A flue gas reinjection parameter optimization method based on differential evolution algorithm

The flue gas return parameters are optimized through differential evolution algorithm and response surface regression model, and the flue gas component optimization problem is solved, efficient reservoir development is achieved, and gas utilization and reservoir recovery effect are improved.

CN116561907BActive Publication Date: 2024-11-19CHINA UNIV OF PETROLEUM (EAST CHINA)
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310390598.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-04-12
Publication Date
2024-11-19
Estimated Expiration
2043-04-12

AI Technical Summary

Technical Problem

The prior art lacks effective gas component optimization methods during the flue gas re-injection process, resulting in high optimization costs, poor adaptability to the results, and unclear oil-driving mechanism for different proportions of gas components.

Method used

The flue gas return parameter optimization method based on differential evolution algorithm is adopted, combined with machine learning and response plane regression model, the gas ratio, gas injection speed, gas injection volume and injection pressure in the flue gas are optimized, and numerical simulation and polynomial response are used. The surface model is constructed to achieve automatic optimization under different reservoir conditions.

Benefits of technology

It improves the accuracy and efficiency of the flue gas return strategy, enhances the displacement effect of CO2 and N2, achieves rapid optimization of different types of reservoirs, and improves the reservoir development effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116561907B_ABST
    Figure CN116561907B_ABST
Patent Text Reader

Abstract

The present invention discloses a flue gas reinjection parameter optimization method based on a differential evolution algorithm, which belongs to the field of petroleum engineering oil and gas field development engineering, and includes the following steps: step 1, determining the gas injection parameters of the flue gas reinjection process; step 2, designing a numerical simulation scheme based on the flue gas reinjection parameters, and establishing a flue gas reinjection strategy optimization sample set by changing different gas ratios; step 3, constructing a polynomial response surface regression model, and establishing the correlation between the gas injection parameters; step 4, establishing a gas injection parameter optimization mathematical model based on a differential evolution algorithm, and realizing the automatic optimization of flue gas components and gas injection parameter combinations under different reservoir conditions through the algorithm to obtain the optimal solution, and finally outputting the flue gas components, gas injection speed, gas injection volume and injection pressure corresponding to the optimal solution. The present invention can realize the rapid optimization of flue gas reinjection strategies for different types of reservoirs, and greatly improve the optimization efficiency and the accuracy of the optimization results.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of petroleum engineering and oil and gas field development engineering, and specifically relates to a flue gas reinjection parameter optimization method based on a differential evolution algorithm. Background Art

[0002] Flue gas is a product generated after the complete combustion of organic matter such as natural gas, crude oil or coal, and its main components are N2 and CO2. As an industrial waste gas, flue gas has the advantages of low cost and stable gas source compared to hydrocarbon gas and CO2. At the same time, the use of flue gas for oil recovery can achieve the purpose of storing harmful gases, saving energy and reducing emissions. Flue gas reinjection can give full play to the dissolution and viscosity reduction effect of CO2 and the elastic performance increase of N2, but the combined effect of gas components in different proportions is still unclear. Therefore, the optimization and evaluation of gas components in the flue gas reinjection process needs to be studied urgently.

[0003] At present, the research and application of flue gas to enhance oil recovery are mainly focused on gas injection and oil recovery. The composition of the injected flue gas is fixed, and the flue gas is used to improve the reservoir development effect by solubilizing, reducing viscosity, miscible phases and gravity separation of crude oil. Most studies focus on the influence of injecting single gases such as CO2 and N2. As a mixed gas of the two, the oil recovery mechanism of flue gas with different proportions of gas components is still unclear, and the optimization method is mainly based on single factor experiments, which makes it difficult to achieve global optimization of the reinjection process. Therefore, it is possible to consider combining machine learning algorithms to optimize the multi-factor collaborative optimization of reinjection parameters and analyze the interactions between the various influencing factors. According to the properties of CO2 and N2, the gas ratio in the flue gas is optimized in combination with different gas injection volumes and injection pressures to further improve the gas drive development effect.

[0004] In summary, there are few studies on the optimization of flue gas reinjection components, and there are defects such as high optimization cost and poor adaptability of optimization results. Summary of the invention

[0005] In order to solve the above problems, the present invention proposes a flue gas reinjection parameter optimization method based on differential evolution algorithm, combines machine learning optimization algorithm and phase characteristics of flue gas reinjection, optimizes the design of development strategies and component optimization schemes for flue gas reinjection, and realizes rapid optimization of reinjection parameters when carrying out flue gas reinjection in different types of reservoirs.

