An automatic history fitting method based on ensemble smoothing and adaptive hybrid gradient

Abstract

The present invention provides a set smoothing automatic history matching method based on adaptive hybrid gradients, which relates to the field of oil reservoir development technology. The method obtains an initialized priori oil reservoir model parameter set and production history dynamic data; dynamically adjusts the preprocessing gradient update direction with adaptive step size; performs covariance preprocessing on the reservoir automatic history matching objective function gradient to construct the preprocessing gradient update direction; calculates parameter sensitivity weights and updates parameters in the preprocessing gradient direction in combination with an adaptive step size dynamic adjustment strategy; constructs a hybrid update direction; introduces a momentum term to accelerate convergence and suppress oscillation; and determines the distribution of reservoir parameters and objective function changes based on the final oil reservoir parameter set obtained after iterative solution of the optimization algorithm, thereby completing the automatic history matching process. The technical solution of the present invention overcomes the problems of the prior art in that it cannot adapt to strongly nonlinear oil reservoir systems, has low parameter inversion accuracy, and has high computational costs.

Description

Technical Field

[0001] The present invention relates to the field of oil reservoir development, and in particular to a set smoothing automatic history matching method based on adaptive mixed gradient. Background Art

[0002] Currently, automated reservoir history matching primarily relies on gradient optimization algorithms, random search algorithms, and ensemble methods. While traditional gradient optimization algorithms converge quickly, they require calculating the objective function gradient in a high-dimensional parameter space. When dealing with systems with nonsmooth objective functions or sudden changes in parameter sensitivity, explicit gradient calculations are susceptible to numerical oscillations, and parameter update paths are easily trapped by nonconvex response surfaces, leading to local optima. Random search algorithms overcome local extrema through global sampling strategies, but they require a superlinearly increasing number of simulations in high-dimensional parameter spaces, posing computational challenges for large-scale reservoir models. Ensemble algorithms, such as the Ensemble Smoother with Multiple Data Assimilation (ES-MDA), implicitly represent gradient directions through the covariance matrix of the parameter set, avoiding explicit gradient calculations. Through multiple data assimilation and expansion factor design, ES-MDA enables global exploration within a limited number of simulations, effectively balancing computational efficiency and inversion accuracy. ES-MDA is currently the mainstream approach for automated reservoir history matching. However, its update formula is based on the Gaussian linear assumption, making it poorly adaptable to strongly nonlinear automatic history fitting problems. Errors in covariance matrix estimation accumulate, and parameter updates are prone to deviating from true physical laws. Furthermore, the strong coupling effects between high-dimensional parameters can undermine the theoretical consistency of the linear update mechanism, making the inversion results prone to systematic deviations.

[0003] Therefore, there is a need for an ensemble smoothing automatic history fitting method that can improve the inversion and index fitting effects of strongly nonlinear reservoir parameters. Summary of the Invention

[0004] The main purpose of the present invention is to propose an ensemble smoothing automatic history matching method based on adaptive hybrid gradient to solve the problems that the existing reservoir automatic history matching method cannot adapt to strongly nonlinear reservoir systems, has low parameter inversion accuracy and high computational cost.

[0005] To achieve the above object, the present invention provides a set smoothing automatic history fitting method based on adaptive hybrid gradient, which specifically includes the following steps:

[0006] S1, obtain the initialization priori reservoir model parameter set and production history dynamic data.

[0007] S2, set the parameters of the AHG-ESMDA algorithm and dynamically adjust the adaptive step size of the preprocessing gradient update direction according to the changes in the automatic history fitting function.

[0008] S3, perform covariance preprocessing on the reservoir automatic history fitting objective function gradient and construct the preprocessing gradient update direction.

[0009] S4 calculates the parameter sensitivity weights and updates the parameters in the preprocessing gradient direction in combination with the adaptive step size dynamic adjustment strategy.

[0010] S5, integrates the ES-MDA collective covariance gradient update direction and the preprocessed gradient update direction to construct a hybrid update direction.

[0011] S6, introduces momentum term in the parameter update process to accelerate convergence and suppress oscillation.

[0012] S7, based on the final reservoir parameter set obtained after iterative solution of the optimization algorithm, the distribution of reservoir parameters and the change of the objective function are determined to complete the automatic history fitting process.

[0013] Furthermore, step S1 specifically includes the following steps:

[0014] S1.1, determine the parameters to be adjusted based on reservoir characteristics.

