Prediction method and system for movable oil content change before and after fracturing

Abstract

The invention discloses a method and system for predicting the change of the movable oil content before and after fracturing, and belongs to the technical field of oilfield development, the method comprises the following steps: firstly, collecting geological data and production data of a target rock, and constructing a standard data set; then establishing and training a movable oil content prediction model based on machine learning; then determining the movable oil content before fracturing by analyzing a sample of a target rock block, obtaining stratum parameters and fracturing well parameters of a target block, and inputting the model to predict the movable oil content after fracturing; and finally, the content change of the target rock is evaluated based on the movable oil content before and after fracturing so as to guide optimization of an oil reservoir development strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of oilfield development, and more particularly to a method and system for predicting the change in movable oil content before and after fracturing. Background Art

[0002] In the technical field of oil and gas field exploration and development, fracturing technology is an important means to improve reservoir permeability and increase the productivity of oil and gas wells. However, due to the characteristics of poor reservoir physical properties, strong heterogeneity, and low natural productivity of oil reservoirs, the productivity prediction before and after fracturing has become a complex and crucial problem.

[0003] Most traditional productivity prediction methods are based on analytical methods. By setting up multiple assumptions and physical models, a corresponding mathematical model is established for the reservoir, and the flow of oil and gas in the reservoir is described by production or pressure, and the production analytical expression is obtained by solving. However, this method has many limitations in predicting the productivity of fractured wells in reservoirs. First, the analytical method cannot comprehensively consider the influence of geological parameters and fracturing construction parameters (such as sand volume, liquid volume, and displacement), resulting in a large gap between the prediction result and the actual productivity. Second, the analytical method does not have a deep understanding of the main controlling factors of the productivity of fractured wells in reservoirs, and cannot accurately identify the key factors affecting productivity, thus affecting the optimization and improvement of the fracturing construction plan.

[0004] In order to overcome the deficiencies of the traditional analytical method, in recent years, researchers have begun to explore new productivity prediction methods. Among them, the numerical simulation method has attracted much attention because it can more accurately describe the complex flow characteristics of reservoirs and fractures. However, the numerical simulation method also faces many challenges, such as grid division, parameter setting, and boundary condition determination in the model establishment process, and these factors will affect the accuracy of the simulation results and the calculation efficiency.

[0005] In addition, some researchers have tried to apply non-linear methods such as fuzzy theory, neural network, and grey analysis model to productivity prediction. These methods have improved the accuracy and reliability of the prediction results to a certain extent, but there are still some problems, such as complex model establishment, difficult parameter selection, and the prediction results being affected by subjective factors.

[0006] Therefore, how to provide a method and system for predicting the change in movable oil content before and after fracturing that can comprehensively consider geological parameters, fracturing construction parameters, and the complex flow characteristics of fluids in reservoirs and fractures is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0007] In view of this, the present invention provides a method and system for predicting the change in movable oil content before and after fracturing. By comprehensively collecting and analyzing the geological data and production data of the target rock block, a standard data set for prediction is constructed, and then a prediction model for movable oil content based on machine learning is established and optimized to help more accurately evaluate the reservoir potential and fracturing effect of the target rock block.

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] On the one hand, the present invention provides a method for predicting the change in movable oil content before and after fracturing, including:

[0010] Collect multi-source data of the target rock block, where the multi-source data includes geological data and production data, and construct a standard data set for predicting the change in movable oil content before and after fracturing based on the geological data and the production data;

[0011] Establish a prediction model for movable oil content based on machine learning, and use the standard data set to train the prediction model for movable oil content;

[0012] Determine the movable oil content before fracturing by analyzing the samples of the target rock block;

[0013] Obtain the formation parameters and fracturing well parameters of the target rock block, and input the formation parameters and the fracturing well parameters into the trained prediction model for movable oil content to obtain the movable oil content after fracturing;

[0014] Evaluate the content change of the target rock block based on the movable oil content before fracturing and the movable oil content after fracturing.

[0015] Preferably, constructing a standard data set for predicting the change in movable oil content before and after fracturing based on the geological data and the production data includes:

[0016] Construct a fracturing simulation model according to the geological data and the production data;

[0017] Construct a fracturing development seepage mathematical model for the reservoir horizontal well according to the geological data;

[0018] Based on the fracturing development seepage mathematical model and the fracturing simulation model, establish a fracturing movable oil content simulation model for the reservoir horizontal well;

[0019] Use the fracturing movable oil content simulation model to obtain the well parameters and movable oil content data after fracturing under different formation parameters and different fracturing operation combinations;

[0020] Normalize the well parameters, formation parameters and corresponding movable oil content data after fracturing to obtain a standard data set for predicting the change in movable oil content before and after fracturing.

[0021] Preferably, according to the geological data, a seepage mathematical model for the fracturing development of horizontal wells in an oil reservoir is constructed, including:

[0022] Standardize the layers in the obtained geological data, and store the layers in a relational database in mdb format to establish a geological database; the layers in the geological data include at least any one of the following: multi-factor raster layer, mark information vector layer, porosity distribution layer, permeability distribution layer, saturation distribution layer, sand body thickness distribution layer, effective sand distribution layer, well position coordinate information vector layer;

[0023] After importing the geological data using an automated tool, automatically divide according to the preset grid size and type, divide the target rock block into grids of the same size, and obtain the initial state information of each grid based on the geological data;

[0024] Construct the seepage motion equation for each grid based on the initial state information;

[0025] Obtain the fluid continuity equation based on the geological data;

[0026] Substitute the fluid continuity equation into the seepage motion equation to obtain the seepage control equation for each grid;

[0027] Combine the seepage control equations of each grid to obtain a seepage mathematical model for the fracturing development of horizontal wells in an oil reservoir.

