Carbonate cave type reservoir prediction method and system based on data mining

Abstract

The invention provides a carbonate cave type reservoir prediction method and system based on data mining, and relates to the technical field of data mining, and the method comprises the steps: carrying out the feature extraction of historical logging data, and screening out key logging parameters and a seismic attribute data set; processing the logging data based on a data mining method to obtain a characteristic mode of reservoir development; in combination with the seismic attributes and the real-time data of the current well drilling, optimizing a reservoir mode by adopting a self-organizing mapping neural network and a Bayesian reasoning method, and constructing an optimized reservoir distribution prediction model; historical oil and gas yield data is utilized, a reservoir prediction rule is constructed through a grey correlation analysis and decision tree method, a high-quality reservoir area is identified, and probability prediction is carried out; and performing spatial interpolation and visualization processing by adopting a Kriging interpolation and connectivity analysis method to generate a connectivity distribution map of the reservoir and a probability density map of the oil and gas enrichment area, and effectively improving the prediction precision of the carbonate cave type reservoir.

Description

Technical Field

[0001] The present invention relates to the technical field of data mining, and in particular, to a method and system for predicting carbonate rock cavernous reservoirs based on data mining. Background Art

[0002] Carbonate rock cavernous reservoirs are important targets for oil and gas exploration and development. Their complex reservoir structures and high heterogeneity pose great challenges to reservoir prediction and evaluation. At present, reservoir prediction in oil and gas exploration mainly relies on well logging data, seismic inversion data, and drilling lithology analysis. However, traditional reservoir prediction methods have certain limitations. For example, empirical judgment methods based on a single well logging curve or seismic attribute are difficult to comprehensively reflect the spatial distribution characteristics of the reservoir, and are prone to large prediction errors. In addition, although existing seismic inversion technologies can provide three-dimensional geological model data, there are problems such as low resolution and large noise interference in the fine characterization of cavernous reservoirs. In recent years, some studies have introduced geostatistical methods, such as ordinary Kriging interpolation and seismic attribute clustering analysis, to improve the accuracy of reservoir prediction. However, these methods still have deficiencies in reservoir connectivity analysis, high-quality reservoir identification, and probability calculation of oil and gas enrichment areas, and it is difficult to comprehensively and accurately identify the reservoir development law, resulting in reduced efficiency of oil and gas exploration and development.

[0003] Therefore, there is an urgent need for a method and system for predicting carbonate rock cavernous reservoirs based on data mining to solve the above problems. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for predicting carbonate rock cavernous reservoirs based on data mining to improve the above problems. To achieve the above purpose, the technical solutions adopted by the present invention are as follows:

[0005] In a first aspect, the present application provides a method for predicting carbonate rock cavernous reservoirs based on data mining, including:

[0006] Obtaining historical well logging data of proven carbonate rock reservoirs, real-time well logging data of current drillings, lithology and sedimentary environment data, historical oil and gas production data, and three-dimensional geological model data after seismic inversion;

[0007] Performing data feature extraction processing on the historical well logging data of the proven carbonate rock reservoirs to obtain a set of key well logging parameters and seismic attribute data representing reservoir development;

[0008] Performing pattern recognition and classification processing on the set of key well logging parameters and seismic attribute data representing reservoir development to obtain a typical feature pattern of cavernous reservoirs;

[0009] Extract seismic attributes and perform clustering analysis based on the three-dimensional geological model data after seismic inversion. Combine the analysis results with the real-time logging data, lithology, and sedimentary environment data of the current well to construct a typical feature pattern of the cavernous reservoir, and optimize and adjust the reservoir pattern to obtain an optimized reservoir distribution prediction model;

[0010] Construct reservoir prediction rules based on the historical oil and gas production data and the optimized reservoir distribution prediction model, and generate prediction results for high-quality reservoir areas based on the reservoir prediction rules;

[0011] Perform spatial interpolation and visualization processing based on the three-dimensional geological model data after seismic inversion and the prediction results of the high-quality reservoir areas to generate a reservoir prediction image, which includes a connectivity distribution map of the cavernous reservoir and a probability density map of the oil and gas enrichment area.

[0012] In a second aspect, the present application also provides a carbonate rock cavernous reservoir prediction system based on data mining, including:

[0013] An acquisition unit for acquiring historical logging data of proven carbonate rock reservoirs, real-time logging data of current wells, lithology and sedimentary environment data, historical oil and gas production data, and three-dimensional geological model data after seismic inversion;

[0014] A processing unit for performing data feature extraction processing based on the historical logging data of the proven carbonate rock reservoirs to obtain a set of key logging parameters and seismic attribute data representing reservoir development;

[0015] A classification unit for performing pattern recognition and classification processing based on the set of key logging parameters and seismic attribute data representing reservoir development to obtain a typical feature pattern of the cavernous reservoir;

[0016] An analysis unit for extracting seismic attributes and performing clustering analysis based on the three-dimensional geological model data after seismic inversion. Combine the analysis results with the real-time logging data, lithology, and sedimentary environment data of the current well to construct a typical feature pattern of the cavernous reservoir, and optimize and adjust the reservoir pattern to obtain an optimized reservoir distribution prediction model;

[0017] A prediction unit for constructing reservoir prediction rules based on the historical oil and gas production data and the optimized reservoir distribution prediction model, and generating prediction results for high-quality reservoir areas based on the reservoir prediction rules;

[0018] A generation unit for performing spatial interpolation and visualization processing based on the three-dimensional geological model data after seismic inversion and the prediction results of the high-quality reservoir areas to generate a reservoir prediction image, which includes a connectivity distribution map of the cavernous reservoir and a probability density map of the oil and gas enrichment area.

[0019] The beneficial effects of the present invention are as follows:

[0020] By combining logging data, seismic inversion data, oil and gas production data, and sedimentary environment characteristics, an optimized reservoir distribution prediction model is constructed. Among them, techniques such as principal component analysis, dynamic time warping, K-means clustering, and wavelet transform are used to extract features from logging data, and data mining methods such as random forest, support vector machine, and Bayesian probability inference are combined to optimize reservoir pattern classification. In addition, through methods such as Kriging interpolation, connectivity analysis, and kernel density estimation, spatial interpolation and visualization processing are performed on high-quality reservoir areas to generate reservoir prediction images, including connectivity distribution maps of cavernous reservoirs and probability density maps of oil and gas enrichment areas. This application can more accurately predict high-quality reservoir areas, improve the accuracy of reservoir connectivity identification, and enhance the probability assessment ability of oil and gas enrichment areas.

