Oil and gas boundary identification method based on statistical learning and connected graph optimization theory

Abstract

The invention discloses a statistical learning and connected graph optimization theory-based oil and gas boundary identification method, and relates to the technical field of oil and gas field development engineering. The problems that an existing gas phase saturation empirical threshold value is not suitable for a heterogeneous stratum, and meanwhile oil and gas boundary line division is not accurate enough are solved. According to the method, oil-gas-water three phases are divided into different categories based on a current statistical learning theory, a dominant area of each category is divided, matching inspection is performed on the dominant areas based on a matching theory, and a maximum significance difference result passing the inspection is selected; and carrying out optimal solution under graph connectivity constraint on the maximized saliency difference result, and the optimal solution is a gas phase saturation threshold value corresponding to the oil and gas boundary and can be used for dividing the oil and gas boundary line. According to the method for identifying the oil-gas boundary based on the statistical learning and connected graph optimization theory, judgment of the position of the oil-gas phase interface can be more accurate and reasonable, and work such as dense well pattern deployment and new well digging development can be guided to a certain extent.

Description

Technical Field

[0001] This invention relates to the field of oil and gas field development engineering technology, and in particular to a method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory. Background Technology

[0002] In oil and gas field development, subsurface multiphase seepage, distribution, and migration are extremely complex. Oil and gas reservoir numerical simulation uses numerical solutions of differential equations to solve for the seepage field distribution at different subsurface locations (spatial and temporal) to study oil and gas field development. Currently, there is extensive research on oil and gas reservoir numerical simulation both domestically and internationally, and mature commercial software such as CMG and Eclipse are available. However, due to the strong heterogeneity of subsurface conditions and the presence of residual oil, gas, and water phases, single-phase oil, gas, and water are almost nonexistent subsurface. Furthermore, due to the discontinuity of the numerical simulation difference, there is also a discontinuity in the saturation of the three phases (oil, gas, and water) in the gridded oil, gas, and water systems.

[0003] The purpose of oil and gas boundary identification is to separate oil-producing and gas-producing areas. Because different production wells (oil and gas wells) have different wellhead equipment and pressure ranges, as well as different specifications for oil-gas separation and gathering pipelines, drilling a gas well in an oil-producing area or an oil well in a gas-producing area not only causes incompatibility with surface equipment and production chaos, but also poses significant safety risks. Furthermore, accurate identification of the oil-gas interface is of great research value for understanding the modes and extent of underground fluid migration. Therefore, accurate identification of the oil-gas interface is of great significance for oil and gas field development.

[0004] Currently, the commonly used method for identifying hydrocarbon phase boundaries relies on empirical thresholds for gas phase saturation determined by reservoir engineering methods. Points exceeding these thresholds are identified as gas phase zones, while those below are identified as oil phase zones. If the grid edges lie on either side of the threshold, a linear inverse distance weighting method is used to determine the location of the critical point. Connecting the critical points gives the obtained continuous hydrocarbon boundary. This method is relatively intuitive and simple to operate, but it requires a high level of experience in reservoir engineering; it is highly subjective, and the results differ significantly between relatively aggressive and relatively conservative strategies. Even within the same hydrocarbon reservoir, due to the heterogeneity of geological conditions, the residual phase saturation varies in different regions, challenging the validity of the same empirical threshold for gas phase saturation. Summary of the Invention

[0005] This invention addresses the problems in the prior art where existing empirical thresholds for gas saturation are inapplicable to heterogeneous formations and result in inaccurate oil-gas boundary delineation. Instead, it provides a method for identifying oil-gas boundaries based on statistical learning and connectivity graph optimization theory. This method enables more accurate and reasonable determination of the oil-gas interface location, and can, to some extent, guide the deployment of infill well networks and the drilling and development of new wells.

[0006] The present invention solves its problem through the following technical solution: This method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory includes the following steps:

[0007] S1: The calculation is performed using numerical simulation methods. The output of the numerical simulation includes the spatial distribution of the pressure field, force field tensor, and saturation field at a certain time point.

