Lithology identification method, apparatus and device, and computer readable storage medium

Abstract

The invention discloses a lithology identification method, apparatus and device, and a computer readable storage medium, which are applied to the technical field of geophysics, and the method comprises the following steps: constructing a lithology identification model based on a KAN network and wavelet transform in advance; an activation function of the lithology identification model is a wavelet function; obtaining logging data of the target area; and inputting the logging data into the lithology identification model to obtain an identification result. According to the method, the lithology identification model is obtained through wavelet transform and KAN network construction, the model can dynamically capture the high-frequency component and the low-frequency component of the logging data by using the multi-scale representation capability of wavelet transform, complex data necessary for accurate lithology classification can be effectively extracted, and manual intervention is not needed any more. Therefore, the model has robustness and accuracy, lithology under various geological conditions can be automatically analyzed, complex geological environments can be coped, various lithology can be accurately predicted on a large scale, and exploration and development risks are reduced.

Description

Technical Field

[0001] The present invention relates to the field of geophysical technology, and in particular to a lithology identification method, device, equipment and computer-readable storage medium. Background Art

[0002] Lithology identification is fundamental in oil and gas exploration and development. Its core objective is to accurately determine the composition and distribution of underground rock formations through a comprehensive analysis of rock physical properties, geochemical characteristics, and sedimentary structures. Existing lithology identification methods include traditional methods and deep learning approaches. Traditional lithology identification methods suffer from drawbacks such as high subjectivity, low efficiency, insufficient resolution, poor accuracy, and technical limitations. While existing deep learning (DL) methods can improve efficiency, their limited adaptability in complex geological conditions limits their accuracy.

[0003] Therefore, how to provide a high-efficiency, high-precision, and highly adaptable lithology identification method is a technical problem that urgently needs to be solved. Summary of the Invention

[0004] In view of this, an object of the present invention is to provide a lithology identification method, device, equipment and computer-readable storage medium, which solve the problems of poor adaptability, low efficiency and poor accuracy in lithology identification in the prior art.

[0005] To solve the above technical problems, the present invention provides a lithology identification method, comprising: A lithology identification model is constructed in advance based on a KAN network and wavelet transform; the activation function of the lithology identification model is a wavelet function; Obtain well logging data for the target area; The logging data is input into the lithology identification model to obtain an identification result.

[0006] Optionally, the well logging data is input into the lithology identification model to obtain an identification result, including: Inputting the well logging data into the embedded layer in the lithology identification model to obtain a mapping result; Inputting the mapping result into a feature extraction module stacked in the lithology identification model to obtain feature components; the feature extraction module utilizes the multi-resolution decomposition and sparse representation characteristics of the wavelet function; Inputting the characteristic components into the pooling layer in the lithology identification model to obtain a prediction result; The prediction result is input into the linear classifier in the lithology identification model to obtain the identification result.

[0007] Optionally, the mapping result is input into a feature extraction module stacked in the lithology identification model to obtain feature components, including: Performing wavelet decomposition on the mapping result to obtain decomposed components; the decomposed components include low-frequency components and high-frequency components; The decomposed components are reconstructed using a learnable activation function and an inverse wavelet transform to obtain the characteristic components.

[0008] Optionally, wavelet decomposition is performed on the mapping result to obtain decomposed components, including: ; ; ; in, It is the intermediate result after the wavelet decomposition of the input sample of the lth layer; is the mapping result corresponding to the input sample of the lth layer; is the wavelet transform function; and They represent the high-frequency component and low-frequency component of the input sample of the lth layer after wavelet transformation respectively; Indicates the number of grids in the feature extraction module; g is the coefficient of the low-pass filter; h is the coefficient of the high-pass filter; Represents the value of the input sample in the nth row and the ith column; H is the number of input samples.

[0009] Optionally, reconstructing the decomposed components using a learnable activation function and an inverse wavelet transform to obtain the characteristic components includes: ; ; ; in, is the characteristic component; is the feature vector reconstructed by inverse wavelet transform; is the learnable activation function bias; is the inverse wavelet transform; is the weight component of the learnable activation function weight in the kth grid; is the intermediate result in the kth grid after the input sample is decomposed by wavelet; is the high-frequency component after wavelet decomposition corresponding to the k-th grid in the l-th layer; is the learnable activation function weight corresponding to the high-frequency component; is the low-frequency component after wavelet decomposition corresponding to the k-th grid in the l-th layer; is the learnable activation function weight corresponding to the low-frequency component; is the learnable weight matrix used to calculate the activation function bias in layer l.