[0028] Preferably, a movable oil content prediction model based on machine learning is established, and the movable oil content prediction model is trained using the standard data set, including:

[0029] Determine the influencing factors of the movable oil content;

[0030] Analyze the influencing factors of the movable oil content based on the analytic hierarchy process to obtain the subjective weight;

[0031] Analyze the influencing factors of the movable oil content based on the coefficient of variation method to obtain the objective weight;

[0032] Use the weighted average method to calculate the combined weight of the influencing factors of the movable oil content, and based on the combined weight, obtain the influencing index of the movable oil content;

[0033] According to the influencing index of the movable oil content, establish a BP neural network prediction model and train the BP neural network prediction model;

[0034] Use the ant colony algorithm to optimize the trained BP neural network prediction model to obtain a movable oil content prediction model.

[0035] Preferably, determining the movable oil content before fracturing by analyzing samples of the target rock mass includes:

[0036] Performing pyrolysis on multiple samples to obtain the pyrolysis oil content of the multiple samples;

[0037] Based on the free oil correction coefficient, correcting the pyrolysis oil content of the multiple samples to obtain the first movable oil content of the multiple samples, where the free oil correction coefficient is used to compensate for the loss content of free oil in the pyrolysis oil;

[0038] Based on the first movable oil content and the adsorbed oil content of the multiple samples, determining the second movable oil content of the multiple samples, where the second movable oil content is used to represent the actual movable oil content of the samples;

[0039] Based on the second movable oil content and the adsorbed oil content of the multiple samples, generating a resource quantity distribution map of the target rock mass, where the resource quantity distribution map is a distribution map of the movable oil content of the target rock mass;

[0040] In the resource quantity distribution map, dividing according to the numerical intervals where the second movable oil content of the multiple samples is located to obtain multiple sample sets, and the second movable oil content of the samples within the sample set corresponds to the same numerical interval;

[0041] Based on the formation area, formation thickness, and second movable oil content of the samples in the multiple sample sets, respectively determining the target formation area, target formation thickness, and target second movable oil content of the multiple sample sets;

[0042] Based on the target formation area, target formation thickness, and target second movable oil content of the multiple sample sets, determining the movable oil content before fracturing of the multiple sample sets.

[0043] Preferably, according to the standard data set, establishing a BP neural network prediction model and training the BP neural network prediction model includes:

[0044] Constructing a BP neural network prediction model, determining the number of nodes in the input layer, the number of nodes in the hidden layer, and the number of nodes in the output layer, and initializing the connection weights between the input layer and the hidden layer, the connection weights between the hidden layer and the output layer, the threshold of the hidden layer, and the threshold of the output layer to obtain the initial structure of the BP neural network prediction model;

[0045] Inputting the data in the standard data set into the input layer of the BP neural network prediction model, and after being processed by the weights and activation functions in the hidden layer, obtaining the predicted value of the movable oil content and outputting it through the output layer. The predicted value output by the BP neural network prediction model is:

[0046]

[0047] Among them, b k represents the threshold from the hidden layer to the k-th output layer node, and O k represents the predicted output value from the hidden layer to the k-th output layer node, and H j represents the output value of the j-th node in the hidden layer;

[0048] Calculate the error between the predicted value and the actual value. The specific formula is:

[0049]

[0050] Among them, E k represents the error of the k-th output layer node, M represents the total number of samples, and O mk represents the predicted value of the m-th sample at the k-th output layer node, represents the actual value of the m-th sample at the k-th output layer node;

[0051] According to the error between the predicted value and the actual value, adjust the weights backward;

[0052] Repeat the process of model training until it stops when the preset number of iterations is reached, and obtain the trained BP neural network prediction model.

[0053] Preferably, according to the error between the predicted value and the actual value, the calculation formulas for backward adjusting the network connection weights and thresholds are as follows:

[0054]

[0055] ω' jk = ω jk + ηH j E k j = 1, 2, …, l; k = 1, 2, …, m

[0056]

[0057] b' k = b k + E k k = 1, 2, …, m

[0058] In the formula: η is the learning rate, ω′ ij represents the updated connection weight between the input layer and the hidden layer, ω′ jk represents the updated connection weight between the hidden layer and the output layer, a′ j represents the updated threshold from the input layer to the j-th hidden layer node, b′ k represents the updated threshold from the hidden layer to the k-th output layer node, ω ijdenotes the connection weight between the input layer and the hidden layer before update, ω jk denotes the connection weight between the hidden layer and the output layer before update, a j denotes the threshold from the input layer to the j-th hidden layer node before update, b k denotes the threshold from the hidden layer to the k-th output layer node before update.

[0059] Preferably, the ant colony algorithm is used to optimize the trained BP neural network prediction model to obtain a movable oil content prediction model, including:

[0060] Set the initial parameters: set the maximum number of ant iterations, the number of ants m, the pheromone concentration, the individual optimal and the global optimal;

[0061] Each weight and threshold in the trained BP neural network prediction model is used as a parameter to be optimized and equally divided to form a parameter set respectively, so that a set of weights and thresholds in the parameter set corresponds to the positions randomly placed by m ants;

[0062] Calculate the fitness value of each ant and set the fitness value as the initial pheromone of the ant; wherein, the fitness value of the ant is calculated by the reciprocal of the sum of squared errors;

[0063] Use a preset optimization function to calculate the transfer probability of each ant, and obtain the optimal path of this time according to the transfer probability of each ant, compare it with the optimal value, if it is optimal, update the optimal value;

[0064] Compare the optimal value of each ant with the optimal value of the entire ant colony. If it is better, it becomes the new optimal value of the entire ant colony, and sort all paths to select the optimal path;

[0065] Update the pheromone concentration of each ant;

[0066] Compare whether the number of iterations reaches the maximum number of iterations or whether all ants converge on one path. If either of them is satisfied, output the weights and thresholds corresponding to each dimension in the global optimal value of the last iteration; if not, continue to calculate the transfer probability of each ant and select the optimal path;

[0067] Update the weights and thresholds in the trained BP neural network model with the weights and thresholds corresponding to each dimension in the global optimal value output in the last iteration to obtain a movable oil content prediction model.

