Logging data layering method and device based on T-KAN structure and medium

Abstract

The invention provides a logging data layering method based on a T-KAN structure, and relates to the field of geology, and the method comprises the steps: obtaining logging curve data of a plurality of wells, and carrying out the preprocessing; windowing the logging curve data and constructing a corresponding label to obtain a training sample; a hierarchical model is constructed on the basis of a Transform encoder; training the hierarchical model by adopting supervised learning in combination with a training sample; preprocessing and windowing logging curve data of a new well, inputting the logging curve data into the trained layered model, and outputting a point-by-point prediction sequence; carrying out smooth filtering on the point-by-point prediction sequence and merging layer boundary candidate points; and alternately determining the layer top and the layer bottom according to a depth sequence, dividing continuous stratum sections, and outputting the top boundary depth and the bottom boundary depth of each stratum section. According to the invention, high automation and intelligentization of the logging layering process are realized, and the dependence on artificial experience is greatly reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of well logging curve data processing, stratum layering and artificial intelligence technology, and particularly relates to a well logging data layering method based on a T-KAN structure. BACKGROUND

[0002] Geological well logging curve data records various formation physical parameters that change with well depth, and is an important basis for formation division, lithology identification and reservoir evaluation in oil and gas exploration and development. Traditional well logging layering mainly relies on manual experience or simple algorithms to identify the characteristics of formation interfaces on well logging curves. For example, common methods include determining layer boundaries according to sudden points, inflection points or peak values of well logging curves, or determining layering positions by analyzing changes in well logging signals using mathematical tools such as continuous wavelet transform and activity function. However, these traditional methods have many shortcomings: first, manual identification requires rich experience, and different interpreters may have different layering results for the same well logging curve, which is low in efficiency and prone to errors. An experienced expert may take twenty minutes to divide a well, while an inexperienced person may take forty minutes or even longer; second, automatic layering methods based on fixed thresholds or specific algorithms often only use a single well logging curve or require manual parameter setting, and have poor adaptability to complex geological conditions and lack a unified standard. The measurement mechanisms and resolutions of different well logging curves are inconsistent, which may cause deviations in layer boundary identification and affect layering accuracy.

[0003] In recent years, machine learning and deep learning technologies have been gradually applied to automatic layering and lithology identification of geological well logging curve data, showing certain potential. For example, early models such as support vector machines (SVM), back propagation (BP) neural networks and self-organizing maps (SOM) have been used for stratum identification, and more advanced deep learning algorithms such as convolutional neural networks (CNN) and long short-term memory networks (LSTM) have also been tried in well logging curve analysis. However, these methods still have many challenges in practical application: first, well logging curve data often has limited samples and high labeling cost, and there are few layering labels available for training, so deep learning models are prone to overfitting and have insufficient generalization ability; second, different stratum categories are unevenly distributed in sample data, and some special lithologies or thin strata have a small proportion in the training set, causing the model to be biased towards the majority class and the recognition rate of the minority class to be low; third, well logging curves have obvious sequential correlation, and traditional neural networks have limited ability to capture long-distance dependencies and global features, making it difficult to fully utilize contextual temporal information to accurately divide layer positions. If the long sequence of the entire well is directly classified as a whole, the model is also easily disturbed by the differences between the upper and lower adjacent strata, reducing the layering accuracy.

