of this invention was applied for lithofacies identification. Compared with subsequent verification drilling results, the Top-1 prediction accuracy reached 84.3%, the Top-2 voting hit rate exceeded 92.6%, and the average confidence score was in the range of 0.71 to 0.93. The prediction, ground truth, and confidence score for a portion of one well section are shown below. Figure 6 As shown, it is important to note that the ground-truth section integrates expert interpretations and core test results, with the core test results taking precedence. The prediction results exhibit clear structural clustering and continuous trends, and the voting results are clearly interpretable.
[0069] This invention also provides a lithofacies identification system based on core tags and similarity, the system comprising: The data acquisition module is used to acquire the logging data to be identified. The feature extraction module is used to extract the first feature vector of the well logging data to be identified using a pre-trained feature extraction model; The feature mapping module is used to linearly map the first feature vector using pre-configured feature mapping parameters to obtain the second feature vector. The similarity calculation module is used to obtain the features of multiple core samples from the core database and calculate the similarity between the second feature vector and each core sample feature. The sample selection module is used to select the top K core samples with the highest similarity to the second feature vector to form a core sample set; where K is a positive integer. The lithofacies identification module is used to output lithofacies identification results by weighting the K core samples in the core sample set according to their similarity to the second feature vector.
[0070] This application provides a petrographic identification system based on core tags and similarity, which is similar to the one described above. Figure 1 The lithofacies identification method based on core tags and similarity shown is a technical solution under the same inventive concept. Through the detailed description of the lithofacies identification method based on core tags and similarity provided in the above embodiments, those skilled in the art can clearly understand the implementation process of each module of the lithofacies identification system based on core tags and similarity in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.
[0071] Accordingly, this embodiment provides a lithofacies identification system based on core labels and similarity, which combines deep feature modeling, feature mapping space optimization, and a similarity voting mechanism to achieve automatic lithofacies classification under weakly supervised or unsupervised conditions. This method fully utilizes the existing expert interpretations and the distributive separability between a small number of core samples to construct a lithofacies classification process that is both accurate and interpretable, possessing advantages such as lightweight deployment, high portability, and strong engineering adaptability.
[0072] In some embodiments, the feature mapping module is used for: Obtain raw well logging data and labels characterizing the lithofacies category of each sampling point at each depth in each well; Sliding sampling is performed along the depth direction of each original logging well according to the preset depth width and step size to obtain the core sample sequence; The core sample sequence and label are combined to obtain a set of core sample pairs; A feature extraction model was obtained by training a pre-constructed bidirectional long short-term memory neural network using a set of core sample pairs. The original feature vector of the core sample sequence was extracted using a feature extraction model; By introducing pre-configured feature mapping parameters, the original feature vector is linearly mapped to obtain the original feature mapping vector; The similarity between core samples is calculated based on the original feature mapping vectors, and the intra-class sample pair set and inter-class sample pair set of core samples are constructed based on the similarity between core samples. Establish a first objective function that maximizes the Wasserstein distance between the set of intra-class sample pairs and the set of inter-class sample pairs; Solve the first objective function to obtain the parameter values of the pre-configured feature mapping parameters.
[0073] In some embodiments, the feature mapping module is further configured to: Obtain raw well logging data and labels characterizing the lithofacies category of each sampling point at each depth in each well; Sliding sampling is performed along the depth direction of each original logging well according to the preset depth width and step size to obtain the core sample sequence; The core sample sequence and label are combined to obtain a set of core sample pairs; A feature extraction model was obtained by training a bidirectional long short-term memory neural network using a core sample set. The original feature vector of the core sample sequence was extracted using a feature extraction model; By introducing pre-configured feature mapping parameters, the original feature vector is linearly mapped to obtain the original feature mapping vector; The similarity between core samples is calculated based on the original feature mapping vectors, and the intra-class sample pair set and inter-class sample pair set of core samples are constructed based on the similarity between core samples. Establish a second objective function that maximizes the Sinkhorn distance between the set of intra-class sample pairs and the set of inter-class sample pairs; Solve the second objective function to obtain the parameter values of the pre-configured feature mapping parameters.
