Rapid automatic history fitting method for polymer oil displacement numerical simulation
Through the fast automatic historical fitting method of adaptive adjustment algorithm parameters, the problem of low parameter fitting efficiency in polymer oil flooding numerical simulation is solved, efficient historical fitting is achieved, parameter adjustment process is simplified, and simulation results are improved. accuracy.
Patent Information
- Application Number
- CN202310568610.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-18
- Publication Date
- 2024-11-22
AI Technical Summary
The existing numerical simulation method of polymer oil flooding is low in parameter fitting efficiency in high-temperature and high-salt reservoirs. Traditional manual adjustment of parameters is cumbersome and difficult. The automatic historical fitting effect is affected by the complexity of algorithm parameter adjustment, making it difficult to achieve fast and accurate history. Fit.
Using the fast automatic historical fitting method of adaptive adjustment algorithm parameters, the parameter adjustment is simplified by determining the objective function, setting algorithm parameters, iteratively calculating the fitting parameters, and updating the algorithm parameters. process.
The fitting efficiency of polymer oil flooding numerical simulation is significantly improved, the difficulty of manually adjusting parameters is reduced, and the rapid and accurate historical fit is achieved, and the accuracy and application of simulation results are improved.
Smart Images

Figure CN119004745A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of theoretical research on numerical simulation of oil reservoirs, technical application and the technical field of oil and gas field development to improve oil reservoir recovery, and in particular to a fast automatic history fitting method for numerical simulation of polymer oil flooding. Background Art
[0002] Polymer flooding is one of the leading technologies for chemical flooding to enhance oil recovery in my country, and is the most mature and effective among various tertiary oil recovery technologies. The main mechanism of polymer flooding technology is to add high-viscosity polymers to the water injected into the reservoir to increase the viscosity of the displacement phase, reduce the mobility ratio of the displacement fluid to the displaced fluid, and thus expand the swept volume. A considerable portion of the oil reservoirs in my country where polymer flooding is carried out are high-temperature and high-salinity reservoirs with temperatures above 75°C, such as Shengli Oilfield and Henan Oilfield. After the polymer is injected into the oil layer, it will undergo thermal degradation under high temperature conditions. This degradation will cause the polymer molecular chain to break and degrade into a small molecular structure, resulting in a decrease in the viscosity of the polymer solution.
[0003] In polymer flooding simulation, mathematical models are needed to describe physical mechanisms such as the relationship between polymer viscosity and concentration, permeability reduction coefficient, polymer adsorption and accessible pore volume. The parameters in these models determine the oil recovery effect. In the polymer flooding model, the parameters can be fitted based on laboratory experimental data, but these data are based on a limited number of cores and cannot reflect the actual situation of the polymer in the actual reservoir. In reservoir numerical simulation, it is usually necessary to fit these parameters through production data. The traditional fitting method is to manually adjust one or more parameters, run the reservoir numerical simulator, observe the difference between the calculated production data and the actual production data, and adjust the parameters again if the difference is large.
[0004] Recently, the data assimilation method based on ensemble Kalman filtering has been introduced into automatic history matching of reservoir numerical simulation. Alexandre A.Emerick et al. proposed the ensemble smoothing iterative method in the paper "Ensemble smoother with multiple data assimilation"; Yan Chen et al. proposed the ensemble randomized maximum likelihood estimation method in the paper "Ensemble Randomized MaximumLikelihood Method as an Iterative Ensemble Smoother". The objective functions of these methods are:
[0005]
[0006] Where m is the fitting parameter, is the prior variance matrix of the parameters, Φ(m) is the simulated production data, including the monthly oil production and water content of the production wells, etc., d obs For actual production data, is the variance of the production data. The parameter m can be determined by the ensemble Kalman filter method to minimize the objective function.
[0007] Obviously, the objective function is a balance between two quantities, and the optimal solution depends on the fitting parameter weights. and observation weights The fitting effect also depends on these two sets of parameters. Reservoir engineers are relatively certain about the range of fitting parameters, but it is difficult to give their variances. In practice, the effect of automatic history fitting is also achieved by adjusting these two sets of algorithm parameters. The algorithm parameters are numerous and difficult to adjust.
