Method and system for evaluating pressure-driving potential of low-permeability tight reservoir

Abstract

The invention discloses a method and system for evaluating the pressure-driving potential of a low-permeability tight reservoir, and relates to the technical field of oil and gas field development, in the method, a pressure-driving potential prediction sample set is constructed, and a DRSN-Adaboost model is trained to serve as a pressure-driving potential prediction model; geological parameters, fluid property parameters and pressure driving construction parameters are obtained, and a corresponding pressure driving potential evaluation index prediction result is obtained through the pressure driving potential prediction model; and finally, according to a pressure driving potential evaluation index prediction result and the weight of each pressure driving potential evaluation index, calculating to obtain a pressure driving potential comprehensive score of each pressure driving well group of the target oil reservoir, thereby realizing pressure driving potential evaluation, and guiding an oil field on-site pressure driving well selection decision. According to the scheme provided by the invention, the pressure-driving effects of different oil reservoirs and different schemes can be predicted, the prediction efficiency is high, the time cost and repetitive work of a traditional numerical simulation method can be reduced, and the method has important guiding significance for evaluating the pressure-driving potential of the low-permeability tight oil reservoir.

Description

Technical Field

[0001] The present application relates to the technical field of oil and gas field development, and in particular to a method and system for evaluating the pressure-drive potential of a low-permeability tight oil reservoir. Background Art

[0002] As the exploration and development of oil and gas resources continues to deepen, the proportion of low-permeability tight oil reservoirs in newly added reserves has gradually increased. Compared with conventional oil reservoirs, low-permeability tight oil reservoirs have the characteristics of low porosity, low permeability, and insufficient natural energy. Traditional water drive development technology generally has problems such as high water injection pressure, inability to inject, and insufficient injection. Pressure drive technology quickly injects a large amount of water at high pressure to form a large number of fracture networks around the water injection wells, improve the seepage capacity of the reservoir, and thus expand the scope of water drive; at the same time, it can quickly replenish formation energy, establish an effective displacement pressure system between oil and water wells, increase the oil production capacity of oil wells, and improve development effects. Pressure drive has become an important means to improve the development effect of low-permeability tight oil reservoirs.

[0003] Pressure-drive construction has high investment and great technical risks, but the reservoir characteristics of low-permeability tight oil reservoirs are significantly different, and different well groups have different pressure-drive potentials. Therefore, it is necessary to carry out pressure-drive potential evaluation for the target well group and select the well group with the best potential for pressure-drive construction, so as to obtain the highest return with limited investment. This has important guiding significance for the efficient development of low-permeability tight oil reservoirs.

[0004] At present, the evaluation of pressure-drive potential mainly relies on the reservoir numerical simulation method, that is, a geological model is established for a certain well group in the reservoir and a pressure-drive numerical simulation is carried out to predict the pressure-drive production increase and recovery effect, and realize the pressure-drive potential evaluation. However, this method has the following limitations: (1) Each time, the pressure-drive potential evaluation can only be carried out for a specific well group. To obtain the pressure-drive potential of other well groups, it is necessary to carry out repeated geological modeling and reservoir numerical simulation, which is labor-intensive, time-consuming and labor-intensive.

[0005] (2) The geological model itself has significant uncertainty, but the results obtained by the pressure-driven numerical simulation prediction are certain. The impact of geological uncertainty on the pressure-driven effect is ignored, which can easily lead to wrong decisions at the oil field site.

[0006] Establishing an efficient and accurate pressure drive potential evaluation method is of great significance for well selection decision-making in pressure drive of low-permeability tight oil reservoirs, reducing the trial-and-error costs caused by blind on-site construction, and achieving quality improvement and efficiency enhancement. Summary of the invention

[0007] The purpose of this application is to provide a method and system for evaluating the pressure-drive potential of low-permeability tight oil reservoirs, which can achieve efficient and accurate evaluation of the pressure-drive potential of different low-permeability tight oil reservoir well groups and under different pressure-drive schemes under the premise of considering geological uncertainties.

[0008] To achieve the above objectives, this application provides the following solutions: In a first aspect, the present application provides a method for evaluating the pressure drive potential of a low permeability tight oil reservoir, comprising the following steps: A pressure-drive potential prediction sample set is constructed; the pressure-drive potential prediction sample set includes geological parameters, fluid property parameters, pressure-drive construction parameters, and pressure-drive potential evaluation index simulation results obtained by pressure-drive numerical simulation based on the geological parameters, fluid property parameters, and pressure-drive construction parameters; the pressure-drive potential evaluation index simulation results include simulation values ​​of several pressure-drive potential evaluation indicators; the pressure-drive potential evaluation indicators include cumulative oil, comprehensive water content, oil change rate per ton of water, pressure recovery coefficient, and pressure increase rate per ton of water.

[0009] The pressure-driven potential prediction sample set is used to train the DRSN-Adaboost model to obtain a pressure-driven potential prediction model; the pressure-driven potential prediction model is used to output corresponding pressure-driven potential evaluation index prediction results according to input geological parameters, fluid property parameters and pressure-driven construction parameters.

[0010] For any well group in the target reservoir, the pressure-drive potential prediction model is used to predict the pressure-drive potential evaluation index prediction result of the well group.

[0011] According to the prediction results of the pressure-drive potential evaluation indicators of multiple well groups in the target oil reservoir and the weights of each pressure-drive potential evaluation indicator, a comprehensive score of pressure-drive potential is calculated; the comprehensive score of pressure-drive potential is a comprehensive evaluation result of pressure-drive potential for the target oil reservoir.

[0012] Optionally, constructing a pressure drive potential prediction sample set specifically includes the following steps: Obtain the value ranges of the target reservoir's geological parameters, fluid property parameters, and pressure-drive construction parameters, as well as the probability distribution form, mean, and variance of the geological parameters' porosity, permeability, oil saturation, Young's modulus, Poisson's ratio, maximum horizontal principal stress, and minimum horizontal principal stress.

[0013] Sampling is performed within the value ranges of various geological parameters, various fluid property parameters and various pressure-driven construction parameters to obtain a number of input parameter combinations; the input parameter combinations include three-dimensional matrix samples corresponding to geological parameters, sampled values ​​of fluid property parameters and sampled values ​​of pressure-driven construction parameters.

[0014] For any input parameter combination, pressure-driven numerical simulation is carried out to obtain the simulation results of pressure-driven potential evaluation index under the corresponding input parameter combination.

[0015] The three-dimensional matrix samples corresponding to the geological parameters of the input parameter combination, the sampled values ​​of the fluid property parameters, the sampled values ​​of the pressure-driving construction parameters and the simulated pressure-driving potential evaluation index simulation results are used as a pressure-driving potential prediction sample; a number of the pressure-driving potential prediction samples are gathered together to construct a pressure-driving potential prediction sample set.

[0016] Optionally, sampling is performed within the value ranges of various geological parameters, various fluid property parameters, and various pressure-driven construction parameters, respectively, specifically including the following steps: For geological parameters, the Monte Carlo method is used to perform random sampling within the value range of the uncertain parameters in the geological parameters to obtain three-dimensional matrix samples corresponding to the geological parameters.