[0074] This invention also provides an electronic device. The electronic device includes a processor, a memory, a communication interface, and at least one communication bus for connecting the processor, the memory, and the communication interface. The memory includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (PROM), or portable read-only memory (CD-ROM), and is used for related instructions and data.
[0075] The communication interface is used to receive and send data. The processor can be one or more CPUs; if the processor is a single CPU, it can be a single-core CPU or a multi-core CPU. The processor in the electronic device reads one or more programs stored in memory and performs the following operations: acquires the logging data to be identified; extracts a first feature vector from the logging data to be identified using a pre-trained feature extraction model; linearly maps the first feature vector using pre-configured feature mapping parameters to obtain a second feature vector; acquires features from multiple core samples in the core database and calculates the similarity between the second feature vector and each core sample feature; selects the K core samples with the highest similarity to the second feature vector to form a core sample set; where K is a positive integer; and for the K core samples in the core sample set, performs a weighted voting based on their similarity to the second feature vector and outputs the lithofacies identification result.
[0076] It should be noted that the specific implementation of each operation can be described above. Figure 1 The corresponding description of the method embodiments shown indicates that the electronic device can be used to execute a lithofacies identification method based on core tags and similarity of the above-described method embodiments of this application, which will not be described in detail here.
[0077] This invention also provides a computer-readable storage medium, which is a memory device in a computer device for storing programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the operating system of the terminal. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the above embodiments regarding a method for identifying petrographic features based on core tags and similarity. Those skilled in the art should understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0078] This invention also provides a computer program product containing program instructions. The computer program product may be software or program products containing program instructions, capable of running on a computing device or stored on any usable medium. When the computer program product is run on at least one electronic device, it causes the at least one electronic device to execute a petrographic identification method based on core tagging and similarity.
[0079] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A lithofacies identification method based on core tags and similarity, characterized in that the method... include: Acquire the logging data to be identified; The first feature vector of the well logging data to be identified is extracted using a pre-trained feature extraction model; The first feature vector is linearly mapped using pre-configured feature mapping parameters to obtain the second feature vector; The features of multiple core samples from the core database are obtained, and the similarity between the second feature vector and the features of each core sample is calculated. The top K core samples with the highest similarity to the second feature vector are selected to form a core sample set; where K is a positive integer. For K core samples in the core sample set, the lithofacies identification result is output by weighted voting based on their similarity to the second feature vector.
2. The method according to claim 1, characterized in that, The pre-configured feature mapping parameter determination process includes: Obtain raw well logging data and labels characterizing the lithofacies category of each sampling point at each depth in each well; Sliding sampling is performed along the depth direction of each original logging well according to the preset depth width and step size to obtain the core sample sequence; The core sample sequence and label are combined to obtain a set of core sample pairs; A feature extraction model was obtained by training a pre-constructed bidirectional long short-term memory neural network using a set of core sample pairs. The original feature vector of the core sample sequence was extracted using a feature extraction model; By introducing pre-configured feature mapping parameters, the original feature vector is linearly mapped to obtain the original feature mapping vector; The similarity between core samples is calculated based on the original feature mapping vectors, and the intra-class sample pair set and inter-class sample pair set of core samples are constructed based on the similarity between core samples. Establish a first objective function that maximizes the Wasserstein distance between the set of intra-class sample pairs and the set of inter-class sample pairs; Solve the first objective function to obtain the parameter values of the pre-configured feature mapping parameters.