[0015] S1.2, generate an initialization priori reservoir model parameter set through the Sequential Gaussian Geological Simulation Generator SGeMS; obtain the historical dynamic data of reservoir production, including the oil production, water production, production pressure of production wells, and the injection volume and bottom hole pressure data of water injection wells.

[0016] Furthermore, the parameters to be adjusted include: grid permeability, porosity, relative permeability curve of the reservoir model and parameters related to the nonlinear reservoir system.

[0017] Furthermore, step S2 specifically includes the following steps:

[0018] S2.1, the parameters of the AHG-ESMDA algorithm include: the number of data assimilation , expansion factor , initial update step size of preprocessing gradient direction , adaptive attenuation factor , momentum term coefficient and blend weights .

[0019] S2.2, according to the changes of the automatic history fitting function, the adaptive step size of the preprocessing gradient update direction is dynamically adjusted. The formula is expressed as:

[0020] ;

[0021] in, The range is (0, 1), For the The set of objective function sizes for sub-data assimilation, is an exponential function, is the step size in the preprocessing gradient direction.

[0022] Furthermore, step S3 specifically includes the following steps:

[0023] S3.1, in gradient descent, the direction of the gradient update after preprocessing for:

[0024] ;

[0025] in, is the reservoir parameter, is a positive definite matrix, Gradient of the objective function for automated reservoir history matching.

[0026] S3.2, Objective Function of Automatic Reservoir History Matching for:

[0027] ;

[0028] in, are the prior reservoir model parameters; is the covariance matrix of the prior reservoir model parameters; is the production observation data of the reservoir; represents the reservoir numerical simulator, represents the observed data error covariance.

[0029] The gradient of the objective function of automatic reservoir history matching:

[0030] ;

[0031] in, is the Jacobian matrix.

[0032] S3.3, let , at this time, the gradient update direction after preprocessing is:

[0033] ;

[0034] After expansion:

[0035] .

[0036] The Jacobian matrix is expressed as:

[0037] ;

[0038] in, is the cross-covariance between reservoir parameters and predicted data, is the variance of the predicted data, and the formula is as follows:

[0039] ;

[0040] ;

[0041] in, is the number of reservoir parameter sets, is the reservoir parameter of the jth set, is the production simulation data of the jth set of reservoir parameters, A collection of simulation data for production.

[0042] S3.4, the simplified preprocessing gradient update direction is:

[0043] .

[0044] Furthermore, step S4 specifically includes the following steps:

[0045] S4.1, decompose the step size of the preprocessing gradient direction into the global step size and parameter sensitivity weights :

[0046] .

[0047] S4.2, using the adaptive step size dynamic adjustment strategy for the preprocessing gradient update direction proposed in step S2, update the global step size in each round of data assimilation process, and at the same time, the parameter sensitivity weight The calculation method is as follows:

[0048] ;

[0049] in, Representative During the round of data assimilation The covariance vector between the reservoir parameters and the observed data, is the number of reservoir parameters in a single set.

[0050] Furthermore, step S5 specifically includes the following steps:

[0051] S5.1, Ensemble covariance gradient update direction for ensemble ES-MDA :

[0052] ;

[0053] in, Indicates the The size of the expansion factor for secondary data assimilation; It is the data disturbance added to the observed data.

[0054] S5.2, introducing mixed direction weights Balance the update ratio of the covariance gradient update direction and the preprocessing gradient direction, and combine the parameter sensitivity weight and global step size Get the preliminary update formula of parameters:

[0055] .

[0056] Furthermore, step S6 specifically includes the following steps:

[0057] S6.1, referring to the momentum optimization algorithm, introduces historical update direction memory:

[0058] .

[0059] The reservoir parameter update formula is:

[0060] .

[0061] S6.2, based on S6.1, the final reservoir parameter update formula is:

[0062] ;

[0063] in, is the momentum term coefficient, and its value ranges from 0 to 0.3.

[0064] Furthermore, step S7 specifically includes the following steps:

[0065] S7.1, monitor the decline rate of the objective function and terminate the calculation when it reaches a preset threshold or the maximum number of iterations to obtain the final reservoir parameter set; the conditions for terminating the calculation are that the relative decline rate of the objective function is less than 1% or the maximum number of iterations is reached.

