Reservoir type classification method, device, electronic device and storage medium

Abstract

The present invention relates to a reservoir type classification method, device, electronic device and storage medium. The method comprises: obtaining an original well logging data set and preprocessing the original well logging data set to obtain a preprocessed well logging data set, wherein the well logging data set includes a training data set; inputting the training data set into an initial hybrid PSO-MKFD model including hyperparameters for training until preset convergence conditions are met, outputting optimal hyperparameters to obtain a fully trained hybrid PSO-MKFD model, wherein the hyperparameters include polynomial degree, radial basis function kernel width and mixing coefficient; obtaining well logging data to be classified, inputting the well logging data to be classified into the fully trained PSO-MKFD model, and outputting a reservoir type classification result. The classification model of the present invention improves the accuracy of reservoir identification.

Description

Technical Field

[0001] The present invention relates to the technical field of reservoir identification, and in particular to a reservoir type classification method, device, electronic equipment and storage medium. Background Art

[0002] Carbonate rocks possess abundant oil and gas resources and production worldwide. In recent years, deep carbonate reservoirs in oil and gas-bearing basins have increasingly become important exploration targets worldwide. Due to the complex diagenetic and tectonic processes during long-term burial, a variety of reservoir spaces, such as pores and fractures, have developed intricately within deep-buried carbonate rocks, forming a variety of reservoir types. Different types of carbonate reservoirs exhibit different storage and seepage capacities at different scales, resulting in significant heterogeneity in reservoir quality and differences in oil and gas enrichment within carbonate reservoirs, posing potential risks to effective development. Accurate identification of reservoir types is one of the key issues in deep carbonate reservoir quality evaluation and rational oil and gas development.

[0003] For subsurface reservoirs, drilling coring and geophysical logging data are commonly used to identify reservoir types. Core characterization is the most direct and effective method for identifying reservoir types, but its low coring rate and time consumption hinder its widespread application. Imaging logging, with its high vertical continuity and resolution, can provide rich information for identifying oil and gas reservoir types. However, this data is not always available from all wells due to its high cost. In contrast, conventional logging data is more widely available, economical, and reliable. Conventional logging measures the physical and chemical properties of the rock surrounding the wellbore using acoustic, electrical, and radioactive detection techniques, including lithology, physical properties, mechanical properties, and hydrocarbon-bearing properties. Traditional methods utilize conventional logging data to identify reservoir types, typically employing techniques such as crossplot analysis, empirical formulas, and interpretive plates. Most methods rely heavily on the experience of geologists, resulting in low efficiency and high subjectivity. Furthermore, due to the complex logging response, these methods cannot achieve the expected accuracy.

[0004] Therefore, how to effectively identify reservoir types based on well logging data is a technical problem that needs to be solved urgently. Summary of the Invention

[0005] In view of this, it is necessary to provide a reservoir type classification method, device, electronic device and storage medium to solve the technical problem of inaccurate reservoir type classification in the prior art.

[0006] In order to solve the above problems, in a first aspect, the present invention provides a reservoir type classification method, comprising:

[0007] Acquiring an original well logging data set, and preprocessing the original well logging data set to obtain a preprocessed well logging data set, wherein the well logging data set includes a training data set;

[0008] Inputting the training data set into an initial hybrid PSO-MKFD model including hyperparameters for training until a preset convergence condition is met, outputting optimal hyperparameters, and obtaining a fully trained hybrid PSO-MKFD model, wherein the hyperparameters include the polynomial degree, the radial basis function kernel width, and the mixing coefficient;

[0009] Acquire the logging data to be classified, input the logging data to be classified into the trained PSO-MKFD model, and output the reservoir type classification result, wherein the reservoir types include porous reservoirs, pore-vuggy reservoirs, fracture-vuggy reservoirs, fracture-type reservoirs and tight reservoirs.

[0010] Furthermore, the well logging data set is obtained, including:

[0011] Acquire multiple well log curves and determine the reservoir type of each well log curve;

[0012] The logging response characteristics in the logging curves are extracted, and a logging data set is constructed by corresponding the reservoir type of each logging curve to the logging response characteristics one by one to serve as the logging data, wherein the logging response characteristics include any one of natural gamma, well diameter, compensated density, acoustic wave time difference, compensated neutron and formation resistivity.