[0021] Other features and advantages of the present invention will be described in the subsequent specification, and part of them will become obvious from the specification or can be understood by implementing the embodiments of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in the written specification, claims, and drawings. Brief Description of the Drawings

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0023] Figure 1 Schematic flow chart of the carbonate rock cavernous reservoir prediction method based on data mining described in the embodiments of the present invention;

[0024] Figure 2 Schematic structural diagram of the carbonate rock cavernous reservoir prediction system based on data mining described in the embodiments of the present invention.

[0025] Reference numerals in the figure: 701, acquisition unit; 702, processing unit; 703, classification unit; 704, analysis unit; 705, prediction unit; 706, generation unit. Detailed Embodiments

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Components of the embodiments of the present invention usually described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0027] It should be noted that like reference numerals and letters denote like items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive distinction and cannot be construed as indicating or implying relative importance.

[0028] Embodiment 1:

[0029] This embodiment provides a method for predicting carbonate rock cavernous reservoirs based on data mining.

[0030] See Figure 1 , which shows that this method includes steps S1, S2, S3, S4, S5, and S6.

[0031] Step S1: Obtain historical logging data of proven carbonate rock reservoirs, real-time logging data of current wells, lithology and sedimentary environment data, historical oil and gas production data, and three-dimensional geological model data after seismic inversion;

[0032] It is understandable that this step collects the historical logging data of the proven carbonate rock reservoirs. These data include key logging curve parameters such as depth, acoustic travel time, resistivity, neutron porosity, and density, which can reflect information such as the lithology, porosity, and fluid properties of the reservoirs. Secondly, by combining the real-time logging data of the current drilling, the reservoir identification model can be dynamically adjusted to adapt to the geological characteristics of the current area and improve the real-time performance and accuracy of prediction. In addition, the lithology and sedimentary environment data provide information on the formation and evolution process of the reservoirs. Through these data, the sedimentary background, lithofacies combination, and structural characteristics of the reservoirs can be analyzed, thus assisting in the construction of the reservoir development model. At the same time, the historical oil and gas production data are used to analyze the productivity performance of different reservoirs, identify the key factors controlling the reservoir productivity, and provide data support for the subsequent prediction of high-quality reservoir areas. Finally, the three-dimensional geological model data after seismic inversion provide the spatial distribution information of the reservoirs. By extracting seismic attributes (instantaneous amplitude, coherence, and spectral attributes), the morphology, connectivity, and reservoir performance of the reservoirs can be characterized in three dimensions. The acquisition and integration of these data not only provide comprehensive input information for subsequent feature extraction, pattern recognition, and reservoir prediction, but also ensure the accuracy and applicability of the reservoir prediction method.

[0033] Step S2: Perform data feature extraction and processing on the historical logging data of the proven carbonate rock reservoirs to obtain a set of key logging parameters and seismic attribute data representing reservoir development;

[0034] It is understandable that this step ensures that the data input into the model is both representative, reduces noise and redundancy, and improves the accuracy of subsequent reservoir pattern recognition and prediction. It enhances the stability and interpretability of the data, lays a solid foundation for reservoir pattern analysis, and effectively reduces the computational amount and improves the prediction efficiency. In this step, step S2 includes step S21, step S22, step S23, and step S24.

[0035] Step S21: Based on the use of the principal component analysis method, perform dimensionality reduction processing on the historical logging data. By calculating the covariance matrix and performing eigenvalue decomposition on it, screen the key logging parameters affecting reservoir development to obtain an optimized set of logging parameters;

[0036] It can be understood that in this step, the logging data is first standardized to eliminate the influence caused by different units or orders of magnitude of different logging parameters. Then, the covariance matrix of the logging data is calculated to characterize the linear correlation between different logging parameters. Subsequently, eigenvalue decomposition is performed on the covariance matrix to obtain a set of eigenvectors (i.e., principal components) and their corresponding eigenvalues, and the principal components are sorted according to the magnitude of the eigenvalues. Usually, the first few principal components with a cumulative contribution rate reaching a preset threshold (such as 95%) can retain most of the information of the data. Therefore, the original logging parameters corresponding to these principal components are selected as the key logging parameters, thereby constructing an optimized set of logging parameters.

[0037] Step S22: According to the optimized set of logging parameters, use the dynamic time warping method to perform time alignment on the logging curve data in the set of logging parameters. Among them, by constructing a cumulative distance matrix and calculating the optimal matching path, the aligned logging curve data is obtained;

[0038] It can be understood that in this step, based on the optimized set of logging parameters, a reference logging curve for alignment is selected (such as the logging data of a standard well). Then, the cumulative distance matrix is calculated for all the logging curves in the set of logging parameters and the reference logging curve to measure the local similarity between the curves. The cumulative distance matrix is calculated based on the Euclidean distance, which represents the matching degree of two logging curves at different depth positions. Next, a dynamic programming algorithm is used to search for the optimal matching path, that is, by finding the shortest path to minimize the total matching error between the logging curves, so as to realize the alignment adjustment of the logging curves. Finally, the logging data is transformed using the optimal matching path to obtain the aligned logging curve data.

[0039] Compared with traditional linear interpolation or static alignment methods, the unique advantage of the dynamic time warping algorithm is that it can adapt to the non-linear deformation of logging curves, that is, it allows elastic deformation of logging data at different depth positions, not limited to rigid translation, so as to more accurately align the key reservoir features. In addition, the dynamic time warping algorithm can also handle the problem of inconsistent logging responses caused by geological changes, making the aligned logging curves more truly reflect the formation structure.

[0040] Step S23: According to the aligned logging curve data, use the K-means clustering method to classify the key logging parameters. Among them, by minimizing the within-class variance method, multiple types of carbonate rock reservoirs are identified to obtain the reservoir classification result;

[0041] It is understandable that in this step, key logging parameters with strong indication for reservoir development are first selected (such as natural gamma ray GR, acoustic travel time DT, resistivity R, density RHOB, and porosity PHI), and then clustering analysis is performed on these logging parameters based on the K-means clustering method. Through iterative optimization, the logging data is divided into K categories, so that the difference between data points within the same category in the high-dimensional parameter space is minimized, while the difference between different categories is maximized.

