Stratum classification interpretation method and device

Abstract

The invention discloses a stratum classification interpretation method and device. The method comprises the steps that a logging data set of a target well section is acquired; inputting the logging data set into a multi-level progressive classification interpretation model obtained by training to obtain a characteristic value corresponding to each logging data in the logging data set, and determining a reservoir type to which the logging data belongs according to the characteristic value; determining a target sub-model corresponding to the logging data from a plurality of sub-models according to the reservoir type and the characteristic value of each logging data; and performing analysis processing on the logging data based on the target sub-model to obtain a target prediction result representing the stratum classification to which the logging data belongs and the corresponding stratum interpretation. The attention degrees of the sub-models corresponding to different reservoir types on the data feature values of the logging data and the data processing modes corresponding to the logging data are different, the situation that traditional models with the same processing methods and the same feature attention emphasis are generally used for conducting classification and interpretation processing on the logging data is avoided, and the accuracy of the stratum prediction result is improved.

Description

Technical Field

[0001] This invention relates to the field of well logging data processing technology, specifically to a method and apparatus for stratigraphic classification and interpretation. Background Technology

[0002] Well logging, acting as the "eyes" penetrating deep into the formation, plays a crucial role in oil and gas field exploration and development. It boasts advantages such as diverse methods, high resolution, and a large amount of information, providing continuous and accurate in-situ physical parameters including electrical, acoustic, and nuclear parameters, offering invaluable data support for reservoir evaluation. Formations are categorized into non-reservoir and reservoir types, with each further subdivided into multiple levels of subcategories. Detailed classification of formations is of great significance in determining the distribution of oil and gas resources, predicting drilling effectiveness, and optimizing development strategies.

[0003] In existing technologies, well logging data is input into a pre-built big data model to obtain corresponding formation prediction results. However, existing technologies do not take into account the different characteristics of well logging data for different formation categories, which prevents the big data model from effectively capturing and utilizing key information in the well logging data, resulting in poor accuracy of the final formation prediction results. Summary of the Invention

[0004] In view of the above problems, embodiments of the present invention are proposed to provide a stratigraphic classification and interpretation method and apparatus that overcomes or at least partially solves the above problems.

[0005] According to one aspect of the present invention, a stratigraphic classification and interpretation method is provided, the method comprising: Obtain the logging dataset for the target well section; the logging dataset includes multiple logging data. The well logging dataset is input into the trained multi-level progressive classification interpretation model to obtain the feature values ​​corresponding to each well logging data in the well logging dataset, and the reservoir type to which the well logging data belongs is determined based on the feature values; wherein, the multi-level progressive classification interpretation model is obtained by integrating multiple sub-models at different levels; different sub-models correspond to different reservoir types; Based on the reservoir type and the characteristic values ​​of each logging data, the target sub-model corresponding to the logging data is determined from multiple sub-models; Based on the target sub-model, well logging data is analyzed and processed to obtain target prediction results that characterize the stratigraphic classification and corresponding stratigraphic interpretation of the well logging data.

[0006] According to another aspect of the present invention, a stratigraphic classification and interpretation apparatus is provided, comprising: The acquisition module is suitable for acquiring the logging dataset of the target well section; the logging dataset includes multiple logging data. The reservoir type module is suitable for inputting well logging datasets into a trained multi-level progressive classification interpretation model to obtain feature values ​​corresponding to each well logging data in the well logging dataset, and to determine the reservoir type to which the well logging data belongs based on the feature values; wherein, the multi-level progressive classification interpretation model is obtained by integrating multiple sub-models at different levels; different sub-models correspond to different reservoir types; The sub-model module is suitable for determining the target sub-model corresponding to the logging data from multiple sub-models based on the reservoir type and the characteristic values ​​of each logging data. The prediction module is suitable for analyzing and processing well logging data based on the target sub-model to obtain target prediction results that characterize the stratigraphic classification and corresponding stratigraphic interpretation of the well logging data.

[0007] According to another aspect of the present invention, a computing device is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus; The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the above-described stratigraphic classification and interpretation method.

[0008] According to another aspect of the present invention, a computer storage medium is provided, the storage medium storing at least one executable instruction that causes a processor to perform operations corresponding to the above-described stratigraphic classification and interpretation method.

[0009] According to another aspect of the present invention, a computer program product is provided, including at least one executable instruction that causes a processor to perform operations corresponding to the stratigraphic classification and interpretation method described above.

[0010] According to the formation classification and interpretation method and apparatus provided in the embodiments of the present invention, the sub-models corresponding to different reservoir types pay different attention to the data feature values ​​of well logging data and have different data processing methods for the corresponding well logging data. This avoids the use of traditional models with the same processing methods and the same focus on features to classify and interpret well logging data, thereby improving the accuracy of formation prediction results.

[0011] The above description is merely an overview of the technical solutions of the embodiments of the present invention. In order to better understand the technical means of the embodiments of the present invention and to implement them in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of the present invention more obvious and understandable, specific implementation methods of the embodiments of the present invention are described below. Attached Figure Description

[0012] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the embodiments of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart of a stratigraphic classification and interpretation method according to an embodiment of the present invention is shown; Figure 2 A schematic diagram of the classification of formation and reservoir types is shown; Figure 3 A schematic diagram of a multi-level progressive classification explanation model is shown; Figure 4 The flowchart of the training process of the multi-level progressive classification explanation model is shown; Figure 5 A schematic diagram of a stratigraphic classification and interpretation apparatus according to an embodiment of the present invention is shown; Figure 6 A schematic diagram of the structure of a computing device according to an embodiment of the present invention is shown. Detailed Implementation

[0013] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0014] Figure 1 A flowchart of a stratigraphic classification and interpretation method according to an embodiment of the present invention is shown, as follows: Figure 1 As shown, the method includes the following steps: Step S101: Obtain the logging dataset for the target well section.

