Oil and gas resource spatial distribution prediction method of tree enhanced naive Bayes classifier

Abstract

The invention discloses an oil and gas resource spatial distribution prediction method of a tree enhanced naive Bayes classifier, and belongs to the field of petroleum geology and artificial intelligence, and the method comprises the steps: S1, collecting exploratory well, earthquake and geological information; s2, oil and gas resource space distribution prediction master control geological factors are extracted, and an exploratory well data set and a grid data set are made; s3, establishing a tree enhanced Bayesian network model according to the exploratory well data set; s4, calculating the oil-gas probability of the grid points by using the model; and S5, drawing an oil and gas resource space distribution probability graph by using an interpolation method. The method can achieve the accurate and efficient prediction of the spatial distribution of oil and gas resources through optimizing the selection strategy of the dependency relationship of the main control geological factors and combining the geological data features, can provide quantitative decision support for oil and gas exploration and deployment, and improves the exploration efficiency and economic benefits.

Description

TECHNICAL FIELD

[0001] The present application relates to the fields of petroleum geology and artificial intelligence, and in particular to a tree-enhanced naive Bayes classifier-based oil and gas resource spatial distribution prediction method. BACKGROUND

[0002] Oil and gas resource spatial distribution prediction refers to a process of quantitatively predicting the occurrence location, scale and probability of underground oil and gas resources in three-dimensional space on the basis of geological, geophysical and geochemical multi-source data, and using mathematical, statistical and artificial intelligence methods. The core goal is to reduce exploration risk and guide efficient development of oil and gas fields. Oil and gas resource spatial distribution prediction is of great significance in oilfield exploration.

[0003] At present, many scholars have conducted research in this field. Chen et al. used the FI method to simulate the oil and gas favorability in the western Sverdrup Basin, Canada. This kind of oil and gas resource assessment method mainly relies on expert experience and geological data, and has the problems of strong subjectivity and insufficient accuracy. In recent years, with the development of data-driven methods, many scholars have proposed prediction models based on machine learning: Hu et al. and Xie et al. integrated multi-dimensional geological attribute variables through Mahalanobis distance, and calculated the oil and gas spatial distribution in the study area by using the Bayes formula. Chen et al. used a support vector machine method for geological risk assessment of oil occurrence in an oil block. Zhu et al. proposed a system method for evaluating geological risk and favorability using a logistic regression algorithm.

[0004] However, the data-driven method is mainly based on discriminant modeling and prediction, and cannot quantitatively express the potential relationship between different geological attributes, and the classification accuracy has a large room for improvement. SUMMARY

[0005] The technical problem to be solved by the present application is to provide a tree-enhanced naive Bayes classifier-based oil and gas resource spatial distribution prediction method, which can provide quantitative decision support for oil and gas exploration deployment, solve the problem of low accuracy of oil and gas resource spatial distribution prediction in the prior art, improve exploration efficiency and economic benefits, and has important practical value and application prospect.

[0006] To solve the above technical problems, the technical solution adopted by the present application is:

[0007] A tree-enhanced naive Bayes classifier-based oil and gas resource spatial distribution prediction method, comprising the following steps:

[0008] S1, collecting exploration wells, seismic and geological information;

[0009] S2, extracting oil and gas resource spatial distribution prediction main control factors from the exploration wells, seismic and geological information collected in S1, and making exploration well data set and grid data set;

[0010] S3. Establish a tree-enhanced Bayesian network model based on the exploration well dataset;

[0011] S4. Calculate the probability of oil and gas in grid points using a tree-enhanced Bayesian network model;

[0012] S5. Use interpolation methods to draw a spatial distribution probability map of oil and gas resources;

[0013] A further improvement of the technical solution of the present invention is that: in S1, dynamic production data of oilfields and field outcrop analysis and testing data related to drilling, seismic and hydrocarbon accumulation are collected. The production data includes the results of drilled exploration wells and logging data, the discovered proven reserves and information on the location and abundance of oilfields and hydrocarbons; the field outcrop analysis and testing data includes analysis and testing results related to hydrocarbon accumulation elements.