[0017] For the fluid property parameters and pressure-driven construction parameters, the Latin hypercube sampling method is used to perform sampling within the value range of the corresponding parameters to obtain the sampling values ​​of the fluid property parameters and pressure-driven construction parameters.

[0018] Optionally, for any input parameter combination, a pressure-driven numerical simulation is carried out to obtain a simulation result of a pressure-driven potential evaluation index under the corresponding input parameter combination, specifically including the following steps: A reservoir geological model is constructed according to three-dimensional matrix samples corresponding to geological parameters generated by sampling; the attribute values ​​of each grid in the reservoir geological model are derived from elements of the three-dimensional matrix samples.

[0019] If the target reservoir has natural fractures, a natural fracture model is obtained through natural fracture modeling according to the length, orientation and linear density of the natural fractures; and the natural fracture model is added to the reservoir geological model to obtain a reservoir geological model containing natural fractures.

[0020] If natural fractures are not developed in the target reservoir, no special treatment is performed and the next step is directly carried out.

[0021] Based on the reservoir geological model or the reservoir geological model containing natural fractures, according to the sampling values ​​of the pressure-driving construction parameters, a pressure-driving fracture network expansion simulation is carried out to obtain a fracture network model after the pressure-driving construction.

[0022] Based on the fracture network model after the pressure-driven construction, the fracture network morphology is finely characterized by using an unstructured grid to obtain a matrix grid model and a fracture grid model.

[0023] Based on the matrix grid model and the fracture grid model, a pressure-driven numerical simulation is performed according to the sampled values ​​of the fluid property parameters to obtain a simulation result of a pressure-driven potential evaluation index; based on the matrix grid model and the fracture grid model, when a pressure-driven numerical simulation is performed according to the sampled values ​​of the fluid property parameters, corresponding stress sensitivity curves are set for different types of grids, and the difference in permeability changes of the matrix grid and the fracture grid under different stresses is simulated through the stress sensitivity curve, thereby simulating different flow characteristics in the matrix grid and the fracture grid during the pressure-driven development process.

[0024] Optionally, the DRSN-Adaboost model includes a data processing layer, a deep residual shrinkage neural network and an Adaboost module; the data processing layer is used to process input data; during the data processing process, through dimensionality increase and fusion operations, a feature channel matrix composed of several three-dimensional matrices is generated according to various parameters in the above-mentioned pressure-driven potential prediction sample as the input of the deep residual shrinkage neural network; the deep residual shrinkage neural network is composed of an input layer, a convolutional layer, a residual module, a soft threshold layer, a pooling layer, a fully connected layer and an output layer; the Adaboost module uses a deep residual shrinkage neural network as a weak learner of the Adaboost module, obtains multiple weak learners through iterative training, and forms a strong learner through weight combination.

[0025] Optionally, through dimensionality increase and fusion operations, a feature channel matrix composed of several three-dimensional matrices is generated according to various parameters in the above pressure-driven potential prediction sample as the input of the deep residual shrinkage neural network, which specifically includes the following steps: The one-dimensional data of various parameters in the above-mentioned pressure-drive potential prediction sample are upgraded to obtain multiple three-dimensional data matrices; specifically, for the one-dimensional parameters and fluid property parameters in the geological parameters, each parameter generates a three-dimensional data matrix consistent with the spatial distribution of the reservoir geological model; all values ​​in the three-dimensional data matrix are the corresponding one-dimensional data parameter values; for the pressure-drive construction parameters, each parameter generates a three-dimensional data matrix consistent with the spatial distribution of the reservoir geological model, and in each three-dimensional data matrix of the pressure-drive construction parameters, the grid positions of all wells in the reservoir model and the well trajectories pass through are filled with the pressure-drive construction parameter values ​​of the wells implementing the pressure-drive at the corresponding matrix element positions in the matrix space, and the values ​​of other positions in the matrix are filled with 0; The multiple three-dimensional data matrices obtained by dimensionality upgrading are fused with the three-dimensional matrix of geological parameters to form a feature channel matrix composed of several three-dimensional matrices as the input of the deep residual shrinkage neural network.

[0026] Optionally, a deep residual shrinkage neural network is used as a weak learner of the Adaboost module, multiple weak learners are obtained through iterative training, and a strong learner is formed through weight combination, which specifically includes the following steps: The deep residual shrinkage neural network is set as the initial weak learner, and the deep residual shrinkage neural network is trained using the pressure-driven potential prediction sample set to generate the first deep residual shrinkage neural network weak learner.

[0027] Iterative update, each iteration obtains a deep residual shrinkage neural network weak learner, completes the training of the preset number of iterations, and obtains a total of M deep residual shrinkage neural network weak learners, and the M deep residual shrinkage neural network weak learners are combined by weights to obtain a deep residual shrinkage neural network strong learner; M is the preset number of iterations, the input of the deep residual shrinkage neural network strong learner is the characteristic channel matrix obtained by the data processing layer, and the output of the deep residual shrinkage neural network strong learner is the prediction result of the pressure-driven potential evaluation index.

[0028] Optionally, M deep residual shrinkage neural network weak learners are weighted together to obtain a deep residual shrinkage neural network strong learner according to the following formula: .

[0029] in, G ( x ) is a deep residual shrinkage neural network strong learner, Delta is the number of iterations, α Delta For the Delta The weights of the weak learners of the optimal deep residual shrinkage neural network for the iteration, T Delta ( x ) is the Delta Optimal deep residual shrinkage neural network weak learner with iterations.

[0030] Optionally, the comprehensive score of the pressure drive potential of multiple well groups in the target reservoir is calculated according to the following formula: .

[0031] in, W i The target reservoir i Comprehensive score of pressure drive potential of each well group, w o , w pi , w pv , w cui , wr is the weight of each pressure drive potential evaluation index, x o , x pi , x pv , x cui , x r is the normalized value of each pressure-driven potential evaluation index.

[0032] In a second aspect, the present application provides a system for evaluating the pressure drive potential of a low-permeability tight oil reservoir, comprising: The pressure-drive potential prediction sample set construction module is used to construct the pressure-drive potential prediction sample set; the pressure-drive potential prediction sample set includes geological parameters, fluid property parameters, pressure-drive construction parameters, and pressure-drive potential evaluation index simulation results obtained by pressure-drive numerical simulation based on the geological parameters, fluid property parameters, and pressure-drive construction parameters; the pressure-drive potential evaluation index simulation results include simulation values ​​of several pressure-drive potential evaluation indicators; the pressure-drive potential evaluation indicators include cumulative oil, comprehensive water content, oil change rate per ton of water, pressure recovery coefficient, and pressure increase rate per ton of water.

[0033] The pressure-driven potential prediction model training module is used to train the DRSN-Adaboost model using the pressure-driven potential prediction sample set to obtain the pressure-driven potential prediction model; the pressure-driven potential prediction model is used to output the corresponding pressure-driven potential evaluation index prediction results based on the input geological parameters, fluid property parameters and pressure-driven construction parameters.

[0034] The target reservoir pressure drive potential prediction module is used to predict the pressure drive potential evaluation index prediction results of any well group in the target reservoir using the pressure drive potential prediction model.