[0042] Step S24: According to the reservoir classification result, the wavelet transform method is used to denoise and perform trend analysis on the logging curve data. Among them, through multi-scale signal decomposition and reconstruction, key reservoir features are extracted to obtain a set of key logging parameters and seismic attribute data representing reservoir development.

[0043] It is understandable that in this step, the logging curve signal is subjected to discrete wavelet transform and decomposed into approximate components (low-frequency part) and detail components (high-frequency part) of different scales. Then, the threshold denoising method is used to set a reasonable threshold to process the high-frequency components, remove the noise that may be introduced by instrument errors, environmental interference, etc., and at the same time retain the key information sensitive to reservoir changes. Next, the low-frequency components are reconstructed to extract the macroscopic trends of reservoir development, such as the evolution of sedimentary environment and lithological changes. Finally, based on denoising and trend analysis, the local extreme points, fluctuation amplitude, and main frequency characteristics of logging parameters are further calculated, and key seismic attribute parameters are extracted in combination with seismic inversion data to form a set of logging and seismic data representing reservoir development.

[0044] Step S3: Perform pattern recognition and classification processing according to the set of key logging parameters and seismic attribute data representing reservoir development to obtain a typical feature pattern of the cavernous reservoir;

[0045] It is understandable that this step can effectively extract the typical features of the cavernous reservoir, reduce the error of subjective human judgment, and improve the automation and accuracy of reservoir classification. Compared with traditional manual discrimination or single logging attribute analysis, the classification method based on multi-source data fusion and machine learning can more precisely characterize reservoir features. In this step, step S3 includes step S31, step S32, step S33, and step S34.

[0046] Step S31: Based on the random forest method, feature selection is performed on the set of key logging parameters and seismic attribute data. Among them, by calculating the feature importance scores of the set of key logging parameters and seismic attribute data, key features affecting reservoir pattern recognition are screened to obtain a reservoir pattern recognition feature set;

[0047] It is understandable that first, an initial feature set is constructed using historical logging data and seismic attribute data. This feature set may contain a large number of redundant or noisy features, which can affect the computational efficiency and prediction accuracy of the model. Subsequently, during the training process of the random forest, each decision tree is trained in a random subsample and a random feature subset, and the importance contributions of each feature in different decision paths are calculated. Common feature importance measurement methods include split gain calculation based on the Gini index and importance evaluation based on the mean squared error. Finally, by comprehensively sorting the feature importance scores of all decision trees, the key logging parameters and seismic attributes that affect reservoir pattern recognition, such as acoustic travel time, resistivity, density, porosity, seismic coherence, etc., are selected to form a reservoir pattern recognition feature set. This step can effectively reduce the data dimension, remove irrelevant or redundant features, improve the accuracy of subsequent pattern recognition and classification, and at the same time improve the computational efficiency, providing more accurate data support for reservoir prediction and identification of oil and gas enrichment areas.

[0048] Step S32: Classify the reservoir patterns based on the support vector machine method for the reservoir pattern recognition feature set. Among them, by constructing an optimal hyperplane to separate multiple reservoir patterns, the reservoir pattern data with preliminary classification is obtained.

[0049] It is understandable that in this step, first, training samples are constructed using the key logging parameters and seismic attribute data selected in the previous stage. Among them, each sample contains multiple feature dimensions and is labeled with the corresponding reservoir type, such as high-porosity cave reservoir, low-porosity fracture reservoir, and tight reservoir. During the training process, the support vector machine method finds a hyperplane that can maximize the interval between categories, so that the data points of different reservoir types are separated as much as possible. For linearly separable data, the support vector machine method directly uses support vectors to determine the optimal hyperplane, while for complex non-linear reservoir pattern distributions, the support vector machine method uses a kernel function (linear kernel function) to map the data to a high-dimensional feature space, and constructs an optimal hyperplane in this space for classification.

[0050] Through this process, the support vector machine method can effectively distinguish reservoir patterns and obtain the reservoir pattern data with preliminary classification. The technical advantages of this method lie in its strong discriminative ability and robustness, which can handle the complex non-linear relationships between reservoir pattern features and avoid strong assumptions about the data distribution of traditional statistical classification methods. In addition, the support vector machine method can also effectively suppress the overfitting problem of high-dimensional data and improve the generalization ability of reservoir pattern classification, thus providing a high-precision data basis for subsequent reservoir feature optimization and reservoir distribution prediction.

[0051] Step S33: Perform pattern similarity measurement on the preliminarily classified reservoir pattern data based on the Euclidean distance calculation method. Specifically, optimize the pattern classification result by calculating the feature distances between each reservoir pattern to obtain the optimized reservoir pattern classification result.

[0052] It can be understood that in this step, the pairwise Euclidean distances between all patterns are first calculated to construct a pattern similarity matrix. Then, based on a preset threshold or clustering method (such as hierarchical clustering), the preliminary classification result is adjusted. For example, if the Euclidean distances of some patterns are close, they may belong to the same class but are misclassified, and in this case, the classification can be merged and adjusted; conversely, if the internal distances of some patterns are large, they may need to be split into finer categories. In addition, the weighted Euclidean distance or Mahalanobis distance can also be combined to enhance the sensitivity to feature differences and improve the accuracy of classification optimization. This step can improve the accuracy of reservoir pattern classification, reduce misclassification cases, and make the final reservoir pattern classification result more in line with geological laws. Through pattern similarity optimization, not only can the reliability of classification be improved, but also more accurate data support can be provided for subsequent reservoir feature analysis and reservoir modeling.

[0053] Step S34: Standardize the reservoir pattern classification result according to the principal component analysis method. Specifically, calculate the principal component weights and normalize the reservoir pattern feature data to obtain the typical feature pattern data of the cavernous reservoir.

