A rapid simulation method for microscopic oil-water movement and water flow path in porous media

Abstract

The present invention discloses a rapid simulation method for microscopic oil-water movement and water flow path in porous media, comprising the following steps: obtaining a weight model, a remaining oil type saturation and position prediction model based on experiments and machine learning methods; segmenting the provided pore-throat structure image into pores and throats; calculating path weights according to pore-throat attributes and prediction models; constructing a weighted directed graph of the pore-throat topological structure; calculating key data such as the path and weighted length connecting the inlet and outlet based on the shortest algorithm; calculating a weighted length threshold according to the target recovery factor to determine the water flow path; and obtaining the distribution results of various types of remaining oil in the model based on the remaining oil type saturation and position prediction model. Based on seepage experiments and graph theory, the technical idea of the present invention conforms to the seepage law, has clear physical meanings, and high result simulation degree; eliminates a large amount of calculations of traditional iterative numerical solutions, and improves the prediction speed by more than a thousand times, making it possible for engineering applications of microscopic-scale research.

Description

Technical Field

[0001] The present invention relates to the field of simulation computing technology, and in particular to a method for quickly simulating microscopic oil-water movement and water flow paths in porous media. Background Art

[0002] As reservoir development enters the later stage, the remaining oil distributed on a macro scale decreases and gradually becomes highly dispersed. Studying the remaining oil distribution problem only at the macro level can no longer meet on-site needs. Therefore, it is necessary to further study the water flow path and the remaining oil occurrence state at the micro pore throat scale to assist in the formulation of potential tapping measures at this stage.

[0003] At present, the description or prediction of oil-water distribution at the microscopic pore scale is mainly obtained through three methods. The first is the physical experimental method, that is, through specific water-displacement oil experiments, observing the experimental results and performing quantitative calculations. The second is to use the ball-and-stick pore throat network model, that is, simplifying the complex microscopic pore network into an idealized ball-and-stick model and performing numerical simulations on this basis. The third is the two-phase flow numerical simulation of the complex microscopic pore throat network model, commonly used methods include the lattice Boltzmann method, dissipative particle dynamics method, level set method, direct solution of the Navier-Stokes equation, etc.

[0004] The above three methods are suitable for describing the oil-water distribution under different conditions, but they all have corresponding defects in specific applications. For the physical simulation experimental method, since it is carried out through actual water-to-oil experiments, the microscopic pore throat model needs to be replaced for each experiment, and the cost of large-scale implementation is extremely high; for the ball-and-stick pore throat network model, an idealized model is provided, which abstracts the real pore structure into an "equivalent" ball-and-stick network model, but it ignores many complexities of real rocks; for the seepage simulation of the complex pore throat network model, it is limited by the hardware computing and solving capabilities and can only be used for small models. Since the above two methods both rely on numerical solutions, a large amount of calculations are required during the experiment, and the calculation speed is very slow.

[0005] In short, the simulation results of physical simulation methods are realistic and precise, the model size is large, and the reservoir is well represented, but the cost is high; the numerical simulation method is low-cost and convenient for quantitative analysis, but the calculation is large and the efficiency is low. At the same time, the model size is small, the reservoir is poorly represented, and it is difficult to simulate the type of microscopic remaining oil. As a result, in practical applications, the conclusions and methods related to the microscopic pore-throat scale water flow path and remaining oil are difficult to effectively connect with the macroscopic scale, which greatly limits the further development of pore-throat scale remaining oil research in engineering applications. Moreover, for previous numerical simulation algorithms, they can often only obtain the range of water flow paths and the distribution of clustered remaining oil, and the results of their numerical simulation cannot be verified by matching the degree of recovery.

[0006] There is a need for a method to predict the water flow path and remaining oil distribution at the micro pore-throat scale that has a large model size, fast calculation and prediction speed, and conforms to the physical laws of oil-water flow in porous media, laying a foundation for the research and scale upgrade of remaining oil at the micro pore-throat size, and providing a feasible solution for its application in macro-scale engineering. Summary of the Invention

[0007] The object of the present invention is to provide a method for quickly simulating micro-scale oil-water movement and water flow path in porous media to solve the technical problems of large calculation amount, low efficiency, small model size in existing prediction methods, and inability to finely characterize complex remaining oil types and occurrence states.

[0008] The present invention provides a method for quickly simulating micro-scale oil-water movement and water flow path in porous media, including the following steps:

[0009] Fabricate pore-throat network physical models with different physical properties, and conduct oil displacement simulation experiments under different injection parameters after saturating the pore-throat network physical models with oil to collect experimental data;

[0010] Perform image binary processing on the experimental data, and convert the pore-throat network physical model containing pore-throat structures into a digital pore-throat structure data volume containing connection relationships and pore-throat attribute parameters;

[0011] Establish a weighted directed graph of the pore-throat network physical model, and use machine learning methods to establish a throat weight model ρ, a remaining oil type saturation prediction model S roi and a remaining oil type pore-throat position prediction model Z;

[0012] Set the recovery factor α, and according to the remaining oil type saturation prediction model S roi Back-calculate the swept area S of the water flow path way to determine the weighted length threshold L threshold to obtain the water flow path;

[0013] Reconstruct the remaining oil types on the water flow path, and within the range of the water flow path, according to the remaining oil type saturation prediction model S roi and the remaining oil type pore-throat position prediction model Z, reconstruct the saturation and occurrence positions of various types of remaining oil within the range of the water flow path.