[0008] S2: Based on the spatial distribution of the pressure field and force field tensor obtained in step 1, mark the connectivity between grids, and form a connectivity tensor according to the properties of the force field tensor. Each element in the tensor is a value in the interval (0,1).

[0009] S3: Based on the matching algorithm, the force field tensor is expanded into a two-dimensional matrix along the i, j, k, and t directions, so that the i, j, k, and t coordinates and connectivity information of each pair of variables are all stored in the matrix;

[0010] S4: Based on the spatial distribution of the saturation field obtained in step 1, the harmonic average value of the control path length of a single well in the oil and gas reservoir is used as the initial value of the path length in the study area. In the study area, the oil and gas phases are determined into two categories based on the saturation near the area.

[0011] S5: Calculate the Euclidean distance between the two classes and the Euclidean distance within the class to make the generalized likelihood ratio as significant as possible. During the process, repeatedly return to step 4 and reclassify. For each element changed, retain the group whose generalized likelihood ratio is more significant until all results have been traversed. Perform optimization on the generalized likelihood ratio significance under connectivity constraints. The obtained optimization results include two optimal solutions. Determine the oil and gas boundary distinction threshold based on the two optimal solutions.

[0012] S6: After obtaining the threshold for distinguishing the oil and gas boundary, a search algorithm is used to find all grids whose gas phase saturation falls on both sides of the gas phase saturation threshold of adjacent connected grids. Statistical learning is then performed based on the degree of connectivity to solve for the inverse distance weighting weights. The optimal value of the inverse distance weighting weights is calculated, and the specific coordinates of the oil and gas phase interface are given accordingly.

[0013] S7: Based on the specific coordinates of the oil and gas interface obtained in step 6, draw a formation image containing oil and gas phase interface information.

[0014] Preferably, the spatial distribution of pressure field, force field tensor and saturation field at a certain time point is calculated using oil and gas reservoir numerical simulation technology.

[0015] Preferably, in step S2, the tensor has four dimensions, namely the i, j, and k coordinates for establishing the mesh model and the time axis t.

[0016] Preferably, the two-dimensional matrix after tensor expansion in step 3 is a 9-column matrix, storing information about the four dimensions of two adjacent points; and it is agreed that spatial coordinates are adjacent to determine connectivity, time coordinates are backtracking to determine connectivity, and all other connectivity is 0.

[0017] Preferably, the length D of the control path for a single well in the oil and gas reservoir in step 4 is defined as follows:

[0018] A and B are two distinct elements in a certain distance space. The path length from A to B is defined as the convolution of the shortest distance moved along the i, j, k coordinate axes and the time length.

[0019] Preferably, the two optimal solutions obtained in step 5 are: the first is the optimal solution for the path length of the study area; the second is the optimal solution for the gas phase saturation threshold corresponding to the oil and gas boundary in the current study area.

[0020] Preferably, the optimization method used in step 5 is to calculate the generalized likelihood ratio of the inter-class Euclidean distance and the intra-class Euclidean distance; let be... For a given grid (i, j, k, t), the neighborhood defined by the path length D is divided into two categories based on gas saturation:

[0021]

[0022] Where M and N are respectively The number of grids in the class; ||·|| is the 2-norm, a· and b· are the class... The elements in the problem; the optimization process of the problem is to minimize each grid within the study area. The process.

[0023] Preferably, in step 6, the input data for solving the inverse distance weighting of the statistical learning weights are: historical production data within the target oil and gas field formation, the grid coordinates of the oil and gas wells, and the corresponding two-dimensional matrix output in step 3.

[0024] Preferably, the method for calculating the optimal weight value of the inverse distance weighted statistical learning solution in step S6 is as follows:

[0025] Solving the loss function Let it be the sum of squared errors of the maximum likelihood function, i.e.:

[0026]

[0027] Where L F L represents the probability value output after the Logistic transformation. I The variable is a binary value of 0-1, with oil wells represented by 1 and gas wells by 0.