[0054] It can be understood that in this step, first, the optimized reservoir pattern data matrix is standardized so that the mean value of each parameter is zero and the variance is one to eliminate the influence of dimension. Then, the covariance matrix of the standardized data is calculated and its eigenvalue decomposition is performed to obtain the eigenvectors and corresponding eigenvalues. The eigenvalues represent the variance contribution rates of each principal component. After sorting, the first few principal components with the cumulative contribution rate reaching a threshold (such as 95%) are selected to reduce the data dimension and retain key information. Then, the original reservoir pattern data is projected onto the new feature space composed of the selected principal components, the principal component weights are calculated, and the data is normalized to enhance the comparability of the data. Finally, the typical feature pattern data of the cavernous reservoir is obtained. Through dimensionality reduction and standardization, redundant information is removed, the most representative reservoir features are retained, and the stability and distinguishability of the cavernous reservoir pattern are improved. At the same time, the normalization process ensures the balanced role of different features in reservoir analysis and reduces the influence of outliers or extreme values on pattern recognition.

[0055] Step S4: Extract seismic attributes and perform clustering analysis based on the three-dimensional geological model data after seismic inversion. Based on the analysis results, combine the real-time logging data, lithology, and sedimentary environment data of the current well to construct a typical feature pattern of the cavernous reservoir, and optimize and adjust the reservoir pattern to obtain an optimized reservoir distribution prediction model.

[0056] It is understandable that in this step, through the multi-source fusion of seismic, logging, lithology, and sedimentary environment data, the identification accuracy of cavernous reservoirs is improved, and the prediction ability of the model is enhanced by using pattern optimization and machine learning methods, making the prediction results of reservoir distribution more reliable. In this step, step S4 includes step S41, step S42, step S43, and step S44.

[0057] Step S41: According to the three-dimensional geological model data after seismic inversion, a multi-scale attribute calculation method is used to extract seismic attributes. Among them, by calculating the instantaneous amplitude, coherence, and spectral attributes at different scales, a set of seismic attributes related to the reservoir is obtained;

[0058] It is understandable that in this step, the three-dimensional geological model data after seismic inversion contains rich underground structure information, but the seismic attributes at a single scale are difficult to comprehensively reflect the complexity of the reservoir. Therefore, a multi-scale attribute calculation method is adopted to extract key seismic attributes at different scales, such as instantaneous amplitude, coherence, and spectral attributes. The instantaneous amplitude reflects the energy intensity of the seismic signal and helps to identify the physical properties of the reservoir; the coherence calculation is used to characterize the continuity of the underground structure and can reveal structures such as reservoir boundaries, fractures, and faults; the spectral attributes are used to analyze the distribution of different frequency components in the reservoir and can characterize the formation thickness and fluid indication characteristics.

[0059] During the calculation process, wavelet transform and multi-resolution Fourier analysis are used to decompose the seismic data at different scales. For example, through wavelet transform, the time-frequency characteristics of the signal can be analyzed simultaneously to capture the local change information of the reservoir at different scales; by using Fourier transform, the main frequency components of the seismic data can be extracted to identify the formation characteristics of different lithologies. Through these methods, seismic attributes are calculated at different scales and fused to obtain a set of seismic attributes related to the reservoir, making the reservoir identification more accurate.

[0060] Step S42: According to the set of seismic attributes related to the reservoir, a self-organizing mapping neural network is used for clustering analysis. Among them, by constructing a high-dimensional seismic attribute topological mapping, all reservoir patterns are identified to obtain a preliminary reservoir pattern classification result;

[0061] It is understandable that in this step, the self-organizing mapping neural network is constructed by first using the extracted reservoir-related seismic attribute set (such as instantaneous amplitude, coherence, and spectral attributes) as input features. The self-organizing mapping network consists of a two-dimensional grid-like neuron array, and each neuron represents a clustering center. Through a competitive learning process, the network maps the high-dimensional seismic attribute data into a low-dimensional grid and adjusts the weights of adjacent neurons through a neighborhood function, so that similar seismic attributes are clustered in adjacent regions. During the training process, the Euclidean distance is used to measure the similarity between the data and the neuron weight vector, and the weights of the neurons are dynamically adjusted, so that the final network can adaptively learn different types of reservoir patterns.

[0062] Step S43: According to the preliminary reservoir pattern classification results, combined with the real-time logging data, lithology, and sedimentary environment data of the current well drilling, the Bayesian probability inference method is used for pattern optimization. By calculating the posterior probability of each pattern, the reservoir classification is optimized to obtain the adjusted reservoir pattern data.

[0063] It is understandable that in this step, based on the preliminary reservoir pattern classification results, prior probabilities are set for each pattern, which usually come from the results of seismic attribute clustering and the statistical data of historical geological research. Subsequently, the real-time logging data of the current well drilling, including key parameters such as acoustic travel time, natural gamma, density, resistivity, etc., as well as lithology characteristics (such as rock composition and porosity) and sedimentary environment data (such as transgression or regression stages and paleo-hydrodynamic conditions) are used to calculate the likelihood probability of each reservoir pattern. Among them, the Bayesian formula is as follows:

[0064]

[0065] Among them, P(H|D) is the posterior probability of a certain reservoir pattern H under the condition of the current logging data D, P(D|H) is the likelihood of observing the logging data when this reservoir pattern is known, P(H) is the prior probability of this reservoir pattern, and P(D) is the comprehensive probability of observing the logging data under all reservoir patterns.

[0066] After calculating the posterior probability, the maximum posterior probability criterion is used to reclassify the reservoir patterns, and the most likely pattern is assigned to the corresponding seismic attribute data points to optimize the reservoir classification results.

[0067] Step S44: According to the adjusted reservoir pattern data, the geostatistical interpolation method is used for reservoir distribution prediction. By calculating the spatial distribution of each reservoir pattern through Kriging interpolation, an optimized reservoir distribution prediction model is obtained.

[0068] First, calculate the variogram based on the adjusted reservoir pattern data. This function reflects the correlation of the reservoir pattern at different spatial positions and is usually fitted using an exponential, spherical, or Gaussian model. The calculation formula for the variogram is as follows:

[0069]

[0070] Among them, γ(h) is the variogram value, h is the spatial distance, N(h) is the number of point pairs at a distance of h, and Z(x i ) is the value of the reservoir pattern parameter at position x i .