[0015] The well logging dataset includes multiple well logging data, such as caliper logging (CAL), natural gamma logging (GR), deep resistivity logging (RD), shallow resistivity logging (RS), neutron logging (CNCF), density logging (ZDEN), spectral logging (PE), spontaneous potential logging (SPDH), sonic logging (DTCR), total hydrocarbon content (Tg), and methane content (C1).

[0016] In well logging, the target interval refers to a specific stratum or rock formation that is planned or expected to be exploited during drilling or oil and gas exploration. This interval typically contains economically valuable oil and gas resources or other mineral resources and is the primary focus of drilling operations. In well logging data acquisition, a "well" refers to the drilled borehole, typically used for the exploration and exploitation of oil, natural gas, geothermal energy, and mineral resources. It is a vertical channel drilled into the ground, through which measurements and samples are taken of the underground rock formations and resources. Various well logging data can be obtained from different logging methods, such as electrical logging, spontaneous potential logging, nuclear logging, radioactive logging, electromagnetic logging, caliper logging, and imaging logging. The specific method chosen depends on the actual situation and is not limited here.

[0017] Step S102: Input the well logging dataset into the trained multi-level progressive classification interpretation model to obtain the feature values ​​corresponding to each well logging data in the well logging dataset, and determine the reservoir type to which the well logging data belongs based on the feature values.

[0018] The multi-level progressive classification interpretation model in this embodiment is obtained by integrating multiple sub-models at different levels. These sub-models are divided into different levels, and each level is constructed based on the reservoir type. Reservoir types can be as follows: Figure 2 As shown, well logging data mainly includes reservoir data and non-reservoir data. Reservoir data includes many important data points for measurement, such as marl layer data, coal seam data, dry layer data, and gas-water layer data. Gas-water layer data further includes Class I gas layer data, Class II gas layer data, water layer data, gas-water layer data, water-gas layer data, and gas-water co-layer data. In the acquired well logging data, non-reservoir data constitutes a large proportion (usually over 70%, even reaching over 90%). This means that existing big data models, when using well logging data to classify formations, often focus more on non-reservoir data. In other words, while big data models are more accurate in predicting the clay content of non-reservoir data when classifying and interpreting the formations represented by well logging data, their accuracy in predicting reservoir content is poor.

[0019] Based on the above problems, and in order to analyze the input well logging data more accurately, the multi-level progressive classification interpretation model in this embodiment integrates multiple sub-models, and... Figure 2Correspondingly, the first level includes: reservoir models and non-reservoir models; the second level includes: marl models, coal seam models, dry seam models, and gas-water layer models related to reservoir models; the third level includes: Class I gas layer models, Class II gas layer models, water layer models, gas-water layer models, water-gas layer models, and gas-water co-layer models related to gas-water layer models. Different reservoir types correspond to different sub-models, thus enabling targeted analysis and interpretation of logging data, thereby improving the accuracy of the final target prediction results. A multi-level progressive classification interpretation model is shown below. Figure 3 As shown, well logging data represents gas-water co-layer data. Because existing big data models focus more on non-reservoir data during training, the output results also tend to focus on non-reservoir related parts, resulting in inaccurate predictions as the final output fails to adequately reflect the gas-water co-layer characteristics of the well logging data. In this embodiment, the multi-level progressive classification interpretation model uses different target sub-models to interpret the well logging data based on its formation classification. By determining the reservoir type of the well logging data layer by layer and using the corresponding sub-model for targeted interpretation, it ensures that the most suitable model for the formation type is used when analyzing well logging data of different formation categories. This accurately reflects the true characteristics of the well logging data and avoids the problem of traditional big data models focusing too much on non-reservoir data when processing gas-water co-layer data, leading to inaccurate target prediction results. Figure 2 and Figure 3 For illustrative purposes only, the specific implementation can be adjusted according to the actual situation. The multi-level progressive classification interpretation model can be set according to the actual stratigraphic classification, and there are no restrictions here.

[0020] Well logging data is input into a multi-level progressive classification and interpretation model. Based on the input well logging data, the characteristic values ​​corresponding to each well logging data can be determined. These characteristic values ​​are then used to determine the corresponding reservoir type, allowing for the step-by-step derivation of the target sub-model. For example... Figure 3 As shown in the diagram, in the multi-level progressive classification and interpretation model of this embodiment, if the current level has a related next level, then the next level includes at least one sub-model. It should be noted that whether each sub-model can be further subdivided into more related sub-models depends on the actual situation. For example, coal seam data can be further subdivided into methane data and gas data; therefore, the coal seam model can be further subdivided into a methane model and a gas model. The specific subdivision depends on the implementation and is not limited here.