[0008] In the Chinese patent application with application number: CN201711108467.8, a method and system for implementing polymer flooding seepage simulation is involved. The method includes: obtaining the state equations of the three phases of oil, water and gas followed by polymer flooding based on a plurality of pre-established mathematical characterization equations; solving the continuity equations of the three phases of oil, water and gas followed by the preset conventional polymer flooding and the mass transfer diffusion equations of the chemical substances in the polymer system when they are transferred in water based on the established state equations to implement polymer flooding seepage simulation. The plurality of mathematical characterization equations include: mathematical characterization equations for polymer mechanical degradation, mathematical characterization equations for polymer dynamic adsorption, mathematical characterization equations for polymer plugging mechanisms and mathematical characterization equations for polymer viscosity enhancement characteristics. Through the embodiment scheme of the invention, the problems that the current polymer flooding simulation method does not fully describe some important polymer flooding seepage mechanisms, cannot meet the needs of field applications, and the availability, pertinence and accuracy of simulation results of some blocks are limited are solved.
[0009] In the Chinese patent application with application number: CN201510135020.4, a chemical flooding reservoir numerical simulation method is involved, which includes: step 1, optimizing the grid model design and establishing a static model; step 2, establishing a dynamic model based on the established static model; and step 3, adjusting the parameters in different time periods based on the established dynamic model to achieve time-based historical fitting, with different parameters being adjusted in different time periods and different fitting accuracy requirements at different levels. The chemical flooding reservoir numerical simulation method improves the application level of numerical simulation software, improves the accuracy of reservoir description, and provides more accurate technical support for the later stage of reservoir development.
[0010] In the Chinese patent application with application number: CN201910831256.X, a method for numerical simulation of water-soluble viscosity reducer flooding in common heavy oil reservoirs is involved. It includes: step 1, collecting basic performance parameters and viscosity reduction experimental data of water-soluble viscosity reducers; step 2, performing trend line fitting on the experiment of reducing crude oil viscosity by water-soluble viscosity reducers to determine the fitting formula; step 3, using the fitted formula to calculate the viscosity of oil-water mixtures at different water-soluble viscosity reducer mass concentrations; step 4, calculating the weight factor of water-soluble viscosity reducers at different mass concentrations in the nonlinear mixing law; step 5, establishing a numerical simulation model of water-soluble viscosity reducer flooding reservoirs; step 6, adjusting the weight factor of the water-soluble viscosity reducer, and fitting the experimental results of core flooding by water-soluble viscosity reducers with the calculation results of reservoir numerical simulation until they meet the engineering calculation accuracy. This method is convenient and fast, has few parameter requirements, and is simple to apply, and realizes the numerical simulation of water-soluble viscosity reducer flooding.
[0011] The above existing technologies are greatly different from the present invention and fail to solve the technical problem we want to solve. Therefore, we have invented a new fast automatic history matching method for polymer flooding numerical simulation. Summary of the invention
[0012] The object of the present invention is to provide a fast automatic history matching method for polymer flooding numerical simulation which realizes fast history matching by adaptively adjusting algorithm parameters.
[0013] The purpose of the present invention can be achieved by the following technical measures: a fast automatic history matching method for polymer flooding numerical simulation, the fast automatic history matching method for polymer flooding numerical simulation comprising:
[0014] Step 1: Determine the objective function of history matching;
[0015] Step 2: Set the introduced algorithm parameters;
[0016] Step 3: Enter iterative calculation to calculate the fitting objective function value
[0017] Step 4: Calculate the weighted average of the fitting parameters at the kth iteration
[0018] Step 5: Update the fitting parameters for each set
[0019] Step 6: Update the introduced algorithm parameters;
[0020] Step 7: When the number of iterations k is greater than the set number of iterations N r , then the iteration is stopped and the difference between the simulated observation and the actual observation is obtained.
[0021] The purpose of the present invention can also be achieved by the following technical measures:
[0022] In step 1, the objective function of history matching is determined:
[0023]
[0024] Where Φ(m) is the simulated observation obtained by the chemical flooding simulator, d obs For actual observation, C D Is a diagonal matrix, the diagonal elements are the weights of each observation, representing the importance of the observation; given C D The values of the diagonal elements in ; the range S for each fitting parameter is given.
