A lithology identification method, device, equipment and computer readable storage medium

Abstract

The application discloses a lithology identification method, device and equipment and a computer readable storage medium, and is applied to the field of geophysical technology, and comprises the following steps: a lithology identification model is constructed in advance based on a KAN network and wavelet transform; an activation function of the lithology identification model is a wavelet function; well logging data of a target area is acquired; and the well logging data is input into the lithology identification model to obtain an identification result. The lithology identification model is constructed through wavelet transform and the KAN network, the model can utilize the multiscale representation capability of the wavelet transform, dynamically capture high-frequency components and low-frequency components of the well logging data, effectively extract complex data necessary for accurate lithology classification, and no longer need manual intervention. Therefore, the model has robustness and precision, can automatically analyze the lithology under various geological conditions, can also cope with complex geological environments, can accurately predict various lithologies on a large scale, and reduces exploration and development risks.

Description

TECHNICAL FIELD

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

[0002] Lithology identification is a basic work in oil and gas exploration and development, and its core is to accurately determine the composition and distribution of underground rock layers through comprehensive analysis of petrophysical properties, geochemical characteristics and sedimentary structures. The existing lithology identification methods include traditional methods and deep learning methods. The traditional lithology identification method has the disadvantages of strong subjectivity, low efficiency, insufficient resolution, poor accuracy and technical limitations. Although the existing deep learning (DL) method can improve efficiency, its adaptability is poor, which limits the recognition accuracy under complex geological conditions.

[0003] Therefore, how to provide a lithology identification method with high efficiency, high accuracy and high adaptability is a technical problem to be solved at present. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a lithology identification method, device, equipment and computer readable storage medium, which solves the problems of poor adaptability, low efficiency and poor accuracy in the prior art.

[0005] To solve the above technical problems, the present application provides a lithology identification method, comprising:

[0006] A lithology identification model is constructed in advance based on a KAN network and wavelet transform; and an activation function of the lithology identification model is a wavelet function;

[0007] Logging data of a target area is obtained;

[0008] The logging data is input into the lithology identification model to obtain an identification result.

[0009] Optionally, inputting the logging data into the lithology identification model to obtain an identification result comprises:

[0010] The logging data is input into an embedding layer in the lithology identification model to obtain a mapping result;

[0011] The mapping result is input into a stacked feature extraction module in the lithology identification model to obtain a feature component; the feature extraction module utilizes multi-resolution decomposition and sparse representation characteristics of the wavelet function;

[0012] The feature component is input into a pooling layer in the lithology identification model to obtain a prediction result;

[0013] inputting the prediction result into a linear classifier in the lithology identification model to obtain the identification result.

[0014] Optionally, the mapping result is input into a stacked feature extraction module in the lithology identification model to obtain a feature component, including:

[0015] wavelet-decomposing the mapping result to obtain decomposed components; the decomposed components include low-frequency components and high-frequency components;

[0016] reconstructing the decomposed components by using a learnable activation function and an inverse wavelet transform to obtain the feature component.

[0017] Optionally, the mapping result is wavelet-decomposed to obtain decomposed components, including:

[0018] ;

[0019] ;

[0020] ;

[0021] wherein, is an intermediate result of the input sample of the lth layer after wavelet decomposition; is the mapping result corresponding to the input sample of the lth layer; is a wavelet transform function; and respectively represent high-frequency components and low-frequency components corresponding to the input sample of the lth layer after wavelet transform; represents the number of grids in the feature extraction module; g is a coefficient of a low-pass filter; h is a coefficient of a high-pass filter; represents a value of the input sample in the nth row and the ith column; H is the number of input samples.

[0022] Optionally, the decomposed components are reconstructed by using a learnable activation function and an inverse wavelet transform to obtain the feature component, including:

[0023] ;

[0024] ;

[0025] ;

[0026] wherein, is the feature component; is a feature vector reconstructed by the inverse wavelet transform; is a learnable activation function bias; is the inverse wavelet transform; is a weight component of a learnable activation function weight in the kth grid; is an intermediate result of the input sample after wavelet decomposition in the kth grid; is a high-frequency component after wavelet decomposition corresponding to the kth grid of the lth layer; is a learnable activation function weight corresponding to the high-frequency component; is a low-frequency component after wavelet decomposition corresponding to the kth grid of the lth layer; is a learnable activation function weight corresponding to the low-frequency component; is a learnable weight matrix for calculating an activation function bias of the lth layer.

[0027] Optionally, the logging data includes: natural gamma 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.