[0035] The pressure-drive potential comprehensive score calculation module is used to calculate the pressure-drive potential comprehensive score based on the pressure-drive potential evaluation index prediction results of multiple well groups in the target oil reservoir and the weights of each pressure-drive potential evaluation index; the pressure-drive potential comprehensive score is the comprehensive evaluation result of the pressure-drive potential for the target oil reservoir.

[0036] According to the specific embodiments provided in this application, this application discloses the following technical effects: The present application provides a method and system for evaluating the pressure-drive potential of a low-permeability tight oil reservoir. In the method, a pressure-drive potential prediction sample set is first constructed, which includes a number of geological parameters, fluid property parameters and pressure-drive construction parameters, and pressure-drive potential evaluation index results simulated according to these parameters; then, a DRSN-Adaboost model is trained using the pressure-drive potential prediction sample set to obtain a pressure-drive potential prediction model; then, the pressure-drive potential prediction model is used to predict a number of pressure-drive potential evaluation index results of each well group in the target oil reservoir; finally, based on the pressure-drive potential evaluation index results of multiple well groups in the target oil reservoir and the weights of each pressure-drive potential evaluation index, a comprehensive score of the pressure-drive potential of multiple well groups in the target oil reservoir is calculated, and the score is the comprehensive evaluation result of the pressure-drive potential for the target oil reservoir.

[0037] The solution provided in this application uses the DRSN-AdaBoost model, which combines the powerful feature extraction capabilities of the deep residual shrinkage neural network and the advantages of AdaBoost in improving model performance, and can effectively capture key information in the input data. The samples used for the training model cover various uncertain parameters in the geological model of low-permeability tight oil reservoirs, and comprehensively consider the influence of different geological factors, fluid properties and pressure-driving construction parameters on the pressure-driving effect, thereby improving the accuracy of the prediction results. The trained DRSN-Adaboost model can predict the pressure-driving effects under different well groups and different pressure-driving schemes, and the prediction time is short, which greatly reduces the time cost and tedious workload of repeated numerical simulations, and has important guiding significance for the evaluation of the pressure-driving potential and well selection strategy of low-permeability tight oil reservoirs. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 A flow chart of a method for evaluating the pressure-drive potential of a low-permeability tight oil reservoir provided in one embodiment of the present application.

[0040] Figure 2 A schematic diagram of the structure of the DRSN-Adaboost model in a method for evaluating the pressure-drive potential of a low-permeability tight oil reservoir provided in one embodiment of the present application.

[0041] Figure 3 A schematic diagram of a natural fracture model simulated in a method for evaluating the pressure drive potential of a low-permeability tight oil reservoir provided in one embodiment of the present application.

[0042] Figure 4 This is a schematic diagram of a pressure-driven fracture network simulated in a method for evaluating pressure-driven potential of a low-permeability tight oil reservoir provided in one embodiment of the present application.

[0043] Figure 5 A schematic diagram of the oil saturation field after one year of simulated production in a method for evaluating the pressure drive potential of a low-permeability tight oil reservoir provided in one embodiment of the present application.

[0044] Figure 6 A schematic diagram of the functional modules of a system for evaluating the pressure-drive potential of a low-permeability tight oil reservoir provided in one embodiment of the present application.

[0045] Figure 7 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0047] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0048] In an exemplary embodiment, Figure 1 As shown, a method for evaluating the pressure drive potential of a low permeability tight oil reservoir is provided, comprising the following steps: A1. Constructing a pressure-drive potential prediction sample set; the pressure-drive potential prediction sample set includes geological parameters, fluid property parameters, pressure-drive construction parameters, and pressure-drive potential evaluation index simulation results obtained by pressure-drive numerical simulation based on the geological parameters, fluid property parameters, and pressure-drive construction parameters; the pressure-drive potential evaluation index simulation results include simulation values ​​of several pressure-drive potential evaluation indicators; the pressure-drive potential evaluation indicators include cumulative oil, comprehensive water content, oil change rate per ton of water, pressure recovery coefficient, and pressure increase rate per ton of water.

[0049] In this embodiment, step A1 specifically includes the following steps: A11. Obtain the value ranges of the geological parameters, fluid property parameters and pressure-driven construction parameters of the target reservoir, as well as the probability distribution form, mean and variance of the porosity, permeability, oil saturation, Young's modulus, Poisson's ratio, maximum horizontal principal stress and minimum horizontal principal stress in the geological parameters.

[0050] The geological parameters are mainly obtained through logging data, coring data and core experiments of oil and water wells, including porosity, permeability, oil saturation, Young's modulus, Poisson's ratio, maximum horizontal principal stress, minimum horizontal principal stress, rock compression coefficient, thickness, length of natural fractures, natural fracture orientation and natural fracture line density.

[0051] Fluid property parameters include oil viscosity, water viscosity, oil density, water density, oil compressibility, water compressibility, and relative permeability curve.

[0052] The pressure drive construction parameters are mainly obtained from the well groups that have been pressure driven in the target oil reservoir, including the pressure drive injection volume, injection displacement, well blocking time, post-pressure work system, pressure drive well location, and pressure drive well body trajectory.

[0053] For some old wells, it is necessary to obtain the historical production dynamics of the old wells, which include daily water production, daily oil production, water content and bottom hole flow pressure. If fracturing construction has been carried out in the old well, the actual fracturing data is also required, which includes the amount of fracturing fluid, type of fracturing fluid, amount of sand added, and pumping pressure curve of fracturing construction.

[0054] For the geological parameters, fluid property parameters and pressure-driven construction parameters obtained above, the value range of each parameter, as well as the probability distribution form, mean and variance of the porosity, permeability, oil saturation, Young's modulus, Poisson's ratio, maximum horizontal principal stress and minimum horizontal principal stress in the geological parameters are statistically determined.

[0055] A12. Sampling is performed within the value ranges of various geological parameters, various fluid property parameters and various pressure-driven construction parameters to obtain a plurality of input parameter combinations; the input parameter combinations include three-dimensional matrix samples corresponding to geological parameters, sampled values ​​of fluid property parameters and sampled values ​​of pressure-driven construction parameters.

[0056] In this embodiment, in step A12, for geological parameters (including porosity, permeability, oil saturation, Young's modulus, Poisson's ratio, maximum horizontal principal stress and minimum horizontal principal stress), based on the probability distribution type, mean and standard deviation of each parameter, a Monte Carlo method is used to perform random sampling within the value range of the uncertain parameters in the geological parameters to obtain a plurality of three-dimensional matrix samples corresponding to the uncertain parameters in the geological parameters. Through this sampling method, a plurality of three-dimensional random distribution fields that conform to the corresponding probability distribution are generated for each uncertain parameter, and these attribute distributions are represented in matrix form, and the matrix size corresponds to the number of three-dimensional grids of the pressure-driven numerical simulation model, so as to ensure the spatial consistency of the parameter distribution and the model grid.

[0057] For the fluid property parameters and pressure-driven construction parameters, the Latin hypercube sampling method is used to perform sampling within the value range of the corresponding parameters to obtain the sampling values ​​of the fluid property parameters and pressure-driven construction parameters.