[0004] To solve the above problems, some automatic layering methods have been proposed in existing research and patents, but there are still obvious deficiencies. Patent CN104793263A proposes a "logging curve automatic layering value and evaluation method", which identifies the layering by multi-curve weighted synthesis, smoothing and inflection point detection. The disadvantage is that it relies on manual setting of weights and thresholds, and lacks self-adaptive ability; at the same time, it mainly relies on local curve mutation, and it is difficult to capture long-range dependence characteristics, so a few thin layers are often not accurately identified. Patent CN102536195A realizes sequence division based on the least square mean variable point model, and identifies the layering point by iterative detection of variable points. Although it can handle multi-scale sequences, it has poor noise resistance and low recognition accuracy when facing complex nonlinear logging curves, and it is difficult to model the long-range correlation across the layer section. Patent CN108562950B uses multi-curve pattern matching to realize automatic layering, and determines the layering point by calculating the similarity between the reference well and the target well. This method is highly dependent on the representativeness of the reference well, and when the geological conditions of the reference well and the target well are significantly different, the recognition result is distorted and the generalization ability is insufficient. Niu et al. (2022) proposed a layering method based on a feature pyramid network combined with boundary guidance to improve layer boundary detection accuracy. This method performs well in image segmentation scenarios, but it lacks long-range modeling for one-dimensional logging sequences, and does not effectively solve the problem of thin layer detection caused by class imbalance.

[0005] In summary, the existing methods either rely on artificial rules, or are only suitable for local features, or are difficult to generalize to new well blocks, and generally have the defects of insufficient layering accuracy for minority classes and thin layers, and inability to model multi-curve nonlinearity and long-range dependence simultaneously. Therefore, it is urgent to propose a new method that can utilize the global modeling capability of deep learning, enhance the nonlinearity fitting capability, and combine with class imbalance optimization strategy, to improve the accuracy and robustness of intelligent layering of logging curve data. SUMMARY

[0006] The purpose of the present application is to solve the technical problems that the existing methods rely on artificial rules, or are only suitable for local features, or are difficult to generalize to new well blocks, and generally have the defects of insufficient layering accuracy for minority classes and thin layers, and inability to model multi-curve nonlinearity and long-range dependence simultaneously, and to provide a logging data layering method based on T-KAN structure.

[0007] The above purpose of the present application is achieved by the following technical solutions: S1: Obtain logging curve data of multiple wells and perform preprocessing; S2: Windowing the logging curve data and constructing corresponding labels to obtain training samples; S3: Construct a layering model based on a Transformer encoder; S4: Train the layering model using supervised learning combined with the training samples; S5: inputting the pretreated and windowed well logging curve data of the new well into the trained layering model to output a point-by-point prediction sequence; S6: performing smoothing filtering and merging layer boundary candidate points on the point-by-point prediction sequence; alternately determining layer top and layer bottom in depth order to divide continuous formation intervals and output top boundary depth and bottom boundary depth of each interval.

[0008] Optionally, step S1 comprises: The well logging curve data comprises natural gamma, density, acoustic time difference, shallow lateral resistivity and deep lateral resistivity. Each well logging curve is normalized, and the normalization formula is wherein is the original curve value, and are the minimum value and the maximum value of the curve, is the normalized value. The well logging curve data is aligned and uniformly sampled in depth, and abnormal values and noises are removed to obtain normalized well logging curve data.

[0009] Optionally, step S2 comprises: S21: slidingly slicing the well logging curve data in the well depth direction using a window length W and a step length S to form a fixed-length well logging curve sequence sample, i.e. a training sample; S22: forming a feature vector of each depth sampling point in each window with the depth value and the corresponding well logging curve data, and marking each depth sampling point in the window as a layer top, a layer bottom or a non-layer boundary according to the existing layering result. The feature vector of each depth sampling point comprises the depth value and the corresponding multi-channel logging attribute value.

[0010] Optionally, step S3 comprises: The classification layer at the end of the layering model is replaced by an output layer of a Kolmogorov-Arnold network to predict the layer position category of each depth sampling point.

[0011] Optionally, step S4 comprises: The training of the layering model uses a class imbalance loss function, and the class imbalance loss function uses a weighted focal loss function, and the mathematical expression is as follows:

[0012] wherein, is the total number of samples, is the total number of categories, is the sample in the category a true label indication of the real label, is a prediction probability of the model to the sample belongs to the class , is a focus modulation coefficient, is a weight coefficient of the class , and the weight set for the top and bottom classes of the layer is greater than the weight set for the non-layer boundary class.