[0068] Preferably, the method further includes: evaluating the change in the movable oil content of the target rock block based on the movable oil content before fracturing and the movable oil content after fracturing.

[0069] Preferably, before prediction, an indoor experiment is conducted on the artificial fracture propagation law and proppant distribution of the target rock formation oil reservoir, including:

[0070] (1) Experiment preparation stage:

[0071] Sample preparation: Core samples are taken from the target rock block oil reservoir and prepared into thinly interbedded rock samples;

[0072] Equipment preparation: Prepare the equipment required for the experiment, including a true triaxial fracturing simulation device, a high-precision CT scanning device, a data acquisition system, etc.;

[0073] (2) Experiment setting stage:

[0074] Sample installation: Install the prepared rock sample into the true triaxial fracturing simulation device to ensure that the sample is fixed stably and evenly stressed;

[0075] Parameter setting: Set the experimental parameters according to the experimental requirements to simulate the actual formation conditions;

[0076] Proppant preparation: Prepare an appropriate amount of proppant and determine the particle size and concentration of the proppant according to the experimental plan;

[0077] (3) The experimental steps include:

[0078] Load stress: Apply triaxial stress to the rock sample according to the set parameters to simulate the in-situ stress state in the actual formation;

[0079] Fracturing fluid injection: Inject the fracturing fluid mixed with fluorescent agent into the wellbore at a displacement of 5 - 20 mL / min, and record the change in wellhead pressure; after the rock fractures, the wellhead pressure will drop rapidly, indicating the end of the preflush injection stage;

[0080] Carrying fluid injection: Increase the displacement to 50 mL / min, open the sand outlet valve of the sand tank containing proppant, and after the proppant is mixed with the fracturing fluid, it enters the fracturing pipeline, entering the carrying fluid injection stage; continuously inject the slurry until the wellhead pressure rises sharply and then drops, and then stop the pump;

[0081] Data acquisition and analysis: Use the data acquisition system to record various data during the experiment in real time;

[0082] Use a micron CT scanner to scan the gray-scale image of the specimen, reconstruct the core through high-precision CT data, and comprehensively analyze and identify the fracture morphology on the surface and inside of the rock sample as well as the proppant distribution situation in combination with the tracer distribution data and the rock sample dissection results;

[0083] The gray-scale images of the specimens were classified using VOLUME GRAPHICS STUDIO MAX software to analyze the structural and morphological characteristics of the specimens. Subsequently, a three-dimensional digital core model was constructed, and the box dimension method was used to calculate the spatial complexity of the fractures.

[0084] On the other hand, the present invention provides a prediction system for the change in movable oil content before and after fracturing, comprising:

[0085] A data set construction module that collects multi-source data of the target rock block, where the multi-source data includes geological data and production data, and constructs a standard data set for predicting the change in movable oil content before and after fracturing based on the geological data and the production data;

[0086] A model generation module for establishing a machine learning-based movable oil content prediction model and training the movable oil content prediction model using the standard data set;

[0087] An input module for obtaining the formation parameters and fracturing well parameters of the target rock block;

[0088] A prediction module for determining the movable oil content before fracturing by analyzing the samples of the target rock block and obtaining the movable oil content after fracturing using the trained movable oil content prediction model;

[0089] An evaluation module for evaluating the change in content of the target rock block based on the movable oil content before fracturing and the movable oil content after fracturing.

[0090] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a prediction method and system for the change in movable oil content before and after fracturing. By collecting multi-source geological and production data and constructing a standard data set, this method uses a fracturing simulation model and a fracturing development seepage mathematical model of a reservoir horizontal well to simulate the fracturing effects under different conditions. Furthermore, based on machine learning techniques, especially the BP neural network, a movable oil content prediction model is established and optimized by the ant colony algorithm to improve the prediction accuracy. Finally, by comparing the movable oil content before and after fracturing, the fracturing effect is evaluated, providing a scientific basis for the efficient development of rock layer reservoirs. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or 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 the embodiments of the present invention, and for those of ordinary skill in the art, other drawings can be obtained based on the provided drawings without creative efforts.

[0092] Figure 1 It is a schematic flow chart of the prediction method provided by the present invention.

[0093] Figure 2 This is the framework diagram of the prediction system provided by the present invention. Specific embodiments

[0094] 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 of 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.

[0095] The embodiments of the present invention disclose a prediction method for the change in movable oil content before and after fracturing. Refer to Figure 1 , including:

[0096] Collect multi-source data of the target rock block. The multi-source data includes geological data and production data. Based on the geological data and production data, construct a standard data set for predicting the change in movable oil content before and after fracturing; the geological data includes reservoir parameters, fluid parameters, and other parameters, and the production data is fracturing parameters.

[0097] The reservoir parameters specifically include:

[0098] Effective reservoir permeability: Permeability determines the flow ability of fluids in the reservoir and is a key parameter for predicting movable oil content.

[0099] Porosity: Porosity reflects the size of the pore space in the reservoir and directly affects the oil storage capacity and movable oil content of the reservoir.