[0010] Optionally, the logging data includes: natural gamma ray logging data, density logging data, compensated neutron logging data, sonic time difference logging data, deep lateral resistivity logging data and shallow lateral resistivity logging data.

[0011] Optionally, after acquiring the well logging data of the target area, the well logging data further includes: Cleaning and standardizing the well logging data; The standardization formula is: ; is the normalized logging data; X is the original logging data; min(X) and max(X) are the minimum and maximum values in the logging data set.

[0012] The present invention also provides a lithology identification device, comprising: A model building module is used to construct a lithology identification model based on a KAN network and wavelet transform in advance; the activation function of the lithology identification model is a wavelet function; Well logging data acquisition module, used to obtain well logging data of the target area; The lithology identification module is used to input the logging data into the lithology identification model to obtain an identification result.

[0013] The present invention also provides a lithology identification device, comprising: memory for storing computer programs; A processor is used to implement the above-mentioned lithology identification method when executing the computer program.

[0014] The present invention also provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are loaded and executed by a processor, the above-mentioned lithology identification method is implemented.

[0015] It can be seen that the present invention obtains a lithologic identification model by pre-constructing it based on the KAN network and wavelet transform; the activation function of the lithologic identification model is a wavelet function; well logging data of the target area is obtained; the well logging data is input into the lithologic identification model to obtain an identification result. The present invention obtains a lithologic identification model by constructing it through wavelet transform and KAN network. The model can utilize the multi-scale representation capability of wavelet transform to dynamically capture the high-frequency and low-frequency components of the well logging data, effectively extracting the complex data necessary for accurate lithologic classification, and eliminating the need for human intervention. Therefore, the model is robust and accurate, capable of automatically analyzing lithologies under various geological conditions, and can also cope with complex geological environments, accurately predict various lithologies on a large scale, and reduce the risks of exploration and development.

[0016] In addition, the present invention also provides a lithology identification device, equipment and computer-readable storage medium, which also have the above-mentioned beneficial effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0018] Figure 1 A flowchart of a lithology identification method provided by an embodiment of the present invention; Figure 2 A schematic diagram of a KAN network; Figure 3 A schematic diagram of a lithology identification model provided in an embodiment of the present invention; Figure 4 A comparison chart of the accuracy of lithology identification based on a volcanic reservoir dataset provided by an embodiment of the present invention; Figure 5 A comparison chart of the accuracy of lithology identification based on a clastic reservoir dataset of gas field A provided in an embodiment of the present invention; Figure 6 A comparison chart of the accuracy of lithology identification based on a clastic reservoir dataset of gas field B provided in an embodiment of the present invention; Figure 7 A schematic structural diagram of a lithology identification device provided by an embodiment of the present invention; Figure 8 A structural schematic diagram of a lithology identification device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0020] Lithology identification is fundamental in oil and gas exploration and development. Its core is to accurately determine the composition and distribution patterns of underground rock formations through a comprehensive analysis of rock physical properties, geochemical characteristics, and sedimentary structures. In the exploration field, lithology identification can effectively reduce the exploration risks brought about by geological uncertainty. During reservoir characterization, lithology identification constructs a reservoir spatial distribution model through techniques such as well logging curve inversion and 3D seismic attribute analysis. In development plan design, lithology distribution characteristics directly determine well pattern deployment and fracturing parameters. The application of this technology runs throughout the entire life cycle of oil and gas reservoir discovery, reserve assessment, and development plan formulation, directly impacting exploration success rates and development benefits.