[0013] Optionally, step S6 comprises: point-by-point prediction sequence is a probability distribution of the horizon class and a predicted label corresponding to each sampling point in the entire depth range of the well; S61: smoothing filtering is performed on the point-by-point prediction sequence to remove isolated pseudo-layer boundary points; S62: a plurality of layer boundary candidate points of the same type within a depth interval less than a preset minimum layer thickness threshold are merged into a single layer boundary; S63: all layer boundary candidate points are traversed in depth order, and the first appearing layer boundary is marked as a layer top and the next appearing layer boundary is marked as a layer bottom to divide a complete stratigraphic interval.

[0014] Optionally, step S6 further comprises a correction processing step: when two layer boundaries of the same type with a distance less than a preset value appear continuously, one of the layer boundaries is retained according to the confidence of the surrounding logging curves; if it is determined according to the thickness information of adjacent stratigraphic intervals that the layering model misses a layer boundary, the missing layer boundary is supplemented through interpolation.

[0015] An electronic device comprises a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to enable the electronic device to perform a logging data layering method based on a T-KAN structure.

[0016] A computer readable storage medium stores instructions, when the instructions are executed, a logging data layering method based on a T-KAN structure is performed.

[0017] The technical scheme provided by the present application has the beneficial effects that: By introducing the Transformer-KAN deep learning structure, the long sequence feature extraction capability of the Transformer and the efficient nonlinear representation capability of the KAN network are effectively fused, so that the model can more accurately identify the weak layer boundary features and complex change patterns in the well logging curve; by using the focal loss and other class imbalance optimization strategies, the recognition accuracy of the model for the minority class stratum is significantly improved, and the model is prevented from being excessively biased towards the main rock stratum, thereby ensuring the comprehensiveness and reliability of the layering result; the well logging layering process is highly automated and intelligent, the dependence on artificial experience is greatly reduced, the layering efficiency is significantly improved compared with the artificial method, and the method has important application value in actual oil and gas field well logging data interpretation. BRIEF DESCRIPTION OF DRAWINGS

[0018] The application will be further described below with reference to the drawings and examples, wherein: Figure 1 is a step diagram in the embodiment of the application; Figure 2 is a model structure diagram in the embodiment of the application; Figure 3 is a well logging curve and layering mark diagram in the embodiment of the application; Figure 4 is a sliding window slicing diagram in the embodiment of the application; Figure 5 is a network architecture diagram in the embodiment of the application; Figure 6 is a confusion matrix diagram in the embodiment of the application; Figure 7 is a layering result diagram in the embodiment of the application; Figure 8 is an electronic device structure diagram in the embodiment of the application. DETAILED DESCRIPTION

[0019] In order to have a clearer understanding of the technical features, objectives and effects of the application, the specific implementation mode of the application will be described in detail with reference to the drawings.

[0020] The embodiment of the application provides a well logging data layering method based on a T-KAN structure.

[0021] Please refer to Figure 1 , Figure 1 is a step diagram of a well logging data layering method based on a T-KAN structure in the embodiment of the application, comprising: S1: acquiring well logging curve data of multiple wells and performing preprocessing; S2: windowing the well logging curve data and constructing corresponding labels to obtain training samples; S3: constructing a hierarchical model based on a Transformer encoder; S4: training the hierarchical model by using supervised learning and combining training samples; S5: inputting the well logging curve data of a new well into the trained hierarchical model after preprocessing and windowing, and outputting a point-by-point prediction sequence; S6: smoothing filtering and merging layer boundary candidate points of the point-by-point prediction sequence, alternately determining layer tops and layer bottoms in depth order, dividing continuous formation intervals, and outputting top boundary depths and bottom boundary depths of each interval.

[0022] As an embodiment, a technology for automatically stratifying well logging curve data by using a Transformer neural network and a Kolmogorov-Arnold network output layer (a Transformer-KAN structure, abbreviated as T-KAN structure, as shown in Figure 2 ).

[0023] Step S1 includes: The well logging curve data includes natural gamma, density, acoustic time difference, shallow lateral resistivity, and deep lateral resistivity. Each well logging curve is normalized, and the normalization formula is wherein is the original curve value, and are the minimum value and the maximum value of the curve, is the normalized value. The well logging curve data is aligned and uniformly sampled in depth, and abnormal values and noises are removed to obtain normalized well logging curve data.