[0013] Furthermore, the preprocessing of the original well logging data set to obtain a preprocessed well logging data set includes:

[0014] The original well logging data set is statistically evaluated based on a preset statistical standard to remove abnormal values in the original well logging data set, wherein the abnormal values are logging values outside the deviation range according to the preset statistical standard.

[0015] Furthermore, the training data set is input into an initial hybrid PSO-MKFD model including hyperparameters for training until a preset convergence condition is met, and the optimal hyperparameters are output to obtain a fully trained hybrid PSO-MKFD model, including:

[0016] Determining initial hyperparameters of the hybrid PSO-MKFD model;

[0017] Based on the particle swarm optimization algorithm, the hyperparameters are iteratively optimized within a preset search range until the average accuracy of the five-fold cross validation converges to the preset fitness value, and the optimal hyperparameters are output to obtain a fully trained hybrid PSO-MKFD model.

[0018] Furthermore, before using the particle swarm optimization algorithm, the method further includes:

[0019] Determine initialization parameters of the particle swarm optimization algorithm, wherein the initialization parameters include a first acceleration factor, a second acceleration factor, a maximum inertia factor, a minimum inertia factor, and a maximum number of iterations.

[0020] Furthermore, the method further comprises:

[0021] Calculating a confusion matrix of the hybrid PSO-MKFD model with respect to training results and test results, wherein the confusion matrix includes the number of correctly classified samples and the number of incorrectly classified samples;

[0022] An evaluation index is calculated based on the number of correctly classified samples and the number of incorrectly classified samples in the confusion matrix, and the classification performance of the hybrid PSO-MKFD model is determined based on the evaluation index, wherein the evaluation index includes accuracy, precision, recall, and the harmonic mean of precision and recall.

[0023] Furthermore, the method further comprises:

[0024] Acquire a first well logging data set, input the first well logging data set into the hybrid PSO-MKFD model, and output a classification result, wherein the first well logging data set is a data set that has not been trained or tested;

[0025] The generalization ability of the hybrid PSO-MKFD model is verified based on the classification results.

[0026] In a second aspect, the present invention further provides a reservoir type classification device, comprising:

[0027] a preprocessing module, configured to obtain an original well logging data set and preprocess the original well logging data set to obtain a preprocessed well logging data set, wherein the well logging data set includes a training data set;

[0028] a parameter optimization module, configured to input the training data set into an initial hybrid PSO-MKFD model including hyperparameters for training until a preset convergence condition is met, and output the optimal hyperparameters to obtain a fully trained hybrid PSO-MKFD model, wherein the hyperparameters include the polynomial degree, the radial basis function kernel width, and the mixing coefficient;

[0029] The classification module is used to input the well logging data to be classified into the fully trained hybrid PSO-MKFD model and output the reservoir type classification results, wherein the reservoir types include porous reservoirs, pore-vuggy reservoirs, fracture-vuggy reservoirs, fracture-type reservoirs and tight reservoirs.

[0030] In a third aspect, the present invention further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps in the above-mentioned reservoir type classification method when executing the computer program.

[0031] In a fourth aspect, the present invention further provides a computer storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned reservoir type classification method are implemented.

[0032] The beneficial effects of adopting the above embodiment are:

[0033] The present invention preprocesses the original logging data set to eliminate abnormal logging data, facilitating subsequent data processing. In view of the complex nonlinear relationship between reservoir type and logging response characteristics, a hybrid PSO-MKFD model is used to analyze the multivariate data set. The training data set is input into an initial hybrid PSO-MKFD model including hyperparameters for training until the preset convergence conditions are met. The optimal hyperparameters are output, the classification performance of the model is enhanced, and the accuracy of the model in reservoir identification is improved. Finally, the fully trained hybrid PSO-MKFD model is applied to improve the accuracy of reservoir type classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A schematic flow chart of an embodiment of a reservoir type classification method provided by the present invention;

[0035] Figure 2 A schematic diagram of characteristics of different reservoir types provided by an embodiment of the present invention;

[0036] Figure 3 A process diagram for searching for optimal hyperparameters in an MKFD model based on a particle swarm optimization algorithm according to one embodiment of the present invention;

[0037] Figure 4 A schematic diagram of a confusion matrix based on the training and testing results of a hybrid PSO-MKFD model provided in one embodiment of the present invention;

[0038] Figure 5 A classification result diagram based on core analysis and MKFD prediction provided by one embodiment of the present invention;

[0039] Figure 6 A schematic structural diagram of an embodiment of a reservoir type classification device provided by the present invention;

[0040] Figure 7 This is a structural diagram of an embodiment of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0041] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, and are not used to limit the scope of the present invention.