[0021] Traditional lithologic identification methods suffer from strong subjectivity, low efficiency, and technical limitations. Under complex geological conditions, they lack resolution for rocks with similar mineral composition or strong heterogeneity (such as volcanic rocks and carbonates), making them ineffective in characterizing microscopic pore structures and impacting the accuracy of unconventional reservoir evaluation. With the advancement of artificial intelligence (AI), deep learning-based lithologic identification methods have significantly improved their interpretability and efficiency, significantly enhancing evaluation accuracy. In the context of energy transition, lithologic identification technology is being deeply integrated with AI, continuously unlocking its core value in the efficient development of oil and gas resources. However, existing deep learning methods suffer from fixed activation functions and limited multi-scale feature extraction. To address these issues, this paper proposes an automatic lithologic identification method that replaces static activation functions with learnable wavelet basis functions (also known as wavelet functions) to dynamically adapt to logging data patterns. Unlike traditional deep learning architectures, this lithologic identification model integrates a wavelet-based feature extraction module to jointly capture high-frequency details (such as lithologic boundaries) and low-frequency trends. Even under complex geological conditions, highly heterogeneous rocks such as volcanic rocks and carbonate rocks can still be efficiently and accurately identified. This method can be widely used in major oil fields in China to accurately identify various complex lithologies, reduce exploration and development risks, and ensure the profitable development of major oil fields.

[0022] Please refer to Figure 1 , Figure 1 A flowchart of a lithology identification method provided in an embodiment of the present invention. The method may include: S101: A lithology recognition model is constructed in advance based on a KAN network and wavelet transform; the activation function of the lithology recognition model is a wavelet function.

[0023] The execution entity of this embodiment is a terminal. This embodiment does not limit the type of terminal; any terminal capable of performing the lithology identification method is sufficient. The KAN (Kolmogorov-Arnold) network is a novel neural network architecture inspired by the Kolmogorov-Arnold Representation Theorem (KART). This states that any multivariate continuous function can be represented as a finite combination of single-variable continuous functions and a binary operation called addition. Figure 2 This diagram illustrates the principles of a KAN model. Unlike previous deep learning models, KAN networks utilize learnable spline parameterized functions rather than fixed activation functions, allowing for more flexible and interpretable high-dimensional function representations. Specifically, KAN networks shift activation functions from nodes (neurons) to edges (weights), replacing each weight parameter with a learnable univariate function (e.g., a spline function). This approach improves the interpretability of deep learning while also enhancing the accuracy and applicability of function fitting tasks.

[0024] Furthermore, given the inherent limitations of B-spline functions, such as their reliance on local approximation and mesh refinement, KAN network training can be complicated and may lead to overfitting or underfitting, particularly when data exhibit abrupt or high-frequency variations. Consequently, their scalability and versatility in well logging interpretation in the field of petroleum geology are significantly limited. Wavelet functions, on the other hand, offer superior localization and multi-resolution analysis capabilities, making them highly effective in capturing local variations in data at multiple scales. These characteristics make wavelet functions particularly well-suited for handling the inherent complexity of well logging data. Therefore, this embodiment replaces B-spline functions with wavelet functions. Specifically, this embodiment utilizes wavelet functions as the basis functions of the wavelet transform, performing multi-scale signal processing and extracting local features. This allows the optimized KAN network (i.e., the LogWKAN model, also referred to as the lithology identification model in this embodiment) to more effectively learn and represent local features, particularly in datasets with sharp transitions or non-stationary behavior. This replacement can significantly improve model performance in such scenarios.

[0025] Figure 3 This diagram illustrates the principles of a lithology identification model provided by an embodiment of the present invention. The lithology identification model is a wavelet-based KAN well logging data network, a novel model that uses learnable wavelet functions to efficiently process and extract frequency domain features. The lithology identification model's workflow includes the following key components: an embedding layer, a Kolmogorov-Arnold (WKA) module (the feature extraction module in this embodiment), a normalization layer, and a linear classifier. This model extracts features by stacking wavelet-based WKA modules.

[0026] S102: Acquire well logging data of the target area.

[0027] Data selection and collection are the core of ensuring the accuracy and reliability of model lithology identification. Reasonable screening of logging parameters can enhance the geological pertinence of the model and avoid redundant interference. In addition, high-quality logging data should be collected to avoid the influence of equipment and human factors. Such logging data can truly reflect the characteristics of the formation and reduce noise and errors. Systematic data collection needs to cover multi-dimensional information of the target layer to ensure the representativeness and integrity of the sample, thereby improving the accuracy of model evaluation. Therefore, in this embodiment, the logging data selected are six types of logging data related to lithology: natural gamma ray logging (GR), density logging (DEN), compensated neutron logging (CNL), acoustic time difference logging (AC), deep lateral resistivity logging (LLD) and shallow lateral resistivity logging (LLS).