[0028] Optionally, after the logging data of the target area is acquired, the logging data further includes:

[0029] cleaning and standardizing the logging data;

[0030] The standardization processing formula is:

[0031] ;

[0032] is the logging data after standardization processing; X is the originally acquired logging data; min(X) and max(X) are the minimum value and the maximum value in the logging data set.

[0033] The application further provides a lithology identification device, which comprises:

[0034] a model construction module, used for constructing a lithology identification model in advance based on a KAN network and wavelet transformation; an activation function of the lithology identification model is a wavelet function;

[0035] a logging data acquisition module, used for acquiring logging data of a target area;

[0036] a lithology identification module, used for inputting the logging data into the lithology identification model to obtain an identification result.

[0037] The application further provides a lithology identification device, which comprises:

[0038] a memory, used for storing a computer program;

[0039] a processor, used for executing the computer program to realize the lithology identification method as described above.

[0040] The application further provides a computer readable storage medium, wherein computer executable instructions are stored in the computer readable storage medium, and the computer executable instructions are loaded and executed by a processor to realize the lithology identification method.

[0041] It can be seen that the lithology identification model is constructed in advance based on the KAN network and the wavelet transform; the activation function of the lithology identification model is a wavelet function; the logging data of a target area is obtained; and the logging data is input into the lithology identification model to obtain an identification result. The lithology identification model is constructed by the wavelet transform and the KAN network, the model can utilize the multi-scale representation capability of the wavelet transform to dynamically capture the high-frequency components and the low-frequency components of the logging data, effectively extract complex data necessary for accurate lithology classification, and no longer need manual intervention. Therefore, the model has robustness and precision, can automatically analyze the lithology under various geological conditions, can also cope with complex geological environments, can accurately predict various lithologies on a large scale, and reduces exploration and development risks.

[0042] In addition, the application further provides a lithology identification device, equipment and computer readable storage medium, which also have the beneficial effects described above. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only constitute the embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of the provided drawings.

[0044] Figure 1 A flow chart of a lithology identification method provided by the embodiment of the application;

[0045] Figure 2 A schematic diagram of a KAN network;

[0046] Figure 3 A schematic diagram of a lithology identification model provided by the embodiment of the application;

[0047] Figure 4 A precision comparison diagram of lithology identification based on a volcanic rock reservoir data set provided by the embodiment of the application;

[0048] Figure 5 A precision comparison diagram of lithology identification based on a clastic rock reservoir data set of an A gas field provided by the embodiment of the application;

[0049] Figure 6The precision comparison chart of lithology identification based on the data set of the clastic rock reservoir of B gas field provided by the embodiment of the present application is provided.

[0050] Figure 7 The structural schematic diagram of the lithology identification device provided by the embodiment of the present application is provided.

[0051] Figure 8 The structural schematic diagram of the lithology identification device provided by the embodiment of the present application is provided. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0053] Lithology identification is a basic work in oil and gas exploration and development, and its core is to accurately determine the composition and distribution of underground rock layers through comprehensive analysis of rock physical properties, geochemical characteristics and sedimentary structure. In the exploration field, lithology identification can effectively reduce the exploration risk caused by geological uncertainty. In the reservoir characterization process, lithology identification constructs a reservoir spatial distribution model through logging curve inversion, three-dimensional seismic attribute analysis and other technical means. In the development plan design, the lithology distribution characteristics directly determine the well pattern deployment and fracturing parameters. The application of this technology runs through the whole life cycle from oil and gas reservoir discovery, reserve evaluation to development plan design, and directly affects the exploration success rate and development benefit.

[0054] Traditional lithology identification methods have the disadvantages of strong subjectivity, low efficiency and technical limitations. Under relatively complex geological conditions, the resolution of lithology identification methods is insufficient for rocks with similar mineral composition or strong heterogeneity (such as volcanic rocks and carbonate rocks), and the micro-pore structure cannot be effectively characterized, which affects the evaluation accuracy of unconventional reservoirs. With the development of artificial intelligence technology, the interpretability and efficiency of lithology identification methods based on deep learning have been significantly improved, and the evaluation accuracy has been greatly improved. Under the background of energy transformation, lithology identification technology is deeply integrated with artificial intelligence technology, and its core value in efficient development of oil and gas resources is continuously released. However, the existing deep learning methods have the problems of fixed activation function and limited multi-scale feature extraction. In order to solve these problems, the present application provides a lithology automatic identification method, which uses a learnable wavelet basis function (also known as a wavelet function) to replace the static activation function, so as to dynamically adapt to the mode of logging data. Unlike traditional deep learning architecture, the lithology identification model integrates a feature extraction module based on a wavelet function to jointly capture high-frequency details (such as lithology boundaries) and low-frequency trends. Under complex geological conditions, volcanic rocks and carbonate rocks with strong heterogeneity can still be efficiently and accurately identified. This method can be widely used in China's major oilfields to accurately identify various complex lithologies, reduce exploration and development risks, and ensure the efficient development of major oilfields.