[0021] The feature values ​​of each logging data point are determined based on multiple components of the logging data. For example, the feature value corresponding to the input logging data is determined to be the mean GR based on the natural gamma ray curve (GR). Multiple feature values ​​can also be obtained by determining the feature values ​​corresponding to each other input component one by one. After obtaining multiple feature values, depending on the level at which the logging data is processed, one or more feature values ​​are selected as the feature values ​​for the current level. This allows for the selection of a suitable sub-model from multiple sub-models at the current level to process the logging data. For example, the GR feature value can be used to determine whether the logging data belongs to reservoir data or non-reservoir data. This helps determine at the first level whether the logging data is processed by a reservoir model or a non-reservoir model. If GR < 90 is defined as reservoir data, and the GR feature value corresponding to the logging data is < 90, then the logging data can be initially processed by the reservoir model, and the target sub-model for the final processing of the logging data can be determined to be one of the sub-models under the reservoir model subdivision.

[0022] Each sub-model corresponds to a reservoir type. Different reservoir types correspond to different sub-models. Since the sub-models at each level focus on different aspects of the logging data, it is necessary to re-determine one or more feature values ​​from multiple feature values ​​in each level to determine the corresponding reservoir type of the logging data at the current level, so as to determine the corresponding sub-model based on the reservoir type.

[0023] Understandably, different reservoir types have different sub-models that focus on the data feature values ​​of well logging data and employ different data processing methods. For example, although the gas-water reservoir model and the water-gas reservoir model are both derived from the gas-water reservoir model, these two sub-models focus on different aspects of the well logging data. This avoids the problem of inaccurate target prediction results caused by using traditional models with the same processing methods and focus on the same features to classify and interpret well logging data in a general way.

[0024] In this embodiment, the multi-level progressive classification explanation model needs to be pre-trained, and the training process is as follows: Figure 4 As shown: Step S401: Obtain the sample logging dataset and the corresponding verification result set.

[0025] In this embodiment, the sample logging dataset includes multiple sample logging data sets, determined based on logging curve data. Logging curve data refers to formation attribute data obtained through logging techniques, such as resistivity, density, and acoustic properties; it is typically presented as a function of depth. After acquiring the logging curve data, data cleaning is performed on each type of categorized logging curve data. This includes supplementing missing values, smoothing noisy data, identifying or removing outliers, and resolving inconsistencies. Alignment processing is then performed on the logging curve data to obtain initial logging data corresponding to each logging curve data set, with each processed initial logging data set having the same data length. Data alignment can employ methods such as interpolation and resampling. Standardization processing is then performed on the initial logging data to ensure comparability between logging data sets of different mud layer types, resulting in standard logging data. Standardization processing includes methods such as mean normalization, z-score normalization, and min-max normalization, thereby ensuring that the logging curve data adopts the same depth or time reference and is converted to a uniform scale, guaranteeing data consistency in the analysis. The standard logging data undergoes dimensionality reduction, specifically by extracting data features for each mud layer type. This can be achieved using relevant feature extraction algorithms to extract features from the input curve data for each mud layer type. Correlation analysis is then performed on the standard logging data based on these features, using methods such as correlation coefficients and covariance matrices to calculate the correlation between the standard logging data. Based on the correlation analysis results, further importance analysis is conducted on the standard logging data using the data features of each type. This can be achieved using relevant algorithms (such as variance selection, chi-square test, and model-based feature selection) to assess the importance of the logging data. The results of the correlation and importance analysis are then used as the processing data for further dimensionality reduction. This results in dimensionality-reduced logging data, which can be obtained using methods such as Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) to transform high-dimensional data into a low-dimensional representation. This reduces the dimensionality of the data, preserving important information and reducing noise and redundancy while improving the efficiency of subsequent model training. Finally, a sample logging dataset is obtained based on all the dimensionality-reduced logging data.

[0026] Step S402: Train the multi-level progressive classification interpretation model based on the sample well logging dataset to obtain the sub-model prediction results corresponding to multiple sample well logging data at different levels.

[0027] Since the sub-models in the multi-level progressive classification explanation model are highly correlated, in order to better determine the final target sub-model, the prediction results of each sub-model at different levels are first obtained so that the parameters of the corresponding sub-models can be adjusted according to the validation result set until all sub-models meet the preset training conditions, thus obtaining the trained multi-level progressive classification explanation model.

[0028] Since different sub-models handle well logging data processing and fitting differently, models constructed with different network structures can be selected as the model architecture for sub-models based on the different processing priorities of sub-models at different levels. Specifically, sub-models at the same level can all be of the same type, such as both reservoir models and non-reservoir models being vector machine models; or, sub-models of different reservoir types at the same level can also have different model types. The specific settings can be determined according to the actual situation and are not limited here.

[0029] Neural network models are computational models inspired by biological nervous systems, primarily used for tasks such as pattern recognition, classification, and regression. They consist of many simple computational units (i.e., neurons or nodes), which are interconnected by connection weights to form a complex network structure. Neural networks include feedforward neural network models, convolutional neural network models, recurrent neural networks, and graph neural networks, among others.

[0030] Decision tree models are supervised learning algorithms, mainly used for classification and regression tasks. They recursively divide the dataset into smaller subsets until each subset (or leaf node) contains only one class or meets certain stopping conditions, thus forming a tree-like structure.

[0031] The Support Vector Machine (SVM) model is a model trained using supervised learning algorithms in machine learning, primarily used for classification and regression analysis. SVM achieves data classification, regression, or anomaly detection tasks by constructing one or more hyperplanes to segment data in a high-dimensional space.