[0028] It should be noted that the collected logging data is input into the model in the form of a sequence. , where H is the sliding window size and P is the number of parameters in the logging sequence. In deep learning, the size of the sliding window is a key parameter that directly affects the model's perception range, feature abstraction ability, and computational efficiency. The sliding window size directly determines the area range of the input data that the model can "see" in a single operation. The sliding window size can be selected by first selecting the default value and then gradually increasing the window to observe the model effect and computational cost, and then selecting the appropriate sliding window size for the study. It can be understood that the sliding window size in this embodiment is the amount of data input to the model.

[0029] Furthermore, data preprocessing is also a key step in ensuring model performance. Raw well logging data often contains noise, missing values, or redundant information. Methods such as cleaning, normalization, and feature engineering can improve data quality and information density, thereby reducing model bias. Appropriate feature selection can also eliminate interference from irrelevant variables and enhance model generalization. Standardization accelerates algorithm convergence and improves prediction efficiency. Preprocessing also involves data balancing and enhancement to avoid overfitting and underfitting. High-quality data preprocessing lays a solid foundation for model construction and directly impacts the accuracy and practical application value of identification results. Because different subsurface lithologies produce different characteristics in the collected well logging data, various logging parameters collected for reservoirs of different lithologies can vary significantly due to different physical principles. To mitigate prediction errors caused by these differences, well logging data must be normalized, as shown in the following formula, to ensure uniform input samples for subsequent analysis.

[0030] The standardization formula is: ; is the normalized logging data; X is the original logging data; min(X) and max(X) are the minimum and maximum values in the logging data set.

[0031] S103: Input the logging data into the lithology identification model to obtain the identification results.

[0032] It should be noted that the lithology identification model in this embodiment is the model constructed in step S101 above, which includes an embedding layer, a feature extraction module, and a linear classifier.

[0033] Furthermore, the above-mentioned inputting of the logging data into the lithologic identification model to obtain the identification result may include the following steps, which may specifically include: Step 11: Input the well logging data into the embedding layer in the lithology identification model to obtain the mapping results.

[0034] Specifically, the well logging data first passes through the model's embedding layer. The embedding layer maps the input data from the P-dimensional space to the model's hidden dimension D, capturing the complex nonlinear relationships and potential features in the well logging data. The mapping process is shown in the following equation: ; in, Represents the transformed data after the embedding layer; represents the weight matrix of the embedding layer; represents the bias of the embedding layer; X in Represents the input data before the embedding layer.

[0035] Step 12: Input the mapping results into the stacked feature extraction module in the lithology identification model to obtain feature components; the feature extraction module utilizes the multi-resolution decomposition and sparse representation characteristics of the wavelet function.

[0036] Specifically, the lithology identification model improves its performance through stacked WKA modules. These modules exploit the multiresolution decomposition and sparse representation properties of wavelet functions. This enhances the lithology identification model's ability to capture both high- and low-frequency components, enabling more flexible processing of complex well logging sequences.

[0037] Furthermore, the above-mentioned inputting of the mapping results into the stacked feature extraction module in the lithology identification model to obtain the feature components may include the following steps: Step 121: performing wavelet decomposition on the mapping result to obtain decomposed components; the decomposed components include low-frequency components and high-frequency components; Step 122: Reconstruct the decomposed components using a learnable activation function and an inverse wavelet transform to obtain feature components.

[0038] Specifically, the WKA module integrates the wavelet transform as a learnable activation function into the KAN network to learn the frequency domain characteristics of well logging data. The WKA module consists of three parts: wavelet decomposition of the well logging sequence, a learnable activation function for the wavelet decomposition result, and inverse wavelet reconstruction based on the activation result.

[0039] The above mapping results are subjected to wavelet decomposition to obtain the decomposed components. The specific formula is: ; ; ; in, It is the intermediate result after the wavelet decomposition of the input sample of the lth layer; is the mapping result corresponding to the input sample of the lth layer; is the wavelet transform function; and They represent the high-frequency component and low-frequency component of the input sample of the lth layer after wavelet transformation respectively; Indicates the number of grids in the feature extraction module; g is the coefficient of the low-pass filter; h is the coefficient of the high-pass filter; Represents the value of the input sample in the nth row and the ith column; H is the number of input samples.