[0014] Furthermore, the pore-throat attribute parameters include equivalent pore body radius, equivalent throat radius, coordination number, pore-throat ratio, shape factor, relative throat radius, tortuosity, and wettability.

[0015] Furthermore, the construction method of the weighted directed graph is as follows:

[0016] Abstract pores and throats as points and lines, and number them;

[0017] An adjacency matrix is established when calculating the pore-throat connectivity. If the pore body numbered i is adjacent to the pore body numbered j, the element a at the i-th row and j-th column of the adjacency matrix ij = 1, otherwise a ij = 0, and an unweighted directed graph is obtained based on the adjacency matrix;

[0018] Based on the attribute parameters of the throats between adjacent pores, the weight of each throat is calculated according to the throat weight model ρ, and thus a weight matrix is established, and a weighted directed graph is obtained by combining with the adjacency matrix.

[0019] Furthermore, the specific method for establishing the throat weight model ρ is as follows:

[0020] The weight of the water flow path is proportional to the flow resistance of the water flow path, then the calculation method of the throat weight model ρ is:

[0021]

[0022] In the formula: τ is the tortuosity of the throat, is the equivalent radius of the throat, l is the length of the throat, and λ is the adjustment coefficient.

[0023] Furthermore, the determination method of the adjustment coefficient λ is:

[0024] Define the fluid inlet and fluid outlet of the pore-throat network physical model, apply Dijkstra's shortest path algorithm to the weighted directed graph, calculate the shortest weighted path pore-throat sequence from the fluid inlet to the fluid outlet and number them sequentially as s0 and the weighted path length L0, the second shortest weighted path pore-throat sequence s1 and the weighted path length L1,..., until the longest weighted path pore-throat sequence s n and the weighted path length L n , arrange them in ascending order of the weighted path length, and obtain the path set s = {s0, s1, s2,..., s n}, and the corresponding weighted path lengths L = {L0, L1, L2,..., L n};

[0025] Calculate the percentage of the covered area of the pores and throats on each path to the total pore-throat area according to the pore-throat numbers in the path set and the corresponding pore-throat attribute parameters, and obtain the area set PA = {PA0, PA1, PA2,..., PA n}; where, the sum of the area percentages of all pores and throats is 1;

[0026] Set a weighted path cut-off value L threshold , and obtain all the path pore and throat numbers, the percentage of the path covered area to the total pore-throat area S within the interval [L0, L threshold ​threshold , at this time L threshold and S threshold have a one-to-one correspondence;

[0027] Calculate the displacement fluid sweep percentage S for different experiments lab , and the swept pore-throat set PR lab ;

[0028] Let S threshold = S lab , determine L threshold , and obtain the set of pores and throats PR threshold included in all paths within the range of [L0, L threshold ;

[0029] Set the coincident pore-throat set as PR, and calculate the difference function between the two:

[0030]

[0031] In the formula, the "·" above the symbol represents the area of the pore-throat with that number;

[0032] Use the minimum value of M as the objective function to train by machine learning method to optimize the adjustment coefficient λ.

[0033] Furthermore, establish a prediction model for the saturation of remaining oil types S roi The specific method is as follows:

[0034] Define the remaining oil types according to the location and morphology of the remaining oil;

[0035] Set the functional relationship between the saturation of remaining oil types and the recovery factor as:

[0036] S ro-i = a·α 4 + b·α 3 + c·α 2 + d·α 1 + e;

[0037] In the formula, a, b, c, d, e are undetermined coefficients, α is the recovery factor, S ro-i is the saturation of remaining oil types, and i represents the remaining oil types;

[0038] Calculate the saturation of each type of remaining oil S at different recovery factors under the steady state of different experiments roi_lab ;

[0039] Train and adjust the coefficients a, b, c, d, e by machine learning method so that the difference between the experimental result S roi_lab and the predicted result S roi is minimized, and obtain the prediction model for the saturation of remaining oil types S roi .

[0040] Furthermore, the specific method for establishing the prediction model Z of the pore throat position of the remaining oil types within the swept range of the water flow path is as follows:

[0041] Define:

[0042]

[0043] In the formula:

[0044] θ is the angle between the central axis of the throat and the displacement pressure direction, τ is the tortuosity of the throat, r is the relative throat radius, d T1 is the shortest distance from the throat to the displacement inlet, d T2 is the shortest distance from the throat to the displacement outlet;

[0045] R is the pore throat ratio of the pore body, G is the shape factor of the pore body, d B1 is the shortest distance from the pore body to the displacement inlet, d B2 is the shortest distance from the pore body to the displacement outlet;

[0046] γ1 and γ2 are weight indices.