[0042] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first" and "second" may explicitly or implicitly include at least one of such features. In addition, "multiple" means two or more, unless otherwise clearly and specifically defined. Reference to "embodiments" in this document means that the specific features, structures or characteristics described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0043] The present invention provides a reservoir type classification method, device, electronic device, and storage medium. This method addresses the complex nonlinear relationship between reservoir type and well logging response characteristics. A hybrid kernel function is used to analyze multivariate datasets and extract meaningful geological features to identify reservoir types.

[0044] The specific embodiments are described in detail below:

[0045] See also Figure 1 , Figure 1 A schematic flow chart of an embodiment of a reservoir type classification method provided by the present invention. A specific embodiment of the present invention discloses a reservoir type classification method, comprising:

[0046] Step S101: obtaining an original well logging data set, and preprocessing the original well logging data set to obtain a preprocessed well logging data set, wherein the well logging data set includes a training data set;

[0047] Step S102: Inputting the training data set into the initial hybrid PSO-MKFD model including hyperparameters for training until the preset convergence conditions are met, outputting the optimal hyperparameters, and obtaining a fully trained hybrid PSO-MKFD model, wherein the hyperparameters include the polynomial degree, the radial basis function kernel width, and the mixing coefficient;

[0048] Step S103: Obtain the well logging data to be classified, input the well logging data to be classified into the fully trained PSO-MKFD model, and output the reservoir type classification results, where the reservoir types include porous reservoirs, pore-vuggy reservoirs, fracture-vuggy reservoirs, fracture-type reservoirs and tight reservoirs.

[0049] First of all, it should be noted that the logging data set is constructed by one-to-one correspondence between the reservoir type and the logging response characteristics of each logging curve. In the embodiment of the present invention, a total of 453 data sets were obtained from the original data for training and testing the initial classification model, including 106 porous reservoirs, 75 pore-vuggy reservoirs, 68 fracture-vuggy reservoirs, 86 fracture-pore reservoirs, and 118 tight reservoirs.

[0050] See also Figure 2 , Figure 2 An embodiment of the present invention provides a schematic diagram of the characteristics of different reservoir types. In porous reservoirs, the storage space is primarily composed of dissolution pores and a small amount of primary pores, with few dissolution vugs and natural fractures (Table 1a). Dissolution pores primarily consist of intercrystalline, intracrystalline, and intergranular dissolution pores. These dissolution pores are generally isolated and dispersed (Table 1b), with pore diameters ranging from 0.6 to 1.3 mm. Core plug measurements indicate that the porosity and permeability of these reservoirs are generally less than 3.0% and 0.01×10⁻³μm², respectively. In vuggy reservoirs, dissolution vugs are the primary storage space, followed by dissolution pores (see Table 1c), which are closely related to karstification. Dissolution vugs and dissolution pores are generally of varying sizes, layered, or irregularly distributed along bedding planes. These dissolution vugs are widespread and densely distributed (see Table 1d), with high porosity and permeability exceeding 3.0% and 0.1×10⁻³μm², respectively. Fracture-cavity reservoirs primarily contain cavities, fractures, and some dissolution pores (Table 1e). High-angle structural fractures are well developed, with some dissolution dilation occurring. Dissolution pores and vugs are typically distributed in a beaded pattern along fractures or parallel to bedding planes, forming a large, well-connected fracture-cavity system (Table 1f). Consequently, fracture-cavity reservoirs generally possess high petrophysical properties, with porosities and permeabilities exceeding 3.0% and 0.5×10-3 μm², respectively. Fracture-cavity reservoirs contain a combination of dissolution pores and structural fractures (Table 1g), with cavities and dissolution dilation fractures being less well developed. Dissolution pores are mostly sporadic (Table 1h), with some interconnected by fractures. Fractures are primarily structural, with different groups of fractures often forming a fracture network (Table 1h). Core analysis indicates that these reservoirs are characterized by low porosity (<3.0%) and high permeability (>0.1×10-3 μm²). Tight reservoirs refer to the carbonate bedrock of the Dengying Formation, where intense compaction and cementation during burial destroyed most of the primary pores (Table 1i). The reservoirs contain very few secondary pores and natural fractures (Table 1j). Consequently, tight reservoirs have virtually no actual water storage or seepage capacity. Core samples typically have porosity and permeability less than 2.0% and 0.001×10⁻³μm², respectively.