[0055] For details, please refer to Figure 1 , Figure 1 A flowchart of a lithology identification method provided by the embodiment of the present application. The method can include:

[0056] S101: A lithology identification model is constructed in advance based on a KAN network and a wavelet transform; and the activation function of the lithology identification model is a wavelet function.

[0057] The execution subject of the embodiment is a terminal. The embodiment does not limit the type of terminal as long as it can complete the operation of the lithology identification method. The KAN (Kolmogorov-Arnold) network is a new type of neural network architecture based on the inspiration of the Kolmogorov-Arnold representation theorem (KART). That is, any multivariate continuous function can be represented as a finite combination of single-variable continuous functions and a binary operation of addition. Figure 2 A principle example diagram of a KAN model. Unlike previous deep learning models, the KAN network can use a learnable spline parameterized function instead of a fixed activation function, so it allows more flexible and interpretable high-dimensional function representation. Specifically, the KAN network transfers the activation function from the node (neuron) to the edge (weight), and replaces each weight parameter with a learnable single-variable function (such as a spline function). This method can improve the interpretability of deep learning, and also improve the accuracy and applicability of function fitting tasks.

[0058] Further, considering the inherent limitations of B-spline functions, such as their dependence on local approximation and grid refinement, which can complicate the training of KAN networks and potentially lead to issues such as overfitting or underfitting, especially when data exhibits abrupt changes or high-frequency variations, their scalability and versatility in well logging interpretation in the field of petroleum geology are greatly limited. However, wavelet functions provide superior locality and multi-resolution analysis capabilities, making them very effective in capturing local variations in data at multiple scales. These characteristics make wavelet functions particularly suitable for handling the inherent complexity of well logging data. Therefore, the present embodiment replaces B-spline functions with wavelet functions. That is, the present embodiment uses wavelet functions as the basis functions for wavelet transform to perform multi-scale processing on signals and extract local features. This allows the optimized KAN network (i.e., the LogWKAN model, which is the lithology identification model in the present embodiment) to more effectively learn and represent local features, especially in data sets with sharp transitions or non-stationary behavior. This replacement can significantly improve model performance in such scenarios.

[0059] Figure 3 A principle example diagram of a lithology identification model provided by the present embodiment. The lithology identification model is a wavelet transform-based KAN well logging data network, which is a new model that uses learnable wavelet functions to efficiently process and extract frequency domain features. The workflow of the lithology identification model includes the following key components: an embedding layer, a WKA module (Kolmogorov-Arnold, which is the feature extraction module in the present embodiment), a normalization layer, and a linear classifier. The model extracts features by stacking WKA modules based on wavelet functions.

[0060] S102: Obtain well logging data of a target area.

[0061] Data selection and collection are the core of ensuring the accuracy and reliability of model lithology identification. Reasonable selection of well logging parameters can enhance the geological relevance of the model and avoid redundant interference. In addition, high-quality well logging data should be collected to avoid the influence of equipment and human factors, such that the well logging data can truly reflect the characteristics of the formation and reduce noise and errors. Systematic data collection should cover multi-dimensional information of the target layer, ensuring the representativeness and completeness of the samples, thereby improving the evaluation accuracy of the model. Therefore, in the present embodiment, six well logging data related to lithology are selected, including natural gamma ray logging (GR), density logging (DEN), compensated neutron logging (CNL), acoustic travel time logging (AC), deep lateral resistivity logging (LLD), and shallow lateral resistivity logging (LLS).

[0062] It should be noted that the collected well logging data is input into the model in the form of a sequence. That is, given an input well logging sequence where H is the size of the sliding window and P is the number of parameters in the logging sequence. In deep learning, the size of the sliding window is a key parameter, which directly affects the perception range, feature abstraction ability and computational efficiency of the model. The size of the sliding window directly determines the range of the area that the model can "see" in the input data in a single operation. The size of the sliding window can be first selected as a default value, then the window is gradually increased to observe the model effect and calculation cost, and then the appropriate sliding window size in the study is selected. It can be understood that the size of the sliding window in the embodiment is the size of the amount of data input into the model.