[0028] Minimize loss function This allows us to obtain the optimal weight values ​​for inverse distance weighting.

[0029] Compared with the above-mentioned background technology, the present invention has the following beneficial effects:

[0030] Based on the numerical simulation results of oil and gas reservoirs, this invention establishes a statistical learning model considering heterogeneous formation conditions, taking into account the formation connectivity obtained from the numerical simulation. Based on this model, different gas saturation thresholds are assigned to different regions of the formation to finely delineate the location of the oil and gas phase interface. This method, grounded in the heterogeneity of the formation, assumes that the oil and gas phase interface should exhibit inconsistent spatiotemporal characteristics to address this heterogeneity. Utilizing statistical learning principles, matching theory, and optimization algorithms based on connected graphs, it solves both the problems of difficulty in identifying oil and gas interfaces under subsurface conditions and the strong subjectivity in judgment, and the current practice of relying on experience to roughly classify oil and gas reservoir areas as homogeneous, leading to a lack of precise description. This makes the determination of the oil and gas phase interface location more accurate and reasonable, demonstrating practicality and innovation, and to a certain extent, can guide the deployment of infill well networks and the drilling and development of new wells. Attached Figure Description

[0031] Figure 1 This is a flowchart of the method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory of the present invention;

[0032] Figure 2 This is a spatial distribution map of the saturation field calculated by reservoir numerical simulation technology in an embodiment of the present invention (the black box line represents the study area);

[0033] Figure 3 A schematic diagram of the production well locations of the SII1+21 oil group in the northern block of the Lamaidian oilfield, on which the statistical learning input data for embodiments of the present invention depends;

[0034] Figure 4 This is a schematic diagram of the production well locations of the SII9 oil group in the northern block of the Lamaidian oilfield, on which the statistical learning input data for this embodiment of the invention depends;

[0035] Figure 5 This is a schematic diagram of the production well locations of the SⅢ1+2 oil group in the northern block of the Lamadian oilfield, on which the statistical learning input data for this embodiment of the invention depends;

[0036] Figure 6 A stratigraphic image of the oil-gas phase interface information of the SII 7+8 oil group in the northern block of the Lamadian oilfield, drawn for an embodiment of the present invention;

[0037] Figure 7 Stratigraphic images of the oil-gas phase interface information of the SII 9 oil group in the northern block of the Lamadian oilfield, according to an embodiment of the present invention;

[0038] Figure 8 Formation images of the oil-gas phase interface information of the SII10+11 oil group in the northern block of the Lamadian oilfield, according to an embodiment of the present invention. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0040] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0041] As attached Figure 1 As shown, this invention discloses a method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory, comprising the following steps:

[0042] S1: The calculation is performed using oil and gas reservoir numerical simulation technology. The output results of the numerical simulation include the spatial distribution of the pressure field, force field tensor and saturation field at a certain time point.

[0043] S2: Based on the spatial distribution of the pressure field and force field tensor obtained in step 1, mark the connectivity between grids, and form a connectivity tensor according to the properties of the force field tensor. Each element in the tensor is a value in the interval (0,1).

[0044] The tensor has four dimensions, namely the i, j, and k coordinates for building the mesh model and the time axis t.

[0045] S3: Based on the matching algorithm, the force field tensor is expanded into a two-dimensional matrix along the i, j, k, and t directions, so that the i, j, k, and t coordinates and connectivity information of each pair of variables are all stored in the matrix;

[0046] The expanded tensor results in a two-dimensional matrix with nine columns, storing information about the four dimensions of two adjacent points. It is agreed that spatial coordinates determine connectivity, time coordinates indicate connectivity, and all other connectivity is 0.

[0047] S4: Based on the spatial distribution of the saturation field obtained in step 1, the harmonic average value of the control path length of a single well in the oil and gas reservoir is used as the initial value of the path length in the study area. In the study area, the oil and gas phases are determined into two categories based on the saturation near the area.

[0048] The control path length D of a single well in the oil and gas reservoir is defined as follows:

[0049] A and B are two distinct elements in a certain distance space. The path length from A to B is defined as the convolution of the shortest distance moved along the i, j, k coordinate axes and the time length.