[0032] The sub-model in this embodiment can be constructed from one of the following: neural network model, decision tree model, or vector machine model.

[0033] The training condition in this embodiment is the number of training iterations. The training conditions for each sub-model can be different. For example, the more iterations a sub-model is at a later level, the more iterations it can train. This is so that the sub-model at a later level can focus on the part of the well logging data with fewer feature values ​​through multiple training iterations. This makes it easier for the trained model to generate the final target prediction result with a high accuracy even when the well logging data contains an unbalanced proportion of feature values.

[0034] During training, when the sub-model at the current level processes the logging data based on the selected feature values, in addition to determining the sub-model at the next level based on the feature values, it also interprets the logging data processing. The interpretation of the sub-model is based on its level. For example, the gas-water layer model only interprets the gas-water related parts of the input logging data, without providing a rigid interpretation of the relevant content at the next level. This is because the next level model has higher performance processing capabilities and can better handle the uninterpreted data; furthermore, the content interpreted at the current level will be passed down to the next level, and a rigid interpretation would further lead to inaccuracies in the interpretation of the next level sub-model. Therefore, during training, sub-models corresponding to different levels are trained to drive the overall system from a local perspective, thus achieving the training of a multi-level progressive classification interpretation model. When sample logging data is input into the multi-level progressive classification interpretation model, it generates sub-models corresponding to different levels based on the selected branch, and obtains the prediction results of the sample logging data at different levels.

[0035] When training the sub-model, multiple first sample feature values ​​are obtained based on the sample logging data of each sample logging dataset to correspond to the intermediate level of the multi-level progressive classification interpretation model. The first sample feature value can be one or more. Based on the multiple first sample feature values, the first sub-model corresponding to the sample logging data in the intermediate level is determined. Based on the first sub-model, the sample logging data is analyzed and processed to obtain the first prediction result corresponding to the first sub-model.

[0036] The use of sub-models emphasizes the selection of sub-models at each level during practical applications. When training sub-models, attention needs to be paid to the interpretation and processing of well logging data to improve the interpretability of each level of sub-model for the current input well logging data. The interpreted results are then shared with the next level of sub-model, and this process is repeated so that the target sub-model can analyze and process the well logging data based on the preprocessing results to obtain the second prediction result. Specifically, based on the reservoir type of the first sub-model, multiple second sample feature values ​​corresponding to the sample well logging data are determined. Based on these second sample feature values, the second sub-model corresponding to the sample well logging data at the next lower level in the intermediate layer is determined. The first prediction result is shared with the second sub-model, allowing it to analyze and process the sample well logging data based on the first prediction result to obtain the second prediction result corresponding to the second sub-model. The second prediction result is obtained based on the analysis and processing of the first prediction result; that is, the first and second prediction results are correlated to connect the data processing results at each level, improving the overall accuracy of the final target prediction result.

[0037] During the training of the multi-level progressive classification interpretation model, joint learning is carried out by passing and sharing information. This allows the sub-models at the next level to use the prediction results of the previous level as an aid to better understand and analyze the sample well logging data, thereby obtaining a more accurate second prediction result. At the same time, it avoids repetitive and identical information calculations, improves computational efficiency and model training speed, and thus improves the overall performance of the multi-level progressive classification interpretation model.

[0038] Step S403: Use the validation result set to validate the prediction results of the sub-models corresponding to different levels, and obtain the performance evaluation value of each sub-model.

[0039] The validation result set comprises multiple validation result groups, each containing multiple validation results. Each validation result corresponds to a sub-model prediction result. The validation results indicate the ideal input result when sample logging data is input into the current level sub-model. Therefore, the validation results are compared with the sub-model prediction results to obtain the performance evaluation value.

[0040] Step S404: Adaptively adjust the model parameters of each sub-model based on the performance evaluation value until the preset training conditions are met, and obtain the trained multi-level progressive classification explanation model.

[0041] Performance evaluation values ​​encompass two aspects: the accuracy of classifying sample well logging data and the accuracy of interpreting the corresponding data. Therefore, adjusting the model parameters of the corresponding sub-models involves these two adjustments to ensure that the trained sub-models can better classify the input sample well logging data, improving the accuracy of determining the next level of sub-models based on the data, and simultaneously improving the accuracy of the current level sub-models in interpreting the sample well logging data. Specifically, this involves adjusting the sample weights of intermediate-level sub-models based on performance evaluation values, and adjusting the learning rate of the last-level sub-model based on performance evaluation values.

[0042] Model parameters related to classification accuracy include the kernel, penalty parameter C, class weight, and kernel parameters; model parameters related to interpretation accuracy include support vector weight, interpretation metric selection, and model complexity. Specific model parameters can be adjusted according to actual circumstances and are not limited here.

[0043] Furthermore, in addition to adjusting model parameters related to classification accuracy and interpretation accuracy, the focus of adjustment for intermediate-level sub-models and target sub-models also differs. Specifically, for intermediate-level sub-models, adjusting the sample weights of different feature values ​​in the input well logging data allows them to focus on different parts of the imbalanced sample data. For example, higher weights can be assigned to fewer feature values ​​in the well logging data so that the corresponding sub-model can better learn the details in a small number of feature values. For the target sub-model, besides the shared information passed from previous levels, it can capture relatively few other feature values. Therefore, it is necessary to adjust its learning rate to improve the accuracy of the final target prediction results obtained by the target sub-model.