[0025] In step 2, N parameter sets are generated using uniform distribution within the range S. j=1,…,N, set the number of iteration steps N r , set the introduced algorithm parameters λ0, σ0, γ0, β0, γ max , γ m in, requirements σ0>0,0<γ min ≤γ0≤γ max , β0>0; different algorithm parameters affect the convergence speed and fitting accuracy.
[0026] In step 3, for each set, a chemical flooding numerical simulator is used to calculate the fitting objective function value in is the fitting parameter of the jth set at the kth iteration step.
[0027] In step 4, collect all and Calculate the weighted average of the fitting parameters at the kth iteration step
[0028]
[0029] in is standard white noise, n is the dimension of m, e i is a column vector whose value is 1 in each direction of the i-th direction and 0 in the other directions; k , σ k , γ k , β k For the algorithm parameters introduced, it is required
[0030] In step 5, update the fitting parameters for each set
[0031]
[0032] In step 6, update the introduced algorithm parameter λ k+1 ,σ k+1 ,γ k+1 ,β k+1 ,
[0033] λ k+1 =λ0ln(k+3);
[0034] σ k+1 =σ0 / (k+2);
[0035]
[0036] β k+1 =β0(k+2).
[0037] In step 7, if the number of iterations is not greater than the set number of iterations N r , the process enters step 3 and continues to iterate.
[0038] At step 8, the difference between the simulated observation and the actual observation is:
[0039]
[0040] Where, d obs is the actual observation, N is the number of parameter sets, Φ(m j ) is the observed value obtained by simulating the jth parameter set, C D is a diagonal matrix.
[0041] The fast automatic history fitting method for polymer flooding numerical simulation in the present invention can automatically search for the optimal parameters within a given fitting parameter range, and achieve fast history fitting by adaptively adjusting the algorithm parameters. The excellent effect of the present invention is that at present, chemical flooding history fitting is mainly completed manually, with many adjustment parameters, great difficulty in adjustment, and low fitting efficiency. By proposing a fast automatic history fitting method for polymer flooding simulation parameters, it is only necessary to fit the range of parameters, further adaptively adjust the algorithm parameters, and achieve fast history fitting. Compared with traditional chemical flooding history fitting methods, the fitting efficiency is significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Figure 1 It is a MSE curve diagram in a specific embodiment of the present invention;
[0043] Figure 2 It is a schematic diagram of the curve fitting effect of the water content and cumulative oil production of No. 5 production well in a specific embodiment of the present invention;
[0044] Figure 3It is a schematic diagram of the fitting effect of the reservoir water content curve and the cumulative oil production curve in a specific embodiment of the present invention;
[0045] Figure 4 The present invention is a flowchart of a specific embodiment of the fast automatic history matching method for polymer flooding numerical simulation. DETAILED DESCRIPTION
[0046] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.
[0047] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations and / or combinations thereof.
[0048] like Figure 4 As shown, Figure 4 The flowchart of the fast automatic history matching method for polymer flooding numerical simulation of the present invention is as follows. The fast automatic history matching method for polymer flooding numerical simulation comprises the following steps:
[0049] Step 101: Determine the objective function of history matching.
[0050]
[0051] Where Φ(m) is the simulated observation obtained by the chemical flooding simulator, d obs For actual observation, C D Is a diagonal matrix, the diagonal elements are the weights of each observation, representing the importance of the observation; given C D The values of the diagonal elements in the middle; the range S of each fitting parameter is given;
[0052] Step 102: Generate N parameter sets using uniform distribution within range S j=1,…,N, set the number of iteration steps N r , set the introduced algorithm parameters λ0, σ0, γ0, β0, γ max , γ m in, requirements σ0>0,0<γ min ≤γ0≤γ max , β0>0; different algorithm parameters affect the convergence speed and fitting accuracy.
[0053] Step 103: For each set, run the chemical flooding numerical simulator to calculate the fitting objective function value in is the fitting parameter of the jth set at the kth iteration step;
[0054] Step 104: Collect All and Calculate the weighted average of the fitting parameters at the kth iteration step
[0055]
[0056] in is standard white noise, n is the dimension of m, e i is a column vector whose value is 1 in each direction of the i-th direction and 0 in the other directions; k , σ k , γ k , β k For the algorithm parameters introduced, it is required
[0057] Step 105: Update the fitting parameters for each set
[0058]
[0059] Step 106: Update the introduced algorithm parameter λ k+1 ,σ k+1 ,γ k+1 ,β k+1 ,
[0060] λ k+1 =λ0ln(k+3);
[0061] σ k+1 =σ0 / (k+2);
[0062]
[0063] β k+1 =β0(k+2);
[0064] Step 107: If the number of iterations k is greater than N r , the iteration stops and the process goes to step 108, otherwise it goes to step 103 and continues to iterate.