[0100] Oil saturation: It refers to the percentage of the pore volume occupied by the oil phase in the reservoir and is an important indicator for evaluating the oil content of the reservoir.

[0101] Effective reservoir thickness: The effective thickness of the reservoir determines the vertical distance of fluid flow and has an important impact on the prediction of movable oil content.

[0102] Reservoir pressure gradient and static pressure: These parameters reflect the fluid pressure distribution in the reservoir and are crucial for predicting fluid flow and movable oil content.

[0103] Reservoir static temperature: Temperature affects the physical properties and fluidity of fluids, thereby affecting the prediction of movable oil content.

[0104] The fluid parameters specifically include:

[0105] Fluid properties: including density, viscosity, compressibility, etc. These parameters determine the flow characteristics of fluids in the reservoir.

[0106] Fluid saturation distribution: Understanding the fluid saturation distribution in the reservoir helps to more accurately predict the movable oil content.

[0107] Other parameters specifically also include:

[0108] Reservoir rock mechanical properties: Such as Poisson's ratio, Young's modulus, compressive strength, etc. These parameters will affect the deformation of the reservoir and the formation of fractures.

[0109] Vertical distribution of reservoir in-situ stress and azimuth of the minimum horizontal principal stress: The in-situ stress distribution will affect the propagation direction and shape of fractures, thus affecting the stimulation effect.

[0110] Lithology, thickness and in-situ stress value of the barrier layer: The properties of the barrier layer will affect the propagation of fractures and the flow of fluids, and thus affect the prediction of movable oil content.

[0111] Fracturing parameters specifically include:

[0112] Fracture geometric dimensions: Including fracture length, width, height, etc. These parameters determine the fracture conductivity and the stimulation potential of the reservoir.

[0113] Fracture conductivity: Fracture conductivity reflects the promotion of fluid flow by fractures and is a key parameter for predicting the stimulation effect after fracturing.

[0114] Fracturing fluid parameters: Including viscosity, flow behavior index, consistency coefficient of the fracturing fluid, etc. These parameters will affect the formation and conductivity of fractures.

[0115] Proppant parameters: Parameters such as proppant type, particle size range, particle density, etc. will affect the propping effect and conductivity of fractures.

[0116] Establish a machine learning-based movable oil content prediction model and train the movable oil content prediction model using a standard data set;

[0117] Determine the movable oil content before fracturing by analyzing samples of the target rock block;

[0118] Obtain the formation parameters and fracturing well parameters of the target rock block, input the formation parameters and the fracturing well parameters into the trained movable oil content prediction model, and obtain the movable oil content after fracturing;

[0119] Based on the movable oil content before fracturing and the movable oil content after fracturing, evaluate the content change of the target rock block.

[0120] Furthermore, construct a standard data set for predicting the change of movable oil content before and after fracturing based on geological data and production data, including:

[0121] Construct a fracturing simulation model based on geological data and production data; the fracturing simulation model can simulate the physical processes in actual fracturing operations, including but not limited to rock fracture, crack propagation, etc. By integrating these data, the geological data before and after fracturing is obtained, facilitating subsequent analysis of fluid penetration.

[0122] Construct a seepage mathematical model for the fracturing development of horizontal wells in the reservoir based on geological data; by deeply understanding the geological characteristics of the reservoir, such as permeability, porosity, fluid properties, etc., and establishing a mathematical model based on these characteristics to describe the flow of fluids in the reservoir, the influence of fracturing operations on the seepage characteristics of the reservoir is simulated.

[0123] Based on the seepage mathematical model for fracturing development and the fracturing simulation model, establish a simulation model for the movable oil content in the fracturing of horizontal wells in the reservoir;

[0124] Use the simulation model for the movable oil content in fracturing to obtain the well parameters and movable oil content data after fracturing under different formation parameters and different fracturing operation combinations; by simulating, a large amount of well parameters and movable oil content data after fracturing are collected, providing a rich sample for subsequent prediction work.

[0125] Normalize the well parameters after fracturing, formation parameters, and the corresponding movable oil content data to obtain a standard data set for predicting the change in movable oil content before and after fracturing.

[0126] Furthermore, construct a seepage mathematical model for the fracturing development of horizontal wells in the reservoir based on geological data, including:

[0127] Standardize the layers in the obtained geological data and store the layers in a relational database in mdb format to establish a geological database; the layers in the geological data include at least any one of the following: multi-factor raster layer, label information vector layer, porosity distribution layer, permeability distribution layer, saturation distribution layer, sand body thickness distribution layer, effective sand distribution layer, well position coordinate information vector layer;

[0128] After importing the geological data using an automated tool, automatically divide according to the preset grid size and type, divide the target rock block into grids of the same size, and obtain the initial state information of each grid based on the geological data; among them, the size of the grid can be set according to the actual situation of the target rock block. For example, use an automated tool (such as Surpac, mView or Petrel) to import the geological data of the target rock block,

[0129] For rock blocks within a range of dozens of kilometers, the initial grid side length can be set to 100 meters or 200 meters. Local refinement can be carried out in areas with complex geological structures (such as near faults). If the geological structure is relatively simple, one-tenth of the rock block size can be selected as the grid side length for division. Use automated tools (such as Surpac, mView, Petrel, or Hypermesh) for grid division. For complex geological structures, unstructured grid division tools (such as UDEC or Flac3D) can be used, and local refinement can be combined with automated scripts.

[0130] Construct the seepage motion equations for each grid based on the initial state information;

[0131] Obtain the fluid continuity equation based on geological data;

[0132] Substitute the fluid continuity equation into the seepage motion equation to obtain the seepage control equations for each grid;

[0133] Combine the seepage control equations for each grid to obtain the seepage mathematical model for the fracturing development of horizontal wells in the oil reservoir.