[0050] The specific calculation method is as follows:

[0051]

[0052] In the formula, ω· represents the connectivity information of the expanded two-dimensional matrix, obtained by searching the coordinates i, j, k, and t of points A and B in the two-dimensional matrix. * j * k * t * Let i, j, k, and t be the coordinates of points A and B.

[0053] S5: Calculate the Euclidean distance between the two classes and the Euclidean distance within the class to make the generalized likelihood ratio as significant as possible. During the process, repeatedly return to step 4 and reclassify. For each element changed, retain the group whose generalized likelihood ratio is more significant until all results have been traversed. Perform optimization on the generalized likelihood ratio significance under connectivity constraints. The obtained optimization results include two optimal solutions. Determine the oil and gas boundary distinction threshold based on the two optimal solutions.

[0054] The two optimal solutions obtained are: the first is the optimal solution for the path length of the study area; the second is the optimal solution for the gas saturation threshold corresponding to the oil and gas boundary in the current study area.

[0055] The optimization method employed is to calculate the generalized likelihood ratio of the inter-class Euclidean distance and the intra-class Euclidean distance; denoted as... For a given grid (i, j, k, t), the neighborhood defined by the path length D is divided into two categories based on gas saturation:

[0056]

[0057] Where M and N are respectively The number of grids in the class; ||·|| is the 2-norm, a· and b· are the class... The elements in the problem; the optimization process of the problem is to minimize each grid within the study area. The process.

[0058] S6: After obtaining the threshold for distinguishing the oil and gas boundary, a search algorithm is used to find all grids whose gas phase saturation falls on both sides of the gas phase saturation threshold of adjacent connected grids. Statistical learning is then performed based on the degree of connectivity to solve for the inverse distance weighting weights. The optimal value of the inverse distance weighting weights is calculated, and the specific coordinates of the oil and gas phase interface are given accordingly.

[0059] The input data for statistical learning to solve the inverse distance weighting of weights are: historical production data within the target oil and gas field formation, the grid coordinates of the oil and gas wells, and the corresponding two-dimensional matrix output in step 3.

[0060] The statistical learning method for calculating the optimal weight value using inverse distance weighting is as follows:

[0061] Solving the loss function Let it be the sum of squared errors of the maximum likelihood function, i.e.:

[0062]

[0063] Where L F L represents the probability value output after the Logistic transformation. I The variable is a binary value of 0-1, with oil wells represented by 1 and gas wells by 0.

[0064] Minimize loss function This allows us to obtain the optimal weight values ​​for inverse distance weighting.

[0065] S7: Based on the specific coordinates of the oil and gas interface obtained in step 6, draw a formation image containing oil and gas phase interface information.

[0066] Example 1

[0067] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings, using the oil and gas distribution of the SaII1-8, SaII9-16, and SaIII1-7 layers in the northern block of the Lamadian Oilfield as an example. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0068] like Figure 1 As shown, a connectivity graph optimization method based on statistical learning and matching theory for identifying the location of the oil-gas phase interface in a subsurface saturation field includes the following steps:

[0069] Step 1: Conduct historical fitting of geological model production data, and determine the spatial distribution of pressure field, force field tensor and saturation field of layers Sa II 1-8, Sa II 9-16 and Sa III 1-7 in the northern block of Lamadian oilfield from 1974 to 2022 on the 1st of each month based on oil and gas reservoir numerical simulation technology.

[0070] Figure 2 This is a spatial distribution map of the saturation field calculated using reservoir numerical simulation technology as described in step 1. The box line indicates the study area. Table 1 shows the time distribution of the force field tensor calculated using reservoir numerical simulation technology as described in step 1.

[0071] Table 1

[0072]

[0073] Step 2: Based on the spatial distribution of the pressure and force field tensors obtained from the numerical simulation of the formation in the Lamadian oil reservoir, mark the connectivity between grids. Form a connectivity tensor based on the properties of the force field tensor, where each element is a value within the interval (0,1).