[0071] Then, use the Kriging equation to calculate the reservoir pattern value at the point to be predicted. The core formula is:

[0072]

[0073] Among them, Z * (x 0 ) is the predicted value at the target position x 0 , λ i is the interpolation weight, and Z(x i ) is the value of the reservoir pattern at the known position x i .

[0074] Step S5: Construct a reservoir prediction rule based on the historical oil and gas production data and the optimized reservoir distribution prediction model, and generate a prediction result for the high-quality reservoir area based on the reservoir prediction rule;

[0075] It can be understood that in this step, through a data-driven approach, geological information and historical production data are organically combined to automatically generate a prediction of the high-quality reservoir area. This not only improves the accuracy of oil and gas resource development but also reduces the error of human judgment, providing a scientific basis for oil and gas exploration and development. In this step, Step S5 includes Step S51, Step S52, Step S53, and Step S54.

[0076] Step S51: Identify the main control factors using the grey relational analysis method based on the historical oil and gas production data. By calculating the correlation degree between the production data and the key reservoir attributes, obtain the main control parameters affecting the reservoir productivity;

[0077] It can be understood that in this step, by calculating the correlation degree between each reservoir characteristic parameter and historical oil and gas production, the influence degree of these parameters on oil and gas production is measured. The calculation of the grey correlation degree is based on the comparison of characteristic values. First, the historical oil and gas production data and reservoir characteristic data are standardized to ensure that data with different dimensions can be compared. Then, by constructing a correlation degree sequence, the grey correlation degree between different parameters and oil and gas production is calculated. The larger the correlation degree value, the closer the relationship between the parameter and oil and gas production, and the more likely it is the main controlling factor affecting oil and gas production. Through grey correlation analysis, the geological characteristic parameters that directly or indirectly affect oil and gas production can be revealed. These main controlling factors include the porosity, permeability, compaction degree, lithology type of the reservoir and the fracture development degree of the formation.

[0078] Step S52: According to the optimized reservoir distribution prediction model, use the decision tree method to construct reservoir prediction rules. By recursively dividing the reservoir characteristic space, establish a classification decision tree for the high-quality reservoir area to obtain a reservoir prediction rule set;

[0079] It can be understood that in this step, by calculating the influence of different characteristics on the high-quality reservoir area, the decision tree will select the optimal characteristic for node division according to the information gain. These division rules gradually refine the reservoir characteristic space into different regions until each leaf node represents a specific reservoir type or high-quality area. The decision tree will continuously perform recursive division, and based on the best characteristic of each node, gradually divide the reservoir data set into multiple subsets. Each division will select those characteristics that can best distinguish different reservoir types according to the existing reservoir data set, so as to maximize the classification purity of each node. Through the divided tree structure, the prediction result of whether the reservoir belongs to the high-quality area can be directly obtained from the leaf node of the decision tree. Each path of the tree represents a set of relationship rules between a group of reservoir characteristics and the high-quality area, and these rules can be effectively used to identify the reservoir quality in new areas.

[0080] Step S53: According to the reservoir prediction rule set, combine the spatial statistical analysis method to conduct high-quality reservoir area division. Identify each reservoir block through cluster analysis to obtain the high-quality reservoir area division result, and the high-quality reservoir area is the area where the reservoir distribution range is greater than the preset threshold;

[0081] It is understandable that in this step, first, by combining the reservoir characteristics obtained from the reservoir prediction rule set, spatial autocorrelation analysis is used to understand the spatial distribution patterns of reservoir characteristics (such as porosity and permeability). Then, based on the results of spatial statistical analysis, the K-means clustering analysis method is adopted to divide the reservoir into several blocks, and the reservoir space is divided into multiple clusters according to the similarity of reservoir characteristics, ensuring that the reservoir characteristics within each cluster are highly similar while the differences in reservoir characteristics between clusters are significant. After the clustering analysis is completed, each reservoir block will be assigned to a specific category according to its characteristics. To determine which areas belong to high-quality reservoirs, among them, the areas where the reservoir distribution range is greater than the preset threshold are identified as high-quality reservoir areas.

[0082] Step S54: According to the division result of the high-quality reservoir areas, adopt a probability prediction modeling method to calculate the probabilities of the high-quality reservoir areas. By establishing a probability model based on a Bayesian network, calculate the probability of each area becoming a high-quality reservoir to obtain the final prediction result of the high-quality reservoir areas.

[0083] It is understandable that in this step, the conditional probabilities of each area becoming a high-quality reservoir are calculated through known historical oil and gas production data, reservoir distribution information, and other influencing factors. This process involves estimating the probabilities of each variable based on the existing data and adjusting the weights of each piece of data in combination with conditional dependencies, so as to obtain the final probability of each area becoming a high-quality reservoir. At this time, the Bayesian network can give the probability prediction of each reservoir area becoming a high-quality area based on historical data and reservoir characteristics, taking into account various uncertainty factors. In reservoir evaluation, different data sources often have incompleteness or noise. The Bayesian network can flexibly handle these uncertainties through the calculation of conditional probabilities. Through Bayesian inference, the system can continuously update and correct the prediction results of the model to further improve the accuracy of the prediction of high-quality reservoir areas.

[0084] Step S6: Perform spatial interpolation and visualization processing based on the three-dimensional geological model data after seismic inversion and the prediction results of the high-quality reservoir areas to generate a reservoir prediction image, and the prediction image includes a connectivity distribution map of cavernous reservoirs and a probability density map of oil and gas enrichment areas.

[0085] It is understandable that the image generated through spatial interpolation and visualization processing in this step provides an efficient and intuitive decision-making basis for subsequent oil and gas exploration, development, and optimization, helping analysts extract key spatial information from a large amount of geological data and convert it into a visualized result, thereby improving the accuracy and production efficiency of oil and gas exploration. In this step, Step S6 includes Step S61, Step S62, Step S63, and Step S64.