[0040] The result of wavelet transform is The learnable activation function of the WKA module is mapped to the nonlinear function space to obtain a high-dimensional feature vector, so that each component corresponds to the activation value under the input, so as to achieve efficient approximation and interpretability of complex continuous multivariate functions.

[0041] Specifically, the above-mentioned learnable activation function and inverse wavelet transform are used to reconstruct the decomposed components to obtain feature components, including: ; ; ; in, is the characteristic component; is the feature vector reconstructed by inverse wavelet transform; is the learnable activation function bias; is the inverse wavelet transform; is the weight component of the learnable activation function weight in the kth grid; is the intermediate result in the kth grid after the input sample is decomposed by wavelet; is the high-frequency component after wavelet decomposition corresponding to the k-th grid in the l-th layer; is the learnable activation function weight corresponding to the high-frequency component; is the low-frequency component after wavelet decomposition corresponding to the k-th grid in the l-th layer; is the learnable activation function weight corresponding to the low-frequency component; is the learnable weight matrix used to calculate the activation function bias in layer l.

[0042] As can be seen, the LogWKAN model leverages the sparse representation properties of the wavelet transform to capture both high- and low-frequency information. By introducing learnable complex-valued weights, the model can dynamically adjust filter parameters, enhancing its ability to approximate complex well logging data. Furthermore, the sparse representation provided by the wavelet transform enhances the expression and generalization of local features.

[0043] Step 13: Input the characteristic components into the pooling layer in the lithology recognition model to obtain the prediction results.

[0044] For details, please refer to Figure 3 , Figure 3 Except for the last WKA module, all features extracted are processed by the BatchNorm (normalization) layer, which can improve the stability of the lithology recognition model and alleviate the gradient disappearance or explosion problem. It is sent to the AvgPooling layer.

[0045] The normalization formula is: ; Where BN(·) represents the normalization operation in the BatchNorm layer; is the output of the (l-1)th WKA module; is the output of layer 1 (including the BatchNorm layer), where .

[0046] The global average pooling GAP(·) formula is: .

[0047] in, Represents the prediction result; GAP(·) represents the pooling layer; D represents the dimension length of the pooling operation, that is, the summation results are averaged to obtain the global representative value of each channel.

[0048] Step 14: Input the prediction results into the linear classifier in the lithology recognition model to obtain the recognition results.

[0049] Specifically, the final output layer is classified by a linear classifier Φ(·): .

[0050] in, Indicates the recognition result; represents the classifier weight, and C is the total number of lithology categories in the well logging dataset.

[0051] The lithology identification method provided in the embodiment of the present invention is applied. A lithology identification model is pre-constructed based on a KAN network and wavelet transform. The activation function of the lithology identification model is a wavelet function. Well logging data of the target area is obtained. The well logging data is input into the lithology identification model to obtain an identification result. The present invention constructs a lithology identification model through wavelet transform and KAN network. The model can utilize the multi-scale representation capability of wavelet transform to dynamically capture the high-frequency and low-frequency components of the well logging data, effectively extracting the complex data necessary for accurate lithology classification without the need for human intervention. Therefore, the model is robust and accurate, capable of automatically analyzing lithologies under various geological conditions, and can also cope with complex geological environments, accurately predict various lithologies on a large scale, and reduce the risks of exploration and development.