[0047] Furthermore, the determination method of the weight indices γ1 and γ2 is as follows:

[0048] Analyze the positions where different types of remaining oil are present in the pores and throats according to different experiments;

[0049] Set the throat determination parameter as R CO , and the pore determination parameter as P MPO ;

[0050] Calculate the P of these pores and throats according to the prediction model Z of the pore throat position of the remaining oil types CO_ro and P MPO_ro ;

[0051] Calculate the P of all pores and throats within the displacement swept range of the model according to the prediction model Z of the pore throat position of the remaining oil types CO_L and P MPO_L , and sort them from large to small;

[0052] Use the first large values of P CO_L and P MPO_L and the minimum error between P CO_ro and P MPO_ro as the objective function to optimize and adjust the weight indices γ1 and γ2 through machine learning methods.

[0053] Furthermore, on the basis of constructing the water flow path, the method for reconstructing the remaining oil types is as follows:

[0054] Define according to the positions of the remaining oil classification:

[0055] The remaining oil outside the water flow path is determined as sheet-like remaining oil;

[0056] The remaining oil within the water flow path is determined as columnar remaining oil and porous remaining oil.

[0057] Further, the method for reconstructing the saturation and occurrence position of each type of remaining oil within the water flow path range is as follows:

[0058] The specific method for adding columnar remaining oil and porous remaining oil to the water flow path range map is as follows:

[0059] Screen all the throats and pore bodies within the water flow path range;

[0060] Calculate R for each throat CO , calculate P for each pore body MPO , and sort the R of all the throats and the P of all the pore bodies within the water flow path range according to the calculation results, and construct vector R CO and vector P MPO ; way and vector P way ;

[0061] Wherein:

[0062] Add columnar remaining oil to the throat with the largest R CO , and add porous remaining oil to the pore body with the largest P MPO ;

[0063] The addition quantity of the columnar remaining oil and the porous remaining oil is based on vector R way and vector P way , and fill the columnar remaining oil and the porous remaining oil into the corresponding pore throats from large to small until the saturation calculated by the remaining oil type saturation prediction model S roi is reached.

[0064] The present invention has the following beneficial effects compared with the prior art:

[0065] First, the calculation speed is fast. Based on graph theory, the present invention quickly obtains the water flow path range, determines the sheet-like remaining oil according to the water flow path, and further adds columnar remaining oil and porous remaining oil within the water flow path range through R_CO and P_MPO. The calculation process is mainly forward calculation, and there is no iterative solution process compared with the traditional numerical simulation method, which greatly speeds up the calculation speed.

[0066] Second, the model is real, fine, larger in size and more representative. Due to the large amount of calculation, traditional numerical simulation methods often modify the pore-throat network model. For example, the pore-throat is simplified to a ball-stick model, which loses many pore-throat attribute characteristics during the simulation process and the simulation results are not easy to interpret; or compress the size of the model and simulate a small number of pore-throat models, resulting in information loss in scale, which hinders the cross-scale application of parameters. The new method changes the simulation idea and greatly improves the calculation speed. On this premise, for complex, fine and large-size pore-throat network models, the water flow path and the distribution results of remaining oil types can also be obtained in a short time, providing the possibility for the upscaling application of pore-throat scale research.

[0067] Third, dual physical driving makes the prediction results more real and reliable. First, the water flow path prediction method based on the weighted directed graph of graph theory conforms to the basic seepage law. Second, the water flow path and the remaining oil type prediction and reconstruction models of the pore-throat network model both come from real experimental results. The specific coefficients are determined through machine learning methods to establish a prediction model, making the prediction results closer to the real physical world. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only exemplary, and for those of ordinary skill in the art, without creative efforts, other implementation drawings can also be obtained according to the provided drawings.

[0069] Figure 1 The flow schematic diagram provided by the embodiment of the present invention;

[0070] Figure 2 The experimental result diagram of the microscopic pore-throat network physical model provided by the embodiment of the present invention;

[0071] Figure 3 The original image of the microscopic pore-throat network model provided by the embodiment of the present invention;

[0072] Figure 4 The distribution diagram of pore-throat attribute parameters - pore radius and throat radius of the microscopic pore-throat network model provided by the embodiment of the present invention;

[0073] Figure 5 The digital image after binary processing of the original microscopic pore-throat network model image provided by the embodiment of the present invention;

[0074] Figure 6 The image of the microscopic pore-throat network digital model after pore-throat identification, segmentation and numbering provided by the embodiment of the present invention;

[0075] Figure 7The weighted directed graph of the embodiment of the present invention;

[0076] Figure 8 The schematic diagram of the shortest water flow path calculated based on Dijkstra's algorithm and the covered pore throats in the embodiment of the present invention;

[0077] Figure 9 The water flow path swept volume (swept pore throats) diagram of the embodiment of the present invention;

[0078] Figure 10 The schematic diagram of the rapid prediction and reconstruction results of the water flow path and the remaining oil type in the pore throat network in the embodiment of the present invention. Detailed implementation manners

[0079] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0080] The technical solution of the present invention includes two major links: the physical model experiment of the pore throat network and the establishment of the prediction model, and the prediction of the water flow path and the reconstruction of the remaining oil type in the digital model of the pore throat network.

[0081] In the first link, the physical model experiment of the pore throat network and the establishment of the prediction model.