[0014] A further improvement to the technical solution of the present invention is that S2 specifically includes the following steps:

[0015] S21. Determine the distribution map of the main controlling geological factors based on the hydrocarbon accumulation mechanism and the collected data on the main controlling geological factors;

[0016] The distribution maps of the main controlling geological factors include the distribution maps of effective source rock thickness, reservoir porosity, and caprock thickness;

[0017] S22. Construction of exploration well dataset;

[0018] Extract the main controlling geological parameters of the well location; the main controlling geological parameters include source rock conditions, reservoir conditions, caprock conditions, trap conditions, migration conditions, and preservation conditions.

[0019] S23, Grid Dataset Construction;

[0020] The study area was divided into evaluation units using rasterization technology, and each evaluation unit contained the main controlling geological factor parameters.

[0021] S24. Discretize the continuous data of the main geological factors based on geological knowledge or expert experience to make them suitable for Bayesian network classifier modeling.

[0022] A further improvement to the technical solution of this invention is that: in S3, a tree-enhanced Bayesian network classifier model is established. The Bayesian network classifier has two types of nodes, including nodes of the main controlling geological factors and nodes of exploration results.

[0023] S3 specifically includes the following steps:

[0024] S31. Calculate the conditional mutual information (CMI) between any major controlling geological factors;

[0025] S32, construct a complete graph with the main controlling geological factors as nodes, and set the weight of the edge between any two nodes as CMI (X i ; X j );

[0026] S33, construct a maximum weight spanning tree with the conditional mutual information as the edge weight;

[0027] S34, add edges from the well result class node to all main controlling geological factor nodes; the calculation method of the conditional mutual information CMI is as follows:

[0028]

[0029] Wherein, X i , X j are different main controlling geological factors, and C represents a class variable.

[0030] The further improvement of the technical scheme of the present application is that in S4, based on the constructed network topology structure, parameter learning is performed using training data; the parameter learning includes calculating a prior probability and a conditional probability, and the specific calculation method is as follows:

[0031] Prior probability:

[0032]

[0033] Wherein, N C is the sample number of the class C=c, and N is the total sample number;

[0034] Conditional probability:

[0035] For each main controlling geological factor X i , the estimation of the conditional probability is:

[0036]

[0037] Wherein, x i is a specific value of the main controlling geological factor X i , is the sample number of the main controlling geological factor X i with the value of the parent node Pa (X i ).

[0038] The further improvement of the technical scheme of the present application is that S5 specifically includes the following steps:

[0039] S51, calculate the oil and gas probability of the grid point;

[0040] Based on the tree-enhanced Bayesian network model, the oil and gas posterior probability of the grid unit is calculated according to the Bayesian formula, and the calculation method is as follows:

[0041]

[0042] wherein, Pa(x i ) represents a parent node of a master control geological factor x i ;

[0043] S52, draw a probability map;

[0044] According to the oil and gas probability value of each grid point, a Kriging interpolation method is used to draw an oil and gas probability map in the work area range; the basic form of Kriging interpolation can be expressed as:

[0045]

[0046] wherein, Z(x) is the value of the point x to be interpolated, Z(x i ) is the value of the known point, and λ i is the Kriging weight, representing the influence of the known point on the interpolation point.

[0047] Due to the adoption of the above technical scheme, the technical progress achieved by the present application is:

[0048] 1. In the present application, the tree-enhanced Bayesian classifier is used as an improved model of the naive Bayesian classifier, allowing each master control geological factor node to select at most one other master control geological factor node as its parent node, forming a tree topology structure, and better depicting the dependency relationship between the master control geological factors.

