Stratigraphic division model visualization method, system and terminal using confidence score

Abstract

This invention discloses a method for visualizing formation segmentation models using confidence scores. The method involves drawing the spatial coordinates of oil wells and selecting similar wells to form a dataset. Outlier removal and standardization are performed on the logging curves in the dataset, and a sliding window is used to generate two-dimensional representation segments. Formation labels are vectorized, and a convolutional neural network model is constructed. The model is trained using the training set data, and the network parameters are optimized through multiple iterations to ensure accurate prediction of formation labels. After training, the logging data segments to be visualized are input into the trained network, and feature maps are output at designated convolutional layers. The feature maps are upsampled and multiplied point-by-point with the input segments, then input again into a convolutional neural network with the same structure to obtain the confidence scores of the feature maps. The confidence scores of each channel's feature map are used as weights and linearly weighted with the corresponding channel's feature map to obtain a class discrimination saliency map.

Description

Technical Field

[0001] This invention relates to the field of oil well logging interpretation technology, specifically to a method, system, and terminal for visualizing formation partitioning models using confidence scores. Background Technology

[0002] Scarce oil resources are an indispensable strategic resource for a nation's survival and development. However, with the emergence of various high-tech advancements, the demand for oil in various countries continues to increase. If this continues, the world's oil reserves will dwindle, eventually reaching depletion. Developing new oil fields is difficult, costly, and risky, while continuing to develop existing oil fields offers relatively lower costs and risks. This is a practical approach that can promote long-term, high, and stable oil production.

[0003] Well logging interpretation is the process of identifying reservoirs in a wellbore profile, classifying their lithology and fluid types, and determining their geological parameters using various well logging data. Among these, stratigraphic delineation is the most fundamental and crucial task. Accurate stratigraphic delineation is essential for subsequent well logging interpretation tasks such as reservoir identification, fluid identification, and lithological classification. When exploring existing oilfields, well logging sensors with specific functions transmit corresponding well logging curve data. This data is the most original data in the well logging data, recording the physical parameters of the strata that change with downhole depth and directly reflecting geological characteristics. Therefore, directly using well logging curve data for stratigraphic delineation is of great significance for the deep development of existing oilfields. Currently, stratigraphic delineation mainly relies on the subjective experience of well logging experts. Due to the significant subjective factors in manual interpretation, stratigraphic delineation conclusions vary considerably. However, constructing a dedicated deep neural network for stratigraphic delineation can reduce errors caused by insufficient subjective experience and other human factors.

[0004] To date, the development of stratigraphic segmentation has largely relied on manual interpretation methods, which have become a complete process. Well logging experts typically use their personal experience to comprehensively analyze well logging curve morphology to achieve stratigraphic segmentation. However, without understanding the specific principles behind these methods, most experts are skeptical of intelligent stratigraphic segmentation, hindering its development. Therefore, interpretability studies of deep learning applications are crucial. These studies can build trust between well logging experts and deep learning methods, enabling their better application in stratigraphic segmentation. Currently, interpretability of deep learning for stratigraphic segmentation is still lacking. Therefore, visualizing deep learning-based stratigraphic segmentation tasks and mapping well logging curve features to a feature space can help well logging interpreters better understand the application principles, thereby accelerating the application of deep learning algorithms in stratigraphic segmentation tasks. Summary of the Invention

[0005] In order to overcome the defects of the existing technology, the purpose of this invention is to provide a method, system and terminal for visualizing stratigraphic models using confidence scores, so as to solve the technical problem that the subjective factors of the existing stratigraphic division manual interpretation method are too large and there are a lot of errors caused by human factors.

[0006] This invention is achieved through the following technical solution:

[0007] A method for visualizing stratigraphic models using confidence scores includes the following steps:

[0008] Step 1: Draw the spatial coordinates of all oil wells within the block, and select oil wells with similar spatial coordinates as the dataset. The dataset includes several logging curve data.

[0009] Step 2: Remove outliers from the logging curve data in the dataset, standardize the logging curve data of each oil well according to the curve, and select a sliding window to obtain a two-dimensional representation segment of the two-dimensional logging data. Convert all the discrete logging curve data in each well into two-dimensional segments of the sliding window size to obtain the formation label corresponding to each segment. Vectorize all the geological layer labels to obtain the dataset, and divide the obtained dataset into training set and test set according to the proportion.

[0010] Step 3: Construct a convolutional neural network model. Input the data in the training set into the convolutional neural network to obtain the network output. Compare the network output with the formation label and calculate the loss function. Then, use the backpropagation algorithm to calculate the gradient of the network parameters with respect to the loss function. Repeat the training process until the preset maximum number of training rounds is reached to complete the training.