[0066] S7.2, saving the final reservoir parameter set that meets the convergence conditions, and verifying the matching degree between the final reservoir parameter set that meets the convergence conditions and the historical production data through a reservoir numerical simulator.

[0067] The present invention has the following beneficial effects:

[0068] The present invention performs automatic history fitting on the reservoir model based on the AHG-ESMDA algorithm. In this process, AHG-ESMDA integrates the global exploration capability of ensemble smoothing with the local correction mechanism of gradient optimization, and controls the updating degree in the preprocessing gradient direction based on the parameter sensitivity matrix and the adaptive step size dynamic adjustment strategy. The momentum term is introduced through the memory of the historical update direction. Therefore, the present invention has good adaptability to nonlinear reservoir systems, and gradually approximates the parameter posterior distribution through multiple rounds of ensemble updates, which can effectively characterize the joint uncertainty of multiple parameters in the nonlinear system and achieve more accurate parameter inversion. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0070] Figure 1 The flowchart of the present invention is a method for automatic history matching based on ensemble smoothing and adaptive mixed gradient.

[0071] Figure 2 A distribution diagram of actual reservoir parameters according to an embodiment of the present invention is shown.

[0072] Figure 3 A graph showing the change in the objective function value of automatic history matching achieved using the AHG-ESMDA algorithm provided by the present invention is shown.

[0073] Figure 4 A permeability field from an initial a priori reservoir model parameter set is shown.

[0074] Figure 5 The final posterior permeability field after solving the AHG-ESMDA algorithm is shown.

[0075] Figure 6 A comparison diagram of the fitting effect of the oil production rate data of the production well P1 according to an embodiment of the present invention is shown.

[0076] Figure 7 A comparison diagram of the fitting effect of water production rate data of the production well P1 according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0077] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0078] like Figure 1 The method for automatic history fitting based on ensemble smoothing and adaptive mixed gradient is shown, which specifically includes the following steps:

[0079] S1, obtain the initialization priori reservoir model parameter set and production history dynamic data.

[0080] S2 sets the parameters of the Adaptive Hybrid Gradient-Based Ensemble Smoother with Multiple Data Assimilation (AHG-ESMDA) algorithm, and dynamically adjusts the adaptive step size of the preprocessing gradient update direction based on changes in the automatic history fitting function.

[0081] S3, perform covariance preprocessing on the reservoir automatic history fitting objective function gradient and construct the preprocessing gradient update direction.

[0082] S4 calculates the parameter sensitivity weights and updates the parameters in the preprocessing gradient direction in combination with the adaptive step size dynamic adjustment strategy.

[0083] S5, integrates the ES-MDA collective covariance gradient update direction and the preprocessed gradient update direction to construct a hybrid update direction.

[0084] S6, introduces momentum term in the parameter update process to accelerate convergence and suppress oscillation.

[0085] S7, based on the final reservoir parameter set obtained after iterative solution of the optimization algorithm, the distribution of reservoir parameters and the change of the objective function are determined to complete the automatic history fitting process.

[0086] Specifically, step S1 includes the following steps:

[0087] S1.1, determine the parameters to be adjusted based on reservoir characteristics.

[0088] S1.2, generate an initialization priori reservoir model parameter set through the Sequential Gaussian Geological Simulation Generator SGeMS; obtain the historical dynamic data of reservoir production, including the oil production, water production, production pressure of production wells, and the injection volume and bottom hole pressure data of water injection wells.

[0089] Specifically, the parameters to be adjusted include: grid permeability, porosity, phase permeability curve of the reservoir model, and nonlinear reservoir system related parameters, such as polymer solution viscosity-concentration relationship parameters.

[0090] In this example, a polymer flooding two-dimensional reservoir model was selected to verify the effectiveness of the AHG-ESMDA algorithm for strong nonlinear systems. The number of grids in the x, y, and z directions of the reservoir model was 60, 60, and 1, respectively, for a total of 3600 grids. The sizes of the grids in the x, y, and z directions were 8m, 8m, and 4m, respectively. The reservoir depth was 4000m. The x-direction permeability distribution of the real model is shown in Figure 2. Figure 2 As shown, Figure 2 Here, mD is the unit of permeability. The model's y-direction permeability is the same as its x-direction permeability, and the z-direction permeability is 0.1 times the x-direction permeability. The model includes one water injection well (I1) and four production wells (P1-P4). The injection well operates at a fixed injection rate of 300 cubic meters per day. The production well operates at a fixed bottomhole pressure of 350 bar. The reservoir model is numerically simulated using both oil and water phases. The production time is 1500 days, divided into 50 time periods. The first 750 days are for water injection into the injection wells, and the last 750 days are for polymer flooding solution injection. The reservoir inversion parameters are the x-direction permeability, the polymer solution viscosity-concentration relationship parameter, the polymer adsorption parameter, and the polymer residual resistance coefficient. The production observation data include the oil and water production rates of the four production wells.