[0058] A13, for any input parameter combination, carry out pressure-driven numerical simulation to obtain the pressure-driven potential evaluation index simulation result under the corresponding input parameter combination. In this embodiment, step A13 specifically includes the following steps: A131. Construct a reservoir geological model based on the three-dimensional matrix samples corresponding to the geological parameters generated by sampling; the attribute values ​​of each grid in the reservoir geological model are derived from the elements of the three-dimensional matrix samples. These attributes include porosity, permeability, oil saturation, formation pressure, Young's modulus, Poisson's ratio, maximum horizontal principal stress field and minimum horizontal principal stress field; If the target oil reservoir has natural fractures, step S132 is executed; if the target oil reservoir has no natural fractures, no special processing is performed and step S134 is directly executed.

[0059] A132. A natural fracture model is obtained by natural fracture modeling according to the natural fracture length, natural fracture orientation and natural fracture line density.

[0060] A133. Add the natural fracture model to the reservoir geological model to obtain the reservoir geological model containing natural fractures.

[0061] A134. Based on the reservoir geological model or the reservoir geological model containing natural fractures, according to the sampling values ​​of the pressure-driving construction parameters, a pressure-driving fracture network expansion simulation is carried out to obtain a fracture network model after the pressure-driving construction.

[0062] A135. Based on the fracture network model after the pressure-driven construction, an unstructured grid is used to finely characterize the fracture network morphology to obtain a matrix grid model and a fracture grid model.

[0063] A136. Based on the matrix grid model and the fracture grid model, a pressure drive numerical simulation is performed according to the fluid property parameters to obtain simulation results of pressure drive potential evaluation indicators; including the cumulative oil, comprehensive water content and oil change rate per ton of water after one year of pressure drive production, as well as the pressure recovery coefficient and pressure increase rate per ton of water after pressure drive.

[0064] When performing pressure-driven numerical simulation based on the matrix grid model and the fracture grid model and according to the sampled values ​​of the fluid property parameters, corresponding stress sensitivity curves are set for different types of grids, and the difference in permeability changes of the matrix grid model and the fracture grid model under different stresses is simulated through the stress sensitivity curves, thereby simulating different flow characteristics in the matrix grid and the fracture grid during the pressure-driven development process.

[0065] A14. The three-dimensional matrix samples corresponding to the geological parameters of the input parameter combination, the sampled values ​​of the fluid property parameters, the sampled values ​​of the pressure-driving construction parameters and the simulated pressure-driving potential evaluation index simulation results are taken as a pressure-driving potential prediction sample; several of the pressure-driving potential prediction samples are combined together to construct a pressure-driving potential prediction sample set.

[0066] The above-mentioned Pressure Recovery Coefficient (PRC) indicates the degree of recovery of formation pressure after pressure drive, and the Pressure Recharge Rate per Ton of Water (PRR) indicates the pressure increase caused by injecting one ton of water. The calculation formula is as follows: .

[0067] .

[0068] in, is the formation pressure after pressure drive, is the original reservoir pressure, is the formation pressure before pressure driving, is the pressure drive water injection volume, is the density of injected water.

[0069] A2. Using the pressure-driven potential prediction sample set, the DRSN-Adaboost model is trained to obtain a pressure-driven potential prediction model; the pressure-driven potential prediction model is used to output the corresponding pressure-driven potential evaluation index prediction results based on the input geological parameters, fluid property parameters and pressure-driven construction parameters. The DRSN-Adaboost model is an Adaboost integrated model based on the deep residual network (DRSN). The model uses the deep residual network (DRSN) as a weak learner and integrates it through the framework of the Adaboost algorithm to improve the overall performance of the model.

[0070] Specifically, if Figure 2 As shown, the DRSN-Adaboost model includes a data processing layer, a deep residual shrinkage neural network and an Adaboost module.

[0071] The data processing layer is used to process the input data. During the data processing, through dimensionality increase and fusion operations, a feature channel matrix composed of several three-dimensional matrices is generated according to various parameters in the above-mentioned pressure-driven potential prediction samples as the input of the deep residual shrinkage neural network.

[0072] The deep residual shrinkage neural network consists of an input layer, a convolutional layer, a residual module, a soft threshold layer, a pooling layer, a fully connected layer, and an output layer.

[0073] The Adaboost module uses a deep residual shrinkage neural network as the weak learner of the Adaboost module. Multiple weak learners are obtained through iterative training, and a strong learner is formed through weight combination.

[0074] Through dimensionality increase and fusion operations, according to various parameters in the above pressure-driven potential prediction sample, a feature channel matrix composed of several three-dimensional matrices is generated as the input of the deep residual shrinkage neural network, which specifically includes the following steps: The one-dimensional data of various parameters in the above-mentioned pressure-driven potential prediction sample are upgraded to obtain multiple three-dimensional data matrices; specifically, for the one-dimensional parameters and fluid property parameters in the geological parameters, each parameter generates a three-dimensional data matrix consistent with the spatial distribution of the reservoir geological model; all values ​​in the three-dimensional data matrix are the corresponding one-dimensional data parameter values; for the pressure-driven construction parameters, each parameter generates a three-dimensional data matrix consistent with the spatial distribution of the reservoir geological model, and in each three-dimensional data matrix of the pressure-driven construction parameters, the grid positions passed by the well locations and well trajectories of all wells in the reservoir model and the corresponding matrix element positions in the matrix space are filled with the pressure-driven construction parameter values ​​of the wells implementing pressure-driven, and the values ​​of other positions in the matrix are filled with 0.

[0075] The multiple three-dimensional data matrices obtained by dimensionality upgrading are fused with the three-dimensional matrix of geological parameters to form a feature channel matrix composed of several three-dimensional matrices as the input of the deep residual shrinkage neural network.

[0076] A deep residual shrinkage neural network is used as the weak learner of the Adaboost module. Multiple weak learners are obtained through iterative training, and a strong learner is formed through weight combination. The specific steps include: The deep residual shrinkage neural network is set as the initial weak learner, and the deep residual shrinkage neural network is trained using the pressure-driven potential prediction sample set to generate the first deep residual shrinkage neural network weak learner.

[0077] Iterative update, each iteration obtains a deep residual shrinkage neural network weak learner, completes the training of the preset number of iterations, and obtains a total of M deep residual shrinkage neural network weak learners, and the M deep residual shrinkage neural network weak learners are combined by weights to obtain a deep residual shrinkage neural network strong learner; M is the preset number of iterations, the input of the deep residual shrinkage neural network strong learner is the characteristic channel matrix obtained in the data processing layer, and the output of the deep residual shrinkage neural network strong learner is the prediction result of the pressure-driven potential evaluation index.

[0078] Specifically in this embodiment, M deep residual shrinkage neural network weak learners are combined by weights according to the following formula to obtain a deep residual shrinkage neural network strong learner: .

[0079] in, G ( x ) is a deep residual shrinkage neural network strong learner, Delta is the number of iterations, α Delta For the Delta The weights of the weak learners of the optimal deep residual shrinkage neural network for the iteration, T Delta ( x ) is the Delta Optimal deep residual shrinkage neural network weak learner with iterations.

[0080] A3. For any well group in the target reservoir, the pressure-drive potential prediction model is used to predict the pressure-drive potential evaluation index prediction results of the well group.