[0134] Furthermore, establish a movable oil content prediction model based on machine learning, and use a standard data set to train the movable oil content prediction model, including:

[0135] Determine the influencing factors of the movable oil content; through literature research, expert interviews, laboratory simulations, and geological data analysis, determine the main factors affecting the movable oil content, including but not limited to reservoir permeability, porosity, oil saturation, crude oil viscosity, formation pressure, formation temperature, rock type, and fluid properties, etc.

[0136] Analyze the influencing factors of the movable oil content based on the analytic hierarchy process to obtain the subjective weights; specifically, take the movable oil content as the target layer, the influencing factors as the criterion layer, and the specific indicators of each influencing factor as the index layer to construct a hierarchical structure model; through expert scoring, compare the factors within the same layer pairwise to construct a judgment matrix; use the eigenvalue method or the sum-product method to calculate the weight vector of the judgment matrix and conduct a consistency test; synthesize the weight vectors of each layer to obtain the subjective weights of each influencing factor.

[0137] Analyze the influencing factors of the movable oil content based on the coefficient of variation method to obtain the objective weights. Specifically, perform dimensionless processing on the original data to eliminate the dimension difference; calculate the coefficient of variation of each influencing factor, that is, the ratio of the standard deviation to the mean, which reflects the degree of data dispersion; determine the objective weights of each influencing factor according to the size of the coefficient of variation. The larger the coefficient of variation, the higher the weight.

[0138] Using the weighted average method, calculate the combined weights of the influencing factors of the movable oil content, and based on the combined weights, obtain the influencing indicators of the movable oil content;

[0139] According to the influencing indicators of the movable oil content, establish a BP neural network prediction model and train the BP neural network prediction model;

[0140] Use the ant colony algorithm to optimize the trained BP neural network prediction model to obtain the movable oil content prediction model.

[0141] In another embodiment, according to the standard data set, establish a BP neural network prediction model and train the BP neural network prediction model, including:

[0142] Construct a BP neural network prediction model. According to the standard data set, determine the number of nodes in the input layer, the number of nodes in the hidden layer, and the number of nodes in the output layer. Initialize the connection weights between the input layer and the hidden layer, the connection weights between the hidden layer and the output layer, the thresholds of the hidden layer, and the thresholds of the output layer to obtain the initial structure of the BP neural network prediction model;

[0143] Input the data in the standard data set into the input layer of the BP neural network prediction model. After being processed by the weights and activation functions in the hidden layer, obtain the predicted value of the movable oil content and output it through the output layer. The predicted value output by the BP neural network prediction model is:

[0144]

[0145] where, b k represents the threshold from the hidden layer to the k-th output layer node, O k represents the predicted output value from the hidden layer to the k-th output layer node, H j represents the output value of the j-th node in the hidden layer;

[0146] Calculate the error between the predicted value and the actual value. The specific formula is:

[0147]

[0148] where, E k represents the error of the k-th output layer node, M represents the total number of samples, O mk represents the predicted value of the m-th sample at the k-th output layer node, represents the actual value of the m-th sample at the k-th output layer node;

[0149] According to the error between the predicted value and the actual value, adjust the weights backward;

[0150] Repeat the process of model training until it stops when the preset number of iterations is reached, and obtain the trained BP neural network prediction model.

[0151] Furthermore, according to the error between the predicted value and the actual value, the calculation formulas for reversely adjusting the network connection weights and thresholds are as follows:

[0152]

[0153] ω' jk = ω jk + ηH j E k j = 1, 2, …, l; k = 1, 2, …, m

[0154]

[0155] b' k = b k + E k k = 1, 2, …, m

[0156] In the formula: η is the learning rate, ω′ ij represents the connection weight between the updated input layer and the hidden layer, ω′ jk represents the connection weight between the updated hidden layer and the output layer, a′ j represents the threshold from the updated input layer to the j-th hidden layer node, b′ k represents the threshold from the updated hidden layer to the k-th output layer node, ω ij represents the connection weight between the input layer and the hidden layer before update, ω jk represents the connection weight between the hidden layer and the output layer before update, a j represents the threshold from the input layer to the j-th hidden layer node before update, b k represents the threshold from the hidden layer to the k-th output layer node before update.

[0157] Determine the movable oil content before fracturing by analyzing the samples of the target rock mass, including:

[0158] Perform pyrolysis on multiple samples to obtain the pyrolysis oil content of multiple samples;

[0159] Among them, the samples are sampled by technicians through the method of drilling and sealing coring, that is, using a coring tool with a special sealing liquid in the inner barrel to drill into the rock oil ore body in the target area to obtain multiple samples in the target area. Through the method of sealing coring, it is possible to prevent mud from infecting the samples and at the same time ensure the integrity of the samples.

[0160] Based on the free oil correction coefficient, correct the pyrolysis oil content of the multiple samples to obtain the first movable oil content of the multiple samples, where the free oil correction coefficient is used to compensate for the loss content of free oil in the pyrolysis oil;

[0161] Based on the first movable oil content and the adsorbed oil content of the multiple samples, determine the second movable oil content of the multiple samples, where the second movable oil content is used to represent the actual movable oil content of the samples;

[0162] Based on the second movable oil content and the adsorbed oil content of the multiple samples, generate a resource quantity distribution map of the target rock block, where the resource quantity distribution map is a movable oil content distribution map of the target rock block;

[0163] In the resource quantity distribution map, divide according to the numerical intervals where the second movable oil content of the multiple samples is located to obtain multiple sample sets, and the second movable oil content of the samples within the sample set corresponds to the same numerical interval;

[0164] Based on the formation area, formation thickness, and second movable oil content of the samples in the multiple sample sets, respectively determine the target formation area, target formation thickness, and target second movable oil content of the multiple sample sets;

[0165] Based on the target formation area, target formation thickness, and target second movable oil content of the multiple sample sets, determine the movable oil content before fracturing of the multiple sample sets.