[0024] Step S2 includes: S21: slidingly slicing the well logging curve data in the depth direction by using a window length W and a step length S to form fixed-length well logging curve sequence samples, i.e., training samples. As an embodiment, the preprocessed well logging data is sequentially divided by using a sliding window in the depth direction to obtain fixed-length samples with a length of Each sample contains depth and multi-channel curve values, and is aligned with the corresponding window label. All window samples are based on multi-channel curve features after well-by-well normalization, ensuring scale consistency between different wells, to construct training samples, and the samples are given corresponding horizon labels based on existing stratification results.

[0025] As an embodiment, the preprocessed well logging data is sequentially divided by using a sliding window in the depth direction to obtain fixed-length samples with a length of fixed-length samples, each containing the depth and multi-channel curve values, and aligned with the corresponding window label, all window samples are based on the normalized multi-channel curve features of each well to ensure scale consistency between different wells, to construct training samples, and assign the corresponding horizon label to the samples based on the existing stratigraphic layering results.

[0026] S22: the depth value of each depth sampling point in each window is combined with the corresponding logging curve data to form a feature vector, and each depth sampling point in the window is marked as a layer top, a layer bottom or a non-layer boundary according to the existing stratigraphic layering results; As an embodiment, the existing artificial layering calibration results are used to construct the corresponding label for each logging window. According to the stratigraphic layering top and bottom depth provided by the geologist, each depth point in the window segment is marked with its layer category: if a certain depth corresponds to the stratigraphic top boundary, it is marked as "layer top", if it corresponds to the stratigraphic bottom boundary, it is marked as "layer bottom", and the rest of the depth points are marked as non-layer boundary.

[0027] The feature vector of each depth sampling point includes the depth value and the corresponding multi-channel logging attribute value.

[0028] Step S3 includes: The classification layer at the end of the stratigraphic layering model is replaced by the output layer of the Kolmogorov-Arnold network, which is used to predict the horizon category of each depth sampling point.

[0029] In a specific implementation of the present application, the output layer of the Kolmogorov-Arnold network is replaced by a combination of several learnable one-dimensional basis functions, replacing the original linear fully connected classification layer, to realize the nonlinear mapping from high-dimensional logging features to horizon category space.

[0030] In a specific implementation of the present application, a standard Transformer encoder-decoder is constructed as the main structure, and the linear classification layer at the end of the original decoder is replaced by a KAN (Kolmogorov-Arnold Network) output layer to realize point-by-point (depth sampling point-by-depth sampling point or window center-by-window center) classification prediction. The encoder: stack multiple layers of multi-head self-attention (MHA) and feedforward network (FFN) to capture the long-range dependencies across depth and the coupling relationship of multi-channel curves; the decoder: receives the encoder representation and obtains the context features through cross-attention. KAN output layer: replace linear mapping with a learnable combination of several one-dimensional basis functions to enhance the nonlinear representation power from high dimension to category space, and output the category scores of each horizon.

[0031] In a specific implementation of the present application, the model structure includes an input layer, several feature extraction layers and an output layer. The input layer converts the data of the logging curve window into a tensor form that can be processed by the model, The multidimensional logging attribute values ​​at each depth point are used to form a feature vector sequence as model input. The main feature extraction mechanism of the model adopts an improved Transformer-KAN encoder, which consists of multiple stacked attention mechanisms and KAN units alternately. Each layer contains a multi-head self-attention sublayer and a feedforward sublayer based on the Kolmogorov-Arnold network to simultaneously capture the depth direction correlation and complex nonlinear feature relationships of the logging curves.