[0081] The scheme formed by combining the geological parameters, fluid property parameters and pressure-drive construction parameters of the target reservoir is used as input, and the pressure-drive potential evaluation index results of each scheme are predicted by the pressure-drive potential prediction model to obtain the cumulative oil, pressure recovery coefficient, pressure increase rate per ton of water, comprehensive water content and oil exchange rate per ton of water. The results of all schemes are statistically analyzed to obtain the probability distribution of each indicator result of the well group, as well as the expected value and confidence interval, so as to quantify the uncertainty and risk of the simulation results and evaluate the pressure-drive potential of the well group.

[0082] A4. Calculate a comprehensive pressure-drive potential score based on the pressure-drive potential evaluation index prediction results of multiple well groups in the target oil reservoir and the weights of each pressure-drive potential evaluation index; the comprehensive pressure-drive potential score is a comprehensive evaluation result of the pressure-drive potential for the target oil reservoir.

[0083] Each pressure-driven potential evaluation index is standardized, and then the weight of each index is determined. The standardized index values ​​are multiplied by the corresponding weights, and the sum is calculated to obtain the comprehensive score of the pressure-driven potential. In this way, an evaluation system covering cumulative oil, pressure recovery coefficient, pressure increase rate per ton of water, comprehensive water content and oil change rate per ton of water is established to calculate the comprehensive score of the pressure-driven potential.

[0084] The above method for determining the weight of each indicator adopts a comprehensive method of hierarchical analysis, entropy weight method and Delphi method. For example, the subjective weight is first determined by the hierarchical analysis method, and the judgment matrix is ​​constructed according to the experience and knowledge of experts. The judgment matrix is ​​checked for consistency to ensure the rationality of the judgment matrix. The weight of each indicator is calculated by the eigenvalue method or normalization to obtain the subjective weight; the subjective weight is corrected by the Delphi method to obtain the corrected subjective weight; then the objective weight is determined by the entropy weight method, the evaluation indicator data is standardized, the standardized proportion matrix is ​​calculated, the information entropy of each indicator is calculated according to the proportion matrix, the redundancy of the information entropy is calculated, and then the objective weight of each attribute is calculated according to the redundancy of the information entropy. The subjective weight and the objective weight are normalized, and the comprehensive weight is calculated by the weighted average method.

[0085] The above comprehensive weight calculation method is as follows: .

[0086] in, is the comprehensive weight of a single indicator, is the subjective weight corrected by the Delphi method, is the objective weight determined by the entropy weight method, is the proportion of subjective weight.

[0087] Specifically in this embodiment, the comprehensive score of the pressure drive potential of multiple well groups in the target reservoir is calculated according to the following formula: .

[0088] in, W i The target reservoir i Comprehensive score of pressure drive potential of each well group, w o , w pi , w pv , w cui , w r is the weight of each pressure drive potential evaluation index, x o , x pi , x pv , x cui , x r is the normalized value of each pressure drive potential evaluation index. According to the above-mentioned comprehensive scores of pressure drive potential of multiple well groups in the target reservoir, the well group for pressure drive can be preferably selected.

[0089] It is worth noting that the above comprehensive moisture content needs to be reversed when normalized. The lower the value, the worse the effect. This ensures that the meaning of each indicator is consistent when conducting a comprehensive evaluation, ensuring the rationality and consistency of the model prediction results. The normalized calculation formula for the comprehensive moisture content is as follows: .

[0090] The effectiveness of the method for evaluating the pressure-driven potential of a low-permeability tight oil reservoir provided in the present application is illustrated below by taking a specific example. The reservoir type to be studied in this embodiment is a low-permeability tight oil reservoir.

[0091] First, the geological parameters, fluid property parameters and pressure-driven construction parameters of a low-permeability tight oil reservoir that implements pressure-driven are collected, as well as the probability distribution form, mean and variance of porosity, permeability, oil saturation, Young's modulus, Poisson's ratio, maximum horizontal principal stress and minimum horizontal principal stress in the geological parameters. The Monte Carlo method is used to randomly sample within the value range of the geological uncertainty parameters, and several three-dimensional random distribution fields that conform to the corresponding probability distribution are generated for each parameter; for other parameters in the geological parameters, including thickness, well spacing, well pattern density, length of natural fractures, natural fracture orientation and natural fracture line density, as well as fluid property parameters and pressure-driven construction parameters, the Latin hypercube sampling method is used to sample within the corresponding value range. The multiple geological parameters, fluid property parameters and pressure-driven construction parameters are combined to form multiple pressure-driven numerical simulation schemes.

[0092] In this embodiment, permeability is taken as an example, and 200 sets of permeability data are collected through oil and water wells. The permeability range is 0.5~65mD, and the data obeys the log-normal distribution. The log-normal distribution has a mean of 2.5 and a standard deviation of 0.75. 1200 three-dimensional permeability distribution field matrices are sampled by the Monte Carlo method, and the three-dimensional data distribution still obeys the log-normal distribution. The one-dimensional parameters, fluid property parameters and pressure-driven construction parameters in the geological parameters are also collected by Latin hypercube sampling. 1200 sets of data are also collected. Under the condition of satisfying the constraints, the geological parameters and pressure-driven construction parameters are combined to generate 1200 sets of different pressure-driven numerical simulation schemes.

[0093] Through the above-mentioned pressure-driven numerical simulation scheme, a geological model is established according to the geological parameters therein, and multiple pressure-driven numerical simulation models are constructed in combination with the pressure-driven construction parameters.

[0094] The above geological model and natural fracture model are established by Petrel software, the expansion of hydraulic fracture and pressure-driven fracture is simulated by Kinetix hydraulic fracturing production integrated plug-in, and the pressure-driven numerical simulation is realized by calling Intersect simulator. The parameters of one of the numerical simulation models in this example are shown in Table 1: Table 1 Numerical model parameters

[0095] The simulated natural fracture model is as follows: Figure 3 The model includes one pressure-driven well: INJ1, and four production wells: PROD1, PROD2, PROD3, and PROD4. The pressure-driven construction parameters in this scheme are: pressure-driven injection volume 20,000 m 3 , injection displacement 1m 3 / min, the well was blocked for 30 days, and the simulated pressure-driven fracture network was as follows Figure 4 shown.

[0096] The original reservoir pressure of INJ1 in the well group was 32MPa. After two years of production of the four oil wells PROD1, PROD2, PROD3, and PROD4, the reservoir pressure became 24MPa. Then, pressure drive construction was carried out on INJ1 well, and 20,000m of water was injected into the pressure drive. 3 After that, the reservoir pressure recovered to 28MPa, the pressure recovery coefficient was 0.875, and the pressure increase rate per ton of water was 0.0002MPa / t. Subsequently, the four production wells in the well group continued to produce for one year. The oil saturation field after one year was as follows: Figure 5 As shown in the figure, the well group has accumulated 4088 tons of oil, a comprehensive water content of 68.7%, and an oil exchange rate of 0.204 tons of water. The index results of other reservoir numerical simulation schemes can also be obtained.

[0097] A pressure-driven numerical simulation is carried out for each scheme to obtain the corresponding cumulative oil, pressure recovery coefficient, pressure increase rate per ton of water, comprehensive water content and oil change rate per ton of water. The geological parameters, fluid property parameters and pressure-driven construction parameters corresponding to each numerical simulation result are sorted out to form a sample set for pressure-driven potential prediction.