[0166] In another embodiment, use the BP neural network prediction model optimized by the ant colony algorithm to obtain a movable oil content prediction model, including:

[0167] Set initialization parameters: set the maximum number of iterations of ants, the number of ants m, the pheromone concentration, individual optimum, and global optimum;

[0168] Take each weight and threshold in the trained BP neural network prediction model as a parameter to be optimized and equally divide them respectively to form a parameter set, so that the positions randomly placed by m ants each correspond to a set of weights and thresholds in the parameter set;

[0169] Calculate the fitness value of each ant and set the fitness value as the initial pheromone of the ant; among them, the fitness value of the ant is calculated by the reciprocal of the sum of squared errors;

[0170] Use a preset optimization function to calculate the transfer probability of each ant, and obtain the optimal path this time according to the transfer probability of each ant, compare it with the optimal value, if it is optimal, then update the optimal value;

[0171] Compare the optimal value of each ant with the optimal value of the entire ant colony. If it is better, it becomes the new optimal value of the entire ant colony. Sort all paths to select the optimal path.

[0172] Update the pheromone concentration of each ant.

[0173] Compare whether the number of iterations reaches the maximum number of iterations or whether all ants converge on one path. If either is satisfied, output the weights and thresholds corresponding to each dimension in the global optimal value of the last iteration. If not satisfied, continue to calculate the transition probability of each ant and select the optimal path.

[0174] Update the weights and thresholds in the trained BP neural network model with the weights and thresholds corresponding to each dimension in the global optimal value output in the last iteration to obtain a movable oil content prediction model.

[0175] Preferably, the method further includes: evaluating the content change of the target rock block based on the movable oil content before fracturing and the movable oil content after fracturing.

[0176] Furthermore, before making predictions, in order to accurately understand the artificial fracture propagation law and proppant distribution in the oil reservoir of the target rock block, indoor experiments were carried out, specifically including:

[0177] I. Experiment preparation

[0178] Sample preparation:

[0179] Take cores from the oil reservoir of the target rock block and prepare them into thinly interbedded rock samples.

[0180] Equipment preparation:

[0181] Prepare the equipment required for the experiment, including a true triaxial fracturing simulation device, a high-precision CT scanning device, a data acquisition system, etc.

[0182] II. Experiment settings

[0183] Sample installation:

[0184] Install the prepared rock sample into the true triaxial fracturing simulation device to ensure that the sample is fixed stably and the stress is evenly distributed.

[0185] Parameter setting:

[0186] According to the experimental requirements, set the experimental parameters, such as fracturing fluid displacement, viscosity, injection pressure, in-situ stress, etc., to simulate the actual formation conditions.

[0187] Proppant preparation:

[0188] Prepare an appropriate amount of proppant (such as quartz sand, etc.) and determine the particle size and concentration of the proppant according to the experimental plan.

[0189] III. Experimental Procedures

[0190] Loading Stress:

[0191] Apply triaxial stress to the rock sample according to the set parameters to simulate the in-situ stress state in the actual formation.

[0192] Fracturing Fluid Injection:

[0193] Inject the fracturing fluid mixed with fluorescent agent into the wellbore at a lower displacement (5 - 20 mL / min), and record the change of wellhead pressure. After the rock fractures, the wellhead pressure will drop rapidly, indicating the end of the preflush injection stage.

[0194] Proppant-Carrying Fluid Injection:

[0195] Increase the displacement to a larger value (50 mL / min), open the sand outlet valve of the sand tank containing proppant. After the proppant is mixed with the fracturing fluid, it enters the fracturing pipeline, and the proppant-carrying fluid injection stage begins. Continuously inject the slurry until the wellhead pressure rises sharply and then drops, and then stop the pump.

[0196] During the whole process of the proppant-carrying fluid injection stage for Specimen No. 1, 200-type proppant is used to investigate the filling situation of small-particle-size proppant when the laminations are developed; for Specimens No. 2 and No. 3, 200-type proppant is used for the first approximately 120 seconds during the proppant-carrying fluid injection stage, and then 1214-type proppant is used. The maximum cumulative pump injection volume for a single group of experiments is 500 mL, and the maximum proppant dosage is 50 g.

[0197] Data Acquisition and Analysis:

[0198] Use the data acquisition system to record various data during the experiment in real time, such as pressure, displacement, time, etc.

[0199] Use a micro-CT scanner to scan the gray-scale image of the specimen, reconstruct the core through high-precision CT data, and comprehensively analyze and identify the fracture morphology on the surface and inside of the rock sample and the proppant distribution situation by combining the tracer distribution data and the rock sample dissection results.

[0200] Use the VOLUME GRAPHICS STUDIO MAX software to classify and process the gray-scale image of the specimen, analyze its structural morphological characteristics (including fracture area, proppant volume, etc.), then construct a three-dimensional digital core model, and calculate the complexity of the fracture space by using the box dimension method.