[0082] The specific steps are as follows:

[0083] In the first step, fabricate the physical models of the pore throat network with different physical properties and design the fluid inlets and outlets;

[0084] In the second step, saturate the model with oil and conduct oil displacement simulation experiments under different injection parameters, and collect the images and data during the experiments;

[0085] In the third step, after binarizing the physical model images for analysis, convert the physical model images containing the pore throat structure into a digital pore throat structure data volume containing connection relationships and pore throat attribute parameters;

[0086] In the fourth step, according to the images during the experiment, obtain parameters such as the recovery factor, the swept range of the water flow path, the remaining oil type, and the distribution pore throat positions at each moment during the experiment;

[0087] In the fifth step, establish a weighted directed graph of the physical model of the pore throat network, and combine the experimental analysis results to optimize the parameters using machine learning methods to establish the throat weight model ρ, the remaining oil type saturation prediction model S roi and the remaining oil type pore throat position prediction model Z.

[0088] The second step is to predict the water flow path and reconstruct the remaining oil types in the digital model of the pore-throat network.

[0089] The specific steps are as follows:

[0090] First step: Provide the image file of the digital model of the pore-throat network to be predicted and reconstructed.

[0091] Second step: Binarize the image of the pore-throat network model. Segment and number the pores and throats through a segmentation algorithm, calculate the pore-throat connectivity and pore-throat attribute parameters, and convert the image of the microscopic pore-throat structure into a digital pore-throat structure data volume containing connection relationships and attribute parameters.

[0092] Third step: Based on the connection relationship between the pores and throats in the original image and the inlet-outlet position relationship, establish an abstract point-line topological structure, that is, establish an unweighted directed graph.

[0093] Fourth step: Calculate the pore-throat attribute parameters. Calculate the throat weight according to the throat weight model ρ, and establish a weighted directed graph.

[0094] Fifth step: Calculate all paths from the inlet to the outlet based on the shortest path algorithms such as Dijkstra. Calculate the weighted path length, the covered pore-throat numbers, and the percentage of the area (volume) of the covered pore-throats (in the present invention, the area can also be the volume, the area reflects the 2D feature, while the volume reflects the 3D feature) of all paths, that is, establish a data volume containing the key information of the paths.

[0095] Sixth step: Set the target recovery rate α. According to the remaining oil type saturation prediction model Sroi, calculate the water flow path swept area (volume) Sway by inverse calculation, determine the weighted length threshold Lthreshold, and obtain the water flow path.

[0096] Seventh step: Outside the range of the water flow path, reconstruct the remaining oil types not swept by the displacement fluid, such as continuous remaining oil.

[0097] Eighth step: Inside the range of the water flow path, reconstruct the saturation and occurrence positions of various types of remaining oil according to the remaining oil type saturation prediction model Sroi and the remaining oil type pore-throat position prediction model Z.

[0098] Explanation of the above terms: The porous medium is a structure composed of a solid material skeleton and a large number of densely grouped tiny voids separated by the skeleton, where the voids are composed of pores and throats; the physical model of the pore-throat network refers to a porous medium model containing pores and throats, such as physical pore-throat network models like solid rock blocks, rock thin sections, chips, etc., and digital pore-throat network models such as virtual pore-throat network images and numerical simulation models.

[0099] The pore throats refer to pores and throats. Pores refer to the enlarged parts in the pore throat network, and throats are the narrow parts connecting the pores. The connection relationship refers to the positional relationship, connection sequence, and network connection relationship between pores and throats. The pore throat property parameters include equivalent pore radius, equivalent throat radius, throat length, coordination number, pore throat ratio, shape factor, tortuosity, wettability, etc.

[0100] The recovery factor refers to the ratio of the oil volume displaced to the original total oil volume in the model (equal to the total pore throat volume of the model). The saturation refers to the ratio of a certain fluid type stored in the model to the total pore throat volume of the model. Among them, the ratio of all fluid types stored in the model to the total pore throat volume of the model is 1.

[0101] The water flow path refers to the path composed of the front and back connections of pores and throats from the inlet to the outlet, which is the range swept (affected) by the displacing phase fluid during the oil displacement process in the experiment. Among them, the inlet and outlet can refer to the injection well and production well of the oilfield in the physical prototype.

[0102] In the embodiment, as Figure 1 shown, a method for rapidly simulating the microscopic oil-water movement and water flow path in a porous medium provided by the present invention predicts the quantitative distribution of microscopic residual oil and displacing phase after oil displacement by the displacing phase under certain conditions according to the microscopic pore throat model, that is, the specific pore throat structure.

[0103] The experimental results of the microscopic pore throat network model are as Figure 2 shown, where the red is the oil phase and the white is the water phase. Through a large number of physical simulation experiments, a throat weight model ρ, a residual oil type saturation prediction model S roi and a residual oil type pore throat position prediction model Z are established.

[0104] The residual oil types refer to the classification of residual oil types based on morphological characterization parameters such as shape factor, contact ratio, aspect ratio, Euler number, etc. and position characteristics. For the convenience of expression, in this method, the residual oil types are simply divided into columnar residual oil, porous residual oil, and sheet-like residual oil. The specific differentiation and definition methods are shown in Table 1:

[0105] Table 1 Quantitative characterization of residual oil characteristic parameters

[0106]

[0107] Among them, columnar residual oil and porous residual oil are the oils affected by the displacing phase but not displaced, that is, the residual oil types within the water flow path range.