[0011] Step 4: Input the two-dimensional representation fragment of the well logging data to be visualized into the trained convolutional neural network, select the convolutional layer to be visualized, and output the feature map of the layer; upsample the output feature map to the size of the input fragment and multiply it point by point with the two-dimensional representation fragment of the input well logging data, and then re-input it into the convolutional neural network with the same structure to obtain the confidence score of each feature map;

[0012] Step 5: Use the confidence score of each channel feature map as a weight and linearly weight it with the feature map of the corresponding channel to obtain the class discrimination saliency map;

[0013] Step 6: Repeat steps 1 to 5 to visualize two-dimensional representation segments of multiple well logging data from the same formation, obtain the corresponding class discrimination saliency map, and obtain the mapping results of the electrical characteristics of the formation in the feature space.

[0014] Preferably, in step 1, hierarchical clustering is used to preferentially select oil wells with similar spatial coordinates as the dataset. Several logging curves in the dataset correspond to the data sources of several features. The logging curves are presented in the form of discrete data of a single depth point, and each depth point corresponds to a formation attribute.

[0015] Preferably, several logging curve data include Depth data, Natural Gamma Regulator (GR) data, Spontaneous Potential (SP) data, Acoustic Transmission Time (AC) data, Array Inductive Resistivity (AT90) data, and Array Inductive Resistivity (AT20) data.

[0016] Preferably, in step 2, the logging curve data of each oil well is standardized according to the curve. The standardization process is Z-Score standardization, and the formula is as follows:

[0017]

[0018] Where μ is the data mean and σ is the data standard deviation.

[0019] Preferably, in step 2, the sliding window size is selected as 96×6 pixels, and it slides along the direction of the logging curve data at intervals of 1 depth point. Each time the window covers an area, a 96×6 two-dimensional logging data segment can be obtained. All the discrete logging curve data in each well are converted into 96×6 two-dimensional segments. The formation label corresponding to each segment is the formation label that appears most frequently among the 96 depth points selected in the window. Finally, all geological layer labels are vectorized. The specific process of vectorizing all geological layer labels is as follows:

[0020] The number of stratigraphic labels n in the training set is statistically analyzed. The stratigraphic labels are numbered from 0, and the number of each label is vectorized into an n-dimensional vector, where one index in the vector is 1 and the rest are 0.

[0021] Preferably, in step 3, the architecture of the convolutional neural network is designed first in the construction of the convolutional neural network model. The architecture of the convolutional neural network includes 3 convolutional layers, 2 fully connected layers, and a softmax classifier. The number of convolutional kernels in the 3 convolutional layers are 64, 128, and 256, respectively. The kernel size is 3×3, the stride is 1, and the padding method is "same". The "same padding" method is to pad the array with a circle of 0s to increase the array size. The number of nodes in the two fully connected layers is 512. The number of nodes in the softmax classifier is determined by the number of statistical stratigraphic label types.

[0022] Preferably, in step 4, the formula for calculating the confidence score of the two-dimensional representation fragment to be visualized is as follows:

[0023]

[0024] in, This is the output feature map of the k-th channel of the l-th convolutional layer. for Global confidence score for the two-dimensional representation segment;

[0025] The calculation formula is:

[0026] The Up function represents the feature map The dimension of the upsampled feature map is determined by the dimension of the input vector. The Norm function represents the normalization of the upsampled feature map.

[0027] Preferably, in step 5, the formula for solving the class discrimination saliency map is as follows:

[0028]

[0029] Among them, L c This is a saliency map for class C strata. The weights corresponding to the feature map of the k-th channel are represented by , where The solution formula is as follows:

[0030]

[0031] in, This is the global confidence score.

[0032] A stratigraphic model visualization system using confidence scores, for implementing the aforementioned stratigraphic model visualization method using confidence scores, includes:

[0033] The drawing module is used to draw the spatial coordinates of all oil wells within the block and filter out oil wells with similar spatial coordinates as a dataset. The dataset includes several logging curve data.

[0034] The first data processing module is used to remove outliers from the logging curve data in the dataset, standardize the logging curve data of each oil well according to the curve, select a sliding window to obtain a two-dimensional representation segment of the two-dimensional logging data, convert all the discrete logging curve data in each well into two-dimensional segments of the sliding window size, obtain the formation label corresponding to each segment, vectorize all the geological layer labels to obtain the dataset, and divide the obtained dataset into training set and test set according to the proportion.

[0035] The model building module is used to build a convolutional neural network model. It inputs the data in the training set into the convolutional neural network to obtain the network's output, compares the network's output with the formation label, calculates the loss function, and then calculates the gradient of the network parameters with respect to the loss function through the backpropagation algorithm. The training process is repeated until the preset maximum number of training rounds is reached to complete the training.

[0036] The second data processing module is used to input the two-dimensional representation fragment of the well logging data to be visualized into the trained convolutional neural network, select the convolutional layer to be visualized, and output the feature map of the layer; after upsampling the output feature map to the size of the input fragment and multiplying it point by point with the two-dimensional representation fragment of the input well logging data, it is re-inputted into the convolutional neural network with the same structure to obtain the confidence score of each feature map;

[0037] The fourth data processing module is used to linearly weight the confidence score of each channel feature map with the corresponding channel feature map to obtain a class discrimination saliency map;

[0038] The execution module is used for repeated execution to visualize two-dimensional representation fragments of multiple well logging data from the same formation and generate corresponding class discrimination saliency maps.