[0091] According to the reservoir model, 100 samples of prior reservoir model parameters were generated using the Sequential Gaussian Geosimulation Generator (SGeMS) for the subsequent automatic history fitting process.

[0092] Specifically, step S2 includes the following steps:

[0093] S2.1, the parameters of the AHG-ESMDA algorithm include: the number of data assimilation , expansion factor , initial update step size of preprocessing gradient direction , adaptive attenuation factor , momentum term coefficient and blending weights.

[0094] In this embodiment, the number of data assimilation times , expansion factor , initial update step size of preprocessing gradient direction , adaptive attenuation factor , momentum term coefficient and blend weights .

[0095] S2.2, according to the changes of the automatic history fitting function, the adaptive step size of the preprocessing gradient update direction is dynamically adjusted. The formula is expressed as:

[0096] ;

[0097] in, The range is (0, 1), For the The set of objective function sizes for sub-data assimilation, is an exponential function, is the step size in the preprocessing gradient direction.

[0098] Specifically, step S3 includes the following steps:

[0099] S3.1, preprocessing is done by introducing a positive definite matrix , transforming the original optimization problem into a new coordinate system so that the transformed problem has a better condition number (i.e., closer to isotropy), thereby accelerating the convergence of gradient-based algorithms. In gradient descent, the direction of the preprocessed gradient update is for:

[0100] ;

[0101] in, is the posterior estimate of the reservoir parameters, is a positive definite matrix, Gradient of the objective function for automated reservoir history matching.

[0102] In the automatic history matching problem, the covariance matrix of the prior reservoir model parameters is Encodes the spatial correlation and uncertainty of the parameters, so the covariance preconditioning choice .

[0103] S3.2, Objective Function of Automatic Reservoir History Matching for:

[0104] ;

[0105] in, are the prior reservoir model parameters; is the covariance matrix of the prior reservoir model parameters; is the production observation data of the reservoir; represents the reservoir numerical simulator, represents the error covariance of the observed data;

[0106] The gradient of the objective function of automatic reservoir history matching:

[0107] ;

[0108] in, is the Jacobian matrix.

[0109] S3.3, let , at this time, the gradient update direction after preprocessing is:

[0110] ;

[0111] After expansion:

[0112] .

[0113] There is no need to explicitly calculate the Jacobian matrix in collection methods , but rather approximate the perturbation through the set, the Jacobian matrix is expressed as:

[0114] ;

[0115] in, is the cross-covariance between reservoir parameters and predicted data, is the variance of the predicted data, and the formula is as follows:

[0116] ;

[0117] ;

[0118] in, is the number of reservoir parameter sets, For the A set of reservoir parameters, For the The production simulation data of a set of reservoir parameters, A collection of simulation data for production;

[0119] S3.4, the simplified preprocessing gradient update direction is:

[0120] .

[0121] Specifically, step S4 includes the following steps:

[0122] S4.1, decompose the step size of the preprocessing gradient direction into the global step size and parameter sensitivity weights :

[0123] .

[0124] S4.2, using the adaptive step size dynamic adjustment strategy for the preprocessing gradient update direction proposed in step S2, update the global step size in each round of data assimilation process, and at the same time, the parameter sensitivity weight The calculation method is as follows:

[0125] ;

[0126] in, Representative During the round of data assimilation The covariance vector between the reservoir parameters and the observed data, is the number of reservoir parameters in a single set.

[0127] Calculated is a diagonal matrix, representing the The parameter sensitivity weight matrix obtained by data assimilation is a matrix with diagonal elements representing parameter sensitivity.

[0128] Specifically, step S5 includes the following steps:

[0129] S5.1, ES-MDA ensemble covariance gradient update direction :

[0130] ;

[0131] in, Indicates the The size of the expansion factor for secondary data assimilation; It is the data disturbance added to the observed data.