[0063] Further, data preprocessing is also a key link to ensure the performance of the model. The original logging data often has noise, missing values or redundant information. Through cleaning, normalization and feature engineering, etc., the data quality and information density can be improved, and the model bias can be reduced. Reasonable feature selection can also eliminate irrelevant variable interference and enhance the generalization ability of the model; standardization processing can accelerate algorithm convergence and improve prediction efficiency. Preprocessing also involves data balancing and enhancement to avoid overfitting or underfitting problems. High-quality data preprocessing lays a reliable foundation for model construction and directly affects the accuracy of the identification result and the actual application value. Because the underground lithology is different, the characteristics of the collected logging data are different, so the various logging parameters of the different lithology reservoirs collected will have significant differences due to different physical principles. In order to reduce the prediction error caused by these differences, the logging data must be standardized, as shown in the following formula, so as to ensure the uniformity of the input samples for subsequent analysis.

[0064] The standardization processing formula is:

[0065] ;

[0066] is the standardized logging data; X is the original logging data obtained; min(X) and max(X) are the minimum and maximum values in the logging data set.

[0067] S103: input the logging data into the lithology identification model to obtain an identification result.

[0068] It should be noted that the lithology identification model in the embodiment is the model constructed in step S101. The model includes an embedding layer, a feature extraction module and a linear classifier.

[0069] Further, the above inputting the logging data into the lithology identification model to obtain an identification result can include the following steps, which can specifically include:

[0070] Step 11: input the logging data into the embedding layer in the lithology identification model to obtain a mapping result.

[0071] Specifically, the logging data is first passed through the embedding layer of the model. The embedding layer maps the input data from a P-dimensional space to a hidden layer dimension D defined by the model, capturing complex nonlinear relationships and potential features in the logging data. The mapping process is shown in the following formula:

[0072] ;

[0073] wherein, 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.

[0074] Step 12: input the mapping result into the stacked feature extraction module of 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.

[0075] Specifically, the lithology identification model improves its performance through stacked WKA modules. These modules utilize the multi-resolution decomposition and sparse representation characteristics of the wavelet function. The ability of the lithology identification model to capture high and low frequency components is enhanced, allowing it to more flexibly handle complex logging sequences.

[0076] Further, the above-mentioned input of the mapping result into the stacked feature extraction module of the lithology identification model to obtain feature components can include the following steps:

[0077] Step 121: wavelet decomposition of the mapping result to obtain decomposed components; the decomposed components include low frequency components and high frequency components;

[0078] Step 122: using a learnable activation function and inverse wavelet transform to reconstruct the decomposed components to obtain feature components.

[0079] Specifically, the WKA module integrates wavelet transform as a learnable activation function into the KAN network to learn the frequency domain features of the logging data. The WKA module consists of three parts: wavelet decomposition of the logging sequence, learnable activation function for the wavelet decomposition results, and inverse wavelet reconstruction based on the activation results.

[0080] The above-mentioned wavelet decomposition of the mapping result to obtain decomposed components has the following specific formula:

[0081] ;

[0082] ;

[0083] ;

[0084] wherein, is the intermediate result of the input sample after wavelet decomposition for the lth layer; is the mapping result corresponding to the input sample of the lth layer; is the wavelet transform function; and respectively represent the high-frequency component and the low-frequency component corresponding to the wavelet transform of the input sample of the lth layer; represents 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 at the nth row and the ith column; H is the number of input samples.

[0085] the obtained wavelet transform result is mapped to the nonlinear function space through the learnable activation function of the WKA module, thereby obtaining a high-dimensional feature vector, so that each component corresponds to the activation value under the input, to realize efficient approximation and explainability of complex continuous multivariate functions.

[0086] Specifically, the above utilizes the learnable activation function and inverse wavelet transform to reconstruct each component after decomposition to obtain the feature component, including:

[0087] ;

[0088] ;

[0089] ;

[0090] wherein, is the feature component; is the feature vector after reconstruction by the inverse wavelet transform; is the bias of the learnable activation function; is the inverse wavelet transform; is the weight component of the learnable activation function weight in the kth grid; is the intermediate result of the input sample after wavelet decomposition in the kth grid; is the high-frequency component after wavelet decomposition corresponding to the kth grid of the lth layer; is the learnable activation function weight corresponding to the high-frequency component; is the low-frequency component after wavelet decomposition corresponding to the kth grid of the lth layer; is the learnable activation function weight corresponding to the low-frequency component; is the learnable weight matrix for calculating the bias of the activation function of the lth layer.