[0039] A mobile terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the above-described method for visualizing a stratigraphic subdivision model using confidence scores.

[0040] Compared with the prior art, the present invention has the following beneficial technical effects:

[0041] This invention provides a method for visualizing formation segmentation models using confidence scores. First, a dataset is formed by plotting the spatial coordinates of oil wells and filtering for similar wells. Then, outlier removal and standardization are performed on the logging curves in the dataset, and a sliding window is used to generate two-dimensional representation segments, while formation labels are vectorized. Next, a convolutional neural network model is constructed and trained using the training set data. The network parameters are optimized through multiple iterations to ensure the network can accurately predict formation labels. After training, the logging data segments to be visualized are input into the trained network, and feature maps are output at designated convolutional layers. Then, the feature maps are upsampled and multiplied point-by-point with the input segments, and then input again into a convolutional neural network with the same structure to obtain the confidence scores of the feature maps. The confidence scores of the feature maps for each channel are used as weights and linearly weighted with the feature maps of the corresponding channels to obtain a class saliency map. Finally, the above techniques are used to visualize multiple logging data segments of the same formation to obtain class saliency maps, thereby revealing the electrical characteristics of the formation and their mapping results in the feature space. The stratigraphic model visualization method using confidence scores eliminates the dependence on gradients during backpropagation in convolutional neural networks, enabling visualization of any convolutional layer in the network. This helps in understanding the electrical characteristics of strata and provides important references in geological research and oil exploration. Attached Figure Description

[0042] Figure 1 This is a flowchart illustrating the visualization method for stratigraphic division models using confidence scores in this invention.

[0043] Figure 2 This is a map showing the geographical coordinates of all wells in a certain block of an oilfield in an embodiment of the present invention;

[0044] Figure 3 This is a geographic coordinate map of the dataset in this invention;

[0045] Figure 4 This is a schematic diagram illustrating the process of converting well logging curve data into a two-dimensional characterization segment in this invention;

[0046] Figure 5 This is a diagram of the convolutional neural network structure constructed in this invention;

[0047] Figure 6 This is a schematic diagram of the visualization mechanism of the stratigraphic division model using confidence scores in this invention;

[0048] Figure 7 This is a visualization result of the Chang6 geological layer in this invention. Detailed Implementation

[0049] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0050] The present invention will now be described in further detail with reference to the accompanying drawings:

[0051] The purpose of this invention is to provide a method, system, and terminal for visualizing stratigraphic models using confidence scores, in order to solve the technical problem that the subjective factors in existing stratigraphic interpretation methods are too large, resulting in a large number of errors caused by human factors.

[0052] See Figure 1 This invention provides a method for visualizing stratigraphic division models using confidence scores, comprising the following steps:

[0053] Step 1: Draw the spatial coordinates of all oil wells within the block, and select oil wells with similar spatial coordinates as the dataset. The dataset includes several logging curve data.

[0054] Among them, hierarchical clustering is used to select oil wells with similar spatial coordinates as the dataset. Several logging curves in the dataset are used as the data source for several features. The logging curves are presented in the form of discrete data of a single depth point, and each depth point corresponds to a formation attribute.

[0055] Among them, several logging curve data include Depth data, Natural Gamma Regulator (GR) data, Spontaneous Potential (SP) data, Acoustic Transit Time (AC) data, Array Inductive Resistivity (AT90) data, and Array Inductive Resistivity (AT20) data.

[0056] Step 2: Remove outliers from the logging curve data in the dataset, standardize the logging curve data of each oil well according to the curve, and select a sliding window to obtain a two-dimensional representation segment of the two-dimensional logging data. Convert all the discrete logging curve data in each well into two-dimensional segments of the sliding window size to obtain the formation label corresponding to each segment. Vectorize all the geological layer labels to obtain the dataset, and divide the obtained dataset into training set and test set according to the proportion.

[0057] The logging curve data for each oil well is standardized using Z-Score standardization, as shown in the following formula:

[0058]

[0059] Where μ is the data mean and σ is the data standard deviation.

[0060] The sliding window is selected to be 96×6 pixels in size. It slides along the direction of the logging curve data at intervals of one depth point. Each time the window covers an area, a 96×6 two-dimensional segment of logging data is obtained. All discrete logging curve data from each well are converted into 96×6 two-dimensional segments. The formation label corresponding to each segment is the formation label that appears most frequently among the 96 depth points selected in the window. Finally, all geological layer labels are vectorized. The specific process for vectorizing all geological layer labels is as follows:

[0061] The number of stratigraphic labels n in the training set is statistically analyzed. The stratigraphic labels are numbered from 0, and the number of each label is vectorized into an n-dimensional vector, where one index in the vector is 1 and the rest are 0.

[0062] Step 3: Construct a convolutional neural network model. Input the data in the training set into the convolutional neural network to obtain the network output. Compare the network output with the formation label and calculate the loss function. Then, use the backpropagation algorithm to calculate the gradient of the network parameters with respect to the loss function. Repeat the training process until the preset maximum number of training rounds is reached to complete the training.