[0044] The above steps complete the training of the multi-level progressive classification explanation model, resulting in the trained multi-level progressive classification explanation model.

[0045] Step S103: Based on the reservoir type and the characteristic values ​​of each logging data, determine the target sub-model corresponding to the logging data from multiple sub-models.

[0046] When determining the target sub-model, the sub-models used at each level can be determined hierarchically until the sub-model of the last level is determined as the target sub-model. Specifically, the intermediate sub-model corresponding to the current level is determined based on the feature values ​​selected at each level, and the target sub-model of the next level up to the last level is determined based on the classification branch in which the intermediate sub-model is located. The target sub-model can perform more detailed interpretation processing of logging data based on the results obtained from the processing of sub-models at previous levels. In this embodiment, the target sub-model is the sub-model in the last level of the multi-level progressive classification interpretation model. The last level refers to the last level of the branch in which the logging data is classified. For example, if the classification path of a certain logging data A in the multi-level progressive classification interpretation model is reservoir model-marl model, then the marl model is determined as the target sub-model corresponding to logging data A; if the classification path of a certain logging data B in the multi-level progressive classification interpretation model is reservoir model-gas-water layer model-gas-water co-layer model, then the gas-water co-layer model is determined as the target sub-model corresponding to logging data B.

[0047] Based on reservoir type, a first feature value and a second feature value can be determined from multiple feature values ​​in the well logging data. The first feature value includes a first proportion of interest, and the second feature value includes a second proportion of interest, with the first proportion of interest being greater than the second proportion of interest. The first feature value is processed based on the first proportion of interest to obtain a first processing result, and the second feature value is processed based on the second proportion of interest to obtain a second processing result. Based on the first and second processing results, the sub-model corresponding to the level of the multi-level progressive classification interpretation model of the well logging data is determined. It is determined whether the level of the sub-model contains a next level. If not, the sub-model is taken as the target sub-model; if so, new first and second feature values ​​are determined and processed again until the level of the obtained sub-model contains no next level, at which point the target sub-model is determined. The new first feature value is different from the original first feature value, and the new second feature value is different from the original second feature value.

[0048] For well logging data, the initial classification determines whether the data represents reservoir or non-reservoir data. Therefore, the well logging data is first preliminarily processed based on the reservoir / non-reservoir model to obtain basic analytical data. More detailed interpretations are then handled by the next-level sub-models. These sub-models aim to address areas not considered in the previous level, especially those with smaller data volumes, to ultimately obtain comprehensive and rich predictive results. To determine the final target sub-model for processing the well logging data, after determining the previous level's sub-model, multiple feature values ​​of the well logging data need to be further analyzed and processed based on it. This process continues to determine the corresponding sub-models for the next level, until the final target sub-model is determined. The target sub-model then completes the final analysis and processing of the well logging data.

[0049] In this embodiment, the first and second feature values ​​can be one or more, used to determine the sub-model corresponding to the current level. For example, after determining that the logging data is processed from reservoir data, more attention needs to be paid to logging data that is more biased towards "reservoir data," i.e., the first feature value is data with a higher correlation to "reservoir data," and the second feature value is data with a lower correlation to "reservoir data." The weight of the first focus is set to be greater than the weight of the second focus. The reservoir model processes the first feature value according to the weight of the first focus to obtain the first processing result, and processes the second feature value according to the weight of the second focus to obtain the second processing result. The sub-model corresponding to the current level is determined by combining the first and second processing results. Different processing weights can be assigned to the first and second processing results respectively, and the total processing result is obtained by combining them. The sub-model that meets the preset matching threshold range of each sub-model at the current level is determined by comparing the total processing result with the preset matching threshold range. Here, the attention weight is used to characterize the degree of attention the sub-models processing well logging data pay to different data, while the processing weight characterizes the proportion of the processed results corresponding to different feature values ​​in the total processing results. These are two different concepts. The attention weight and processing weight can be set to specific values ​​according to actual applications, and are not limited here. The preset matching threshold interval is preset by each sub-model according to the implementation situation, and the setting of the preset matching threshold interval can correspond to the first feature value and the second feature value, used to characterize the corresponding sub-model's ability to further process well logging data. For example, the preset matching threshold intervals corresponding to the marlstone model, coal seam model, dry layer model, and gas-water layer model can be matching interval A, matching interval B, matching interval C, and matching interval D, respectively. The total processing result obtained by processing the well logging data based on the previous level sub-model is 'a'. When the total processing result 'a' falls within matching interval A, the sub-model corresponding to the current level can be determined to be the marlstone model.

[0050] After the sub-model at the current level is determined, since the multi-level progressive classification and interpretation model contains multiple levels, it can be determined whether the level where the sub-model is located has a next level, i.e., whether the level where the sub-model is located is the last level of the multi-level progressive classification and interpretation model. If not, new first and second feature values ​​are determined and processed until the level where the sub-model is located has no next level, and the target sub-model is determined. The new first feature value is different from the original first feature value, and the new second feature value is different from the original second feature value. For ease of understanding, here, the new first feature value is the third feature value, and the new second feature value is the fourth feature value. The third and fourth feature values ​​are determined from multiple feature values ​​according to the reservoir type. The third feature value includes the corresponding third attention weight, and the fourth feature value includes the corresponding fourth attention weight. The third attention weight is greater than the fourth attention weight. The first and third feature values ​​are different, and the second and fourth feature values ​​are different. Considering that different sub-models at different levels focus on different feature values ​​when processing well logging data, for example, if a well logging data is initially processed by a reservoir model, to determine the corresponding second-level sub-model for that data, the GR and CAL feature values ​​need to be considered. However, to determine the sub-model corresponding to the next level of well logging data, the RD (deep resistivity logging curve) and RS (shallow resistivity logging curve) feature values ​​may need to be considered. Therefore, the GR and CAL feature values ​​are determined as the first feature values, and the other feature values ​​in the well logging data besides GR and CAL are determined as the second feature values; the RD and RS feature values ​​are determined as the third feature values, and the other feature values ​​in the well logging data besides RD and RS are determined as the fourth feature values. The above is an example; the specific settings should be determined according to the implementation situation and are not limited here.