[0049] 2. The present application realizes accurate and efficient prediction of the spatial distribution of oil and gas resources by optimizing the selection strategy of the dependency relationship of the master control geological factors and combining the characteristics of the geological data. DETAILED DESCRIPTION

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings;

[0051] Figure 1 is a flowchart of the prediction of the spatial distribution of oil and gas resources of the present application;

[0052] Figure 2 is a display example of the prediction result of the work area. DETAILED DESCRIPTION

[0053] It is to be understood that the terms "including", "comprising", "having" and "with" used in the specification and the aforementioned claims are used in the sense of "including" and are not used in the sense of "consisting of only" or "consisting of", for example, a process, method, system, product or apparatus that comprises a list of steps or elements as "including" but not limited to those specifically recited.

[0054] The application will be further described in detail below with reference to the accompanying drawings and examples:

[0055] As shown in the figure, a method for predicting the spatial distribution of oil and gas resources based on a Bayesian network classifier includes the following steps: Figure 1

[0056] S1, collecting exploration wells, seismic and geological information;

[0057] Collecting drilling, seismic and oil and gas reservoir-related oilfield dynamic production data and field outcrop analysis test data, including drilled exploration well results and logging data, discovered proven reserves and oilfield location information, and reservoir-related data such as source rock distribution, reservoir distribution, cap rock distribution, trap evaluation, target layer burial depth, etc. Field outcrop analysis test data includes total organic carbon (TOC) value of source rock organic matter abundance index.

[0058] S2, extracting the main control geological factors of the spatial distribution of oil and gas resources from the exploration wells, seismic and geological information collected in S1, and making an exploration well dataset and a grid dataset;

[0059] S2 specifically includes the following steps:

[0060] S21, determining the main control geological factor distribution map based on the reservoir forming mechanism and the collected main control geological factor data;

[0061] Oil and gas go through multiple interrelated links from source to reservoir. Each link can become a bottleneck restricting oil and gas accumulation. Due to differences in the geological characteristics of oil and gas accumulation in different types of basins, the key link controlling accumulation is also different. In order to comprehensively and objectively evaluate the research object, it is necessary to find the geological characteristics and main control factors of accumulation, and to develop a relatively unified evaluation standard.

[0062] The main control geological factor distribution map includes effective source rock thickness distribution map, source rock organic matter abundance distribution map, cap rock thickness distribution map, reservoir thickness distribution map, elevation distribution map, and trap evaluation value distribution map.

[0063] S22, construction of the exploration well dataset;

[0064] ​Extracting various main control geological factor parameters of the location where the exploration well is located; the main control geological factor parameters include source rock thickness, source rock organic matter abundance, reservoir thickness, cap rock thickness, trap condition information, and elevation information.

[0065] S23, grid dataset construction;

[0066] The research area is divided into evaluation units using rasterization technology, and each evaluation unit contains the main control geological factor parameters described in S22;

[0067] S24, discretizing the continuous data of the main control geological factors based on geological knowledge or expert experience to adapt to the modeling of the Bayesian network classifier.

[0068] S3, establishing a tree-enhanced Bayesian network model according to the exploration well dataset;

[0069] The tree-enhanced Bayesian network classifier model is established, and there are two types of nodes in the Bayesian network classifier, including main control geological factor nodes and exploration well result nodes (i.e. class nodes).

[0070] S3 specifically includes the following steps:

[0071] S31, calculating the conditional mutual information CMI between any main control geological factors;

[0072] S32, constructing a complete graph with main control geological factors as nodes, and setting the weight of the edge between any two nodes as CMI(X i ;X j |C);

[0073] S33, constructing a maximum weight spanning tree with conditional mutual information as edge weight;

[0074] S34, adding an edge from the exploration well result class node to all main control geological factor nodes; the calculation method of conditional mutual information CMI is as follows:

[0075]

[0076] Wherein, X i , X j are different main control geological factors, and C represents the class variable.

[0077] S4, calculating the oil and gas probability of the grid point using the tree-enhanced Bayesian network model;

[0078] Based on the constructed network topology structure, parameter learning is performed using training data; parameter learning includes calculating prior probability and conditional probability, and the specific calculation method is as follows:

[0079] Prior probability:

[0080]

[0081] where N C is the number of samples in class C=c, and N is the total number of samples;

[0082] Conditional probability:

[0083] For each master geological factor X i , the estimate of the conditional probability is:

[0084]

[0085] where x i is a specific value of the master geological factor X i , is the number of samples in class C=c in which the master geological factor X i takes the value of the parent node Pa(X i ).