[0098] The input data is processed through the data processing layer in the DRSN-Adaboost model.

[0099] The input layer fuses the input one-dimensional data with the three-dimensional matrix corresponding to each uncertain parameter of the geological parameter by dimensional upgrading to form a three-dimensional data matrix. Specifically, the one-dimensional data dimensional upgrading method generates a three-dimensional data matrix consistent with the grid model space for the one-dimensional parameters and fluid property parameters in the geological parameters. In this embodiment, the matrix form is 30×30×10, and all the values ​​of the matrix are the corresponding parameter values; for the pressure-driven construction parameters, each parameter generates a three-dimensional data matrix consistent with the spatial distribution of the reservoir geological model. In each three-dimensional data matrix of the pressure-driven construction parameters, the grid positions of all wells in the reservoir model and the grid positions passed by the well trajectories are filled with the pressure-driven construction parameter values ​​of the well INJ1 where the pressure drive is implemented, and the values ​​of other positions in the matrix are filled with 0. These three-dimensional data matrices will jointly constitute the characteristic channel matrix as the input of the deep residual shrinkage network.

[0100] The results of the pressure-driven potential evaluation index are predicted by training the deep residual shrinkage network in the DRSN-Adaboost model. First, the input feature channel matrix that integrates geological parameters, fluid property parameters and pressure-driven construction parameters is extracted through the convolution layer. The convolution layer uses multiple convolution kernels to perform sliding calculations on the input data to capture high-order features at different spatial scales. In order to enhance the expressiveness of the model and avoid the simple linear combination generated by the convolution operation, the network introduces nonlinear activation functions (such as ReLU, Sigmoid or Tanh) after the convolution layer. These nonlinear functions are used to activate the features after convolution to enhance the nonlinear expression ability of the model. By directly adding the input to the output, a "jump connection" is realized, so that the network can still retain the characteristic information of the input geological parameters when performing multi-layer calculations, thereby improving the training efficiency of the deep network and enhancing the expressiveness of the model. Then the data enters the soft threshold layer, which further strengthens the effective features and reduces the influence of invalid features by screening the features and reducing the weights; in order to reduce the amount of calculation and model complexity, the network introduces a pooling layer at a specific layer. The pooling layer reduces the dimension of the data matrix output by the convolution layer through downsampling operations; based on the data matrix after the pooling layer reduces the dimension, the data is flattened into a one-dimensional vector through the fully connected layer. Finally, in the prediction process, the network processes the input data in sequence through forward propagation, and directly outputs the pressure-driven potential evaluation index results through the linear layer in the output layer after the fully connected layer. During the training process, the network learns feature patterns through forward propagation and back propagation, and uses loss functions such as mean square error (MSE) to calculate the difference between the predicted value and the true value. The back propagation algorithm continuously optimizes the network weights to gradually reduce the loss function and improve the prediction ability. After sufficient training, the network can output the final pressure-driven potential evaluation index results based on the new sample data input.

[0101] The deep residual shrinkage network is used as a weak learner of Adaboost. Multiple weak learners are obtained through training iterations, and a stronger prediction model is obtained by combining them. The specific steps include: setting the deep residual shrinkage network as the initial weak learner, training the deep residual shrinkage network with a pressure-driven potential prediction sample set including geological parameters, fluid property parameters, pressure-driven construction parameters and pressure-driven potential index results, and generating the first weak learner; iterative updating, obtaining a weak learner in each iteration, and presetting the number of iterations to 3 times, obtaining a total of 3 weak learners, and combining the 3 weak learners through weights to obtain a strong learner; the input of the strong learner is the feature channel matrix obtained by the data processing layer, and the output of the strong learner is the pressure-driven potential evaluation index result.

[0102] In this example, a pressure drive construction is carried out on an oil field, and its pressure drive potential is evaluated by the method of this embodiment. The scheme formed by combining geological parameters, fluid property parameters and pressure drive construction parameters is used as input, and the results of each scheme are predicted by the trained DRSN-Adaboost model to obtain the cumulative oil, pressure recovery coefficient, pressure replenishment index, comprehensive water content and ton water oil change rate index results of each scheme. Affected by the uncertainty of geological parameters, the sample scheme results obtained have certain probability distribution characteristics. All results are statistically analyzed to obtain the probability distribution of each indicator result of the oil field.

[0103] Under different geological and construction parameter conditions, the expected values ​​and confidence intervals of key indicators such as cumulative oil, pressure recovery coefficient, pressure increase rate per ton of water, comprehensive water content and oil change rate per ton of water can reflect the statistical characteristics and uncertainty of the pressure drive effect. Through the probability distribution of each indicator result of the oil field, the pressure drive potential of the oil field can be effectively evaluated, thus providing a scientific basis for the optimization and decision-making of the pressure drive construction plan.

[0104] In this embodiment, the prediction results of 800 samples from 32 different well groups in the oil field are standardized, and the comprehensive weight of each indicator is calculated using the hierarchical analysis method, entropy weight method and Delphi method. In this embodiment, based on the experience and knowledge of experts, a judgment matrix is ​​constructed as shown below: .

[0105] The maximum eigenvalue and corresponding eigenvector of the judgment matrix are calculated, and after consistency test, it can be known that the judgment matrix is ​​acceptable. After normalizing the eigenvector, the subjective weight of each indicator is obtained. w sub , as shown below: .

[0106] The weights were modified and normalized by the Delphi method to obtain the subjective weights. w sub , as shown below: .

[0107] The objective weight of each indicator is calculated by entropy weight method w obj , as shown below: .

[0108] The comprehensive weight of the cumulative oil, pressure recovery coefficient, pressure increase rate per ton of water, comprehensive water content and oil exchange rate per ton of water of the oil field is calculated. w j , as shown below: .

[0109] That is, the comprehensive weights of cumulative oil, pressure recovery coefficient, pressure increase rate per ton of water, comprehensive water content and oil exchange rate per ton of water are 0.425, 0.325, 0.125, 0.075 and 0.05 respectively. The average values ​​of the cumulative oil, pressure recovery coefficient, pressure increase rate per ton of water, comprehensive water content and oil exchange rate per ton of water predicted by the oil field obtained by the previous steps of this embodiment are 0.75, 0.65, 0.22, 0.41 and 0.5 respectively after standardization. Substituting them into the comprehensive index calculation formula, the comprehensive score result of this scheme is 0.613.

[0110] Similarly, based on the geological parameters and fluid property parameters of other well groups and the same pressure-driven construction parameter combination as input, the trained DRSN-Adaboost prediction model is used to quickly predict the result indicators of each well group under the same construction plan, and the comprehensive scoring results of other well groups are calculated. By comparing the comprehensive scoring results, the well group with the best pressure-driven effect in the reservoir under the same pressure-driven construction plan can be selected.

[0111] In addition, after the pressure-driven well group is optimized, the geological parameters and fluid property parameters of the well group are combined with different pressure-driven construction parameters to form input data, and the pressure-driven potential evaluation index results under different pressure-driven construction parameter combinations are predicted and the comprehensive score is calculated, which can be used for the optimization of the pressure-driven construction plan.