[0108] Sheet-like residual oil is the oil not affected by the displacing phase, that is, the residual oil type outside the water flow path range.

[0109] The sum of the saturations of all remaining oil types and the non - oil - phase saturation is 1, and the remaining oil types can be more finely divided under this principle.

[0110] In the embodiment, the pore - throat network model to be predicted is as Figure 3 , where the white part represents pore - throats and the blue part represents the rock skeleton; its pore radius distribution and throat radius distribution are as Figure 4 , and the specific attribute parameters of each pore and throat are automatically extracted and calculated during the calculation process.

[0111] The simulation experiment conditions are as follows: the model is a water - wet model with a wetting angle of 65°; the crude oil viscosity is 30 mPa·s, and the displacement velocity is 0.1 μL / min.

[0112] A method for quickly predicting and reconstructing the water flow path and remaining oil types in the pore - throat network includes the following steps:

[0113] Input the digital model image of the pore - throat network, and perform image binarization, as Figure 5 , where the white part represents pore - throats and the black part represents the rock skeleton;

[0114] Perform pore - throat segmentation on the microscopic pore - throat network through a fast pore - throat segmentation algorithm, and perform pore - throat connectivity calculation and pore - throat parameter calculation, converting the image of the microscopic pore - throat structure into a digital pore - throat structure, as Figure 6 , where the white part represents pores, the gray part represents throats, and the black part represents the rock skeleton; the pore - throat parameters to be calculated include pore radius, throat radius, coordination number, pore - throat ratio, shape factor, relative throat radius, and tortuosity.

[0115] Among them:

[0116] Pore - throat ratio

[0117] Relative throat radius

[0118] Shape factor

[0119] Tortuosity

[0120] In the formula: r B is the pore radius, r T is the throat radius, is the radius of the throat connected by the pores adjacent to the throat, A is the area, P is the perimeter, L t is the curve length of the throat central axis, and L0 is the straight - line distance between the two endpoints of the throat central axis.

[0121] Construct a weighted directed graph of the pore - throat network structure, and calculate the shortest water flow path based on the weighted directed graph.

[0122] Furthermore, the method for establishing a weighted directed graph based on the pore-throat network model is as follows:

[0123] Based on graph theory, pores and throats are abstracted as points and lines, numbered, the pore-throat connectivity is calculated and an adjacency matrix is established. If the pore numbered i is adjacent to the pore numbered j, then the element a at the i-th row and j-th column of the adjacency matrix ij = 1, otherwise a ij = 0, and an unweighted directed graph is obtained based on the adjacency matrix;

[0124] Based on the attribute parameters of the throats between adjacent pores, the weight of each throat is calculated according to the throat weight model ρ, and thus a weight matrix is established. Combining with the adjacency matrix, a weighted directed graph is obtained. In view of the fact that a complex weighted directed graph is not easy to display in a two-dimensional plane, for the sake of easy understanding, Figure 7 FIG. is a schematic diagram of a weighted directed graph, where pores are abstracted as nodes, throats are abstracted as lines, the numbers on the lines are weights, and the arrows indicate the direction of the lines.

[0125] Among them, the specific method for establishing the throat weight model ρ is as follows:

[0126] The weight model of the throat is defined as ρ, and the weight ρ is obtained by calculating the pore-throat attribute parameters that affect the flow resistance. The weight indicates the ease of fluid flowing through the throat, that is, it is proportional to the flow resistance of the throat. In this method, this coefficient is defined as the following formula, but based on the above principle, the weight calculation formula can be further optimized and enriched:

[0127]

[0128] In the formula: τ is the tortuosity of the throat, is the equivalent radius of the throat, l is the length of the throat, λ is an adjustment coefficient, which is related to other pore-throat attribute parameters and injection parameters, and is determined by experimental results and machine learning methods, with an initial value of 1.

[0129] The specific method for determining λ is as follows:

[0130] Define the fluid inlet and outlet of the pore-throat network model, apply the Dijkstra shortest path algorithm to the weighted directed graph, calculate the shortest weighted path pore-throat sequence (number) s0 from the inlet to the outlet and the weighted path length L0, the second shortest weighted path pore-throat sequence s1 and the weighted path length L1,..., until the longest weighted path pore-throat sequence s n and the weighted path length L n , arrange them in ascending order of the weighted path length, and obtain the path set s = {s0, s1, s2,..., s n} from the inlet to the outlet, and the corresponding weighted path lengths L = {L0, L1, L2,..., L n};

[0131] As shown in Figure 8 , where the blue color represents the entrance (starting point), the red color represents the exit (ending point), and the yellow color represents the shortest weighted path (pores and throats on the path).

[0132] Calculate the percentage of the pore-throat coverage area (volume) on each path to the total pore-throat area (volume) according to the pore-throat numbers and corresponding pore-throat attribute parameters on the path, and obtain the set PA = {PA0, PA1, PA2,..., PA n}; Among them, the sum of the area (volume) percentages of all pores and throats is 1;

[0133] Thus, the different weighted length paths, the pore and throat numbers on the path, and the percentage of the pore-throat area (volume) to the total pore-throat area (volume) can be statistically obtained, providing a data set for subsequent path analysis.