[0032] The formula for calculating the vector feature sequence is as follows:

[0033] In the Transformer-KAN structure, the multi-head self-attention sublayer extracts relevant patterns of geological layers by calculating the attention weights between depth points in the sequence; the KAN feedforward sublayer uses a combination of learnable one-dimensional basis functions to replace the traditional fully connected network, performing a nonlinear transformation on the channel features output by each attention sublayer. Each Transformer-KAN unit employs residual connections and layer normalization strategies to ensure model convergence stability.

[0034] The output layer and classification principle: The model's output layer uses a fully connected classifier to predict features at each depth point point by point. Specifically, a fully connected layer is applied to the latent vector hi at each position output by the last layer of the Transformer-KAN encoder to calculate its score belonging to each category, and the softmax function is used to convert the score into a probability.

[0035] in The weight matrix is ​​the weight matrix of the corresponding category c of the output layer, where C is the total number of categories (in this invention, C=3 corresponds to the three categories of "layer top", "layer bottom" and "non-layer boundary"). This represents the score for category c. Through the above output layer calculations, the probability distribution of each depth point in the logging curve window belonging to each category can be obtained, and its most likely stratigraphic category can be determined accordingly.

[0036] Step S4 includes: The hierarchical model is trained using a class imbalance loss function, which is a weighted focus loss function, and its mathematical expression is as follows:

[0037] in, The total number of samples, The total number of categories, It is a sample In category The true label indicates, It is the model on the sample Category The predicted probability, For focus modulation coefficients, For category The weighting coefficients, and the weights set for the top and bottom layer categories. Greater than the weight set for non-layer boundary categories.

[0038] In one specific implementation of this application, a class imbalance loss function is introduced to improve the accuracy of identifying minority class samples at the "top" and "bottom" of a layer. (Focus modulation coefficients) The range of values ​​is Preferred The weighting coefficients of the "top" and "bottom" categories. The value range is the weight coefficient of the "non-layer boundary" category. times, preferably times.

[0039] In one specific implementation of this application, supervised learning is used to train the Transformer-KAN model. To mitigate the bias risk caused by imbalanced class distribution across layers, a class imbalance loss is introduced for multiple optimizations, preferably using focal loss or weighted cross-entropy, which can be combined with L2 regularization. By using appropriate balance factors and focal exponents, the model achieves a better trade-off between overall accuracy and minority class recall; and by combining learning rate scheduling, early stopping, and data augmentation techniques, generalization performance is improved.

[0040] Step S6 includes: The point-by-point prediction sequence is the probability distribution and prediction label of the stratigraphic category corresponding to each sampling point within the entire depth range of the well. As one embodiment, the logging curve data of the well to be analyzed is processed using the same preprocessing and windowing strategy as during training, and then input into the trained Transformer-KAN model. The model outputs the corresponding layer category prediction result for each window of input logging data, that is, it gives the probability or label of each sampling point within the depth range of the well as belonging to "layer top", "layer bottom", or "non-layer boundary". Based on the point-by-point prediction sequence output by the model, the layer identification result for each sampling point from shallow to deep can be obtained. The model output is usually given in probability form; for ease of interpretation, the category with the highest probability can be used as the determination category for that depth point. For example, when a depth point is determined by the model to be the layer top with the highest probability and exceeds a preset threshold, then that point is considered to correspond to the top boundary of a layer segment. Overlapping results generated during the sliding window prediction process are merged to obtain a complete wellbore layer sequence. Since adjacent windows may repeatedly predict the same depth point, their results should be consistent or merged through voting; for portions at the beginning and end of the well segment that are less than a window, appropriate edge processing strategies can be adopted (e.g., discarding the insufficient portion or using filled data for prediction). Through the above steps, the predicted sequence of stratigraphic categories for each point within the entire depth range of the target well is finally obtained.

[0041] S61: Perform smoothing filtering on the point-by-point prediction sequence to remove isolated pseudo-layer boundaries; S62: Set the depth interval to the preset minimum layer thickness threshold. Multiple candidate points of the same type within a given range are merged into a single layer boundary; S63: Traverse all candidate boundary points in depth order, and alternately mark the first boundary that appears as the top of the layer and the next boundary that appears as the bottom of the layer to divide the complete stratigraphic segment.