[0086] Step S61: Based on the 3D geological model data after seismic inversion, spatial interpolation is performed using the Kriging interpolation method. By constructing a semi-variogram and performing optimal weight estimation, the spatial distribution data of the cavernous reservoir is obtained;

[0087] It can be understood that an important step in the Kriging interpolation in this step is to construct a semi-variogram. The semi-variogram measures the similarity or difference of geological data at different spatial distances. By calculating the differences between different spatial points, a curve or model reflecting the data variability is constructed. This function can capture the spatial dependence of reservoir data. For example, the data between measurement points that are closer may have a stronger correlation, while the data that are farther apart may have a weaker correlation. By fitting and optimizing the semi-variogram, the variation law of reservoir characteristics in space can be better described. In Kriging interpolation, the semi-variogram is used to calculate the influence degree of each data point on the unknown point. This process involves the estimation of optimal weights, that is, weights are assigned according to spatial distance and data variability to ensure that the interpolation result is both consistent in space and can truly reflect reservoir characteristics. The optimal weight estimation determines the most suitable interpolation weights by minimizing the prediction error, thereby achieving a more accurate estimation of the reservoir spatial distribution.

[0088] Step S62: Based on the spatial distribution data of the cavernous reservoir, a connectivity analysis method is used to calculate the reservoir connectivity. By constructing a 3D seepage network and calculating the connected paths, the reservoir connectivity distribution data is obtained;

[0089] It can be understood that this step regards the reservoir space as a 3D network, where each pore unit is connected to other adjacent units to form a seepage network. Among them, parameters such as the geometric shape, porosity, and permeability of each pore unit are extracted from the spatial distribution data of the reservoir, and then this information is converted into nodes and edges in the seepage network to ensure that the seepage network can reflect the physical characteristics of the reservoir. After the 3D seepage network is constructed, the next task is to calculate the connectivity inside the reservoir. The core of the connectivity analysis is to evaluate the fluid flow paths between different regions of the reservoir, usually by calculating the connectivity metrics between pore units. In this process, the shortest path algorithm is used to determine whether there are fluid channels in each reservoir region. Furthermore, the system can identify which regions inside the reservoir are interconnected and which regions may form isolated reservoir blocks. This step can provide an effective basis for improving the oil and gas recovery rate, thereby optimizing the development plan and increasing the production efficiency of oil and gas. In addition, the reservoir connectivity distribution data can provide accurate data support for subsequent dynamic reservoir simulation and adjustment of the development plan.

[0090] Step S63: According to the prediction results of the high-quality reservoir areas, the kernel density estimation method is used to calculate the hydrocarbon enrichment probability. By performing probability density estimation on the spatial point pattern of the high-quality reservoir areas, the probability density data of the hydrocarbon enrichment areas is obtained;

[0091] It can be understood that after estimating the spatial points within the high-quality reservoir areas through the kernel density estimation method in this step, the probability density value of each point is obtained. This probability density value represents the possibility of hydrocarbon enrichment at that point, and further reflects the distribution of the hydrocarbon enrichment areas. In this way, it is possible to reveal which areas have a high probability of hydrocarbon enrichment, and thus provide accurate hydrocarbon distribution maps for developers. The kernel density estimation method avoids overly simplified assumptions, thereby improving the accuracy of predicting hydrocarbon enrichment areas.

[0092] Step S64: According to the reservoir connectivity distribution data and the probability density data of the hydrocarbon enrichment areas, the geological modeling and GIS visualization method is used for image rendering. By constructing a three-dimensional visualization model and color mapping, a connectivity distribution map of the cavernous reservoir and a probability density map of the hydrocarbon enrichment areas are generated.

[0093] It can be understood that in this step, first, the reservoir connectivity distribution data and the probability density data of the hydrocarbon enrichment areas are converted into a three-dimensional model through geological modeling software (such as Petrel, GOCAD, etc.). These modeling tools form a set of visual datasets by arranging reservoir units in space and quantifying the characteristics (such as porosity, permeability, etc.) of each unit. The Geographic Information System (GIS) will be used to further render and display the three-dimensional model of the reservoir. GIS technology can combine the geological data in the three-dimensional model with spatial information, and through establishing a spatial analysis framework, realize the display and analysis of complex data. Finally, the color mapping technology is adopted to map different data values of reservoir connectivity and hydrocarbon enrichment areas to different colors. For example, higher connectivity areas may be represented by red, while lower connectivity areas may be represented by blue; in the probability density map of hydrocarbon enrichment areas, the enrichment areas can be colored dark red or yellow, while the poor areas are colored light blue or green. This color mapping method enables geological engineers to see at a glance the spatial distribution characteristics of the reservoir and the degree of hydrocarbon enrichment, helping them quickly locate potential areas. In this step, intuitive image rendering can clearly display the spatial distribution of the reservoir, connectivity characteristics, and the probability density of hydrocarbon enrichment, greatly improving the usability of the data.

[0094] Embodiment 2:

[0095] This embodiment provides a carbonate rock cavernous reservoir prediction system based on data mining. The system includes:

[0096] An acquisition unit 701, configured to acquire historical logging data of an explored carbonate reservoir, real-time logging data of a current well drilling, lithology and sedimentary environment data, historical oil and gas production data, and three-dimensional geological model data after seismic inversion;

[0097] A processing unit 702, configured to perform data feature extraction processing according to the historical logging data of the explored carbonate reservoir to obtain key logging parameters representing reservoir development and a set of seismic attribute data;

[0098] A classification unit 703, configured to perform pattern recognition and classification processing according to the set of key logging parameters and seismic attribute data representing reservoir development to obtain a typical feature pattern of a cave-type reservoir;

[0099] An analysis unit 704, configured to perform seismic attribute extraction and clustering analysis according to the three-dimensional geological model data after seismic inversion, construct a typical feature pattern of a cave-type reservoir based on the analysis results in combination with the real-time logging data, lithology and sedimentary environment data of the current well drilling, and optimize and adjust the reservoir pattern to obtain an optimized reservoir distribution prediction model;

[0100] A prediction unit 705, configured to construct a reservoir prediction rule according to the historical oil and gas production data and the optimized reservoir distribution prediction model, and generate a prediction result of a high-quality reservoir area based on the reservoir prediction rule;

[0101] A generation unit 706, configured to perform spatial interpolation and visualization processing according to the three-dimensional geological model data after seismic inversion and the prediction result of the high-quality reservoir area to generate a reservoir prediction image, where the prediction image includes a connectivity distribution map of a cave-type reservoir and a probability density map of an oil and gas enrichment area.

[0102] It should be noted that regarding the system in the above embodiments, the specific manners in which each unit performs operations have been described in detail in the embodiments related to the method, and will not be elaborated herein.