[0112] In summary, in the above scheme provided by the present embodiment, a large number of random geological parameter attribute fields are generated by Monte Carlo simulation sampling, the uncertainty of geological parameters is fully considered, a sample set is established and the DRSN-Adaboost model is trained to obtain a trained DRSN-Adaboost model, and the prediction of the pressure-driven potential evaluation index under different geological parameters and pressure-driven construction parameters is realized. Through the rapid prediction results, the development effect of the oil field under different schemes can be obtained, including indicators such as cumulative oil, pressure recovery coefficient, ton water pressure increase rate, comprehensive water content and ton water oil change rate. Through statistical analysis, the probability distribution of each indicator of the entire reservoir pressure-driven development can be obtained, and a comprehensive scoring index evaluation system is established to evaluate the pressure-driven potential. Compared with conventional reservoir numerical simulation methods, the technology provided by this embodiment not only has a wider range of applications, can provide prediction effects under different development schemes, and has a short prediction time, while fully considering the uncertainty of geological parameters, improving the reliability of the prediction results, and can greatly reduce time costs and repetitive work. Therefore, it has important guiding significance for the potential evaluation of pressure-driven low-permeability tight reservoirs.

[0113] Based on the same inventive concept, the embodiment of the present application also provides a system for implementing the above-mentioned method for evaluating the pressure-driven potential of low-permeability tight oil reservoirs. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the system for evaluating the pressure-driven potential of low-permeability tight oil reservoirs provided below can refer to the limitations of the method for evaluating the pressure-driven potential of low-permeability tight oil reservoirs above, and will not be repeated here.

[0114] In an exemplary embodiment, Figure 6 As shown, a system for evaluating the pressure drive potential of a low permeability tight oil reservoir is provided, comprising: The pressure-drive potential prediction sample set construction module is used to construct the pressure-drive potential prediction sample set; the pressure-drive potential prediction sample set includes geological parameters, fluid property parameters, pressure-drive construction parameters, and pressure-drive potential evaluation index simulation results obtained by pressure-drive numerical simulation based on the geological parameters, fluid property parameters, and pressure-drive construction parameters; the pressure-drive potential evaluation index simulation results include simulation values ​​of several pressure-drive potential evaluation indicators; the pressure-drive potential evaluation indicators include cumulative oil, comprehensive water content, oil change rate per ton of water, pressure recovery coefficient, and pressure increase rate per ton of water.

[0115] The pressure-driven potential prediction model training module is used to train the DRSN-Adaboost model using the pressure-driven potential prediction sample set to obtain the pressure-driven potential prediction model; the pressure-driven potential prediction model is used to output the corresponding pressure-driven potential evaluation index prediction results based on the input geological parameters, fluid property parameters and pressure-driven construction parameters.

[0116] The target reservoir pressure drive potential prediction module is used to predict the pressure drive potential evaluation index prediction results of any well group in the target reservoir using the pressure drive potential prediction model.

[0117] The pressure-drive potential comprehensive score calculation module is used to calculate the pressure-drive potential comprehensive score based on the pressure-drive potential evaluation index prediction results of multiple well groups in the target oil reservoir and the weights of each pressure-drive potential evaluation index; the pressure-drive potential comprehensive score is the comprehensive evaluation result of the pressure-drive potential for the target oil reservoir.

[0118] certainly, Figure 6 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different functions. Figure 6 One or at least two components of the system shown.

[0119] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 7As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for evaluating the pressure drive potential of a low-permeability tight oil reservoir is implemented.

[0120] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0121] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0122] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0123] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for evaluating the pressure drive potential of a low permeability tight oil reservoir, characterized in that: include: Constructing a pressure-drive potential prediction sample set; the pressure-drive potential prediction sample set includes geological parameters, fluid property parameters, pressure-drive construction parameters, and pressure-drive potential evaluation index simulation results obtained by pressure-drive numerical simulation based on the geological parameters, fluid property parameters, and pressure-drive construction parameters; the pressure-drive potential evaluation index simulation results include simulation values ​​of several pressure-drive potential evaluation indicators; the pressure-drive potential evaluation indicators include cumulative oil, comprehensive water content, oil change rate per ton of water, pressure recovery coefficient, and pressure increase rate per ton of water; The pressure-driven potential prediction sample set is used to train the DRSN-Adaboost model to obtain a pressure-driven potential prediction model; the pressure-driven potential prediction model is used to output corresponding pressure-driven potential evaluation index prediction results according to input geological parameters, fluid property parameters and pressure-driven construction parameters; For any well group in the target oil reservoir, using the pressure-drive potential prediction model, predicting a pressure-drive potential evaluation index prediction result of the well group; According to the prediction results of the pressure-drive potential evaluation indicators of multiple well groups in the target oil reservoir and the weights of each pressure-drive potential evaluation indicator, a comprehensive score of pressure-drive potential is calculated; the comprehensive score of pressure-drive potential is a comprehensive evaluation result of pressure-drive potential for the target oil reservoir.

2. The method for evaluating the pressure-drive potential of a low-permeability tight oil reservoir according to claim 1, characterized in that: Construct a sample set for pressure-driven potential prediction, including: Obtain the value ranges of the target reservoir's geological parameters, fluid property parameters, and pressure-driven construction parameters, as well as the probability distribution form, mean, and variance of the geological parameters' porosity, permeability, oil saturation, Young's modulus, Poisson's ratio, maximum horizontal principal stress, and minimum horizontal principal stress; Sampling is performed within the value ranges of various geological parameters, various fluid property parameters and various pressure-driven construction parameters to obtain a plurality of input parameter combinations; the input parameter combinations include three-dimensional matrix samples corresponding to the geological parameters, sampled values ​​of the fluid property parameters and sampled values ​​of the pressure-driven construction parameters; For any input parameter combination, a pressure-driven numerical simulation is carried out to obtain the simulation results of the pressure-driven potential evaluation index under the corresponding input parameter combination; The three-dimensional matrix samples corresponding to the geological parameters of the input parameter combination, the sampled values ​​of the fluid property parameters, the sampled values ​​of the pressure-driving construction parameters and the simulated pressure-driving potential evaluation index simulation results are used as a pressure-driving potential prediction sample; a number of the pressure-driving potential prediction samples are gathered together to construct a pressure-driving potential prediction sample set.

3. The method for evaluating the pressure-drive potential of a low-permeability tight oil reservoir according to claim 2, characterized in that: Sampling is carried out within the range of values ​​of various geological parameters, various fluid property parameters and various pressure drive construction parameters, including: For geological parameters, the Monte Carlo method is used to perform random sampling within the value range of the uncertain parameters in the geological parameters to obtain three-dimensional matrix samples corresponding to the geological parameters; For the fluid property parameters and pressure-driven construction parameters, the Latin hypercube sampling method is used to perform sampling within the value range of the corresponding parameters to obtain the sampling values ​​of the fluid property parameters and pressure-driven construction parameters.