[0063] Specifically, in constructing the convolutional neural network model, the architecture of the convolutional neural network is first designed. The architecture of the convolutional neural network includes 3 convolutional layers, 2 fully connected layers, and a softmax classifier. The number of convolutional kernels in the 3 convolutional layers are 64, 128, and 256, respectively. The kernel size is 3×3, the stride is 1, and the padding method is "same". The "same padding" method is to fill the array with a circle of 0s to increase the array size. The number of nodes in the two fully connected layers is 512 each. The number of nodes in the softmax classifier is determined by the number of statistically analyzed stratigraphic label types.

[0064] Step 4: Input the two-dimensional representation fragment of the well logging data to be visualized into the trained convolutional neural network, select the convolutional layer to be visualized, and output the feature map of the layer; upsample the output feature map to the size of the input fragment and multiply it point by point with the two-dimensional representation fragment of the input well logging data, and then re-input it into the convolutional neural network with the same structure to obtain the confidence score of each feature map;

[0065] Specifically, the formula for calculating the confidence score of the two-dimensional representation fragment to be visualized is as follows:

[0066]

[0067] in, This is the output feature map of the k-th channel of the l-th convolutional layer. for Global confidence score for the two-dimensional representation segment;

[0068] The calculation formula is:

[0069] The Up function represents the feature map The dimension of the upsampled feature map is determined by the dimension of the input vector. The Norm function represents the normalization of the upsampled feature map.

[0070] Step 5: Use the confidence score of each channel feature map as a weight and linearly weight it with the feature map of the corresponding channel to obtain the class discrimination saliency map;

[0071] The formula for solving the class discrimination saliency map is as follows:

[0072]

[0073] Among them, L c This is a saliency map for class C strata. The weights corresponding to the feature map of the k-th channel are represented by , where The solution formula is as follows:

[0074]

[0075] in, This is the global confidence score.

[0076] Step 6: Repeat steps 1 to 5 to visualize two-dimensional representations of multiple well logging data segments from the same formation, derive corresponding class discrimination saliency maps, analyze their common patterns, and summarize the mapping of the formation's electrical characteristics in the feature space. The results show that the deep-level features of the well logging curves are dominated by the curve features in the middle region, and the curve interactions exhibit a clear periodic vertical arrangement. The visualization results show that the regions of interest to the network exhibit uneven color distribution, indicating that different well logging curves contribute differently to the network's decision-making.

[0077] Example

[0078] This embodiment provides a method for visualizing stratigraphic division models using confidence scores. The specific process is as follows:

[0079] Step 1: Block data filtering.

[0080] In this embodiment, a geographic coordinate map of all wells in a certain block of an oil field was first drawn, such as... Figure 2As shown, 228 wells with similar geographical locations were selected as the dataset using hierarchical clustering. 28 of these wells were randomly selected as test wells, and the remaining 200 wells were used as training wells. The well location map of the dataset is shown below. Figure 3 As shown in Table 1, each well in the dataset contains six logging curves: Depth, Natural Gamma (GR), Spontaneous Potential (SP), Acoustic Transit Time (AC), Array Inductive Resistivity (AT90), and Array Inductive Resistivity (AT20). The logging data for each well is presented as discrete data at a single depth point, with each depth point corresponding to a geological stratum. There are 10 geological strata in total: 'K1z2+1', 'J2z', 'J1y', 'J2a', 'J1f', 'Chang1', 'Chang2', 'Chang3', 'Chang4+5', and 'Chang6'.

[0081] Table 1. Examples of dataset formats

[0082] well name Depth(m) GR(gAPI) SP(mv) AC (us / m) AT90 (Ω·m) AT20(Ω·m) Label W100 648.000 50.83 40.87 263.05 14.93 15.40 J1a W100 648.125 53.03 41.73 269.79 13.44 14.54 J1a … … … … … … … … W100 1270.000 97.21 66.91 243.65 18.82 20.33 J1f W100 1270.125 96.60 66.73 244.11 15.71 17.84 Chang1 … … … … … … … … W100 1807.625 0.00 78.83 239.94 21.72 8.12 Chang6 W100 1807.750 0.00 78.70 237.83 21.47 8.08 Chang6

[0083] Step 2: Two-dimensional characterization of well logging data.

[0084] First, outliers such as 0, -1, and 9999 are removed from the logging curve data of each well in the dataset due to factors such as logging sensor jitter, to prevent outliers from affecting the network training effect. Then, the mean and standard deviation of each logging curve for each well are calculated according to the standardization formula:

[0085] The logging curve data for each well is standardized according to the curve to avoid excessive data deviation. Then, a 96×6 sliding window is selected, and the window is slid along the direction of the logging curve data at intervals of one depth point. Each time the window covers an area, a 96×6 two-dimensional representation segment of the logging data is obtained. The process of converting logging curve data into a two-dimensional representation segment is as follows: Figure 4 As shown. Finally, all six discrete logging curves from each well were converted into 96×6 two-dimensional representation segments of logging data. The formation label corresponding to each segment was the formation label that appeared most frequently among the 96 depth points selected in the window. Finally, all geological layer labels were vectorized. That is, the types of geological layer labels in the training set were counted as 10, and the geological layer labels were numbered from 0. Then, the number of each label was vectorized into a 10-dimensional vector, where one index in the vector was 1 and the rest were 0.