[0091] It can be seen that the LogWKAN model captures high-frequency and low-frequency information by using the sparse representation characteristics of wavelet transform. By introducing learnable complex-valued weights, the model can dynamically adjust the filter parameters and enhance its ability to approximate complex logging data. In addition, the sparse representation provided by wavelet transform enhances the expression and generalization ability of local features.

[0092] Step 13: input the feature components into the pooling layer in the lithology identification model to obtain a prediction result.

[0093] For details, please refer to Figure 3 , Figure 3 In addition to the features extracted in the last WKA module, the features are processed by a BatchNorm (normalization) layer, which can improve the stability of the lithology identification model and alleviate the problem of gradient vanishing or explosion. The features output by the last WKA module are fed into an AvgPooling (pooling layer).

[0094] The normalization formula is:

[0095] ;

[0096] where BN(·) represents the normalization operation in the BatchNorm layer; is the output of the (l-1)th WKA module; is the output of the first layer (including the BatchNorm layer), where .

[0097] The global average pooling GAP(·) formula is:

[0098] .

[0099] where represents the prediction result; GAP(·) represents the pooling layer; D represents the dimension length of the pooling operation, i.e., taking the average of the summation result to obtain the global representative value of each channel.

[0100] Step 14: input the prediction result into the linear classifier in the lithology identification model to obtain an identification result.

[0101] Specifically, the last output layer is classified by a linear classifier Φ(·):

[0102] .

[0103] where represents the identification result; represents the classifier weight, and C is the total number of lithology classes in the logging data set.

[0104] The lithology identification method provided by the embodiment of the application comprises the following steps: a lithology identification model is constructed in advance based on a KAN network and wavelet transform; an activation function of the lithology identification model is a wavelet function; well logging data of a target area is acquired; and the well logging data is input into the lithology identification model to obtain an identification result. The lithology identification model is constructed by the wavelet transform and the KAN network, the model can utilize the multi-scale representation capability of the wavelet transform, dynamically capture high-frequency components and low-frequency components of the well logging data, effectively extract complex data necessary for accurate lithology classification, and no longer need manual intervention. Therefore, the model has robustness and precision, can automatically analyze the lithology under various geological conditions, can also cope with complex geological environments, can accurately predict various lithologies on a large scale, and can reduce exploration and development risks.

[0105] To verify the feasibility of the above method, the data of clastic rock reservoirs are from A and B gas fields, and the data set of volcanic rock reservoirs is from C gas field. The data set records six logging data related to lithology, including natural gamma ray (GR), density logging (DEN), compensated neutron logging (CNL), acoustic time difference logging (AC), deep lateral resistivity logging (LLD), and shallow lateral resistivity logging (LLS). Four deep learning methods, including LogWKAN model, convolutional neural network (CNN), Transformers (a deep neural network structure based on self-attention mechanism), multi-layer perceptron (MLP), and recurrent neural network (RNN), are used for testing and comparison. The branch algorithm of convolutional neural network (Convolutional Neural Network, CNN) evolved in processing sequence data, one-dimensional convolutional neural network (1DCNN) and time series convolutional network (Temporal Convolutional Network, TCN), Transformers (a deep neural network structure based on self-attention mechanism) branch algorithm vision transformer (Vision Transformer, ViT) and information model (Informer), multi-layer perceptron (Multi-Layer Perceptron, MLP) branch algorithm multi-layer perceptron hybrid model (MLP-Mixer), time series hybrid model (Time-Series Mixer, TSMixer) and neural basis expansion analysis for interpretable time series (N-BEATS), and recurrent neural network (Recurrent Neural Network, RNN) branch algorithm long short-term memory network (Long Short-Term Memory, LSTM) and gated recurrent unit network (Gated Recurrent Unit, GRU) are compared with LogWKAN model, and the results are shown in Tables 1, 2 and 3, and Figure 4 、 Figure 5 and Figure 6 It can be seen that the method is better in the lithology prediction task of various types of rock reservoirs in oilfields. It can maintain robustness for unbalanced data sets, cope with complex geological environments, and accurately predict various lithologies on a large scale, reducing the risk of exploration and development.