[0042] As one implementation, the depths of all locations marked as tops or bottoms are extracted from the stratigraphic sequence results predicted by the model, initially identifying candidate boundary points for stratigraphic segments. These candidate boundary depth values ​​are arranged in depth order to form an initial segment division list. The initial list of candidate boundary points is then filtered and matched to eliminate false positives and determine the top and bottom boundaries of each segment. Methods such as moving window smoothing or median filtering can be used to smooth the predicted sequence and remove isolated noise points. Then, the candidate points are traversed according to stratigraphic order, merging multiple candidate boundaries that are too close together into a single boundary. The first boundary is marked as the top boundary of the first segment, and the next boundary is marked as the bottom boundary of that segment, while also serving as the top boundary of the next segment. This process is repeated to alternately determine the top / bottom boundaries of each segment.

[0043] As one example, the trained model is deployed to the well logging interpretation system to automatically stratify the logging curves of new wells: Inference: The preprocessed and windowed sequences are input into the model to obtain the point-by-point category probability distribution and predicted labels; Post-processing: Formation continuity constraints and interlayer smoothing rules are applied to the point-by-point predicted sequences to suppress pseudo-layers with "short jumps / rebounds"; Result mapping: Adjacent depth points of the same continuous category are grouped into segments, and the top / bottom depth of the segment and the layer name are output. Excel format export is supported for comparison with manual interpretation.

[0044] Step S6 also includes a correction process: When two consecutive boundary layers of the same type with a distance less than a preset value appear, one of the boundary layers is retained based on the confidence level of the surrounding logging curves. If the layering model misses a layer boundary based on the thickness information of adjacent layers, the missing layer boundary is supplemented by interpolation.

[0045] As one example, to address the potential for misjudgment of layer boundaries in model predictions, if two consecutive top or bottom layer labels are too close together, contradicting common geological knowledge, one of them can be selectively retained or discarded based on changes in surrounding well logging curves. For layer boundaries not predicted by the model but which should exist based on adjacent layer thickness information, they can be manually supplemented through interpolation or by referencing adjacent well data. Through the above post-processing optimization, the final output is a list of top and bottom depth data for each stratum segment, achieving intelligent stratification output of geological logging data.

[0046] As an example, taking the logging data obtained from nine wells in a certain area, the specific implementation scheme of the present invention is as follows: Step 110: Data preparation and sample construction; Obtain conventional logging data from nine wells, including at least natural gamma (GR), density (DEN), sonic transit time (DT), and shallow / deep lateral resistivity (RS, RD). The raw data table includes depth, multichannel curve columns, layer (if applicable), and well fields. Perform depth alignment and standardize the sampling step size (e.g., 0.125m) for each well, and remove or interpolate outliers / missing values. Perform a safe logarithmic transformation on resistivity curves to eliminate the influence of non-physical negative values. Subsequently, standardize the data at each wellhead to reduce instrument and regional differences.

[0047] Specifically, conventional logging curves were collected from nine wells, including at least natural gamma (GR), density (DEN), sonic transit time (DT), shallow lateral resistivity (RS), and deep lateral resistivity (RD). Figure 3The original table structure should at least include: depth, multiple curve columns, layer labels, and well numbers. Automatic field identification can be achieved using the `infer_column` function. Then, the start and end depths and sampling step distances of each curve for each well are calculated, common overlapping segments are extracted and resampled to a uniform step distance (e.g., 0.125m); resampling uses linear interpolation or piecewise cubic Hermite interpolation. Outlier / missing values ​​and safe logarithmic transformation are performed. For outliers (e.g....)... 9999) and missing points are removed or interpolated for repair; a safe logarithmic transformation (safe_log_transform) is performed on resistivity-type RS / RD: severely negative values ​​(≤ threshold) are set to NaN; slightly negative values ​​are shifted as a whole and then the common logarithm is taken to eliminate the adverse effects of non-physical negative values ​​on training.