[0051] The first and third eigenvalues, and the second and fourth eigenvalues, differ to ensure that different levels of sub-models focus on different aspects of the logging data. This allows the current level to further focus on previously unaddressed eigenvalues, building upon the focus of the previous level's sub-model. By selectively focusing on multiple eigenvalues, the most suitable target sub-model for processing the logging data can be selected from among several sub-models. Alternatively, the first eigenvalue can be a subset of the third eigenvalue. This allows for more detailed analysis of the first eigenvalue by the current level's sub-model, after the previous level's sub-model has analyzed and processed it. In this case, the current level's processing results can be reused from the previous level, accelerating data analysis efficiency. The above are illustrative examples; the specific eigenvalues ​​corresponding to the first, second, third, and fourth eigenvalues ​​can be set according to actual circumstances and are not limited here.

[0052] This embodiment determines the reservoir type to which the logging data belongs layer by layer, thereby enabling the use of corresponding sub-models to interpret the logging data in a targeted manner. This ensures that in practical applications, logging data belonging to different reservoir types can be interpreted using the target sub-model most suitable for that type, accurately reflecting the true characteristics of the logging data. This allows the target prediction results to serve as supporting data for subsequent formation exploration and development, providing a highly accurate scientific basis for the actual development of the formation.

[0053] Step S104: Analyze and process the logging data based on the target sub-model to obtain the target prediction results that characterize the formation classification and corresponding formation interpretation of the logging data.

[0054] After determining the target sub-model, the logging data can be cleaned and corrected based on the target sub-model to remove outliers, noise and other interference factors that exist in the data processing, so as to ensure the accuracy and reliability of the logging data.

[0055] Further interpretation and analysis of the processed logging data reveals rock properties, such as sandstone and mudstone. It also identifies the boundaries and characteristics of different strata, determining formation parameters such as thickness, lithology, porosity, and permeability. When evaluating reservoirs, it determines their reservoir capacity, hydrocarbon properties, and distribution characteristics, providing a basis for subsequent exploration and development. Furthermore, based on the interpretation and processing results of the logging data, subsurface geological models, including reservoir models and structural models, are constructed, providing a geological foundation and basis for oil and gas exploration and development.

[0056] According to the formation classification and interpretation method provided in the embodiments of the present invention, the sub-models corresponding to different reservoir types pay different attention to the data feature values ​​of well logging data and have different data processing methods for the corresponding well logging data. This avoids the use of traditional models with the same processing methods and the same focus on features to classify and interpret well logging data, thereby improving the accuracy of formation prediction results.

[0057] Figure 5 A schematic diagram of the stratigraphic classification and interpretation apparatus provided in an embodiment of the present invention is shown. Figure 5 As shown, the device includes: The acquisition module 510 is suitable for acquiring the logging dataset of the target well section; the logging dataset includes multiple logging data. The reservoir type module 520 is suitable for inputting well logging datasets into a trained multi-level progressive classification interpretation model to obtain feature values ​​corresponding to each well logging data in the well logging dataset, and to determine the reservoir type to which the well logging data belongs based on the feature values; wherein, the multi-level progressive classification interpretation model is obtained by integrating multiple sub-models at different levels; different sub-models correspond to different reservoir types; Sub-model module 530 is suitable for determining the target sub-model corresponding to the logging data from multiple sub-models based on the reservoir type and the characteristic values ​​of each logging data. The prediction module 540 is suitable for analyzing and processing well logging data based on the target sub-model to obtain target prediction results that characterize the stratigraphic classification and corresponding stratigraphic interpretation of the well logging data.

[0058] Optionally, sub-model module 530 is further adapted to: Based on the reservoir type, a first characteristic value and a second characteristic value are determined from multiple characteristic values ​​of the well logging data; the first characteristic value includes a first proportion of interest, the second characteristic value includes a second proportion of interest, and the proportion of the first proportion of interest is greater than the proportion of the second proportion of interest; The first feature value is processed based on the first attention weight to obtain the first processing result, and the second feature value is processed based on the second attention weight to obtain the second processing result. Based on the first and second processing results, determine the sub-model corresponding to the level of the multi-level progressive classification interpretation model for well logging data; Determine if the sub-model exists at a lower level; If not, then the sub-model will be used as the target sub-model; If so, re-determine the new first feature value and the new second feature value for processing and judgment until the level of the obtained sub-model does not have a lower level, and determine the target sub-model.

[0059] Optionally, the prediction module 540 is further adapted to: Based on the target sub-model, the well logging data is cleaned and corrected, and rock properties, formation parameters, reservoir evaluation, and underground geological models are determined based on the well logging data; formation parameters include thickness, lithology, porosity and / or permeability.