[0103] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

[0104] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present invention, and all should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for predicting carbonate cave reservoirs based on data mining, characterized in that: include: Obtain historical logging data of proven carbonate reservoirs, real-time logging data of current drilling, lithology and sedimentary environment data, historical oil and gas production data, and three-dimensional geological model data after seismic inversion; Performing data feature extraction processing based on the historical logging data of the proven carbonate reservoir to obtain a key logging parameter and seismic attribute data set representing reservoir development; Performing pattern recognition and classification processing based on the key logging parameters and seismic attribute data set representing reservoir development to obtain a typical characteristic pattern of cave-type reservoirs; Seismic attribute extraction and cluster analysis are performed based on the three-dimensional geological model data after seismic inversion, and a typical characteristic pattern of cave-type reservoirs is constructed based on the analysis results combined with real-time logging data, lithology and sedimentary environment data of the current drilling, and the reservoir pattern is optimized and adjusted to obtain an optimized reservoir distribution prediction model; Constructing reservoir prediction rules according to the historical oil and gas production data and the optimized reservoir distribution prediction model, and generating prediction results of high-quality reservoir areas based on the reservoir prediction rules; Spatial interpolation and visualization processing are performed based on the three-dimensional geological model data after seismic inversion and the prediction results of the high-quality reservoir areas to generate a reservoir prediction image, which includes a connectivity distribution map of cave-type reservoirs and a probability density map of oil and gas enrichment areas.

2. The method for predicting carbonate cave reservoirs based on data mining according to claim 1, characterized in that: The data feature extraction process is performed based on the historical logging data of the proven carbonate reservoir, including: Based on the principal component analysis method, the historical logging data is processed for dimensionality reduction. By calculating the covariance matrix and performing eigenvalue decomposition, the key logging parameters that affect reservoir development are screened to obtain the optimized logging parameter set. According to the optimized logging parameter set, the dynamic time warping method is used to time align the logging curve data in the logging parameter set, wherein the aligned logging curve data is obtained by constructing a cumulative distance matrix and calculating the optimal matching path; According to the aligned logging curve data, the K-means clustering method is used to classify the key logging parameters. Among them, multiple types of carbonate reservoirs are identified by minimizing the intra-class variance method to obtain the reservoir classification results. According to the reservoir classification results, the wavelet transform method is used to denoise and analyze the trend of the logging curve data. The key features of the reservoir are extracted through multi-scale signal decomposition and reconstruction, and the key logging parameters and seismic attribute data sets representing the reservoir development are obtained.

3. The method for predicting carbonate cave reservoirs based on data mining according to claim 1, characterized in that: Pattern recognition and classification processing are performed based on the key logging parameters and seismic attribute data sets representing reservoir development, including: Based on the random forest method, feature selection is performed on the key logging parameter and seismic attribute data set, wherein the key features affecting reservoir pattern recognition are screened by calculating the feature importance scores of the key logging parameter and seismic attribute data set, and a reservoir pattern recognition feature set is obtained; Based on the support vector machine method, reservoir pattern classification is performed on the reservoir pattern recognition feature set, wherein multiple reservoir patterns are separated by constructing an optimal hyperplane to obtain preliminary classified reservoir pattern data; The pattern similarity measurement is performed on the initially classified reservoir pattern data based on the Euclidean distance calculation method, wherein the pattern classification result is optimized by calculating the characteristic distance between each reservoir pattern, and the optimized reservoir pattern classification result is obtained; The optimized reservoir pattern classification results are standardized according to the principal component analysis method, wherein the typical characteristic pattern data of cave-type reservoirs are obtained by calculating the principal component weights and normalizing the reservoir pattern characteristic data.

4. The method for predicting carbonate cave reservoirs based on data mining according to claim 1, characterized in that: Seismic attribute extraction and cluster analysis are performed based on the three-dimensional geological model data after seismic inversion. Based on the analysis results and combined with the real-time logging data, lithology and sedimentary environment data of the current drilling, a typical characteristic pattern of cave-type reservoirs is constructed, and the reservoir pattern is optimized and adjusted to obtain an optimized reservoir distribution prediction model, including: Based on the three-dimensional geological model data after seismic inversion, a multi-scale attribute calculation method is used to extract seismic attributes, wherein a set of reservoir-related seismic attributes is obtained by calculating the instantaneous amplitude, coherence and spectrum attributes at different scales; Based on the set of reservoir-related seismic attributes, a self-organizing map neural network is used for cluster analysis. By constructing a high-dimensional seismic attribute topological map, all reservoir patterns are identified and preliminary reservoir pattern classification results are obtained. Based on the preliminary reservoir pattern classification results, combined with the real-time logging data, lithology and sedimentary environment data of the current drilling, the Bayesian probabilistic reasoning method is used to optimize the pattern. By calculating the posterior probability of each pattern, the reservoir classification is optimized to obtain the adjusted reservoir pattern data; According to the adjusted reservoir pattern data, the geostatistical interpolation method is used to predict the reservoir distribution. The spatial distribution of each reservoir pattern is calculated by Kriging interpolation to obtain the optimized reservoir distribution prediction model.

5. The method for predicting carbonate cave reservoirs based on data mining according to claim 4, characterized in that: Constructing a reservoir prediction rule according to the historical oil and gas production data and the optimized reservoir distribution prediction model, and generating a prediction result of a high-quality reservoir area based on the reservoir prediction rule, including: Based on historical oil and gas production data, the grey correlation analysis method is used to identify the main controlling factors. By calculating the correlation between production data and key reservoir attributes, the main control parameters affecting reservoir productivity are obtained. According to the optimized reservoir distribution prediction model, the decision tree method is used to construct reservoir prediction rules. By recursively dividing the reservoir feature space, a classification decision tree for high-quality reservoir areas is established to obtain a reservoir prediction rule set. According to the reservoir prediction rule set, combined with the spatial statistical analysis method, the reservoir high-quality area is divided, and each reservoir block is identified by cluster analysis to obtain the reservoir high-quality area division result, wherein the reservoir high-quality area is an area where the reservoir distribution range is greater than a preset threshold; According to the results of reservoir high-quality area division, the probability of high-quality reservoir areas is calculated by using the probability prediction modeling method. By establishing a probability model based on the Bayesian network, the probability of each area becoming a high-quality reservoir is calculated to obtain the final prediction result of the high-quality reservoir area.