[0106] Table 1 Precision comparison table for lithology identification based on volcanic rock reservoir data set

[0107]

[0108] Table 2 Precision comparison table of lithology identification based on the data set of the clastic rock reservoir of A gas field

[0109]

[0110] Table 3 Precision comparison table of lithology identification based on the data set of the clastic rock reservoir of B gas field

[0111]

[0112] On the basis of three different data sets, the lithology prediction effects of different models are tested by using different models to predict, and the common model evaluation method OA (precision) is used here. And on the single well logging profile, it can be seen directly that no matter what kind of geological condition and what kind of lithology, the identification performance of the method is better than that of the other four deep learning models, and the precision is also better than that of the other models. Therefore, the lithology identification model provided by the present application is more safe and reliable than the other models. In the lithology prediction task of various types of rock reservoirs in oil fields, it shows excellent effect. It can maintain robustness for unbalanced data sets, can cope with complex geological environment, can accurately predict various lithologies on a large scale, and can reduce exploration and development risks.

[0113] The lithology identification device provided by the embodiment of the present application is introduced below, and the lithology identification device described below can be correspondingly referred to the lithology identification method described above.

[0114] For details, please refer to Figure 7 , Figure 7 The structure diagram of the lithology identification device provided by the embodiment of the present application can include:

[0115] The model construction module 100 is used to construct the lithology identification model based on the KAN network and the wavelet transform in advance; the activation function of the lithology identification model is a wavelet function;

[0116] The logging data acquisition module 200 is used to acquire logging data of a target area;

[0117] The lithology identification module 300 is used to input the logging data into the lithology identification model to obtain an identification result.

[0118] Based on the above embodiment, the lithology identification module 300 can include:

[0119] The embedding unit is used to input the logging data into the embedding layer of the lithology identification model to obtain a mapping result;

[0120] 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 multi-resolution decomposition and sparse representation characteristics of the wavelet function;

[0121] a pooling unit configured to input the feature component into a pooling layer in the lithology identification model to obtain a prediction result;

[0122] a classification unit configured to input the prediction result into a linear classifier in the lithology identification model to obtain the identification result.

[0123] Based on the above embodiment, the feature extraction unit can include:

[0124] 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;

[0125] a reconstruction subunit configured to reconstruct the decomposed components by using a learnable activation function and an inverse wavelet transform to obtain the feature component.

[0126] Based on the above embodiment, the decomposition subunit can include:

[0127] ;

[0128] ;

[0129] ;

[0130] wherein, is an intermediate result of the input sample of the lth layer after wavelet decomposition; is the mapping result corresponding to the input sample of the lth layer; is a wavelet transform function; and respectively represent high-frequency components and low-frequency components corresponding to the input sample of the lth layer after wavelet transform; represents the number of grids in the feature extraction module; g is a coefficient of a low-pass filter; h is a coefficient of a high-pass filter; represents a value of the input sample in the nth row and the ith column; H is the number of input samples.

[0131] Based on the above embodiment, the reconstruction subunit can include:

[0132] ;

[0133] ;

[0134] ;

[0135] wherein, is a feature component; is a feature vector reconstructed by inverse wavelet transform; is a learnable activation function bias; is inverse wavelet transform; is a weight component of a learnable activation function weight in the kth grid; is an intermediate result of the input sample after wavelet decomposition in the kth grid; is a high-frequency component corresponding to the kth grid of the lth layer after wavelet decomposition; is a learnable activation function weight corresponding to the high-frequency component; is a low-frequency component corresponding to the kth grid of the lth layer after wavelet decomposition; is a learnable activation function weight corresponding to the low-frequency component; is a learnable weight matrix for calculating the activation function bias of the lth layer.

[0136] Based on the above embodiment, wherein the well logging data acquisition module can comprise:

[0137] An acquisition unit, configured to acquire 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.

[0138] Based on the above embodiment, wherein the lithology identification device can comprise:

[0139] A preprocessing unit, configured to clean and standardize the well logging data;

[0140] The standardization processing formula is:

[0141] ;

[0142] is the well logging data after standardization processing; X is the original acquired well logging data; min(X) and max(X) are the minimum and maximum values in the well logging data set.

[0143] It should be noted that the order of the modules and units in the above lithology identification device can be changed without affecting the logic.

[0144] The lithology identification device provided in this embodiment of the invention uses a model building module 100 to pre-construct 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 of the target area; and a lithology identification module 300 is used to input the well logging data into the lithology identification model to obtain the identification result. This device constructs a lithology identification model using wavelet transform and a KAN network. This model can utilize the multi-scale representation capability of wavelet transform to dynamically capture high-frequency and low-frequency components of well logging data, effectively extracting the complex data necessary for accurate lithology classification without requiring manual intervention. Therefore, this model is robust and accurate, capable of automatically analyzing lithology under various geological conditions, and can also cope with complex geological environments, accurately predicting various lithologies on a large scale, thus reducing exploration and development risks.