[0201] On the other hand, the present invention provides a prediction system for the change of movable oil content before and after fracturing, refer to Figure 2 , including:

[0202] The dataset construction module collects multi-source data of the target rock block. The multi-source data includes geological data and production data, and constructs a standard dataset for predicting the change in the content of movable oil before and after fracturing based on the geological data and production data;

[0203] The model generation module is used to establish a machine learning-based movable oil content prediction model and train the movable oil content prediction model using the standard dataset;

[0204] The input module is used to obtain the formation parameters and fracturing well parameters of the target rock block;

[0205] The prediction module is used to determine the movable oil content before fracturing by analyzing the samples of the target rock block, and use the trained movable oil content prediction model to obtain the movable oil content after fracturing;

[0206] The evaluation module is used to evaluate the content change of the target rock block based on the movable oil content before fracturing and the movable oil content after fracturing.

[0207] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0208] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the change of movable oil content before and after fracturing, characterized in that: include: Collect multi-source data of the target rock block, the multi-source data including geological data and production data, and construct a standard data set for predicting the change of movable oil content before and after fracturing based on the geological data and the production data; Establishing a movable oil content prediction model based on machine learning, and using the standard data set to train the movable oil content prediction model; Determine the movable oil content before fracturing by analyzing samples of the target rock mass; Obtaining target rock formation parameters and fracturing well parameters, inputting the formation parameters and the fracturing well parameters into a trained movable oil content prediction model to obtain movable oil content after fracturing; Based on the movable oil content before fracturing and the movable oil content after fracturing, a content change of the target rock block is evaluated.

2. The method for predicting the change of movable oil content before and after fracturing according to claim 1, characterized in that: A standard data set for predicting the change of movable oil content before and after fracturing is constructed based on the geological data and the production data, including: Constructing a fracturing simulation model according to the geological data and the production data; According to the geological data, a mathematical model of seepage flow in the horizontal well of the oil reservoir is constructed; Based on the hydraulic fracturing development seepage mathematical model and the hydraulic fracturing simulation model, a hydraulic fracturing movable oil content simulation model of a horizontal well in an oil reservoir is established; The fracturing movable oil content simulation model is used to obtain post-fracture well parameters and movable oil content data under different formation parameters and different fracturing operation combinations; The well parameters, formation parameters and corresponding movable oil content data after fracturing are normalized to obtain a standard data set for predicting the change of movable oil content before and after fracturing.

3. The method for predicting the change of movable oil content before and after fracturing according to claim 2, characterized in that: Based on the geological data, a mathematical model of seepage flow in the horizontal well of the oil reservoir is constructed, including: The layers in the acquired geological data are standardized and stored in a relational database in mdb format to establish a geological database; the layers in the geological data include at least any one of the following: a multi-factor grid layer, a label information vector layer, a porosity distribution layer, a permeability distribution layer, a saturation distribution layer, a sand body thickness distribution layer, an effective sand distribution layer, and a well location coordinate information vector layer; After the geological data is imported by an automated tool, the target rock block is automatically divided into grids of the same size according to a preset grid size and type, and initial state information of each grid is obtained based on the geological data; Constructing the infiltration motion equation of each grid based on the initial state information; obtaining a fluid continuity equation based on the geological data; Substituting the fluid continuity equation into the seepage motion equation to obtain the seepage control equation of each grid; The seepage control equations of each grid are combined to obtain the mathematical model of seepage in the fracturing development of horizontal wells in the reservoir.

4. The method for predicting the change of movable oil content before and after fracturing according to claim 1, characterized in that: Establishing a movable oil content prediction model based on machine learning, and using the standard data set to train the movable oil content prediction model, including: Determine the factors affecting movable oil content; The factors affecting movable oil content were analyzed based on the analytic hierarchy process, and the subjective weights were obtained. The factors affecting movable oil content were analyzed based on the coefficient of variation method to obtain objective weights; Calculating the combined weights of the factors affecting the movable oil content by using a weighted average method, and obtaining the movable oil content influencing index based on the combined weights; According to the movable oil content influencing index, a BP neural network prediction model is established, and the BP neural network prediction model is trained; The ant colony algorithm was used to optimize the trained BP neural network prediction model to obtain the movable oil content prediction model.

5. The method for predicting the change of movable oil content before and after fracturing according to claim 4, characterized in that: According to the standard data set, a BP neural network prediction model is established, and the BP neural network prediction model is trained, including: Constructing a BP neural network prediction model, determining the number of nodes in the input layer, the number of nodes in the hidden layer, and the number of nodes in the output layer, initializing the connection weights between the input layer and the hidden layer, the connection weights between the hidden layer and the output layer, the threshold of the hidden layer, and the threshold of the output layer, and obtaining the initial structure of the BP neural network prediction model; The data in the standard data set is input into the input layer of the BP neural network prediction model, and after being processed by the weight and activation function in the hidden layer, the predicted value of the movable oil content is obtained and output through the output layer. The predicted value of the output of the BP neural network prediction model is: Among them, b k represents the threshold from the hidden layer to the kth output layer node, O k represents the predicted output value from the hidden layer to the kth output layer node, H j Represents the output value of the jth node in the hidden layer; Calculate the error between the predicted value and the actual value. The specific formula is: Among them, E k represents the error of the kth output layer node, M represents the total number of samples, O mk represents the predicted value of the mth sample at the kth output layer node, Represents the actual value of the mth sample at the kth output layer node; According to the error between the predicted value and the actual value, the weight is adjusted inversely; The model training process is repeated until the preset number of iterations is reached, and the trained BP neural network prediction model is obtained.