[0051] The present invention preprocesses the original logging data set to eliminate abnormal logging data, facilitating subsequent data processing. In view of the complex nonlinear relationship between reservoir type and logging response characteristics, a hybrid PSO-MKFD model is used to analyze the multivariate data set. The training data set is input into an initial hybrid PSO-MKFD model including hyperparameters for training until the preset convergence conditions are met. The optimal hyperparameters are output, the classification performance of the model is enhanced, and the accuracy of the model in reservoir identification is improved. Finally, the fully trained hybrid PSO-MKFD model is applied to improve the accuracy of reservoir type classification.

[0052] In one embodiment of the present invention, obtaining a well logging data set includes:

[0053] Acquire multiple well log curves and determine the reservoir type of each well log curve;

[0054] The logging response features in the logging curves are extracted, and the reservoir type of each logging curve is mapped to the logging response features to construct a logging data set. The logging response features include any one of natural gamma, well diameter, compensated density, acoustic wave time difference, compensated neutron and formation resistivity.

[0055] It is understood that well logs can be obtained from collected cores and thin sections, as well as geophysical logging data. Core samples are depth-matched to well logs based on core description reports. In this embodiment, six well logs and corresponding reservoir type datasets were collected from seven cored wells in the Dengying Formation. Log responses vary among reservoir types. On average, compensated density and formation resistivity logs decrease in the order of tight reservoirs, porous reservoirs, fractured reservoirs, pore-vuggy reservoirs, and fracture-vuggy reservoirs, while natural gamma ray, wellbore diameter, acoustic transit time, and compensated neutron logs show an increasing trend. The abundance of pores and vugs in deep carbonate reservoirs can reduce the weight of rock per unit volume, resulting in extended acoustic wave travel time at low compensated density and high acoustic transit time and compensated neutron values, providing ample space for natural gas accumulation. Because fractures provide a certain amount of storage space, the development of fractures can also produce the same compensated density, acoustic transit time, and compensated neutron log responses as pores and caves. However, the magnitude of fault-related log responses is far less pronounced than in the latter two categories. At the same time, well-developed open fractures in dense rock formations serve as effective seepage pathways, facilitating the invasion of drilling fluids and leading to relatively low resistivity values in fractured rock formations. Furthermore, when pores, caves, and fractures are abundant, the mechanical strength of the rock decreases and the potential for pore expansion increases. Well-developed open fractures facilitate drilling fluid invasion because they serve as effective seepage pathways in dense rock, leading to relatively low formation resistivity values in fractured rock. Furthermore, when pores, caves, and fractures are abundant, the mechanical strength of the rock decreases and the potential for pore expansion increases. The logging response characteristics of these carbonate reservoirs are generally consistent with previous research results. Well-developed open fractures facilitate drilling fluid invasion because they serve as effective seepage pathways in dense rock, leading to relatively low formation resistivity values in fractured rock. Furthermore, when pores, caves, and fractures are abundant, the mechanical strength of the rock decreases and the potential for pore expansion increases. The logging response characteristics of these carbonate reservoirs are generally consistent with previous research results. Despite this, there is considerable overlap in well-log values between different reservoir types, indicating that features relevant to classification are being obscured by less relevant but more robust features. Some reservoir types exhibit similar well-log responses. Significant overlap in well-log values remains, blurring the decision boundaries between different reservoir types. This overlap can lead to multiple interpretation issues in reservoir type identification. A complex, nonlinear relationship exists between reservoir type and well-log curves.

[0056] In one embodiment of the present invention, preprocessing the original well logging data set to obtain the preprocessed well logging data set includes:

[0057] The original well logging data set is statistically evaluated based on a preset statistical standard, and outliers in the original well logging data set are removed, wherein the outliers are logging values outside the deviation range according to the preset statistical standard.