[0086] Step 3: Construct and train a convolutional neural network model. First, construct a convolutional neural network containing 3 convolutional layers, 2 fully connected layers, and a softmax classifier for stratum classification. The convolutional neural network structure is as follows: Figure 5As shown, the number of convolutional kernels in the three convolutional layers of the convolutional neural network are 64, 128, and 256, respectively. The kernel size is 3×3, the stride is 1, and the padding method is "same padding," which means padding the array with zeros around it to increase the array size. During convolution, the kernel can extend beyond the original array boundary to prevent the loss of edge information during convolution and ensure that the size of the result after convolution is consistent with that before convolution. A batch normalization layer is added between each convolutional layer to prevent overfitting during network training. The number of nodes in both fully connected layers is 512. Since there are 10 geological layers, the number of nodes in the softmax classifier is set to 10. The activation function of the convolutional layers and fully connected layers is ReLU, and the Adam algorithm is used for network training optimization. The preprocessed two-dimensional well logging data representation fragments from step two are input into the convolutional neural network. The network parameters are updated according to the stratigraphic labels corresponding to the two-dimensional fragments. Finally, the network model is trained and saved.

[0087] Step 4: Calculate the confidence score of the two-dimensional representation fragment to be visualized. Input the two-dimensional representation fragment of the well logging data to be visualized into the trained convolutional neural network. Select the convolutional layer to be visualized and output the feature map of that layer. Upsample the output feature map to the size of the input fragment and multiply it point-by-point with the input two-dimensional representation fragment of the well logging data. Then, re-input it into the convolutional neural network with the same structure to obtain the confidence score of each feature map. The formula is:

[0088]

[0089] in, This is the output feature map of the k-th channel of the l-th convolutional layer. for Global confidence score for the two-dimensional representation fragment. The calculation formula is:

[0090]

[0091] The Up function represents the feature map The dimension of the upsampled feature map is determined by the dimension of the input vector. The Norm function represents the normalization of the upsampled feature map.

[0092] Step 5: Visualize the 2D representation fragment. The confidence score of each channel feature map obtained in Step 4 is treated as a weight and linearly weighted with the corresponding channel feature map to obtain a class discrimination saliency map, thus visualizing the 2D representation fragment. The visualization mechanism is as follows: Figure 6 As shown. The formula for solving the class discrimination saliency map of the corresponding segment is:

[0093]

[0094] Where L cThis is a saliency map for class C strata. The weight corresponding to the feature map of the k-th channel is calculated using the following formula:

[0095]

[0096] in, This is the global confidence score obtained in step four.

[0097] Step Six: Analyze and summarize the mapping rules of formation electrical characteristics in the feature space. Visualize two-dimensional representation segments of multiple well logging data from the same formation to obtain corresponding class discrimination saliency maps. Analyze the common patterns and summarize the mapping of the formation's electrical characteristics in the feature space. Considering that the Chang6 formation has the highest hydrocarbon content, this invention mainly focuses on visualizing two-dimensional representation segments of well logging data from the Chang6 geological layer. The visualization results are as follows: Figure 7 As shown. Figure 7 The visualization results of three wells (W615, W189, and W634) in the test set are presented. Two two-dimensional representation segments belonging to the Chang6 layer were randomly selected from each well for visualization. Considering the specialization of the logging field, this paper feeds back the visualization results to the logging curve level for comprehensive analysis. The leftmost part of each figure shows the standardized curve shape of five logging curves: GR, SP, AC, AT90, and AT20, with the vertical axis representing the depth of the sampling point. To achieve the observability of the curves, each figure shows a 36-meter logging curve segment, with a vertical axis depth interval of 12 meters, each interval corresponding to a two-dimensional representation segment of logging data. The right side shows the visualization results of the Chang6 layer data. The non-blue areas in the figure are the deep features of the Chang6 layer extracted by the network. The redder the color, the greater the positive influence of the area on the network decision. The blue ellipse on the far right of each figure is a magnified view of the deep features. It can be seen from the figures that the deep features of the Chang6 layer logging curves are mainly the curve features of the middle region, and the interaction of the curves shows a clear periodic vertical arrangement. The visualization results show that the regions of interest to the network exhibit uneven color distribution, indicating that different logging curves contribute differently to the network's decision-making.

[0098] In this embodiment, the Chang6 geological layer with the highest oil and gas content is visualized using the stratigraphic division model visualization method based on confidence scores. The regularity of its electrical characteristics in the feature space can be clearly observed, demonstrating the effectiveness of the present invention and the feasibility of the method.