3. The method according to claim 1, characterized in that, The pre-configured feature mapping parameter determination process also includes: Obtain raw well logging data and labels characterizing the lithofacies category of each sampling point at each depth in each well; Sliding sampling is performed along the depth direction of each original logging well according to the preset depth width and step size to obtain the core sample sequence; The core sample sequence and label are combined to obtain a set of core sample pairs; A feature extraction model was obtained by training a bidirectional long short-term memory neural network using a core sample set. The original feature vector of the core sample sequence was extracted using a feature extraction model; By introducing pre-configured feature mapping parameters, the original feature vector is linearly mapped to obtain the original feature mapping vector; The similarity between core samples is calculated based on the original feature mapping vectors, and the intra-class sample pair set and inter-class sample pair set of core samples are constructed based on the similarity between core samples. Establish a second objective function that maximizes the Sinkhorn distance between the set of intra-class sample pairs and the set of inter-class sample pairs; Solve the second objective function to obtain the parameter values of the pre-configured feature mapping parameters.
4. The method according to any one of claims 1 to 3, characterized in that, The similarity is cosine similarity.
5. The method according to claim 1, characterized in that, To obtain the characteristics of multiple core samples from the core database, specifically: The third feature vector of each core sample is extracted using a pre-trained feature extraction model; The third feature vector is linearly mapped according to the pre-configured feature mapping parameters to obtain the core sample features.
6. A lithofacies identification system based on core tags and similarity, characterized in that, The system includes: The data acquisition module is used to acquire the logging data to be identified. The feature extraction module is used to extract the first feature vector of the well logging data to be identified using a pre-trained feature extraction model; The feature mapping module is used to linearly map the first feature vector using pre-configured feature mapping parameters to obtain the second feature vector. The similarity calculation module is used to obtain the features of multiple core samples from the core database and calculate the similarity between the second feature vector and each core sample feature. The sample selection module is used to select the top K core samples with the highest similarity to the second feature vector to form a core sample set; where K is a positive integer. The lithofacies identification module is used to output lithofacies identification results by weighting the K core samples in the core sample set according to their similarity to the second feature vector.
7. The system according to claim 6, characterized in that, The feature mapping module is used for: Obtain raw well logging data and labels characterizing the lithofacies category of each sampling point at each depth in each well; Sliding sampling is performed along the depth direction of each original logging well according to the preset depth width and step size to obtain the core sample sequence; The core sample sequence and the tag are combined to obtain a set of core sample pairs; A feature extraction model was obtained by training a pre-constructed bidirectional long short-term memory neural network using a set of core sample pairs. The original feature vectors of the core sample sequence were extracted using a feature extraction model; By introducing pre-configured feature mapping parameters, the original feature vector is linearly mapped to obtain the original feature mapping vector; The similarity between core samples is calculated based on the original feature mapping vectors, and the intra-class sample pair set and inter-class sample pair set of core samples are constructed based on the similarity between core samples. Establish a first objective function that maximizes the Wasserstein distance between the set of intra-class sample pairs and the set of inter-class sample pairs; Solve the first objective function to obtain the parameter values of the pre-configured feature mapping parameters.
8. In one implementation, the feature mapping module is further used for: Obtain raw well logging data and labels characterizing the lithofacies category of each sampling point at each depth in each well; Sliding sampling is performed along the depth direction of each original logging well according to the preset depth width and step size to obtain the core sample sequence; The core sample sequence and the tag are combined to obtain a set of core sample pairs; A feature extraction model was obtained by training a bidirectional long short-term memory neural network using a core sample set. The original feature vectors of the core sample sequence were extracted using a feature extraction model; By introducing pre-configured feature mapping parameters, the original feature vector is linearly mapped to obtain the original feature mapping vector; The similarity between core samples is calculated based on the original feature mapping vectors, and the intra-class sample pair set and inter-class sample pair set of core samples are constructed based on the similarity between core samples. Establish a second objective function that maximizes the Sinkhorn distance between the set of intra-class sample pairs and the set of inter-class sample pairs; Solve the second objective function to obtain the parameter values of the pre-configured feature mapping parameters.
9. An electronic device, characterized in that, Including memory and processor; A memory for storing computer programs, the computer programs including program instructions; A processor is configured to execute the program instructions to cause the electronic device to perform the steps of a petrographic identification method based on core tags and similarity as described in any one of claims 1 to 5.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a computer program that, when executed by one or more processors, implements a lithofacies identification method based on core tags and similarity as described in any one of claims 1 to 5.