6. A carbonate cave reservoir prediction system based on data mining, characterized in that: include: An acquisition unit is used to acquire historical logging data of proven carbonate reservoirs, real-time logging data of current drilling, lithology and sedimentary environment data, historical oil and gas production data, and three-dimensional geological model data after seismic inversion; A processing unit, configured to perform data feature extraction processing based on the historical logging data of the proven carbonate reservoir to obtain a key logging parameter and seismic attribute data set representing reservoir development; A classification unit, used for performing pattern recognition and classification processing based on the key logging parameters and seismic attribute data set representing reservoir development to obtain a typical characteristic pattern of cave-type reservoirs; An analysis unit is used to extract seismic attributes and perform cluster analysis based on the three-dimensional geological model data after seismic inversion, construct a typical characteristic pattern of cave-type reservoirs based on the analysis results combined with real-time logging data, lithology and sedimentary environment data of the current drilling, and optimize and adjust the reservoir pattern to obtain an optimized reservoir distribution prediction model; A prediction unit, configured to construct a reservoir prediction rule according to the historical oil and gas production data and the optimized reservoir distribution prediction model, and generate a prediction result of a high-quality reservoir area based on the reservoir prediction rule; A generating unit is used to perform spatial interpolation and visualization processing based on the three-dimensional geological model data after seismic inversion and the prediction results of the high-quality reservoir area to generate a reservoir prediction image, wherein the prediction image includes a connectivity distribution map of the cave-type reservoir and a probability density map of the oil and gas enrichment area.

7. The carbonate cave reservoir prediction system based on data mining according to claim 6 is characterized in that: The processing unit comprises: The first processing subunit is used to perform dimensionality reduction processing on the historical logging data based on the principal component analysis method, and to screen the key logging parameters that affect the reservoir development by calculating the covariance matrix and performing eigenvalue decomposition on it, so as to obtain an optimized logging parameter set; The second processing subunit is used to perform time alignment on the logging curve data in the logging parameter set according to the optimized logging parameter set by using a dynamic time warping method, wherein the aligned logging curve data is obtained by constructing a cumulative distance matrix and calculating an optimal matching path; The third processing subunit is used to classify key logging parameters according to the aligned logging curve data by using the K-means clustering method, wherein multiple types of carbonate reservoirs are identified by minimizing the intra-class variance method to obtain reservoir classification results; The fourth processing subunit is used to denoise and perform trend analysis on the logging curve data using the wavelet transform method according to the reservoir classification results, wherein the key features of the reservoir are extracted through multi-scale signal decomposition and reconstruction to obtain a set of key logging parameters and seismic attribute data representing reservoir development.

8. The carbonate cave reservoir prediction system based on data mining according to claim 6, characterized in that: The classification unit includes: The first classification subunit is used to perform feature selection on the key logging parameter and seismic attribute data set based on the random forest method, wherein the key features affecting the reservoir pattern recognition are screened by calculating the feature importance scores of the key logging parameter and seismic attribute data set to obtain the reservoir pattern recognition feature set; The second classification subunit is used for classifying reservoir patterns based on the reservoir pattern recognition feature set based on the support vector machine method, wherein the reservoir pattern data of preliminary classification is obtained by constructing an optimal hyperplane to separate multiple reservoir patterns; The third classification subunit is used to measure the pattern similarity of the initially classified reservoir pattern data based on the Euclidean distance calculation method, wherein the pattern classification result is optimized by calculating the characteristic distance between each reservoir pattern to obtain the optimized reservoir pattern classification result; The fourth classification subunit is used to perform pattern standardization on the optimized reservoir pattern classification results according to the principal component analysis method, wherein the typical characteristic pattern data of the cave-type reservoir is obtained by calculating the principal component weights and normalizing the reservoir pattern characteristic data.

9. The carbonate cave reservoir prediction system based on data mining according to claim 6, characterized in that: The analysis unit comprises: The first analysis subunit is used to extract seismic attributes based on the three-dimensional geological model data after seismic inversion by using a multi-scale attribute calculation method, wherein a reservoir-related seismic attribute set is obtained by calculating the instantaneous amplitude, coherence and spectrum attributes at different scales; The second analysis subunit is used to perform cluster analysis using a self-organizing map neural network based on a set of reservoir-related seismic attributes, wherein all reservoir patterns are identified by constructing a high-dimensional seismic attribute topological map to obtain a preliminary reservoir pattern classification result; The third analysis subunit is used to optimize the model based on the preliminary reservoir model classification results, combined with the real-time logging data, lithology and sedimentary environment data of the current drilling, and adopt the Bayesian probability reasoning method to optimize the reservoir classification by calculating the posterior probability of each model to obtain the adjusted reservoir model data; The fourth analysis subunit is used to predict reservoir distribution using a geostatistical interpolation method based on the adjusted reservoir pattern data, and to calculate the spatial distribution of each reservoir pattern through Kriging interpolation to obtain an optimized reservoir distribution prediction model.

10. The carbonate cave reservoir prediction system based on data mining according to claim 9, characterized in that: The prediction unit comprises: The first prediction subunit is used to identify the main control factors based on the historical oil and gas production data by using the grey correlation analysis method, and obtain the main control parameters affecting the reservoir production capacity by calculating the correlation between the production data and the key attributes of the reservoir; The second prediction subunit is used to construct reservoir prediction rules by using a decision tree method according to the optimized reservoir distribution prediction model, and to establish a classification decision tree for high-quality reservoir areas by recursively dividing the reservoir feature space to obtain a reservoir prediction rule set; The third prediction subunit is used to divide the reservoir high-quality area according to the reservoir prediction rule set in combination with the spatial statistical analysis method, identify each reservoir block through cluster analysis, and obtain the reservoir high-quality area division result, wherein the reservoir high-quality area is an area where the reservoir distribution range is greater than a preset threshold; The fourth prediction subunit is used to calculate the probability of high-quality reservoir areas based on the results of reservoir high-quality area division using a probability prediction modeling method. By establishing a probability model based on a Bayesian network, the probability of each area becoming a high-quality reservoir is calculated to obtain the final prediction result of the high-quality reservoir area.

Images