[0060] Optionally, the device further includes: a training module 550, comprising: Sample unit 551 is suitable for acquiring sample logging datasets and corresponding validation result sets; the sample logging dataset includes multiple sample logging data; the multi-level progressive classification interpretation model is obtained by integrating multiple sub-models at different levels, and each sub-model is constructed from one of the neural network model, decision tree model, and vector machine model; the sub-models correspond to different reservoir types; Sub-model prediction result unit 552 is suitable for training a multi-level progressive classification interpretation model based on a sample well logging dataset to obtain sub-model prediction results corresponding to multiple sample well logging data at different levels. Evaluation unit 553 is suitable for using the validation result set to validate the prediction results of the sub-models corresponding to different levels, and to obtain the performance evaluation value of each sub-model. The adjustment unit 554 is adapted to adaptively adjust the model parameters of each sub-model based on the performance evaluation value until the preset training conditions are met, so as to obtain the trained multi-level progressive classification explanation model; wherein, the sample weight values ​​of the intermediate-level sub-models are adjusted based on the performance evaluation value, and the learning rate of the last-level sub-model is adjusted based on the performance evaluation value.

[0061] Optionally, sample unit 551 is further adapted to: Acquire multiple well logging curve data; Data alignment is performed on multiple logging curves to obtain initial logging data; The initial logging data is standardized to obtain standard logging data; Dimensionality reduction processing is performed on standard logging data to obtain dimensionality-reduced logging data; A sample logging dataset was obtained based on all the dimensionality-reduced logging data.

[0062] Optionally, the sub-model prediction result unit 552 is further adapted to: Based on the sample logging data of each sample logging dataset, multiple first sample feature values ​​are obtained for the intermediate level of the multi-level progressive classification interpretation model. Based on the multiple first sample feature values, the first sub-model corresponding to the sample logging data in the intermediate level is determined. Based on the first sub-model, the sample logging data is analyzed and processed to obtain the first prediction result corresponding to the first sub-model. Based on the reservoir type of the first sub-model, multiple second sample feature values ​​corresponding to the sample logging data are determined. Based on the multiple second sample feature values, the second sub-model corresponding to the lower layer of the sample logging data in the intermediate layer is determined. The first prediction result is transmitted and shared to the second sub-model so that the second sub-model can analyze and process the sample logging data based on the first prediction result to obtain the second prediction result corresponding to the second sub-model.

[0063] The descriptions of the above modules refer to the corresponding descriptions in the method embodiments, and will not be repeated here.

[0064] This invention also provides a non-volatile computer storage medium storing at least one executable instruction that can perform the operation corresponding to the stratigraphic classification and interpretation method in any of the above method embodiments.

[0065] This application provides a computer program product, which includes at least one executable instruction or computer program that enables a processor to perform operations corresponding to the stratigraphic classification and interpretation method in any of the above method embodiments.

[0066] Figure 6The diagram illustrates the structure of a computing device according to an embodiment of the present invention. The specific embodiments of the present invention do not limit the specific implementation of the computing device.

[0067] like Figure 6 As shown, the computing device may include: a processor 602, a communication interface 604, a memory 606, and a communication bus 608.

[0068] in: The processor 602, communication interface 604, and memory 606 communicate with each other via communication bus 608.

[0069] Communication interface 604 is used to communicate with other network elements such as clients or other servers.

[0070] The processor 602 is used to execute program 610, which can specifically execute the relevant steps in the above-described stratigraphic classification and interpretation method embodiment.

[0071] Specifically, program 610 may include program code that includes computer operation instructions.

[0072] Processor 602 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The computing device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0073] Memory 606 is used to store program 610. Memory 606 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0074] Specifically, program 610 can be used to cause processor 602 to execute the stratigraphic classification and interpretation method in any of the above method embodiments. The specific implementation of each step in program 610 can be found in the corresponding descriptions of the steps and units in the above stratigraphic classification and interpretation embodiments, and will not be repeated here. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the devices and modules described above can be referred to the corresponding process descriptions in the foregoing method embodiments, and will not be repeated here.

[0075] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, the embodiments of the present invention are not directed to any particular programming language. It should be understood that the embodiments of the present invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing preferred embodiments of the present invention.

[0076] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0077] Similarly, it should be understood that, in order to streamline the embodiments of the invention and aid in understanding one or more of the various inventive aspects, features of the embodiments of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the above description of exemplary embodiments of the invention. However, this disclosure should not be construed as reflecting an intention that the claimed embodiments of the invention require more features than are expressly recited in each claim. Rather, as reflected in the following claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the invention.

[0078] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0079] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments can be used in any combination.

[0080] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components according to the embodiments of the present invention. The embodiments of the present invention can also be implemented as device or apparatus programs (e.g., computer programs and computer program products) for performing part or all of the methods described herein. Such programs implementing the embodiments of the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0081] It should be noted that the above embodiments are illustrative of the present invention and not restrictive of the invention, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. Embodiments of the present invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the order of execution.