6. The method for predicting the change of movable oil content before and after fracturing according to claim 5, characterized in that: The movable oil content before fracturing is determined by analyzing samples of the target rock, including: Performing thermal decomposition on a plurality of samples to obtain the pyrolysis oil content of the plurality of samples; Based on a free oil correction coefficient, the pyrolysis oil content of the plurality of samples is corrected to obtain a first movable oil content of the plurality of samples, wherein the free oil correction coefficient is used to compensate for the loss content of free oil in the pyrolysis oil; determining a second movable oil content of the plurality of samples based on the first movable oil content and the adsorbed oil content of the plurality of samples, the second movable oil content being used to represent an actual movable oil content of the samples; Based on the second movable oil content and the adsorbed oil content of the plurality of samples, generating a resource distribution map of the target rock block, wherein the resource distribution map is a movable oil content distribution map of the target rock block; In the resource distribution map, the plurality of samples are divided according to the numerical intervals in which the second movable oil contents of the plurality of samples are located, to obtain a plurality of sample sets, wherein the second movable oil contents of the samples in the sample sets correspond to the same numerical interval; Determining target formation areas, target formation thicknesses, and target second movable oil contents of the plurality of sample sets, respectively, based on the formation areas, formation thicknesses, and second movable oil contents of the samples in the plurality of sample sets; The pre-fracture movable oil content of the plurality of sample sets is determined based on the target formation area, the target formation thickness, and the target second movable oil content of the plurality of sample sets.

7. The method for predicting the change of movable oil content before and after fracturing according to claim 5, characterized in that: The ant colony algorithm is used to optimize the trained BP neural network prediction model to obtain the movable oil content prediction model, including: Set initialization parameters: set the maximum number of ant iterations, the number of ants m, pheromone concentration, individual optimality and global optimality; Each weight and threshold in the trained BP neural network prediction model is divided equally as the parameters to be optimized to form a parameter set, so that the positions where the m ants are randomly placed correspond to a set of weights and thresholds in the parameter set; Calculating the fitness value of each ant, and setting the fitness value as the initial pheromone of the ant; wherein the fitness value of the ant is calculated by the inverse of the sum of squares of the errors; The preset optimization function is used to calculate the transfer probability of each ant, and the optimal path is obtained based on the transfer probability of each ant. It is compared with the optimal value. If it is optimal, the optimal value is updated; Compare the optimal value of each ant with the optimal value of the entire ant colony. If it is better, it becomes the new optimal value of the entire ant colony. Sort all paths and select the optimal path. Update the pheromone concentration of each ant; Compare whether the number of iterations reaches the maximum number of iterations or whether all ants converge on one path. If any of the conditions are met, output the weight and threshold corresponding to each dimension in the global optimal value of the last iteration; if not, continue to calculate the transfer probability of each ant and select the optimal path; The weights and thresholds corresponding to each dimension in the global optimal value outputted in the last iteration are used to update the weights and thresholds in the trained BP neural network model to obtain a movable oil content prediction model.

8. The method for predicting the change of movable oil content before and after fracturing according to claim 1, characterized in that: Before prediction, indoor experiments are conducted on the propagation law of artificial fractures and proppant distribution in the target rock oil reservoir, including: (1) Experimental preparation stage: Sample preparation: cores are taken from the target rock block oil reservoir to prepare thin interbedded rock samples; Equipment preparation: Prepare the equipment required for the experiment, including true triaxial fracturing simulation device, high-precision CT scanning equipment, data acquisition system, etc.; (2) Experimental setup phase: Sample installation: Install the prepared rock sample into the true triaxial fracturing simulation device to ensure that the sample is fixed stably and evenly stressed; Parameter setting: set the experimental parameters according to the experimental requirements to simulate the actual formation conditions; Proppant preparation: prepare an appropriate amount of proppant and determine the particle size and concentration of the proppant according to the experimental plan; (3) The experimental steps include: Loading stress: Apply triaxial stress to the rock sample according to the set parameters to simulate the ground stress state in the actual stratum; Fracturing fluid injection: Inject fracturing fluid mixed with fluorescent agent into the wellbore at a displacement of 5-20 mL / min, and record the change in wellhead pressure; after the rock breaks, the wellhead pressure will drop rapidly, that is, the pre-fluid injection stage is over; Sand-carrying fluid injection: Increase the displacement to 50mL / min, open the sand outlet valve of the sand tank filled with proppant, and the proppant and fracturing fluid are mixed and enter the fracturing pipeline, entering the sand-carrying fluid injection stage; continue to inject the mixed slurry until the wellhead pressure rises sharply and then drops, then stop the pump; Data collection and analysis: Use the data collection system to record various data in the experimental process in real time; Use a micrometer CT scanner to scan the grayscale image of the sample, reconstruct the core through high-precision CT data, and combine the tracer distribution data with the rock sample segmentation results to comprehensively analyze and identify the fracture morphology on the surface and inside of the rock sample and the distribution of proppant; VOLUME GRAPHICS STUDIO MAX software was used to classify the grayscale images of the samples and analyze the structural morphological characteristics of the samples. Subsequently, a three-dimensional digital core model was constructed, and the box dimension method was used to calculate the spatial complexity of the fractures.

9. A prediction system for the change of movable oil content before and after fracturing, characterized in that: include: A data set construction module collects multi-source data of the target rock block, wherein the multi-source data includes geological data and production data, and constructs a standard data set for predicting the change of movable oil content before and after fracturing based on the geological data and the production data; A model generation module, used to establish a movable oil content prediction model based on machine learning, and to train the movable oil content prediction model using the standard data set; An input module is used to obtain the formation parameters of the target rock block and the parameters of the fracturing well; A prediction module is used to determine the movable oil content before fracturing by analyzing samples of the target rock block, and to obtain the movable oil content after fracturing by using a trained movable oil content prediction model; The evaluation module is used to evaluate the change of movable oil content of the target rock block based on the movable oil content before fracturing and the movable oil content after fracturing.

Images