[0134] At this time, given a weighted path cut-off value L threshold , all the pore and throat numbers on the paths within the interval [L0, L threshold , and the percentage S of the path coverage area (volume) to the total pore-throat area (volume) can be obtained threshold ; At this time, there is a one-to-one correspondence between L threshold and S threshold ;

[0135] Collect a large number of experimental images of different pore-throat network physical models and different injection parameters, select the stable final displacement state, and calculate the displacement fluid sweep percentage S lab , and the swept pore-throat set PR lab ;

[0136] Let S threshold = S lab , determine L threshold , and obtain the set PR threshold of pores and throats included in all paths within the range of [L0, L threshold ;

[0137] Assume that the overlapping pore-throat set of the two is PR, and calculate the difference function between the two through the following formula:

[0138]

[0139] In the formula, the "·" above the symbol represents the area (volume) of the pore-throat with this number;

[0140] During the training process of the machine learning method, continuously optimize the value of λ until M is the smallest, and thus establish the throat weight model ρ;

[0141] Furthermore, establish the remaining oil type saturation prediction model Sroi The specific method is as follows:

[0142] Define the remaining oil types according to the position and shape of the remaining oil, such as columnar remaining oil, porous remaining oil, and continuous remaining oil;

[0143] Assume that the functional relationship between the saturation of each type of remaining oil and the recovery factor is a polynomial relationship. For the convenience of expression and calculation, it is recommended to simplify it to a quartic function relationship, that is:

[0144] S roi = a·α 4 + b·α 3 + c·α 2 + d·α 1 + e, (i = 1, 2, 3)

[0145] In the formula, a, b, c, d, and e are undetermined coefficients, α is the recovery factor, and i = 1, 2, 3 represent columnar remaining oil, porous remaining oil, and continuous remaining oil types respectively.

[0146] Collect a large number of experimental images under different pore-throat network physical models and different injection parameters, and analyze the saturation S of each type of remaining oil at different recovery factors in its stable state roi_lab(i＝1,2,3) ;

[0147] Through the machine learning method, continuously optimize the coefficients a, b, c, d, and e to minimize the difference between the experimental result S roi_lab and the predicted result S roi value, and obtain the prediction model S of the remaining oil type saturation roi ;

[0148] Furthermore, the specific method for establishing the prediction model Z of the pore-throat position of the remaining oil type within the swept range of the water flow path is as follows:

[0149] Define:

[0150]

[0151] In the formula:

[0152] θ is the angle between the central axis of the throat and the displacement pressure direction, τ is the tortuosity of the throat, r is the equivalent throat radius, d T1 is the shortest distance from the throat to the displacement inlet, d T2 is the shortest distance from the throat to the displacement outlet;

[0153] R is the pore-throat ratio of the pore, G is the shape factor of the pore, d B1 is the shortest distance from the pore to the displacement inlet, d B2 is the shortest distance from the pore to the displacement outlet;

[0154] γ1 and γ2 are weight indices, which are related to injection parameters and determined based on the results of physical simulation experiments and using machine learning methods.

[0155] Among them, the determination methods of γ1 and γ2 are as follows:

[0156] Collect a large number of experimental images of different pore-throat network physical models under different injection parameters, analyze the pore-throat positions where various types of remaining oil are stored, and calculate the P of these pore-throats according to the prediction model Z CO_ro and P MPO_ro ;

[0157] Calculate the P of all pore-throats within the swept range of the displacement phase of the model according to the prediction model Z CO_L and P MPO_L , sort them from large to small; optimize γ1 and γ2 through machine learning methods so that the first few large-value elements of P CO_L and P MPO_L have the smallest error with P CO_ro and P MPO_ro , and obtain the prediction model Z for the pore-throat positions of the remaining oil types within the swept range of the water flow path.

[0158] Furthermore, set the target recovery rate α, use the remaining oil type saturation prediction model S roi , calculate the saturations of various types of remaining oil at the target recovery rate α, and determine the swept volume S of the water flow path according to the saturation of the remaining oil types outside the water flow path way , and the specific formula is:

[0159] S way = 1 - S ro3 (α), i = 3 (referring to continuous remaining oil)

[0160] Determine the weighted path length threshold L threshold , so that the percentage of the pore and throat areas (volumes) passed by the paths within the interval [L0, L threshold from the shortest weighted path length to the weighted path length threshold is equal to the swept volume S of the water flow path way ;

[0161] At this time, all the pores and throats passed by the paths within [L0, L threshold are the water flow paths under the set target recovery rate α.

[0162] For example, set the target recovery rate to α = 0.3, and according to the remaining oil type saturation prediction model S roi , calculate that under these pore-throat property parameters and injection parameters, the swept volume of the water flow path should be S way = 0.37, let to obtain L x = L threshold, at this time, {s0, s1, s2,..., s x} is the water flow path affecting pore throats we are looking for. As Figure 9 shown, where the white part is the predicted water flow path, the black part is the continuous remaining oil, and the gray part is the rock skeleton.