[0145] The lithology identification device provided in the embodiments of the present invention will be described below. The lithology identification device described below can be referred to in correspondence with the lithology identification method described above.

[0146] Please refer to Figure 8 , Figure 8 A schematic diagram of a lithology identification device provided in an embodiment of the present invention may include:

[0147] Memory 10 is used to store computer programs;

[0148] The processor 20 is used to execute computer programs to implement the above-described lithology identification method.

[0149] The memory 10, processor 20, and communication interface 31 all communicate with each other through the communication bus 32.

[0150] In this embodiment of the invention, the memory 10 is used to store one or more programs. The programs may include program code, which includes computer operation instructions. In this embodiment of the invention, the memory 10 may store programs for implementing the following functions:

[0151] A lithology identification model was pre-constructed based on a KAN network and wavelet transform; the activation function of the lithology identification model was a wavelet function.

[0152] Acquire well logging data for the target area;

[0153] The well logging data is input into the lithology identification model to obtain the identification results.

[0154] In a possible implementation, the memory 10 can include a program storage area and a data storage area, wherein the program storage area can store an operating system and application programs required by at least one function, etc.; and the data storage area can store data created during use.

[0155] In addition, the memory 10 can include a read-only memory and a random access memory, and provide instructions and data for the processor. A part of the memory can also include an NVRAM. The memory stores an operating system and operation instructions, executable modules or data structures, or a subset of them, or an extended set of them, wherein the operation instructions can include various operation instructions for implementing various operations. The operating system can include various system programs for implementing various basic tasks and processing hardware-based tasks.

[0156] The processor 20 can 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, and the processor 20 can be a microprocessor or any conventional processor, etc. The processor 20 can invoke a program stored in the memory 10.

[0157] The communication interface 31 can be an interface of a communication module, used for connecting with other devices or systems.

[0158] It should be noted that, Figure 8 The structures shown do not constitute a limitation on the lithology identification device in the embodiments of the present application, and in actual applications, the lithology identification device can include more or fewer components than Figure 8 shown, or combine certain components.

[0159] The computer readable storage medium provided by the embodiments of the present application is described below, and the computer readable storage medium described below can be referred to in conjunction with the lithology identification method described above.

[0160] The present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the lithology identification method described above.

[0161] The computer readable storage medium can include a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0162] The various embodiments described in the specification are progressive in nature, and each embodiment highlights the differences from other embodiments. The same or similar parts between the various embodiments can be mutually referred to each other. For the apparatus disclosed by the embodiments, since it corresponds to the method disclosed by the embodiments, the description is relatively simple, and the relevant parts can be referred to the method part.

[0163] Those skilled in the art will further appreciate that the individual steps of the example units and algorithm described in connection with the embodiments disclosed herein can be embodied directly in hardware, in computer software, or in combinations thereof. To clearly illustrate the interchangeability of hardware and software, the examples have been described generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the particular application and design constraints imposed on the overall system. Those skilled in the art can implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

[0164] Finally, it should be noted that the terms "first", "second", and the like, merely denote different categories, and do not necessarily imply a sequence or order to the entities or operations described. Also, the terms "comprises", "comprising", or other any variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0165] The above describes in detail the lithology identification method, device, equipment and computer readable storage medium provided by the present application. The principles and implementation manners of the present application are described by using specific examples. The above embodiment description is only used to help understand the method of the present application and its core idea. For those skilled in the art, according to the idea of the present application, the specific implementation manner and application range can be changed. The above description should not be understood as a limitation of the present application.