[0058] The preset statistical standard can be a three-sigma standard. To ensure validity, a statistical evaluation is performed on the original well logging dataset to remove outliers. According to the three-sigma standard, all logging values outside the ±3.0 standard deviation range are considered outliers and removed. Furthermore, to improve computational speed and reduce estimation errors during the modeling process, the mapminmax function is used to normalize all input parameters to [0, 1]. This normalized raw value, minimum value, and maximum value are obtained to facilitate subsequent processing.

[0059] In one embodiment of the present invention, a training dataset is input into an initial hybrid PSO-MKFD model including hyperparameters for training until a preset convergence condition is met, and the optimal hyperparameters are output to obtain a fully trained hybrid PSO-MKFD model, including:

[0060] Determine the initial hyperparameters of the hybrid PSO-MKFD model;

[0061] Based on the particle swarm optimization algorithm, the hyperparameters are iteratively optimized within the preset search range until the average accuracy of the five-fold cross validation converges to the preset fitness value. The optimal hyperparameters are output and a fully trained hybrid PSO-MKFD model is obtained.

[0062] Understandably, the performance of the MKFD model is highly dependent on hyperparameters, including the RBF kernel width σ, the polynomial degree d, and the mixing coefficient μ. The mixing coefficient μ represents the weight of the two kernels and is used to adjust the extrapolation and interpolation capabilities of the mixed kernel. For a single kernel, the kernel parameters control the kernel's complexity, which in turn determines the distribution of data in the mapped feature space. For the RBF kernel, a smaller value of σ increases the model complexity and is prone to overfitting. Conversely, if the value of σ is too large, the model becomes overly restricted and fails to capture the complexity of the dataset, making underfitting inevitable. For the polynomial kernel, a larger value of d results in higher computational and learning complexity, making the model prone to overfitting. Therefore, to balance the classification accuracy, generalization ability, and computational cost of the MKFD model, the kernel parameter search ranges are set to [0.01, 10] for σ and [1, 6] for d, respectively.

[0063] See also Figure 3 , Figure 3A diagram illustrating the process of searching for optimal hyperparameters in an MKFD model based on a particle swarm optimization algorithm, according to one embodiment of the present invention. It can be seen that as the iteration process progresses, the fitness value, i.e., the average accuracy of the five-fold cross-validation, increases accordingly. Within the maximum number of iterations, the average accuracy eventually converges to a stable value, indicating that the parameter search has achieved the optimal result. Therefore, the corresponding hyperparameters are defined as optimal values. Through PSO optimization, the optimal hyperparameters of the MKFD model are determined to be [σ, d, μ] = [1.572, 2.385, 0.736].

[0064] In addition, it should be noted that after the classification model is trained, the test data set needs to be input into the trained classification model for further testing to improve the performance of the classification model.

[0065] In one embodiment of the present invention, before using the particle swarm optimization algorithm, the method further includes:

[0066] Determine the initialization parameters of the particle swarm optimization algorithm, wherein the initialization parameters include a first acceleration factor, a second acceleration factor, a maximum inertia factor, a minimum inertia factor, and a maximum number of iterations.

[0067] First, it's important to note that the particle swarm optimization algorithm (PSO) is a population-based heuristic search technique that originated from the study of social coordination in animal groups. The PSO algorithm involves randomly initializing several particles in a search space, each with its own position and velocity. These particles represent potential solutions to an extreme optimization problem and are used to compute the global optimal solution to the fitness function.

[0068] The initialization parameters for the particle swarm optimization algorithm include the first acceleration factor, the second acceleration factor, the maximum inertia factor, the minimum inertia factor, and the maximum number of iterations. To balance model performance and operational efficiency, based on extensive experimental results, in one embodiment of the present invention, the first acceleration factor is set to 1.5, the second acceleration factor is set to 1.7, the maximum inertia factor is set to 0.9, the minimum inertia factor is set to 0.4, and the maximum number of iterations is set to 100.

[0069] During the particle swarm optimization algorithm's search for optimal hyperparameters, the particle's position is continuously updated as its velocity changes, ultimately converging to the global optimum in the search space. The average accuracy of the five-fold cross-validation is then used as the fitness function to evaluate the performance of the optimization process. The PSO algorithm terminates when the minimum error threshold or the maximum number of iterations is reached.