4. The method for evaluating the pressure-drive potential of a low-permeability tight oil reservoir according to claim 2, characterized in that: For any input parameter combination, a pressure-driven numerical simulation is carried out to obtain the simulation results of the pressure-driven potential evaluation index under the corresponding input parameter combination, including: Constructing a reservoir geological model according to the three-dimensional matrix samples corresponding to the geological parameters generated by sampling; the attribute values ​​of each grid in the reservoir geological model are derived from the elements of the three-dimensional matrix samples; If the target reservoir has natural fractures, a natural fracture model is obtained by natural fracture modeling according to the length, orientation and linear density of the natural fractures; and the natural fracture model is added to the reservoir geological model to obtain a reservoir geological model containing natural fractures; If the target reservoir has no natural fractures, no special treatment is performed and the next step is carried out directly; Based on the reservoir geological model or the reservoir geological model containing natural fractures, according to the sampled values ​​of the pressure-driving construction parameters, a pressure-driving fracture network expansion simulation is carried out to obtain a fracture network model after the pressure-driving construction; Based on the fracture network model after the pressure-driven construction, an unstructured grid is used to finely characterize the fracture network morphology to obtain a matrix grid model and a fracture grid model; Based on the matrix grid model and the fracture grid model, a pressure-driven numerical simulation is performed according to the sampled values ​​of the fluid property parameters to obtain a simulation result of a pressure-driven potential evaluation index; based on the matrix grid model and the fracture grid model, when a pressure-driven numerical simulation is performed according to the sampled values ​​of the fluid property parameters, corresponding stress sensitivity curves are set for different types of grids, and the difference in permeability changes of the matrix grid and the fracture grid under different stresses is simulated through the stress sensitivity curve, thereby simulating different flow characteristics in the matrix grid and the fracture grid during the pressure-driven development process.

5. The method for evaluating the pressure-drive potential of a low-permeability tight oil reservoir according to claim 1, characterized in that: The DRSN-Adaboost model includes a data processing layer, a deep residual shrinkage neural network and an Adaboost module; the data processing layer is used to process input data; during the data processing process, through dimensionality increase and fusion operations, a feature channel matrix composed of several three-dimensional matrices is generated according to various parameters in the above-mentioned pressure-driven potential prediction sample as the input of the deep residual shrinkage neural network; the deep residual shrinkage neural network is composed of an input layer, a convolutional layer, a residual module, a soft threshold layer, a pooling layer, a fully connected layer and an output layer; the Adaboost module uses a deep residual shrinkage neural network as a weak learner of the Adaboost module, obtains multiple weak learners through iterative training, and forms a strong learner through weight combination.

6. The method for evaluating the pressure-drive potential of a low-permeability tight oil reservoir according to claim 5, characterized in that: Through dimensionality increase and fusion operations, according to various parameters in the above pressure-driven potential prediction samples, a feature channel matrix composed of several three-dimensional matrices is generated as the input of the deep residual shrinkage neural network, specifically including: The one-dimensional data of various parameters in the above-mentioned pressure-drive potential prediction sample are upgraded to obtain multiple three-dimensional data matrices; specifically, for the one-dimensional parameters and fluid property parameters in the geological parameters, each parameter generates a three-dimensional data matrix consistent with the spatial distribution of the reservoir geological model; all values ​​in the three-dimensional data matrix are the corresponding one-dimensional data parameter values; for the pressure-drive construction parameters, each parameter generates a three-dimensional data matrix consistent with the spatial distribution of the reservoir geological model, and in each three-dimensional data matrix of the pressure-drive construction parameters, the grid positions of all wells in the reservoir model and the well trajectories pass through are filled with the pressure-drive construction parameter values ​​of the wells implementing the pressure-drive at the corresponding matrix element positions in the matrix space, and the values ​​of other positions in the matrix are filled with 0; The multiple three-dimensional data matrices obtained by dimensionality upgrading are fused with the three-dimensional matrix of geological parameters to form a feature channel matrix composed of several three-dimensional matrices as the input of the deep residual shrinkage neural network.

7. The method for evaluating the pressure-drive potential of a low-permeability tight oil reservoir according to claim 5, characterized in that: A deep residual shrinkage neural network is used as the weak learner of the Adaboost module. Multiple weak learners are obtained through iterative training, and a strong learner is formed through weight combination, which includes: The deep residual shrinkage neural network is set as the initial weak learner, and the deep residual shrinkage neural network is trained using the pressure-driven potential prediction sample set to generate the first deep residual shrinkage neural network weak learner; Iterative update, each iteration obtains a deep residual shrinkage neural network weak learner, completes the training of the preset number of iterations, and obtains a total of M deep residual shrinkage neural network weak learners, and the M deep residual shrinkage neural network weak learners are combined by weights to obtain a deep residual shrinkage neural network strong learner; M is the preset number of iterations, the input of the deep residual shrinkage neural network strong learner is the characteristic channel matrix obtained by the data processing layer, and the output of the deep residual shrinkage neural network strong learner is the prediction result of the pressure-driven potential evaluation index.

8. The method for evaluating the pressure-drive potential of a low-permeability tight oil reservoir according to claim 7, characterized in that: According to the following formula, M deep residual shrinkage neural network weak learners are combined by weights to obtain a deep residual shrinkage neural network strong learner: ； in, G ( x ) is a deep residual shrinkage neural network strong learner, Delta is the number of iterations, α Delta For the Delta The weights of the weak learners of the optimal deep residual shrinkage neural network for the iteration, T Delta ( x ) is the Delta Optimal deep residual shrinkage neural network weak learner with iterations.

9. The method for evaluating the pressure-drive potential of a low-permeability tight oil reservoir according to claim 1, characterized in that: The comprehensive score of the pressure drive potential of multiple well groups in the target reservoir is calculated according to the following formula: ； in, W i The target reservoir i Comprehensive score of pressure drive potential of each well group, w o , w pi , w pv , w cui , w r is the weight of each pressure drive potential evaluation index, x o , x pi , x pv , x cui , x r is the normalized value of each pressure-driven potential evaluation index.

10. A system for evaluating the pressure drive potential of low permeability tight oil reservoirs, characterized in that: include: A pressure-drive potential prediction sample set construction module is used to construct a pressure-drive potential prediction sample set; the pressure-drive potential prediction sample set includes geological parameters, fluid property parameters, pressure-drive construction parameters, and pressure-drive potential evaluation index simulation results obtained by pressure-drive numerical simulation based on the geological parameters, fluid property parameters, and pressure-drive construction parameters; the pressure-drive potential evaluation index simulation results include simulation values ​​of several pressure-drive potential evaluation indicators; the pressure-drive potential evaluation indicators include cumulative oil, comprehensive water content, oil change rate per ton of water, pressure recovery coefficient, and pressure increase rate per ton of water; A pressure-driven potential prediction model training module is used to train the DRSN-Adaboost model using the pressure-driven potential prediction sample set to obtain a pressure-driven potential prediction model; the pressure-driven potential prediction model is used to output corresponding pressure-driven potential evaluation index prediction results according to input geological parameters, fluid property parameters and pressure-driven construction parameters; A target reservoir pressure drive potential prediction module is used to predict the pressure drive potential evaluation index prediction result of any well group in the target reservoir by using the pressure drive potential prediction model; The pressure-drive potential comprehensive score calculation module is used to calculate the pressure-drive potential comprehensive score according to the pressure-drive potential evaluation index prediction results of multiple well groups in the target oil reservoir and the weights of each pressure-drive potential evaluation index; the pressure-drive potential comprehensive score is the comprehensive evaluation result of the pressure-drive potential for the target oil reservoir.

Images