[0163] Furthermore, on the basis of constructing the water flow path, the method for reconstructing the remaining oil type is:

[0164] Combined with the position definition of the remaining oil classification:

[0165] Determine the remaining oil outside the water flow path as continuous remaining oil;

[0166] There are columnar remaining oil and porous remaining oil within the water flow path;

[0167] Furthermore, within the range of the water flow path, reconstruct the types of columnar remaining oil and porous remaining oil, including the reconstruction of the saturation and position of each remaining oil type. The specific determination method is:

[0168] According to the set target recovery factor α, use the remaining oil type saturation prediction model S roi(i＝1,2,3) to calculate the saturation of each type of remaining oil;

[0169] Screen all the throats and pores within the range of the water flow path, apply the remaining oil type pore throat position prediction model Z, calculate R for each throat within the range of the water flow path CO , calculate P for each pore MPO , and sort the R of the throats and the P of the pores from large to small according to the calculation results, and construct the vector R CO and the vector P MPO ; way ; way ;

[0170] Among them, add columnar remaining oil in the throat with the largest R CO , add porous remaining oil in the pore with the largest P MPO , according to the calculated saturation of columnar remaining oil and porous remaining oil, add columnar remaining oil in the throat with the largest R CO , add porous remaining oil in the pore with the largest P MPO , and overall identify and optimize the remaining oil type, classify the remaining oil types that cannot be accurately identified inside the water flow path, and obtain the final results of the microscopic pore throat water flow path and the occurrence state of the remaining oil type. As Figure 10 shown, where the white is the water phase, the black is the continuous remaining oil, the red is the porous remaining oil, and the yellow is the columnar remaining oil.

[0171] Based on the vector R way and the vector Pway , fill columnar residual oil and porous residual oil into the corresponding pore throats from large to small until the saturation calculated by each type of residual oil through model S roi is reached.

[0172] In the present invention, the prediction result of the new method is more refined and realistic. The water flow path and the distribution of multi-type residual oil after water flooding in the microscopic pore throat model can be quickly obtained directly by the forward calculation method, such as columnar residual oil and porous residual oil. New types of residual oil can also be defined. Therefore, it can be more consistent with the details of the real experimental results. The traditional numerical simulation method can only obtain the range of the water flow path and the distribution of continuous residual oil, and involves a large number of calculations for solving the seepage equation, with extremely slow speed.

[0173] In the present invention, the new method greatly improves its calculation efficiency and reaches the engineering application level. Through comparative simulation experiments, it is proved that the traditional method is applicable to small-size models with fewer pore throats and has a slow calculation speed. For example, the direct solution method of the Navier-Stokes equation, when tested on a model with 621 pores, this algorithm takes 5 - 7 days. This technical solution has a faster calculation speed and can be further applied to large-size models with a large number of pore throats. For example, when tested on the AMD Ryzen7Pro 5850U platform, for a small model with 700 pores, the result can be obtained in only 55 seconds, and for a large model with 5800 pores, it takes 20 minutes, with a speed increase of nearly more than 7000 times.

[0174] The present invention can be applied to two-dimensional, three-dimensional and multi-dimensional situations. The above embodiments are the applications of the method in a 2D model and are only exemplary embodiments of the present application, not used to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art can make various modifications or equivalent replacements within the essence and protection scope of the present application, and such modifications or equivalent replacements should also be regarded as falling within the protection scope of the present application.

Claims

1. A method for rapid simulation of microscopic oil-water movement and water flow path in porous media, characterized in that: The steps include: Making a pore-throat network physical model with different physical property characteristics, and carrying out an oil displacement simulation experiment under different injection parameters after saturating the pore-throat network physical model with oil, so as to collect experimental data; The experimental data are subjected to image binarization processing, and the pore-throat network physical model containing the pore-throat structure is converted into a digital pore-throat structure data body containing connection relations and pore-throat attribute parameters; Using machine learning to determine adjustment coefficients , Remaining oil type saturation prediction model S roi And the remaining oil type pore throat location prediction model Z, based on the adjustment coefficient Establishing a throat weight model ρ, and establishing a weighted directed graph of the pore-throat network physical model through the throat weight model ρ; The adjustment factor The method to determine is: The Dijkstra shortest path algorithm is applied to the weighted directed graph, and the weighted path pore throat sequence and weighted path length are arranged from small to large according to the weighted path length, and the path set s = {s0, s1, s2, ..., s n }, and the corresponding weighted path length L = {L0, L1, L2, ..., L n }; According to the pore throat numbers and corresponding pore throat attribute parameters in the path set, the percentage of the pore throat coverage area on each path to the total pore throat area is calculated, and the area set PA = {PA0, PA1, PA2, ..., PA n }; where the sum of the area percentages of all pores and pore throats is 1; Set a weighted path cutoff value L threshold , we get [L0, L threshold ] The number of all path pores and throats in the interval, and the percentage of path coverage area to the total pore throat area S threshold , at this time L threshold With S threshold There is a one-to-one correspondence; Calculate the displacement fluid sweep percentage S for different experiments lab , and the affected pore throat set PR lab ; Let S threshold = S lab , determine L threshold , we get [L0, L threshold The set of pores and throats contained in all paths within the range PR threshold ; Assume that the set of overlapping pore throats is PR, and calculate the difference function between the two: In the formula, the "·" above the symbol represents the area of ​​the pore throat with that number; The machine learning method is trained with the minimum M value as the objective function to optimize the adjustment coefficient ; Set the recovery factor α, and according to the remaining oil type saturation prediction model S roi Back-calculate the flow path and affected area S way , determine the weighted length threshold L threshold , get the water flow path; The remaining oil type is reconstructed on the water flow path, and within the water flow path, the remaining oil type saturation prediction model S is used. roi And the remaining oil type pore throat position prediction model Z is used to reconstruct the saturation and occurrence location of each type of remaining oil within the water flow path.