[0070] In one embodiment of the present invention, the above method further includes:

[0071] Calculating a confusion matrix of the hybrid PSO-MKFD model with respect to the training results and the test results, wherein the confusion matrix includes the number of correctly classified samples and the number of incorrectly classified samples;

[0072] An evaluation index is calculated based on the number of correctly classified samples and the number of incorrectly classified samples in the confusion matrix, and the classification performance of the hybrid PSO-MKFD model is determined based on the evaluation index, wherein the evaluation index includes accuracy, precision, recall, and the harmonic mean of precision and recall.

[0073] It is understandable that for multi-classification problems, confusion matrices and related parameters calculated based on them are often used to evaluate model performance. Common evaluation metrics include accuracy, precision, recall, and the F1 value, which is the harmonic mean of precision and recall. Specifically, accuracy refers to the ratio of the total number of correctly classified samples to the total number of samples in all categories. Precision is the ratio of correctly classified samples to all samples classified into that class, which reflects the accuracy specific to that class. Recall is defined as the ratio of correctly classified samples to the total number of samples in that class, indicating how comprehensively the model classifies each class. The F1 value is defined as the harmonic mean of precision and recall. Higher values, i.e., values closer to 1, indicate better model classification performance. All of these metrics provide complementary information for evaluating model performance.

[0074] To demonstrate the improved performance of the proposed hybrid PSO-MKFD model, in one embodiment of the present invention, two KFD models with different basic kernels—a radial basis function kernel and a polynomial kernel—were applied while maintaining the training and test datasets unchanged. During the modeling process, a five-fold cross-validation technique was used to optimize the kernel parameters. Furthermore, the search ranges for the radial basis function kernel and the polynomial kernel were set to σ[0.01, 10] and d[1, 6], respectively. The optimal kernel parameters were determined when the model achieved the highest accuracy.

[0075] See also Figure 4 , Figure 4 This figure shows a confusion matrix based on the training and testing results of a hybrid PSO-MKFD model, according to one embodiment of the present invention. The diagonal lines represent correct classifications, while the rest represent incorrect classifications. Based on this confusion matrix, the classification accuracy of the hybrid PSO-MKFD model can be calculated. The training accuracy of the hybrid PSO-MKFD model is 93.8%, while the corresponding test data accuracy is 91.2%.

[0076] In one embodiment of the present invention, the above method further includes:

[0077] Obtain a first well logging data set, input the first well logging data set into the hybrid PSO-MKFD model, and output a classification result, wherein the first well logging data set is a data set that has not been trained or tested;

[0078] The generalization ability of the hybrid PSO-MKFD model is verified based on the classification results.

[0079] It is understood that in order to further demonstrate the superiority of the generalization ability of the classification model in the embodiment of the present invention, specifically, the PSO-MKFD model is applied to another set of data with a total of 41.5m of two core intervals in the M15 well, in which all 322 data sets have not been used for modeling before. Figure 6 , Figure 6 This diagram shows a classification result based on core analysis and MKFD prediction, provided by one embodiment of the present invention. Specifically, it includes logging response characteristics, namely, gamma ray (GR), well bore (CAL), compensated density (DEN), acoustic transit time (AC), compensated neutron (CNL), and formation resistivity (RT) logging curves, as well as the reservoir type classification results obtained by applying multiple classification models to these logging curves. It can be seen that the classification accuracy of blind test data based on the MKFD model is 92.7%, which is very close to the overall accuracy of the modeling process of 93.2%. In addition, it can be seen that the misclassification phenomena are similar to those in the modeling data.

[0080] It should be noted that misclassification cannot be entirely attributed to the discriminative ability of the MKFD model. The performance of supervised learning models depends largely on the comprehensiveness and reliability of the input data. On the one hand, big data helps the model learn more information and extract more key features for further discrimination. When the data used for modeling is insufficient, the accuracy and reliability of supervised learning methods deteriorate. In this work, although all available data were used, the data used in the modeling process may not provide sufficient feature information for MKFD. On the other hand, the accuracy of well logging data is inevitably affected by factors such as ground pressure, ground temperature, and drilling fluid during drilling. Log data with subtle errors may still exist, even though some outliers in the original data have been cleaned up. Nevertheless, the successful application of the M15 blind test well shows that the proposed MKFD method can be used as a reliable tool for classifying deep carbonate reservoir types.