[0065] Step 108, output the difference between the simulated observation and the actual observation (MSE, dimensionless),
[0066]
[0067] The following are several specific embodiments of the present invention.
[0068] Embodiment 1:
[0069] In a specific embodiment 1 of the present invention, a reservoir conceptual model is established, the grid step size of the model in the x, y, and z directions is 5m×5m×5m, the grid scale is 21×21×1, the permeability is 800 millidarcy, there is one injection well and five production wells, the plane grid position of the injection well is (11,11), and the four production wells are located at the four corners of the reservoir, namely (1,1), (21,1), (1,21), (21,21). The injection-production ratio is 1:1. The simulation time is 1500 days, and the injection slug is injected: a polymer with a concentration of 1000 mg / L is injected from 500 days to 550 days, and the rest of the time is water flooding.
[0070] The fitting parameters are the accessible pore volume of the polymer, adsorption model parameters, permeability reduction factor and other model parameters, a total of 9. The initial values of the fitting parameters are different from the values that produce the "real" observed data. Model description: When running the initial model, "real" data is generated; random perturbations are added to the initial model parameters, and parameters are generated based on the perturbed model. Calculate the parameters by formula (2) Run the model to obtain the pre-fitting model; iterate the algorithm in the present invention 3 times to obtain the post-fitting model; the number of sets is N = 30, the initial values of the parameters are σ0 = 5.1, λ0 = 250, γ0 = 0.01, β0 = 30, γ max =0.1,γ min =0.001, C D =1.
[0071] Attached Figure 1 The variation of MSE with the number of iterations is described. In the numerical experiment, after three iterations, MSE dropped from 832 to 66, showing an exponential downward trend and fast convergence. MSE dropped to 87 in one iteration, a decrease of about 90%, indicating that a good fitting effect can be achieved with only one iteration.
[0072] Figure 2-3 The changes of the simulated observation data before and after fitting are described. It can be seen that after 3 iterations, the simulated calculated observation values of the fitted model are close to the "actual" values, and the fitting effect is good for the time when the "funnel" appears, indicating that the proposed algorithm can be used for fitting polymer flooding parameters.
[0073] Compared with existing algorithms, this algorithm does not need to give the distribution of fitting parameters, but only needs the range of fitting parameters. In addition, the objective function does not consider the prior error, but directly minimizes the difference between simulated observations and actual observations. It further achieves the purpose of rapid convergence by automatically adjusting the size of the introduced algorithm parameters.
[0074] Embodiment 2:
[0075] Based on Example 1, the parameters λ0=250, γ0=0.01, β0=30, γ max =0.1,γ min =0.001, C D =1, the number of sets is 30, and the iteration is 3 steps; when the parameter σ0 value is changed, when σ0=7.1, the MSE is reduced to 70; when σ0=4.1, the MSE is reduced to 88, indicating that increasing this parameter affects the fitting accuracy. When this parameter is between 5 and 6, a higher fitting accuracy can be obtained.
[0076] Embodiment 3:
[0077] Based on Example 1, the parameters are σ0=5.1, γ0=0.01, β0=30, γ max =0.1,γ min =0.001, C D =1, the number of sets is 30, and the iteration is 3 steps; when the value of parameter λ0 is changed, when λ0=50, MSE drops to 180, and when λ0=150, MSE drops to 70, which shows that this parameter affects the fitting accuracy. Increasing this parameter can improve the accuracy, but when this parameter increases to a certain value, the accuracy improvement becomes smaller; when this parameter is set to about 250, a higher fitting accuracy can be obtained.