[0132] S5.2, introducing mixed direction weights Balance the update ratio of the covariance gradient update direction and the preprocessing gradient direction, and combine the parameter sensitivity weight and global step size Get the preliminary update formula of parameters:

[0133] .

[0134] Specifically, step S6 includes the following steps:

[0135] S6.1, referring to the momentum optimization algorithm, introduces historical update direction memory:

[0136] .

[0137] The reservoir parameter update formula is:

[0138] .

[0139] S6.2, based on S6.1, the final reservoir parameter update formula is:

[0140] ;

[0141] in, is the momentum term coefficient, and its value ranges from 0 to 0.3.

[0142] Specifically, step S7 includes the following steps:

[0143] S7.1, monitor the rate of decrease of the objective function (observation residual norm), terminate the calculation when it reaches a preset threshold or the maximum number of iterations, and obtain the final reservoir parameter set; the condition for termination of the calculation is that the relative rate of decrease of the objective function is less than 1% or the maximum number of iterations is reached.

[0144] S7.2, saving the final reservoir parameter set that meets the convergence conditions, and verifying the matching degree between the final reservoir parameter set that meets the convergence conditions and the historical production data through a reservoir numerical simulator.

[0145] The AHG-ESMDA algorithm demonstrates significant advantages in reservoir parameter inversion and dynamic index fitting for highly nonlinear reservoir systems. By integrating the global exploration capabilities of ensemble smoothing with the local correction mechanism of gradient optimization, the algorithm maintains the computational efficiency advantages of ensemble methods in complex heterogeneous reservoir systems while overcoming the adaptability limitations of linear update mechanisms for highly nonlinear systems. This provides reliable theoretical support for high-precision dynamic prediction of reservoir models.

[0146] Figure 3 The figure shows the change in the objective function of each sample during the AHG-ESMDA optimization iteration. The x-axis ranges from 0 to 100, representing the objective function values of each prior reservoir model in the initial set. Each 100 steps thereafter represents the change in the objective function values of the model samples after data assimilation. Initially, the sample points are dispersed and have high values, reflecting the large initial uncertainty of the parameters. As data assimilation progresses, the sample points rapidly converge to low-value areas, with the distribution range shrinking significantly. Later, they become dense and stable, indicating that the algorithm effectively suppresses parameter perturbations through adaptive mechanisms and achieves systematic optimization of the objective function. This smooth transition from global exploration to targeted convergence verifies the robustness and adaptability of the algorithm in strongly nonlinear systems.

[0147] Figure 4-Figure 5 This is the inversion of the x-direction permeability field after efficient solution by the AHG-ESDMA algorithm. Figure 4 is the permeability field of the initial set of parameters, Figure 5This is the final posterior permeability map after the AHG-ESMDA solution. It can be seen that the uncertainty of the model parameter field is reduced after the efficient AHG-ESMDA solution, and it is closer to the true permeability field than the initial permeability field, reflecting the parameter distribution characteristics of the true permeability field.

[0148] Figure 6 、 Figure 7 Figure 1 compares the results of fitting the observed data (abbreviated as "observed" in the figure) for the oil and water production rates of production well 1. The curves in the figure represent the initial model set (abbreviated as "initial" in the figure) and the final model set obtained using the AHG-ESMDA method (abbreviated as "final" in the figure). It can be seen that the simulated data from the initial model set have a large error compared to the observed data, while the optimized data almost coincide with the observed data, demonstrating a good fit.

[0149] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the scope of protection of the present invention.