[0099] The present invention also provides a stratigraphic model visualization system using confidence scores, for implementing the stratigraphic model visualization method using confidence scores described above, including a drawing module, a first data processing module, a model building module, a second data processing module, a fourth data processing module, and an execution module;

[0100] The drawing module is used to draw the spatial coordinates of all oil wells within the block and filter out oil wells with similar spatial coordinates as a dataset. The dataset includes several logging curve data.

[0101] The first data processing module is used to remove outliers from the logging curve data in the dataset, standardize the logging curve data of each oil well according to the curve, select a sliding window to obtain a two-dimensional representation segment of the two-dimensional logging data, convert all the discrete logging curve data in each well into two-dimensional segments of the sliding window size, obtain the formation label corresponding to each segment, vectorize all the geological layer labels to obtain the dataset, and divide the obtained dataset into training set and test set according to the proportion.

[0102] The model building module is used to build a convolutional neural network model. It inputs the data in the training set into the convolutional neural network to obtain the network's output, compares the network's output with the formation label, calculates the loss function, and then calculates the gradient of the network parameters with respect to the loss function through the backpropagation algorithm. The training process is repeated until the preset maximum number of training rounds is reached to complete the training.

[0103] The second data processing module is used to input the two-dimensional representation fragment of the well logging data to be visualized into the trained convolutional neural network, select the convolutional layer to be visualized, and output the feature map of the layer; after upsampling the output feature map to the size of the input fragment and multiplying it point by point with the two-dimensional representation fragment of the input well logging data, it is re-inputted into the convolutional neural network with the same structure to obtain the confidence score of each feature map;

[0104] The fourth data processing module is used to linearly weight the confidence score of each channel feature map with the corresponding channel feature map to obtain a class discrimination saliency map;

[0105] The execution module is used for repeated execution to visualize two-dimensional representation fragments of multiple well logging data from the same formation and generate corresponding class discrimination saliency maps.

[0106] The present invention also provides a mobile terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor, such as a stratigraphic partitioning model visualization program using confidence scores.

[0107] When the processor executes the computer program, it implements the steps of the above-described method for visualizing stratigraphic models using confidence scores, including the following steps:

[0108] Step 1: Draw the spatial coordinates of all oil wells within the block, and select oil wells with similar spatial coordinates as the dataset. The dataset includes several logging curve data.

[0109] Step 2: Remove outliers from the logging curve data in the dataset, standardize the logging curve data of each oil well according to the curve, and select a sliding window to obtain a two-dimensional representation segment of the two-dimensional logging data. Convert all the discrete logging curve data in each well into two-dimensional segments of the sliding window size to obtain the formation label corresponding to each segment. Vectorize all the geological layer labels to obtain the dataset, and divide the obtained dataset into training set and test set according to the proportion.

[0110] Step 3: Construct a convolutional neural network model. Input the data in the training set into the convolutional neural network to obtain the network output. Compare the network output with the formation label and calculate the loss function. Then, use the backpropagation algorithm to calculate the gradient of the network parameters with respect to the loss function. Repeat the training process until the preset maximum number of training rounds is reached to complete the training.

[0111] Step 4: Input the two-dimensional representation fragment of the well logging data to be visualized into the trained convolutional neural network, select the convolutional layer to be visualized, and output the feature map of the layer; upsample the output feature map to the size of the input fragment and multiply it point by point with the two-dimensional representation fragment of the input well logging data, and then re-input it into the convolutional neural network with the same structure to obtain the confidence score of each feature map;

[0112] Step 5: Use the confidence score of each channel feature map as a weight and linearly weight it with the feature map of the corresponding channel to obtain the class discrimination saliency map;

[0113] Step 6: Repeat steps 1 to 5 to visualize two-dimensional representation segments of multiple well logging data from the same formation, obtain the corresponding class discrimination saliency map, and obtain the mapping results of the electrical characteristics of the formation in the feature space.

[0114] Alternatively, when the processor executes the computer program, it implements the functions of each module in the above system. For example, the drawing module is used to draw the spatial coordinates of all oil wells in the block and select oil wells with similar spatial coordinates as a dataset. The dataset includes several logging curve data.

[0115] The first data processing module is used to remove outliers from the logging curve data in the dataset, standardize the logging curve data of each oil well according to the curve, select a sliding window to obtain a two-dimensional representation segment of the two-dimensional logging data, convert all the discrete logging curve data in each well into two-dimensional segments of the sliding window size, obtain the formation label corresponding to each segment, vectorize all the geological layer labels to obtain the dataset, and divide the obtained dataset into training set and test set according to the proportion.

[0116] The model building module is used to build a convolutional neural network model. It inputs the data in the training set into the convolutional neural network to obtain the network's output, compares the network's output with the formation label, calculates the loss function, and then calculates the gradient of the network parameters with respect to the loss function through the backpropagation algorithm. The training process is repeated until the preset maximum number of training rounds is reached to complete the training.