[0078] Embodiment 4:
[0079] Based on Example 1, the parameters are σ0=5.1, λ0=250, β0=30, γ max =0.1,γ min =0.001, C D =1, the number of sets is 30, and the iteration is 3 steps; when the value of γ0 is changed, when γ0=0.02, the MSE is reduced to 62, which is better than the effect of Example 1, indicating that increasing this parameter improves the convergence speed. However, if the parameter is too large, it will cause oscillation or non-convergence, so the value of this parameter cannot be too large.
[0080] Embodiment 5:
[0081] Based on Example 1, the parameters are σ0=5.1, λ0=250, γ0=0.01, γ max =0.1,γ min =0.001, C D=1. The number of sets is 30, and the iteration is 3 steps; the value of β0 is changed. When β0=80, the MSE is reduced to 63, which is better than the effect of Example 1, indicating that increasing the value of this parameter can improve the accuracy. Similar to the parameter λ0, further increasing this value will reduce the improvement of the accuracy; when this parameter is about 80, a higher fitting accuracy can be achieved.
[0082] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions recorded in the aforementioned embodiments or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
[0083] Except for the technical features described in the specification, all other technical features are known technologies to those skilled in the art.
Claims
1. A fast automatic history matching method for polymer flooding numerical simulation, characterized in that: The fast automatic history matching method for polymer flooding numerical simulation includes: Step 1: Determine the objective function of history matching; Step 2: Set the introduced algorithm parameters; Step 3: Enter iterative calculation to calculate the fitting objective function value Step 4: Calculate the weighted average of the fitting parameters at the kth iteration Step 5: Update the fitting parameters for each set Step 6: Update the introduced algorithm parameters; Step 7: When the number of iterations k is greater than the set number of iterations N r , then the iteration is stopped and the difference between the simulated observation and the actual observation is obtained.
2. The rapid automatic history matching method for polymer flooding numerical simulation according to claim 1, characterized in that: In step 1, the objective function of history matching is determined: Where Φ(m) is the simulated observation obtained by the chemical flooding simulator, d obs For actual observation, C D Is a diagonal matrix, the diagonal elements are the weights of each observation, representing the importance of the observation; given C D The values of the diagonal elements in ; the range S for each fitting parameter is given.
3. The rapid automatic history matching method for polymer flooding numerical simulation according to claim 1, characterized in that: In step 2, N parameter sets are generated using uniform distribution within the range S. Set the number of iterations N r , set the introduced algorithm parameters λ0, σ0, γ0, β0, γ max , γ min ,Require σ0>0,0<γ min ≤γ0≤γ max , β0>0; different algorithm parameters affect the convergence speed and fitting accuracy.
4. The rapid automatic history matching method for polymer flooding numerical simulation according to claim 1, characterized in that: In step 3, for each set, a chemical flooding numerical simulator is used to calculate the fitting objective function value in is the fitting parameter of the jth set at the kth iteration step.
5. The rapid automatic history matching method for polymer flooding numerical simulation according to claim 4, characterized in that: In step 4, collect all and Calculate the weighted average of the fitting parameters at the kth iteration step in is standard white noise, E[w i ]=0, n is the dimension of m, e i is a column vector whose value is 1 in each direction of the i-th direction and 0 in the other directions; k , σ k , γ k , β k For the algorithm parameters introduced, it is required 6. The rapid automatic history matching method for polymer flooding numerical simulation according to claim 5, characterized in that: In step 5, update the fitting parameters for each set 7. The rapid automatic history matching method for polymer flooding numerical simulation according to claim 6, characterized in that: In step 6, update the introduced algorithm parameter λ k+1 ,σ k+1 ,γ k+1 ,β k+1 , l k+1 =λ0ln(k+3); s k+1 =σ0 / (k+2); β k+1 =β0(k+2)。 8. The rapid automatic history matching method for polymer flooding numerical simulation according to claim 1, characterized in that: In step 7, if the number of iterations is not greater than the set number of iterations N r , the process enters step 3 and continues to iterate.
9. The rapid automatic history matching method for polymer flooding numerical simulation according to claim 1, characterized in that: At step 8, the difference between the simulated observation and the actual observation is: Where, d obs is the actual observation, N is the number of parameter sets, Φ(m j ) is the observed value obtained by simulating the jth parameter set, C D is a diagonal matrix.
Citation Information
Cited By
Geothermal reservoir productivity decline prediction method, device, equipment and medium
CN122434008A