[0074] Step 3: Based on the matching algorithm, the tensor is expanded into a two-dimensional matrix along the i, j, k, and t directions. Table 2 shows a partial excerpt of the expanded two-dimensional matrix of the connectivity tensor. From left to right, the meanings are: the t coordinate, i coordinate, j coordinate, and k coordinate of point A; the t coordinate, i coordinate, j coordinate, and k coordinate of point B; and the connectivity between points A and B as defined by the coordinates of points A and B.

[0075] Table 2

[0076] TA IA JA KA TB IB JB KB PERV 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9657329 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9588449 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9319854 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9195508 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9144935 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9740131 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9849303 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9151611 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9182541 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9787836 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9521440 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9282444 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9961933 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9579105 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9430531 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9272900 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9144871 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9731716 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9984792 1988 / 7 / 1 71 16 25 1988 / 6 / 1 71 16 25 0.9520380

[0077] Step 4: Based on the spatial distribution of the saturation field in the numerical simulation results of the Lamadian reservoir, the harmonic average value of the control path length of a single well in the oil and gas reservoir, 79 (rounded down to 79.1142), is used as the initial value of the path length in the study area. Within the study area, the gas phase saturation near the oil and gas phase is determined to be of two types.

[0078] Step 5: Calculate the inter-class Euclidean distance and intra-class Euclidean distance after classification based on gas saturation, making the generalized likelihood ratio as significant as possible. During this process, repeatedly return to and reclassify as in Step 4. Perform optimization under connectivity constraints on the significance level to obtain the optimal solution for the path length of the study partition and the gas saturation threshold corresponding to the oil and gas boundary in the current study area.

[0079] Step 6: After obtaining the oil and gas boundary distinction threshold, a search algorithm is used to find all grids whose gas phase saturation falls on both sides of the gas phase saturation threshold of adjacent connected grids. The weights are then calculated by statistical learning based on the degree of connectivity, and the formation xyz coordinates of the oil and gas phase interface are given accordingly.

[0080] Figure 3-5 This is a schematic diagram of the production well locations upon which the statistical learning input data in step 6 depends. Figure 3 A schematic diagram of the production well locations of the SII1+21 oil group in the northern block of the Lamaidian oilfield, on which the statistical learning input data for embodiments of the present invention depends; Figure 4 This is a schematic diagram of the production well locations of the SII9 oil group in the northern block of the Lamaidian oilfield, on which the statistical learning input data for this embodiment of the invention depends; Figure 5 This is a schematic diagram of the production well locations of the SⅢ1+2 oil group in the northern block of the Lamadian oilfield, on which the statistical learning input data for this embodiment of the invention depends;

[0081] Step 7: Based on the specific coordinates of the oil and gas interface obtained in Step 6, draw a formation image containing oil and gas phase interface information.

[0082] Figure 6-8 Here is an example of a formation image containing the hydrocarbon phase interface information generated in step 7. Wherein: Figure 6 A stratigraphic image of the oil-gas phase interface information of the SII 7+8 oil group in the northern block of the Lamadian oilfield, drawn for an embodiment of the present invention; Figure 7 Stratigraphic images of the oil-gas phase interface information of the SII 9 oil group in the northern block of the Lamadian oilfield, according to an embodiment of the present invention; Figure 8 Formation images of the oil-gas phase interface information of the SII10+11 oil group in the northern block of the Lamadian oilfield, according to an embodiment of the present invention.

[0083] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the implementation methods of the present invention, and should be understood that the scope of protection of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of the present invention.