Claims

1. A lithology identification method, characterized in that, include: A lithology identification model was pre-constructed based on KAN network and wavelet transform; The activation function of the lithology identification model is a learnable wavelet function; Acquire logging data for the target area; the logging data includes: natural gamma logging data, density logging data, compensated neutron logging data, sonic transit time logging data, deep lateral resistivity logging data, and shallow lateral resistivity logging data; The well logging data is input into the lithology identification model to obtain the identification results; The well logging data is input into the lithology identification model to obtain identification results, including: The well logging data is input into the embedding layer of the lithology identification model to obtain the mapping result; The mapping result is input into the 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; The feature components are input into the pooling layer in the lithology identification model to obtain the prediction results; The prediction result is input into the linear classifier in the lithology identification model to obtain the identification result; The mapping result is input into the feature extraction module stacked in the lithology identification model to obtain feature components, including: The mapping result is decomposed by wavelet decomposition to obtain the decomposed components; the decomposed components include low-frequency components and high-frequency components. The decomposed components are reconstructed using a learnable activation function and inverse wavelet transform to obtain the feature components; The mapping result is decomposed by wavelet decomposition to obtain the decomposed components, including: ； ； ； in, This is the intermediate result of wavelet decomposition of the input samples at layer l; This represents the mapping result corresponding to the input sample of the l-th layer; It is a wavelet transform function; and These represent the high-frequency and low-frequency components of the wavelet transform corresponding to the input sample of layer l, respectively. The grid number represents the number of grid cells in the feature extraction module; g represents the coefficients of the low-pass filter; h represents the coefficients of the high-pass filter. H represents the value of the input sample in the nth row and ith column; H is the number of input samples. The decomposed components are reconstructed using a learnable activation function and inverse wavelet transform to obtain the feature components, including: ； ； ； in, For characteristic components; The eigenvectors are reconstructed using inverse wavelet transform; Bias for learnable activation functions; This is the inverse wavelet transform; The weights of the learnable activation function are the weight components in the k-th grid. This represents the intermediate result of the input sample in the k-th grid after wavelet decomposition. The high-frequency components are the wavelet decomposition results of the k-th grid in the l-th layer. These are the learnable activation function weights corresponding to the high-frequency components; The low-frequency component after wavelet decomposition of the k-th grid in the l-th layer; These are the learnable activation function weights corresponding to the low-frequency components; This is the learnable weight matrix used in layer l to compute the activation function bias.

2. The lithology identification method according to claim 1, characterized in that, The logging data, after acquiring the logging data for the target area, also includes: The well logging data is cleaned and standardized. The standardization formula is as follows: ； X is the standardized logging data; X is the original logging data; min(X) and max(X) are the minimum and maximum values ​​in the logging dataset.

3. A lithology identification device, characterized in that, include: The model building module is used to pre-build a lithology identification model based on a KAN network and wavelet transform; the activation function of the lithology identification model is a learnable wavelet function. The well logging data acquisition module is used to acquire well logging data for the target area. The lithology identification module is used to input the well logging data into the lithology identification model to obtain the identification result; The well logging data acquisition module includes: an acquisition unit for acquiring natural gamma logging data, density logging data, compensated neutron logging data, sonic transit time logging data, deep lateral resistivity logging data, and shallow lateral resistivity logging data; The lithology identification module includes: An embedding unit is used to input the well logging data into the embedding layer of the lithology identification model to obtain a mapping result; The feature extraction unit is used to input the mapping result into the 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; A pooling unit is used to input the feature components into the pooling layer in the lithology identification model to obtain prediction results; A classification unit is used to input the prediction result into the linear classifier in the lithology identification model to obtain the identification result; The feature extraction unit includes: The decomposition subunit is used to perform wavelet decomposition on the mapping result to obtain the 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 inverse wavelet transform to obtain the feature components; Decomposed into sub-units, including: ； ； ； in, This is the intermediate result of wavelet decomposition of the input samples at layer l; This represents the mapping result corresponding to the input sample of the l-th layer; It is a wavelet transform function; and These represent the high-frequency and low-frequency components of the wavelet transform corresponding to the input sample of layer l, respectively. The grid number represents the number of grid cells in the feature extraction module; g represents the coefficients of the low-pass filter; h represents the coefficients of the high-pass filter. H represents the value of the input sample in the nth row and ith column; H is the number of input samples. Reconstructing sub-units, including: ； ； ； in, For characteristic components; The eigenvectors are reconstructed using inverse wavelet transform; Bias for learnable activation functions; This is the inverse wavelet transform; The weights of the learnable activation function are the weight components in the k-th grid. This represents the intermediate result of the input sample in the k-th grid after wavelet decomposition. The high-frequency components are the wavelet decomposition results of the k-th grid in the l-th layer. These are the learnable activation function weights corresponding to the high-frequency components; The low-frequency component after wavelet decomposition of the k-th grid in the l-th layer; These are the learnable activation function weights corresponding to the low-frequency components; This is the learnable weight matrix used in layer l to compute the activation function bias.

4. A lithology identification device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the lithology identification method as described in any one of claims 1 to 2.

5. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when loaded and executed by a processor, implement the lithology identification method as described in any one of claims 1 to 2.

Images