[0117] The second data processing module is used to input the two-dimensional representation fragment of the well logging data to be visualized into the trained convolutional neural network, select the convolutional layer to be visualized, and output the feature map of the layer; after upsampling the output feature map to the size of the input fragment and multiplying it point by point with the two-dimensional representation fragment of the input well logging data, it is re-inputted into the convolutional neural network with the same structure to obtain the confidence score of each feature map;

[0118] The fourth data processing module is used to linearly weight the confidence score of each channel feature map with the corresponding channel feature map to obtain a class discrimination saliency map;

[0119] The execution module is used for repeated execution to visualize two-dimensional representation fragments of multiple well logging data from the same formation and generate corresponding class discrimination saliency maps.

[0120] For example, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the mobile terminal. For example, the computer program can be divided into a drawing module, a first data processing module, a model building module, a second data processing module, a fourth data processing module, and an execution module;

[0121] The specific functions of each module are as follows:

[0122] The drawing module is used to draw the spatial coordinates of all oil wells within the block and filter out oil wells with similar spatial coordinates as a dataset. The dataset includes several logging curve data.

[0123] The first data processing module is used to remove outliers from the logging curve data in the dataset, standardize the logging curve data of each oil well according to the curve, select a sliding window to obtain a two-dimensional representation segment of the two-dimensional logging data, convert all the discrete logging curve data in each well into two-dimensional segments of the sliding window size, obtain the formation label corresponding to each segment, vectorize all the geological layer labels to obtain the dataset, and divide the obtained dataset into training set and test set according to the proportion.

[0124] The model building module is used to build a convolutional neural network model. It inputs the data in the training set into the convolutional neural network to obtain the network's output, compares the network's output with the formation label, calculates the loss function, and then calculates the gradient of the network parameters with respect to the loss function through the backpropagation algorithm. The training process is repeated until the preset maximum number of training rounds is reached to complete the training.

[0125] The second data processing module is used to input the two-dimensional representation fragment of the well logging data to be visualized into the trained convolutional neural network, select the convolutional layer to be visualized, and output the feature map of the layer; after upsampling the output feature map to the size of the input fragment and multiplying it point by point with the two-dimensional representation fragment of the input well logging data, it is re-inputted into the convolutional neural network with the same structure to obtain the confidence score of each feature map;

[0126] The fourth data processing module is used to linearly weight the confidence score of each channel feature map with the corresponding channel feature map to obtain a class discrimination saliency map;

[0127] The execution module is used for repeated execution to visualize two-dimensional representation fragments of multiple well logging data from the same formation and generate corresponding class discrimination saliency maps.

[0128] The mobile terminal can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. The mobile terminal may include, but is not limited to, a processor and a memory.

[0129] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the mobile terminal, connecting various parts of the mobile terminal via various interfaces and lines.

[0130] The memory can be used to store the computer program and / or module. The processor implements various functions of the mobile terminal by running or executing the computer program and / or module stored in the memory and calling the data stored in the memory.

[0131] The memory may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function (such as sound playback or image playback). The data storage area may store data created based on the use of the phone (such as audio data or a phonebook). Furthermore, the memory may include high-speed random access memory (RAM) and non-volatile memory, such as hard disks, RAM, plug-in hard disks, SmartMediaCards (SMC), Secure Digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0132] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for visualizing stratigraphic models using confidence scores, characterized in that, Includes the following steps: Step 1: Draw the spatial coordinates of all oil wells within the block, and select oil wells with similar spatial coordinates as the dataset. The dataset includes several logging curve data. Step 2: Remove outliers from the logging curve data in the dataset, standardize the logging curve data of each oil well according to the curve, and select a sliding window to obtain a two-dimensional representation segment of the two-dimensional logging data. Convert all the discrete logging curve data in each well into two-dimensional segments of the sliding window size to obtain the formation label corresponding to each segment. Vectorize all the geological layer labels to obtain the dataset, and divide the obtained dataset into training set and test set according to the proportion. Step 3: Construct a convolutional neural network model. Input the data in the training set into the convolutional neural network to obtain the network output. Compare the network output with the formation label and calculate the loss function. Then, use the backpropagation algorithm to calculate the gradient of the network parameters with respect to the loss function. Repeat the training process until the preset maximum number of training rounds is reached to complete the training. Step 4: Input the two-dimensional representation fragment of the well logging data to be visualized into the trained convolutional neural network, select the convolutional layer to be visualized, and output the feature map of the layer; upsample the output feature map to the size of the input fragment and multiply it point by point with the two-dimensional representation fragment of the input well logging data, and then re-input it into the convolutional neural network with the same structure to obtain the confidence score of each feature map; Step 5: Use the confidence score of each channel feature map as a weight and linearly weight it with the feature map of the corresponding channel to obtain the class discrimination saliency map; Step 6: Repeat steps 1 to 5 to visualize two-dimensional representation segments of multiple well logging data from the same formation, obtain the corresponding class discrimination saliency map, and obtain the mapping results of the electrical characteristics of the formation in the feature space.