Claims

1. A method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory, characterized in that: Includes the following steps: S1: The calculation is performed using numerical simulation methods. The output of the numerical simulation includes the spatial distribution of the pressure field, force field tensor, and saturation field at a certain time point. S2: Based on the spatial distribution of the pressure field and force field tensor obtained in step 1, mark the connectivity between grids, and form a connectivity tensor according to the properties of the force field tensor. Each element in the tensor is a value in the interval (0,1). S3: Based on the matching algorithm, the force field tensor is expanded into a two-dimensional matrix along the i, j, k, and t directions, so that the i, j, k, and t coordinates and connectivity information of each pair of variables are all stored in the matrix; S4: Based on the spatial distribution of the saturation field obtained in step 1, the harmonic average value of the control path length of a single well in the oil and gas reservoir is used as the initial value of the path length in the study area. Within the study area, the oil and gas facies are determined to be of two types based on the saturation near the area. S5: Calculate the Euclidean distance between the two classes and the Euclidean distance within the class to make the generalized likelihood ratio as significant as possible. During the process, repeatedly return to step 4 and reclassify. For each element changed, retain the group whose generalized likelihood ratio is more significant until all results have been traversed. Perform optimization on the generalized likelihood ratio significance under connectivity constraints. The obtained optimization results include two optimal solutions. Determine the oil and gas boundary distinction threshold based on the two optimal solutions. S6: After obtaining the threshold for distinguishing the oil and gas boundary, a search algorithm is used to find all grids whose gas phase saturation falls on both sides of the gas phase saturation threshold of adjacent connected grids. Statistical learning is then performed based on the degree of connectivity to solve for the inverse distance weighting weights. The optimal value of the inverse distance weighting weights is calculated, and the specific coordinates of the oil and gas phase interface are given accordingly. S7: Based on the specific coordinates of the oil and gas interface obtained in step 6, draw a formation image containing oil and gas phase interface information.

2. The method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory according to claim 1, characterized in that, The spatial distribution of pressure field, force field tensor and saturation field at a certain point in time is calculated using numerical simulation technology for oil and gas reservoirs.

3. The method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory according to claim 1, characterized in that, In step S2, the tensor has four dimensions, namely the i, j, and k coordinates for establishing the mesh model and the time axis t.

4. The method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory according to claim 1, characterized in that, In step 3, the two-dimensional matrix after tensor expansion is a 9-column matrix that stores information about the four dimensions of two adjacent points. It is agreed that spatial coordinates are adjacent to determine connectivity, time coordinates are backtracking to determine connectivity, and all other connectivity is 0.

5. The method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory according to claim 1, characterized in that, The control path length D of a single well in the oil and gas reservoir in step 4 is defined as follows: A and B are two distinct elements in a certain distance space. The path length from A to B is defined as the convolution of the shortest distance moved along the i, j, k coordinate axes and the time length.

6. The method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory according to claim 1, characterized in that, The two optimal solutions obtained in step 5 are: the first is the optimal solution for the path length of the study area; the second is the optimal solution for the gas saturation threshold corresponding to the oil and gas boundary in the current study area.

7. The method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory according to claim 1, characterized in that, The optimization method used in step 5 is to calculate the generalized likelihood ratio of the inter-class Euclidean distance and the intra-class Euclidean distance; let be... For a given grid (i,j,k,t), the neighborhood defined by the path length D described above is divided into two categories based on gas saturation: Where M and N are respectively The number of grids in the class; ||·|| is the 2-norm, a· and b· are the class... The elements in the problem; the optimization process of the problem is to minimize each grid within the study area. The process.

8. The method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory according to claim 1, characterized in that, In step 6, the input data for the inverse distance weighting of the statistical learning solution are: historical production data within the target oil and gas field formation, the grid coordinates of the oil and gas wells, and the corresponding two-dimensional matrix output in step 3.

9. The method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory according to claim 1, characterized in that, The method for calculating the optimal weight value of the inverse distance weighted system in step S6 using statistical learning is as follows: Solving the loss function It is set to be the sum of squared errors of the maximum likelihood function; Minimize loss function This allows us to obtain the optimal weight values ​​for inverse distance weighting.

10. The method for identifying oil and gas boundaries based on statistical learning and connected graph optimization theory according to claim 9, characterized in that, The solution loss function The formula is: In the formula: L F This represents the probability value output after the Logistic transformation. L I The variable is a binary variable of 0-1, with oil wells being 1 and gas wells being 0.

Images