[0081] In order to better implement the reservoir type classification method in the embodiment of the present invention, based on the reservoir type classification method, please refer to Figure 6 , Figure 6 This is a schematic structural diagram of an embodiment of a reservoir type classification device provided by the present invention. The embodiment of the present invention provides a reservoir type classification device 600, comprising:

[0082] A preprocessing module 601 is used to obtain an original well logging data set and preprocess the original well logging data set to obtain a preprocessed well logging data set, wherein the well logging data set includes a training data set;

[0083] a parameter optimization module 602 for inputting a training dataset into an initial hybrid PSO-MKFD model including hyperparameters for training until a preset convergence condition is met, and outputting the optimal hyperparameters to obtain a fully trained hybrid PSO-MKFD model, wherein the hyperparameters include the polynomial degree, the radial basis function kernel width, and the mixing coefficient;

[0084] The classification module 603 is used to input the well logging data to be classified into the fully trained hybrid PSO-MKFD model and output the reservoir type classification results, where the reservoir types include porous reservoirs, pore-vuggy reservoirs, fracture-vuggy reservoirs, fracture-type reservoirs and tight reservoirs.

[0085] It should be noted here that the device 600 provided in the above embodiment can implement the technical solutions described in the above method embodiments. The specific implementation principles of the above modules or units can be found in the corresponding contents in the above method embodiments, which will not be repeated here.

[0086] Based on the above-mentioned reservoir type classification method, an embodiment of the present invention also provides an electronic device, including: a processor and a memory, and a computer program stored in the memory and executable on the processor; when the processor executes the computer program, it implements the steps in the reservoir type classification method of the above-mentioned embodiments.

[0087] Figure 7 7 shows a schematic diagram of the structure of an electronic device 700 suitable for implementing an embodiment of the present invention. The electronic devices in the embodiments of the present invention may include, but are not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 7 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0088] The electronic device includes: a memory and a processor, wherein the processor here may be referred to as a processing device 701 hereinafter, and the memory may include at least one of a read-only memory (ROM) 702, a random access memory (RAM) 703, and a storage device 708 hereinafter, as specifically shown below:

[0089] like Figure 7As shown, the electronic device 700 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage device 708 into a random access memory (RAM) 703. Various programs and data required for the operation of the electronic device 700 are also stored in the RAM 703. The processing device 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0090] Typically, the following devices may be connected to the I / O interface 705: an input device 706 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 707 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 708 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 709. The communication device 709 may allow the electronic device 700 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 7 The electronic device 700 is shown with various devices, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead.

[0091] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication device 709, or installed from the storage device 708, or installed from the ROM 602. When the computer program is executed by the processing device 701, the above-mentioned functions defined in the method of the embodiment of the present invention are performed.

[0092] Based on the above-mentioned reservoir type classification method, an embodiment of the present invention also provides a computer-readable storage medium, which stores one or more programs. The one or more programs can be executed by one or more processors to implement the steps in the reservoir type classification method of the above-mentioned embodiments.

[0093] Those skilled in the art will appreciate that all or part of the process steps of the above-described embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, such as a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0094] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed by the present invention should be covered by the scope of protection of the present invention.

Claims

1. A reservoir type classification method, characterized in that: include: Acquiring an original well logging data set, and preprocessing the original well logging data set to obtain a preprocessed well logging data set, wherein the well logging data set includes a training data set; Inputting the training data set into an initial hybrid PSO-MKFD model including hyperparameters for training until a preset convergence condition is met, outputting optimal hyperparameters, and obtaining a fully trained hybrid PSO-MKFD model, wherein the hyperparameters include the polynomial degree, the radial basis function kernel width, and the mixing coefficient; Acquire well logging data to be classified, input the well logging data to be classified into the trained PSO-MKFD model, and output reservoir type classification results, wherein the reservoir types include porous reservoirs, pore-vuggy reservoirs, fracture-vuggy reservoirs, fracture-type reservoirs, and tight reservoirs; Acquire well logging data sets, including: Acquire multiple well log curves and determine the reservoir type of each well log curve; Extracting logging response features from the logging curves, and constructing a logging data set by mapping the reservoir type of each logging curve to the logging response features, wherein the logging response features include any one of natural gamma, well diameter, compensated density, acoustic wave travel time, compensated neutron, and formation resistivity; The training data set is input into the initial hybrid PSO-MKFD model including hyperparameters for training until a preset convergence condition is met, and the optimal hyperparameters are output to obtain a fully trained hybrid PSO-MKFD model, including: Determining initial hyperparameters of the hybrid PSO-MKFD model; The hyperparameters are iteratively optimized within a preset search range based on the particle swarm optimization algorithm until the average accuracy of the five-fold cross validation converges to a preset fitness value, and the optimal hyperparameters are output to obtain a fully trained hybrid PSO-MKFD model; The method further comprises: Calculating a confusion matrix of the hybrid PSO-MKFD model with respect to training results and test results, wherein the confusion matrix includes the number of correctly classified samples and the number of incorrectly classified samples; An evaluation index is calculated based on the number of correctly classified samples and the number of incorrectly classified samples in the confusion matrix, and the classification performance of the hybrid PSO-MKFD model is determined based on the evaluation index, wherein the evaluation index includes accuracy, precision, recall, and the harmonic mean of precision and recall.