[0052] To verify the feasibility of the above method, clastic reservoir data from Gas Fields A and B, and a volcanic reservoir dataset from Gas Field C were used. This dataset contains six logging data items related to lithology: gamma ray logging (GR), density logging (DEN), compensated neutron logging (CNL), acoustic transit time (AC), deep lateral resistivity logging (LLD), and shallow lateral resistivity logging (LLS). The LogWKAN model was tested and compared with four currently most accurate and mainstream deep learning methods: convolutional neural networks (CNNs), Transformers (deep neural network architectures based on self-attention mechanisms), multilayer perceptrons (MLPs), and recurrent neural networks (RNNs). The branching algorithms evolved from Convolutional Neural Network (CNN) in processing sequence data, namely, one-dimensional convolutional neural network (1DCNN) and temporal convolutional network (TCN), Transformers (deep neural network structure based on self-attention mechanism) branching algorithm vision transformer (ViT) and information model (Informer), multi-layer perceptron (MLP) branching algorithm multi-layer perceptron hybrid model (MLP-Mixer), time series mixture model (Time-Series Mixer, TSMixer) and neural basis expansion time series analysis model (Neural Basis Expansion Analysis for Interpretable Time Series, N-BEATS), and recurrent neural network (RNN) branching algorithm long short-term memory network (Long Short-Term Memory, LSTM) and gated recurrent unit network (GRU) and LogWKAN model were compared. The results are shown in Tables 1, 2 and 3, as well as Figure 4 、 Figure 5 and Figure 6 As shown in the figure, this method is more effective in predicting lithology across various rock reservoir types in oil fields. It is robust to unbalanced datasets, can handle complex geological environments, accurately predict various lithologies on a large scale, and reduce exploration and development risks.

[0053] Table 1 Comparison of lithology identification accuracy based on volcanic reservoir dataset

[0054]

[0055] Table 2 Comparison of lithology identification accuracy based on the clastic reservoir dataset of gas field A

[0056]

[0057] Table 3 Comparison of lithology identification accuracy based on the clastic reservoir dataset of gas field B

[0058]

[0059] Based on three different datasets, different models were used to predict lithology, testing their performance. The common model evaluation method, OA (accuracy), was used. Analysis was also conducted on single-well logging profiles. It is intuitively clear that, regardless of geological conditions or lithology, this method outperforms the other four deep learning models in recognition performance and accuracy. Therefore, the lithology recognition model provided by this invention is more secure and reliable than other models. It has demonstrated excellent results in lithology prediction tasks for various rock reservoir types in oil fields. It maintains robustness against unbalanced datasets, can cope with complex geological environments, and accurately predicts various lithologies on a large scale, reducing exploration and development risks.

[0060] The following is an introduction to a lithology identification device provided by an embodiment of the present invention. The lithology identification device described below and the lithology identification method described above can be referenced to each other.

[0061] Please refer to Figure 7 , Figure 7 A schematic structural diagram of a lithology identification device provided in an embodiment of the present invention may include: The model construction module 100 is used to construct a lithology identification model based on the KAN network and wavelet transform in advance; the activation function of the lithology identification model is a wavelet function; The well logging data acquisition module 200 is used to acquire well logging data of the target area; The lithology identification module 300 is used to input the well logging data into the lithology identification model to obtain an identification result.

[0062] Based on the above embodiment, the lithology identification module 300 may include: an embedding unit, configured to input the well logging data into an embedding layer in the lithology identification model to obtain a mapping result; A feature extraction unit, configured to input the mapping result into a feature extraction module stacked in the lithology identification model to obtain a feature component; the feature extraction module utilizes the multi-resolution decomposition and sparse representation characteristics of the wavelet function; A pooling unit, configured to input the characteristic components into a pooling layer in the lithology identification model to obtain a prediction result; The classification unit is used to input the prediction result into the linear classifier in the lithology identification model to obtain the identification result.

[0063] Based on the above embodiment, the feature extraction unit may include: A decomposition subunit, configured to perform wavelet decomposition on the mapping result to obtain decomposed components; the decomposed components include low-frequency components and high-frequency components; The reconstruction subunit is used to reconstruct the decomposed components using a learnable activation function and an inverse wavelet transform to obtain the characteristic components.

[0064] Based on the above embodiment, the decomposition subunits may include: ; ; ; in, It is the intermediate result after the wavelet decomposition of the input sample of the lth layer; is the mapping result corresponding to the input sample of the lth layer; is the wavelet transform function; and They represent the high-frequency component and low-frequency component of the input sample of the lth layer after wavelet transformation respectively; Indicates the number of grids in the feature extraction module; g is the coefficient of the low-pass filter; h is the coefficient of the high-pass filter; Represents the value of the input sample in the nth row and the ith column; H is the number of input samples.