2. A method for rapid simulation of microscopic oil-water movement and water flow path in porous media according to claim 1, characterized in that: The pore throat property parameters include equivalent pore body radius, equivalent throat radius, coordination number, pore throat ratio, shape factor, relative throat radius, tortuosity, and wettability.

3. A method for rapid simulation of microscopic oil-water movement and water flow path in porous media according to claim 2, characterized in that: The method for constructing a weighted directed graph is: Abstract the pores and throats into points and lines and number them; When calculating the pore throat connectivity, an adjacency matrix is ​​established. If the pore body numbered i is adjacent to the pore body numbered j, then the element in the i-th row and j-th column of the adjacency matrix is , otherwise , obtaining an unweighted directed graph based on the adjacency matrix; Based on the attribute parameters of the throat between adjacent pores, The weight of each throat is calculated, and a weight matrix is ​​established based on it. Combined with the adjacency matrix, a weighted directed graph is obtained.

4. A method for rapid simulation of microscopic oil-water movement and water flow path in porous media according to claim 3, characterized in that: in, The throat The specific method of establishment is: The weight of the water flow path is proportional to the flow resistance of the water flow path, so the throat The calculation method is: ； Where: τ is the tortuosity of the throat, is the equivalent radius of the throat, is the length of the throat, is the adjustment coefficient.

5. A method for rapid simulation of microscopic oil-water movement and water flow path in porous media according to claim 1, characterized in that: Establishment of remaining oil type saturation prediction model S roi The specific method is: Define the type of residual oil according to its location and form; The functional relationship between the remaining oil type saturation and the recovery factor is set as: ； In the formula, a, b, c, d, e are unknown coefficients. is the recovery factor, i represents the type of residual oil; Calculate the remaining oil saturation S of various types at different recovery factors under stable conditions in different experiments roi_lab ; Through machine learning methods, we train and adjust the coefficients a, b, c, d, and e so that the experimental results S roi_lab And the prediction result S roi The value difference is the smallest, and the remaining oil type saturation prediction model S is obtained. roi .

6. A method for rapid simulation of microscopic oil-water movement and water flow path in porous media according to claim 1, characterized in that: The specific method for establishing the prediction model Z of the pore throat position of the remaining oil type within the range of the water flow path is as follows: definition: ； Where: The throat determination parameters are , the pore determination parameter is θ is the angle between the throat centerline and the displacement pressure direction, τ is the throat tortuosity, r is the relative throat radius, is the shortest distance from the throat to the displacement inlet, The shortest distance from the throat to the displacement outlet; R is the pore throat ratio of the pore body, G is the shape factor of the pore body, is the shortest distance from the pore to the displacement entrance, The shortest distance from the pore to the displacement outlet; and is the weight index.

7. A method for rapid simulation of microscopic oil-water movement and water flow path in porous media according to claim 6, characterized in that: Weight Index and The method to determine is: According to different experiments, the locations of various types of residual oil in pores and throats are analyzed; Set the throat judgment parameters to , the pore determination parameter is ; According to the remaining oil type pore throat position prediction model Z, the judgment parameters of these throats and pores are calculated and recorded as and ; According to the pore throat position prediction model Z of the remaining oil type, the judgment parameters of all throats and pores within the displacement range of the model are calculated and recorded as and by and The maximum value of the front and The minimum error is used as the objective function to train the machine learning method to optimize the weight index and .

8. A method for rapid simulation of microscopic oil-water movement and water flow path in porous media according to claim 7, characterized in that: On the basis of constructing the water flow path, the method for reconstructing the residual oil type is: Definition of location according to residual oil classification: The remaining oil outside the water flow path is judged as continuous sheet-like remaining oil; The residual oil in the water flow path is classified into columnar residual oil and porous residual oil.

9. A method for rapid simulation of microscopic oil-water movement and water flow path in porous media according to claim 8, characterized in that: The method for reconstructing the saturation and location of various types of remaining oil within the water flow path is: The specific method of adding columnar residual oil and porous residual oil to the water flow path range map is: Screen all throats and orifices within the water flow path; Calculate for each throat , calculate for each hole , and based on the calculation results, the throats of all the throats within the water flow path are and all pores Sort and construct vector R way and vector P way ; in: exist Add columnar residual oil in the largest throat. Add porous residual oil in the largest pores; The amount of columnar residual oil and porous residual oil added is based on the vector R way and vector P way , fill the corresponding pore throats with columnar remaining oil and porous remaining oil from large to small until each type of remaining oil is reached through the remaining oil type saturation prediction model S roi Calculated saturation.

Images