2. The reservoir type classification method according to claim 1, characterized in that: The preprocessing of the original well logging data set to obtain a preprocessed well logging data set includes: The original well logging data set is statistically evaluated based on a preset statistical standard to remove abnormal values in the original well logging data set, wherein the abnormal values are logging values outside the deviation range according to the preset statistical standard.

3. The reservoir type classification method according to claim 1, characterized in that: The method further comprises: Determine initialization parameters of the particle swarm optimization algorithm, wherein the initialization parameters include a first acceleration factor, a second acceleration factor, a maximum inertia factor, a minimum inertia factor, and a maximum number of iterations.

4. The reservoir type classification method according to claim 1, characterized in that: The method further comprises: Acquire a first well logging data set, input the first well logging data set into the hybrid PSO-MKFD model, and output a classification result, wherein the first well logging data set is a data set that has not been trained or tested; The generalization ability of the hybrid PSO-MKFD model is verified based on the classification results.

5. A reservoir type classification device, characterized in that: include: a preprocessing module, configured to obtain an original well logging data set and preprocess the original well logging data set to obtain a preprocessed well logging data set, wherein the well logging data set includes a training data set; a parameter optimization module, configured to input the training data set into an initial hybrid PSO-MKFD model including hyperparameters for training until a preset convergence condition is met, and output the optimal hyperparameters to obtain a fully trained hybrid PSO-MKFD model, wherein the hyperparameters include the polynomial degree, the radial basis function kernel width, and the mixing coefficient; A classification module is used to input the well logging data to be classified into the fully trained hybrid PSO-MKFD model and output a reservoir type classification result, wherein the reservoir types include porous reservoirs, pore-vuggy reservoirs, fracture-vuggy reservoirs, fracture-type reservoirs and tight reservoirs; Acquire well logging data sets, including: Acquire multiple well log curves and determine the reservoir type of each well log curve; Extracting logging response features from the logging curves, and constructing a logging data set by mapping the reservoir type of each logging curve to the logging response features, wherein the logging response features include any one of natural gamma, well diameter, compensated density, acoustic wave travel time, compensated neutron, and formation resistivity; The training data set is input into the initial hybrid PSO-MKFD model including hyperparameters for training until a preset convergence condition is met, and the optimal hyperparameters are output to obtain a fully trained hybrid PSO-MKFD model, including: Determining initial hyperparameters of the hybrid PSO-MKFD model; The hyperparameters are iteratively optimized within a preset search range based on the particle swarm optimization algorithm until the average accuracy of the five-fold cross validation converges to a preset fitness value, and the optimal hyperparameters are output to obtain a fully trained hybrid PSO-MKFD model; The method further comprises: Calculating a confusion matrix of the hybrid PSO-MKFD model with respect to training results and test results, wherein the confusion matrix includes the number of correctly classified samples and the number of incorrectly classified samples; An evaluation index is calculated based on the number of correctly classified samples and the number of incorrectly classified samples in the confusion matrix, and the classification performance of the hybrid PSO-MKFD model is determined based on the evaluation index, wherein the evaluation index includes accuracy, precision, recall, and the harmonic mean of precision and recall.

6. An electronic device, characterized in that: It comprises a memory and a processor, wherein the memory is used to store a program; the processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the reservoir type classification method described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps of the reservoir type classification method according to any one of claims 1 to 4.

Images