[0065] Based on the above embodiment, the reconstruction subunit may include: ; ; ; in, is the characteristic component; is the feature vector reconstructed by inverse wavelet transform; is the learnable activation function bias; is the inverse wavelet transform; is the weight component of the learnable activation function weight in the kth grid; is the intermediate result in the kth grid after the input sample is decomposed by wavelet; is the high-frequency component after wavelet decomposition corresponding to the k-th grid in the l-th layer; is the learnable activation function weight corresponding to the high-frequency component; is the low-frequency component after wavelet decomposition corresponding to the k-th grid in the l-th layer; is the learnable activation function weight corresponding to the low-frequency component; is the learnable weight matrix used to calculate the activation function bias in layer l.

[0066] Based on the above embodiment, the well logging data acquisition module may include: The acquisition unit is used to acquire natural gamma ray logging data, density logging data, compensated neutron logging data, sonic time difference logging data, deep lateral resistivity logging data and shallow lateral resistivity logging data.

[0067] Based on the above embodiment, the lithology identification device may include: A pre-processing unit, configured to clean and standardize the well logging data; The standardization formula is: ; is the normalized logging data; X is the original logging data; min(X) and max(X) are the minimum and maximum values in the logging data set.

[0068] It should be noted that the order of the modules and units in the above-mentioned lithology identification device can be changed without affecting the logic.

[0069] The lithology identification device provided by the embodiment of the present invention utilizes a model construction module 100 for pre-constructing a lithology identification model based on a KAN network and wavelet transform; the activation function of the lithology identification model is a wavelet function; a well logging data acquisition module 200 is used to acquire well logging data from a target area; and a lithology identification module 300 is used to input the well logging data into the lithology identification model to obtain an identification result. This device constructs a lithology identification model using wavelet transform and a KAN network. This model can leverage the multi-scale representation capabilities of the wavelet transform to dynamically capture high-frequency and low-frequency components of the well logging data, effectively extracting the complex data necessary for accurate lithology classification without the need for human intervention. Therefore, the model is robust and accurate, capable of automatically analyzing lithologies under various geological conditions. It can also cope with complex geological environments, accurately predict various lithologies on a large scale, and reduce exploration and development risks.

[0070] The following is an introduction to a lithology identification device provided by an embodiment of the present invention. The lithology identification device described below and the lithology identification method described above can be referenced to each other.

[0071] Please refer to Figure 8 , Figure 8 A structural diagram of a lithology identification device provided in an embodiment of the present invention may include: Memory 10, for storing computer programs; The processor 20 is configured to execute a computer program to implement the above-mentioned lithology identification method.

[0072] The memory 10 , the processor 20 , and the communication interface 31 all communicate with each other via the communication bus 32 .

[0073] In the embodiment of the present invention, the memory 10 is used to store one or more programs. The program may include program code, and the program code includes computer operation instructions. In the embodiment of the present invention, the memory 10 may store programs for implementing the following functions: A lithology recognition model is constructed in advance based on the KAN network and wavelet transform; the activation function of the lithology recognition model is a wavelet function; Obtain well logging data for the target area; The logging data is input into the lithology identification model to obtain the identification results.

[0074] In one possible implementation, the memory 10 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and applications required for at least one function, etc.; the data storage area may store data created during use.

[0075] In addition, the memory 10 may include a read-only memory and a random access memory, and provides instructions and data to the processor. A portion of the memory may also include NVRAM. The memory stores an operating system and operating instructions, executable modules or data structures, or a subset or an extended set thereof. The operating instructions may include various operating instructions for implementing various operations. The operating system may include various system programs for implementing various basic tasks and processing hardware-based tasks.

[0076] The processor 20 may be a central processing unit (CPU), an application-specific integrated circuit, a digital signal processor, a field programmable gate array, or other programmable logic device. The processor 20 may be a microprocessor or any conventional processor. The processor 20 may call a program stored in the memory 10 .

[0077] The communication interface 31 may be an interface of a communication module, used for connecting to other devices or systems.

[0078] Of course, it needs to be explained that Figure 8The structure shown does not constitute a limitation on the lithology identification device in the embodiment of the present invention. In actual application, the lithology identification device may include Figure 8 More or fewer components than shown, or combinations of certain components.

[0079] The computer-readable storage medium provided by an embodiment of the present invention is introduced below. The computer-readable storage medium described below and the lithology identification method described above can be referenced to each other.