[0086] S5, using an interpolation method to draw an oil and gas resource spatial distribution probability map;

[0087] S5 specifically includes the following steps:

[0088] S51, calculating the oil and gas probability of a grid point;

[0089] Based on the tree-enhanced Bayesian network model, the posterior probability of oil and gas of a grid cell is calculated according to the Bayesian formula, and the calculation method is as follows:

[0090]

[0091] where Pa(x i ) represents the parent node of the geological attribute x i ;

[0092] S52, drawing a probability map;

[0093] According to the oil and gas probability value of each grid point, a Kriging interpolation method is used to draw an oil and gas probability map within the work area. Kriging interpolation is a spatial interpolation method based on statistics, widely used in geographic information systems (GIS), geological exploration and other fields. Kriging method assumes that the spatial data to be interpolated can be regarded as a random field, that is, the data has certain randomness and correlation in space. The value of each data point is related not only to its own characteristics, but also to the values of surrounding data points. The basic form of Kriging interpolation can be expressed as:

[0094]

[0095] where Z(x) is the value of the point x to be interpolated, Z(x i ) is the value of the known point, and λ iis the Kriging weight, indicating the influence of the known point on the interpolation point.

[0096] Embodiment

[0097] In order to have a clearer understanding of the technical features, objectives and beneficial effects of the present application, we use the spatial distribution prediction of a certain block of oil and gas resources to make the following detailed description of the technical solutions of the present application. It should be noted that this example description cannot be understood as a limitation on the scope of the present application.

[0098] The well data set and the grid data set are formed according to the foregoing data set construction method.

[0099] There are 149 known wells in the example area, of which 49 are oil and gas wells and 100 are non-oil and gas wells. Table 1 is the well data set, and Table 2 is the grid data set.

[0100] Table 1

[0101] Serial number SE RE TE GT ST SA class 1 2 1 1 2 2 2 2 2 2 1 1 2 2 2 1 3 2 1 2 2 2 2 1 4 2 1 2 2 2 2 2 5 2 1 1 3 3 2 2 … … … … … … … … 142 2 1 3 2 2 2 2 143 2 1 3 2 2 2 1 144 2 1 2 3 2 2 1 145 2 1 1 2 3 1 1 146 3 1 1 3 3 2 1 147 3 1 2 3 3 2 1 148 1 1 1 3 3 3 1 149 1 1 1 2 3 3 1

[0102] Table 2

[0103] Serial number SE RE TE GT ST SA 1 1 2 1 1 1 1 2 1 2 1 1 1 1 3 1 1 1 1 1 1 4 1 1 1 1 1 1 5 2 1 1 1 1 1 … … … … … … … 9993 2 1 1 2 2 2 9994 1 2 1 1 1 1 9995 1 2 1 1 1 1 9996 2 1 2 1 2 1 9997 2 1 2 1 2 1 9998 1 2 1 1 1 1 9999 2 1 1 1 1 1 10000 2 1 1 1 1 1

[0104] Wherein, SE is the elevation; RE is the reservoir thickness; TE is the trap evaluation value; GT is the caprock thickness; ST is the source rock thickness; SA is the source rock organic matter abundance; class is the well type, 1 indicating a non-oil and gas well and 2 indicating an oil and gas well.

[0105] The well data set is used for model training, and then the trained model is used to predict the virtual points in the grid data set, and the prediction results are shown in FIG. 2. Figure 2 As can be seen from the figure, the oil and gas probability predicted by our model in the known reserve area is mostly red, i.e. the oil and gas occurrence probability is relatively high, which is consistent with the existing exploration results. In addition, some oil and gas resource favorable areas are still predicted outside the known reserve area, which can show that our model has an important guiding role for the next step of oil well exploration and deployment, and has great practical application value and broad application prospect.