Claims

1. A method for automatic history fitting based on ensemble smoothing and adaptive hybrid gradient, characterized in that: The specific steps include: S1, obtaining the initialization priori reservoir model parameter set and production history dynamic data; S2, set the parameters of the AHG-ESMDA algorithm and dynamically adjust the adaptive step size of the preprocessing gradient update direction according to the changes of the automatic history fitting function; S3, performs covariance preprocessing on the reservoir automatic history fitting objective function gradient and constructs the preprocessing gradient update direction; S4, calculates the parameter sensitivity weights and updates the parameters in the preprocessing gradient direction in combination with the adaptive step size dynamic adjustment strategy; S5, integrates the ES-MDA collective covariance gradient update direction and the preprocessed gradient update direction to construct a hybrid update direction; S6, introduces momentum term in the parameter update process to accelerate convergence and suppress oscillation; S7, based on the final reservoir parameter set obtained after iterative solution of the optimization algorithm, the distribution of reservoir parameters and the change of the objective function are determined to complete the automatic history fitting process; Step S2 specifically includes the following steps: S2.1, the parameters of the AHG-ESMDA algorithm include: the number of data assimilation , expansion factor , initial update step size of preprocessing gradient direction , adaptive attenuation factor , momentum term coefficient and blend weights ; S2.2, according to the changes of the automatic history fitting function, the adaptive step size of the preprocessing gradient update direction is dynamically adjusted. The formula is expressed as: ； in, The range is (0, 1), For the The set of objective function sizes for sub-data assimilation, is an exponential function, is the step size in the preprocessing gradient direction; Step S3 specifically includes the following steps: S3.1, in gradient descent, the direction of the gradient update after preprocessing for: ； in, is the reservoir parameter, is a positive definite matrix, The gradient of the objective function of automatic reservoir history matching; S3.2, Objective Function of Automatic Reservoir History Matching for: ； in, are the prior reservoir model parameters; is the covariance matrix of the prior reservoir model parameters; is the production observation data of the reservoir; represents the reservoir numerical simulator, represents the error covariance of the observed data; The gradient of the objective function of automatic reservoir history matching: ； in, is the Jacobian matrix; S3.3, let , at this time, the gradient update direction after preprocessing is: ； After expansion: ； The Jacobian matrix is expressed as: ； in, is the cross-covariance between reservoir parameters and predicted data, is the variance of the predicted data, and the formula is as follows: ； ； in, is the number of reservoir parameter sets, For the A set of reservoir parameters, For the The production simulation data of a set of reservoir parameters, A collection of simulation data for production; S3.4, the simplified preprocessing gradient update direction is: ； Step S4 specifically includes the following steps: S4.1, decompose the step size of the preprocessing gradient direction into the global step size and parameter sensitivity weights : ； S4.2, using the adaptive step size dynamic adjustment strategy for the preprocessing gradient update direction proposed in step S2, update the global step size in each round of data assimilation process, and at the same time, the parameter sensitivity weight The calculation method is as follows: ； in, Representative During the round of data assimilation The covariance vector between the reservoir parameters and the observed data, is the number of reservoir parameters in a single set; Step S5 specifically includes the following steps: S5.1, ES-MDA ensemble covariance gradient update direction : ； in, Indicates the The size of the expansion factor for secondary data assimilation; It is the data disturbance added to the observed data; S5.2, introducing mixed direction weights Balance the update ratio of the covariance gradient update direction and the preprocessing gradient direction, and combine the parameter sensitivity weight and global step size Get the preliminary update formula of parameters: ； Step S6 specifically includes the following steps: S6.1, referring to the momentum optimization algorithm, introduces historical update direction memory: ； The reservoir parameter update formula is: ； S6.2, based on S6.1, the final reservoir parameter update formula is: ； in, is the momentum term coefficient, and its value ranges from 0 to 0.

3.

2. The method of automatic history fitting based on adaptive hybrid gradient ensemble smoothing according to claim 1, characterized in that: Step S1 specifically includes the following steps: S1.1, determine the parameters to be adjusted based on reservoir characteristics; S1.2, generate an initialized priori reservoir model parameter set through the Sequential Gaussian Geological Simulation Generator SGeMS; obtain the historical production dynamic data of the reservoir, including the oil production, water production, production pressure of the production wells, and the injection volume and bottom hole pressure data of the water injection wells.

3. The method of automatic history fitting based on adaptive mixed gradient ensemble smoothing according to claim 2, characterized in that: The parameters to be adjusted include: permeability, porosity, relative permeability curve of the reservoir model and parameters related to the nonlinear reservoir system.

4. The method of automatic history fitting based on adaptive mixed gradient ensemble smoothing according to claim 1, characterized in that: Step S7 specifically includes the following steps: S7.1, monitor the decline rate of the objective function. When the preset threshold or the maximum number of iterations is reached, terminate the calculation to obtain the final reservoir parameter set. The termination condition is that the relative decline rate of the objective function is less than 1% or the maximum number of iterations is reached. S7.2, saving the final reservoir parameter set that meets the convergence conditions, and verifying the matching degree between the final reservoir parameter set that meets the convergence conditions and the historical production data through a reservoir numerical simulator.

Images