[0048] Step 120: Windowing and Tag Construction; Specifically, along the depth direction, with the window length and step length For each well, multiple curves are sliced ​​using a sliding method to form a fixed-length sequence of samples. To improve location sensitivity, depth is used as a one-dimensional feature and concatenated into the feature vector of each sampling point. Point-by-point labels are generated for each depth point within the window using the existing top / bottom depth of the layer, defining three categories: 0 = non-layer boundary, 1 = layer top, and 2 = layer bottom. If necessary, one-hot encoding is used for supervised training. To improve cross-well comparability, the window length can be interpolated to the target length corresponding to the average layer thickness of the training wells. In this embodiment, the window length is 5, and the step size is 5 (…). Figure 4 ).

[0049] Among them, windowed samples (C is the number of channels)

[0050] Step 130: Model Building and Training Optimization Specifically, a standard Transformer encoder-decoder structure is adopted: positional encoding + multi-head self-attention (MHA) + feedforward network (FFN), and residual connections and layer normalization are used to ensure stable training; the input is a W×C sequence tensor. The linear classification head at the end of the original decoder is replaced with the KAN (Kolmogorov-Arnold Network) output layer ( Figure 5 This approach uses a combination of learnable one-dimensional basis functions to achieve a non-linear mapping from high-dimensional features to the class space, resulting in more economical parameters and stronger expressive power. A grouped KAN is used to share some basis function parameters to control computational cost and overfitting risk. Subsequently, point-by-point classification and probabilistic mapping are performed, with the decoder at each position... The latent vectors are processed by the KAN output layer to obtain the class scores. Three types of probabilities (non-layer boundary / layer top / layer bottom) are obtained through Softmax:

[0051] Training employs a 90% "leave one well" cross-validation method to evaluate generalization ability: each iteration, one well is fixed as the test well, while the remaining eight wells participate in both training and validation (the validation set comprises approximately 20% of the training samples), for a total of nine iterations. The optimizer prioritizes AdamW (learning rate 1×10⁻⁶). 4) Combine ReduceLROnPlateau learning rate scheduling and early stopping to prevent overfitting, and weight decay (L2) to suppress model complexity. To address the problem of significantly fewer samples at the top / bottom of layers, introduce weighted focal loss (FocalLoss) or a combination of it and weighted cross-entropy, give higher weights to boundary classes, and control the focusing intensity of hard examples with γ (e.g., γ=2).

[0052] After training, the script automatically outputs: training / validation loss curves, confusion matrix and classification report for each test fold, and selects two test wells from the nine folds to generate a comparison chart of "true and predicted" layers for intuitive verification.

[0053] Step 140: Reasoning, Continuity Post-processing, and Result Output Specifically, the new well data is preprocessed and windowed using the same process as steps 110–120, and then input into the trained and converged Transformer-KAN model to obtain point-by-point class probabilities and label sequences. Multiple predictions from overlapping windows are fused through voting or probability averaging; continuity constraints and a minimum layer thickness threshold are applied, isolated short segments are merged, and pseudo-interfaces with low confidence are removed; subsequently, top / bottom boundaries are paired alternately according to depth, with wellhead / bottom boundary completion used when no pairing is found at the beginning / end. Finally, adjacent points of the same type are merged into continuous segments, outputting a list of top / bottom depths / thicknesses for each segment, and generating a comparison and confusion matrix of true / predicted segments. Figure 6 , Figure 7 ).

[0054] This application also discloses an electronic device. (See reference...) Figure 8 , Figure 8 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application. The electronic device 500 may include: at least one processor 501, at least one network interface 504, a user interface 503, a memory 505, and at least one communication bus 502.

[0055] The communication bus 502 is used to enable communication between these components.

[0056] The user interface 503 may include a display screen, and optionally, the user interface 503 may also include a standard wired interface or a wireless interface.

[0057] The network interface 504 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0058] This application also discloses a computer-readable storage medium storing multiple instructions adapted for loading by a processor to execute the above-described well logging data layering method based on a T-KAN structure.

[0059] The above are merely exemplary embodiments of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure.