[0106] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent substitutions for part or all of the technical features; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting the spatial distribution of oil and gas resources by tree-enhanced naive Bayes classifier, characterized in that: The method comprises the following steps: S1, collecting well, seismic and geological information; S2, extracting main geological factors for predicting spatial distribution of oil and gas resources from the well, seismic and geological information collected in S1, and making well dataset and grid dataset; S3, establishing a tree-enhanced Bayesian network model according to the well dataset; S4, calculating the oil and gas probability of the grid points by using the tree-enhanced Bayesian network model; S5, drawing the probability map of the spatial distribution of oil and gas resources by using the interpolation method.

2. The method of claim 1, wherein the tree-augmented Naive Bayes classifier is used for spatial distribution prediction of oil and gas resources. In S1, the collected information includes drilling, seismic and oil and gas accumulation related oilfield dynamic production data and field outcrop analysis test data, the production data includes drilled well results, logging data, discovered proven reserves and oilfield location and oil and gas abundance information, and the field outcrop analysis test data includes analysis and test results related to accumulation elements.

3. The method of claim 1, wherein the tree-augmented Naive Bayes classifier is used for spatial distribution prediction of oil and gas resources. S2 specifically comprises the following steps: S21, determining the main geological factor distribution map based on the accumulation mechanism and the collected main geological factor data; the main geological factor distribution map includes distribution maps related to hydrocarbon source rock, reservoir, cap rock, trap, migration and preservation and other accumulation conditions; S22, well dataset construction; extracting each main geological factor parameter at the location of the well; the main geological factor parameters include hydrocarbon source rock condition information, reservoir condition information, cap rock condition information, trap condition information, migration condition information and preservation condition information; S23, grid dataset construction; using gridding technology to divide the study area into evaluation units, each evaluation unit containing the main geological factor parameters described in S22; S24, discretizing the continuous data in the main geological factors based on geological knowledge or expert experience to adapt to the Bayesian network classifier modeling.

4. The method of claim 1, wherein the tree-augmented Naive Bayes classifier is used for spatial distribution prediction of oil and gas resources. In S3, a tree-enhanced Bayesian network classifier model is established. There are two types of nodes in the Bayesian network classifier, including main geological factor nodes and well result nodes; S3 specifically comprises the following steps: S31, calculating the conditional mutual information CMI between any main geological factors; S32, construct a complete graph with the master geological factors as the nodes, and set the weight of the edge between any two nodes as CMI(X i ; X j |C); S33, constructing the maximum weight spanning tree with the conditional mutual information as the edge weight; S34, adding edges from the well result class node to all main geological factor nodes; the calculation method of the conditional mutual information CMI is as follows: where X i , X j are different master geological factors, and C represents a class variable.

5. The method of claim 1, wherein the tree-augmented Naive Bayes classifier is used for spatial distribution prediction of oil and gas resources. In S4, based on the established network topology structure, parameter learning is performed using training data; parameter learning includes calculating prior probability and conditional probability, and the specific calculation method is as follows: Prior probability: where N C is the number of samples of class C=c, N is the total number of samples; Conditional probability: For each master geologic factor X i The estimate of the conditional probability is: where x i is the specific value of the master geologic factor X i , is the number of samples in category C=c with master geologic factor X i with Pa(X i ) as parent node value.

6. The method of claim 1, wherein the tree-augmented Naive Bayes classifier is used for spatial distribution prediction of oil and gas resources. S5 specifically comprises the following steps: S51, calculating the oil and gas probability of the grid points; Based on the tree-enhanced Bayesian network model, the posterior probability of the grid unit containing oil and gas is calculated according to the Bayesian formula, and the calculation method is as follows: wherein Pa(x i ) denotes the parent node of the master geologic factor x i . S52, drawing the probability map; According to the oil and gas probability value of each grid point, the kriging interpolation method is used to draw the oil and gas probability map in the work area; the basic form of kriging interpolation can be expressed as: where Z(x) is the value of the point x to be interpolated, Z(x i ) is the value of the known point, λ i is the Kriging weight, indicating the influence of the known point on the interpolated point.

Images