2. The method for visualizing stratigraphic division models using confidence scores according to claim 1, characterized in that, In step 1, hierarchical clustering is used to prioritize oil wells with similar spatial coordinates as the dataset. Several logging curves in the dataset are used as the data sources for several features. The logging curves are presented in the form of discrete data at a single depth point, and each depth point corresponds to a formation attribute.

3. The method for visualizing stratigraphic division models using confidence scores according to claim 1, characterized in that, Several logging curve data include Depth data, Natural Gamma Regulator (GR) data, Spontaneous Potential (SP) data, Acoustic Transit Time (AC) data, Array Inductive Resistivity (AT90) data, and Array Inductive Resistivity (AT20) data.

4. The method for visualizing stratigraphic division models using confidence scores according to claim 1, characterized in that, In step 2, the logging curve data of each oil well is standardized according to the curve. The standardization process is Z-Score standardization, and the formula is as follows: Where μ is the data mean and σ is the data standard deviation.

5. The method for visualizing stratigraphic division models using confidence scores according to claim 1, characterized in that, In step 2, a sliding window of 96×6 pixels is selected and slid along the direction of the logging curve data at intervals of one depth point. Each time the window covers an area, a 96×6 two-dimensional segment of logging data is obtained. All the discrete logging curve data in each well are converted into 96×6 two-dimensional segments. The formation label corresponding to each segment is the formation label that appears most frequently among the 96 depth points selected in the window. Finally, all geological layer labels are vectorized. The specific process of vectorizing all geological layer labels is as follows: The number of stratigraphic labels n in the training set is statistically analyzed. The stratigraphic labels are numbered from 0, and the number of each label is vectorized into an n-dimensional vector, where one index in the vector is 1 and the rest are 0.

6. The method for visualizing stratigraphic division models using confidence scores according to claim 1, characterized in that, In step 3, the architecture of the convolutional neural network is designed first in the construction of the convolutional neural network model. The architecture of the convolutional neural network includes 3 convolutional layers, 2 fully connected layers, and a softmax classifier. The number of convolutional kernels in the 3 convolutional layers are 64, 128, and 256, respectively. The kernel size is 3×3, the stride is 1, and the padding method is "same". The "same padding" method is to pad the array with a circle of 0s to increase the array size. The number of nodes in the two fully connected layers is 512 each. The number of nodes in the softmax classifier is determined by the number of statistical stratigraphic label types.

7. The method for visualizing stratigraphic division models using confidence scores according to claim 1, characterized in that, In step 4, the formula for calculating the confidence score of the two-dimensional representation fragment to be visualized is as follows: in, This is the output feature map of the k-th channel of the l-th convolutional layer. for Global confidence score for the two-dimensional representation segment; The calculation formula is: The Up function represents the feature map The dimension of the upsampled feature map is determined by the dimension of the input vector. The Norm function represents the normalization of the upsampled feature map.

8. The method for visualizing stratigraphic division models using confidence scores according to claim 1, characterized in that, In step 5, the formula for solving the class discrimination saliency map is as follows: Among them, L c This is a saliency map for class C strata. The weights corresponding to the feature map of the k-th channel are represented by , where The solution formula is as follows: in, This is the global confidence score.

9. A stratigraphic model visualization system using confidence scores, used to implement the stratigraphic model visualization method using confidence scores as described in any one of claims 1-8, characterized in that, include: The drawing module is used to draw the spatial coordinates of all oil wells within the block and filter out oil wells with similar spatial coordinates as a dataset. The dataset includes several logging curve data. The first data processing module is used to remove outliers from the logging curve data in the dataset, standardize the logging curve data of each oil well according to the curve, select a sliding window to obtain a two-dimensional representation segment of the two-dimensional logging data, convert all the discrete logging curve data in each well into two-dimensional segments of the sliding window size, obtain the formation label corresponding to each segment, vectorize all the geological layer labels to obtain the dataset, and divide the obtained dataset into training set and test set according to the proportion. The model building module is used to build a convolutional neural network model. It inputs the data in the training set into the convolutional neural network to obtain the network's output, compares the network's output with the formation label, calculates the loss function, and then calculates the gradient of the network parameters with respect to the loss function through the backpropagation algorithm. The training process is repeated until the preset maximum number of training rounds is reached to complete the training. The second data processing module is used to input the two-dimensional representation fragment of the well logging data to be visualized into the trained convolutional neural network, select the convolutional layer to be visualized, and output the feature map of the layer; after upsampling the output feature map to the size of the input fragment and multiplying it point by point with the two-dimensional representation fragment of the input well logging data, it is re-inputted into the convolutional neural network with the same structure to obtain the confidence score of each feature map; The fourth data processing module is used to linearly weight the confidence score of each channel feature map with the corresponding channel feature map to obtain a class discrimination saliency map; The execution module is used for repeated execution to visualize two-dimensional representation fragments of multiple well logging data from the same formation and generate corresponding class discrimination saliency maps.

10. A mobile terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for visualizing a stratigraphic partitioning model using confidence scores as described in any one of claims 1 to 8.

Images