[0060] This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A method for layering well logging data based on T-KAN structure, characterized in that, The method comprises the following steps: S1: obtaining and preprocessing well logging curve data of multiple wells; S2: windowing the well logging curve data and constructing corresponding labels to obtain training samples; S3: constructing a hierarchical model based on a Transformer encoder; S4: training the hierarchical model based on supervised learning and the training samples; S5: inputting the well logging curve data of a new well after preprocessing and windowing into the trained hierarchical model to output a point-by-point prediction sequence; S6: smoothing filtering and merging layer boundary candidate points of the point-by-point prediction sequence; alternately determining layer tops and layer bottoms in depth order to divide continuous formation intervals and output top and bottom boundary depths of each interval.

2. The well logging data stratification method based on T-KAN structure as described in claim 1, characterized in that, Step S1 comprises: The well logging curve data comprises natural gamma, density, acoustic time difference, shallow lateral resistivity, and deep lateral resistivity; The normalization formula for each well logging curve is wherein is the original curve value, and is the minimum and maximum value of the curve, is the normalized value; The well logging curve data is aligned and uniformly sampled in depth, and abnormal values and noise are removed to obtain normalized well logging curve data; the well logging curve data is multi-channel well logging curve data.

3. The method of claim 1, wherein the T-KAN structure is a T-KAN structure of a well logging data layering method, characterized in that, Step S2 comprises: S21: slidingly slicing the well logging curve data in the well depth direction using a window length W and a step S to form fixed-length well logging curve sequence samples, i.e., training samples; S22: forming a feature vector from the depth value of each depth sampling point in the window and the corresponding well logging curve data, and marking each depth sampling point in the window as a layer top, a layer bottom, or a non-layer boundary according to the existing stratification results; The feature vector of each depth sampling point comprises the depth value and corresponding multi-channel well logging attribute values.

4. The method of claim 1, wherein the T-KAN structure is a T-KAN structure of a well logging data layering method, characterized in that, Step S3 comprises: The classification layer at the end of the hierarchical model is replaced with an output layer of a Kolmogorov-Arnold network.

5. The well logging data stratification method based on T-KAN structure as described in claim 1, characterized in that, Step S4 comprises: The training of the hierarchical model uses a class imbalance loss function, and the class imbalance loss function uses a weighted focal loss function, which has the following mathematical expression: wherein, is the total number of samples, is the total number of classes, is a sample in a class with a true label indication, is the predicted probability of a sample belonging to a class by a model, is the focal modulation coefficient, is the weight coefficient of a class with the weight for the top and bottom classes of a layer being greater than the weight for the non-boundary classes of a layer.

6. The method of claim 1, wherein the T-KAN structure is based on a well log data layering method, characterized by, Step S6 comprises: The point-by-point prediction sequence is the layer position category probability distribution and prediction label corresponding to each sampling point in the entire depth range of the well; S61: smoothing filtering the point-by-point prediction sequence to remove isolated pseudo-layer boundary points; S62: interval the depth at a preset minimum layer thickness threshold merge the multiple same type of layer boundary candidate points within the inner portion into a single layer boundary; S63: traversing all layer boundary candidate points in depth order and alternately marking the first appearing layer boundary as a layer top and the next appearing layer boundary as a layer bottom to divide complete formation intervals.

7. A well logging data stratification method based on a T-KAN structure as described in claim 6, characterized in that, Step S6 further comprises a correction processing step: When two layer boundaries of the same type appear continuously and the distance between them is less than a preset value, one of the two layer boundaries is retained according to the confidence of the surrounding well logging curves; If the hierarchical model misses a layer boundary according to the thickness information of adjacent intervals, the missing layer boundary is supplemented by interpolation.

8. An electronic device, comprising: The electronic device comprises a processor, a memory, a user interface, and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to enable the electronic device to perform the well logging data stratification method based on the T-KAN structure according to any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores instructions, when the instructions are executed by a computer, a method for layering logging data based on a T-KAN structure according to any one of claims 1-7 is executed.

Images