[0080] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the above-mentioned lithology identification method are implemented.

[0081] The computer-readable storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., which can store program codes.

[0082] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from the other embodiments. Reference can be made to the descriptions of the identical or similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and the relevant parts can be referred to the descriptions of the methods.

[0083] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0084] Finally, it should be noted that, in this document, relationships such as first and second, etc., are used solely to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0085] The above is a detailed introduction to the lithologic identification method, device, equipment and computer-readable storage medium provided by the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the ideas of the present invention, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A lithology identification method, characterized in that: include: A lithology identification model is constructed in advance based on the KAN network and wavelet transform; The activation function of the lithology identification model is a wavelet function; Obtain well logging data for the target area; The logging data is input into the lithology identification model to obtain an identification result.

2. The lithology identification method according to claim 1, characterized in that: Inputting the logging data into the lithology identification model to obtain identification results includes: Inputting the well logging data into the embedded layer in the lithology identification model to obtain a mapping result; Inputting the mapping result into a feature extraction module stacked in the lithology identification model to obtain feature components; the feature extraction module utilizes the multi-resolution decomposition and sparse representation characteristics of the wavelet function; Inputting the characteristic components into the pooling layer in the lithology identification model to obtain a prediction result; The prediction result is input into the linear classifier in the lithology identification model to obtain the identification result.

3. The lithology identification method according to claim 2, characterized in that: The mapping results are input into the feature extraction module stacked in the lithology identification model to obtain feature components, including: Performing wavelet decomposition on the mapping result to obtain decomposed components; the decomposed components include low-frequency components and high-frequency components; The decomposed components are reconstructed using a learnable activation function and an inverse wavelet transform to obtain the characteristic components.

4. The lithology identification method according to claim 3, characterized in that: Perform wavelet decomposition on the mapping result to obtain the decomposed components, including: ； ； ； in, It is the intermediate result after the wavelet decomposition of the input sample of the lth layer; is the mapping result corresponding to the input sample of the lth layer; is the wavelet transform function; and They represent the high-frequency component and low-frequency component of the input sample of the lth layer after wavelet transformation respectively; Indicates the number of grids in the feature extraction module; g is the coefficient of the low-pass filter; h is the coefficient of the high-pass filter; Represents the value of the input sample in the nth row and the ith column; H is the number of input samples.

5. The lithology identification method according to claim 3, characterized in that: The decomposed components are reconstructed using a learnable activation function and an inverse wavelet transform to obtain the characteristic components, including: ； ； ； in, is the characteristic component; is the feature vector reconstructed by inverse wavelet transform; is the learnable activation function bias; is the inverse wavelet transform; is the weight component of the learnable activation function weight in the kth grid; is the intermediate result in the kth grid after the input sample is decomposed by wavelet; is the high-frequency component after wavelet decomposition corresponding to the k-th grid in the l-th layer; is the learnable activation function weight corresponding to the high-frequency component; is the low-frequency component after wavelet decomposition corresponding to the k-th grid in the l-th layer; is the learnable activation function weight corresponding to the low-frequency component; is the learnable weight matrix used to calculate the activation function bias in layer l.

6. The lithology identification method according to claim 1, characterized in that: The logging data includes natural gamma ray logging data, density logging data, compensated neutron logging data, acoustic time difference logging data, deep lateral resistivity logging data and shallow lateral resistivity logging data.

7. The lithology identification method according to claim 1, characterized in that: The well logging data, after acquiring the well logging data of the target area, further includes: Cleaning and standardizing the well logging data; The standardization formula is: ； is the normalized logging data; X is the original logging data; min(X) and max(X) are the minimum and maximum values in the logging data set.

8. A lithology identification device, characterized in that: include: A model building module is used to construct a lithology identification model based on a KAN network and wavelet transform in advance; the activation function of the lithology identification model is a wavelet function; Well logging data acquisition module, used to obtain well logging data of the target area; The lithology identification module is used to input the logging data into the lithology identification model to obtain an identification result.

9. A lithology identification device, characterized in that: include: memory for storing computer programs; A processor, configured to implement the lithology identification method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are loaded and executed by the processor, the lithology identification method according to any one of claims 1 to 7 is implemented.

Images