Claims

1. A method of formation classification interpretation, characterized by, The method comprises: obtaining a well logging data set of a target interval; the well logging data set comprises a plurality of well logging data; inputting the well logging data set into a trained multi-level progressive classification interpretation model to obtain a feature value corresponding to each well logging data in the well logging data set, and determining a reservoir type to which the well logging data belongs according to the feature value; wherein the multi-level progressive classification interpretation model is obtained by integrating a plurality of sub-models of different levels; different sub-models correspond to different reservoir types; determining a target sub-model corresponding to the well logging data from a plurality of sub-models according to the reservoir type and the feature value of each well logging data; based on the target sub-model, analyzing and processing the well logging data to obtain a target prediction result representing the formation classification and corresponding formation interpretation to which the well logging data belongs.

2. The method of claim 1, wherein, The method further comprises: determining a first feature value and a second feature value from a plurality of feature values of the well logging data according to the reservoir type; the first feature value comprises a first attention proportion, and the second feature value comprises a second attention proportion, wherein the first attention proportion is greater than the second attention proportion; processing the first feature value based on the first attention proportion to obtain a first processing result, and processing the second feature value based on the second attention proportion to obtain a second processing result; determining a sub-model corresponding to a level of the multi-level progressive classification interpretation model according to the first processing result and the second processing result; determining whether there is a next level in the level of the sub-model; if not, the sub-model is used as the target sub-model; if yes, new first feature value and new second feature value are determined for processing and determination until there is no next level in the level of the obtained sub-model, and the target sub-model is determined.

3. The method of claim 1, wherein, The method further comprises: based on the target sub-model, performing data cleaning and correction processing on the well logging data, and determining rock properties, formation parameters, reservoir evaluation, and constructing an underground geological model according to the well logging data; the formation parameters include thickness, lithology, porosity, and / or permeability.

4. The method of claim 1, wherein, The training process of the multi-level progressive classification interpretation model comprises: obtaining a sample well logging data set and a verification result set corresponding to the sample well logging data set; the sample well logging data set comprises a plurality of sample well logging data; the multi-level progressive classification interpretation model is obtained by integrating a plurality of sub-models of different levels, each sub-model is constructed by one of a neural network model, a decision tree model, and a vector machine model; and the sub-models correspond to different reservoir types; based on the sample well logging data set, training the multi-level progressive classification interpretation model to obtain a sub-model prediction result corresponding to different levels of a plurality of sample well logging data; and The verification result set is used to verify the sub-model prediction results corresponding to different levels, and performance evaluation values of each sub-model are obtained; The model parameters of each sub-model are adaptively adjusted based on the performance evaluation values until a preset training condition is reached, and the multi-level progressive classification interpretation model after training is obtained; wherein, the sample weight value of the sub-model of the intermediate level is adjusted based on the performance evaluation value, and the learning rate of the sub-model of the last level is adjusted based on the performance evaluation value.

5. The method of claim 4, wherein, The obtained sample logging data set further comprises: Obtain a plurality of logging curve data; Align the data of the plurality of logging curve data to obtain initial logging data; The initial logging data is standardized to obtain standard logging data; The standard logging data is dimensionally reduced to obtain dimensionally reduced logging data; The sample logging data set is obtained based on all the dimensionally reduced logging data.

6. The method of claim 4, wherein, The training of the multi-level progressive classification interpretation model based on the sample logging data set to obtain a plurality of sample logging data at different levels corresponding to the sub-model prediction result further comprises: Based on each sample logging data of the sample logging data set, a plurality of first sample feature values corresponding to the intermediate level of the multi-level progressive classification interpretation model are obtained, a first sub-model corresponding to the intermediate level of the sample logging data is determined based on the plurality of first sample feature values, and the sample logging data is analyzed and processed based on the first sub-model to obtain a first prediction result corresponding to the first sub-model; According to the reservoir type of the first sub-model, a plurality of second sample feature values corresponding to the sample logging data are determined, a second sub-model corresponding to the lower level of the intermediate level of the sample logging data is determined based on the plurality of second sample feature values, and the first prediction result is shared to the second sub-model to enable the second sub-model to analyze and process the sample logging data based on the first prediction result to obtain a second prediction result corresponding to the second sub-model.

7. A formation classification interpretation device characterized by, The device comprises: An acquisition module adapted to acquire a logging data set of a target well section; The logging data set comprises a plurality of logging data; A reservoir type module adapted to input the logging data set into a multi-level progressive classification interpretation model trained to obtain feature values corresponding to each logging data in the logging data set, and determine the reservoir type to which the logging data belongs according to the feature values; wherein, the multi-level progressive classification interpretation model is integrated by a plurality of sub-models of different levels; different sub-models corresponding to different reservoir types; A sub-model module adapted to determine a target sub-model corresponding to the logging data from a plurality of sub-models according to the reservoir type and the feature values of each logging data; A prediction module adapted to analyze and process the logging data based on the target sub-model to obtain a target prediction result representing the formation classification and corresponding formation interpretation to which the logging data belongs.

8. A computing device, comprising: It comprises: A processor, a memory, a communication interface and a communication bus, the processor, the memory and the communication interface complete communication with each other through the communication bus; The memory is configured to store at least one executable instruction, and the executable instruction is configured to enable the processor to perform operations corresponding to the formation classification interpretation method according to any one of claims 1-6.

9. A computer storage medium, characterized in that The storage medium stores at least one executable instruction, and the executable instruction is configured to enable the processor to perform operations corresponding to the formation classification interpretation method according to any one of claims 1-6.

10. A computer program product, characterised in that, The computer program product comprises at least one executable instruction, and the executable instruction is configured to enable the processor to perform operations corresponding to the formation classification interpretation method according to any one of claims 1-6.

Images