[0006] The technical solution of the present invention is as follows:

[0007] A flue gas reinjection parameter optimization method based on differential evolution algorithm comprises the following steps:

[0008] Step 1, determining the gas injection parameters of the flue gas reinjection process;

[0009] Step 2: design a numerical simulation scheme based on the flue gas reinjection parameters, and establish a flue gas reinjection strategy optimization sample set by changing different gas ratios;

[0010] Step 3: construct a polynomial response surface regression model to establish the correlation between gas injection parameters;

[0011] Step 4: Establish a mathematical model for gas injection parameter optimization based on the differential evolution algorithm. The algorithm is used to automatically select the flue gas components and gas injection parameter combinations under different reservoir conditions to obtain the optimal solution. Finally, the flue gas components, gas injection rate, gas injection volume and injection pressure corresponding to the optimal solution are output.

[0012] Furthermore, the gas injection parameters of the flue gas reinjection process are flue gas composition, gas injection rate, gas injection volume and injection pressure; the evaluation index of the reinjection development effect is the cumulative oil production.

[0013] Furthermore, the specific process of step 2 is as follows: design the N2:CO2 ratio to be reduced from the initial ratio of 3:1 to 1:3, set the parameter range of gas injection rate, injection volume and injection pressure according to the actual reservoir conditions, use the Latin hypercube sampling method to generate a large number of numerical simulation schemes within the value range, and establish a flue gas reinjection strategy optimization sample set.

[0014] Furthermore, the specific process of step 3 is:

[0015] Step 3.1, construct a polynomial response surface regression model;

[0016] A polynomial function response surface equation based on Taylor series expansion is established. The response surface equation is as follows:

[0017] f(d)=f(d1,d2,…,d i ,…,d m ) (1);

[0018] Where f(d) represents the cumulative oil production of the reservoir, d i represents the i-th flue gas reinjection parameter, and m represents the total number of flue gas reinjection parameters;

[0019] Based on the polynomial function of Taylor series expansion, various forms of polynomial response surface regression models are constructed by the progression of orders. Formula (1) is expanded into seven different forms of polynomial functions according to Taylor series, and seven different forms of response surface regression models are constructed as follows:

[0020] Second-order model without interaction terms:

[0021]

[0022] Where f1(d) represents the cumulative oil production of the reservoir in the second-order model without cross terms; α0 represents the constant term to be determined; α i represents the undetermined coefficient of the i-th flue gas reinjection parameter; α ii represents the undetermined coefficient of the square term of the i-th flue gas reinjection parameter;

[0023] Full second-order model:

[0024]

[0025] Where f2(d) represents the cumulative oil production of the reservoir in the complete second-order model; α ij represents the undetermined coefficient of the cross term between the i-th and j-th flue gas reinjection parameters;

[0026] The third-order model without interaction terms:

[0027]

[0028] Where f3(d) represents the cumulative oil production of the reservoir in the third-order model without cross terms; α iii represents the undetermined coefficient of the cubic term of the i-th flue gas reinjection parameter;

[0029] The third-order model with cubic interactions:

[0030]

[0031] Where f4(d) represents the cumulative oil production of the reservoir in the third-order model with cubic cross terms; α ijk represents the unknown coefficients of the cross terms of the i-th, j-th and k-th flue gas reinjection parameters;

[0032] Third-order model with squared cross terms:

[0033]

[0034] Where f5(d) represents the cumulative oil production of the reservoir in the third-order model with square cross terms; α iij represents the undetermined coefficient of the cross term between the square of the i-th flue gas reinjection parameter and the j-th flue gas reinjection parameter;

[0035] Full third-order model:

[0036]

[0037] Where, f6(d) represents the cumulative oil production of the reservoir in the complete third-order model;

[0038] The fourth-order model without interaction terms:

[0039]

[0040] Where f6(d) represents the cumulative oil production of the reservoir in the fourth-order model without cross terms;

[0041] Step 3.2, optimizing the response surface model form;

[0042] First, the F test is performed on the response surface regression model of the same form to determine the rationality of the response surface regression model of different forms; the F statistic of each response surface equation is calculated using the variance analysis method:

[0043]

[0044] Among them, F represents the statistical quantity of the influence of flue gas reinjection parameters on the cumulative oil production of the reservoir, Q R represents the regression sum of squares, Q E represents the residual sum of squares, n represents the number of flue gas reinjection parameters, and p represents the number of non-constant terms in the response surface equation;

[0045] For a given significance level γ, look up the F statistic critical value table F γ (p,n--1), compare F with it, if F≥F γ The model is considered reasonable, otherwise it is unreasonable;

[0046] Secondly, using the coefficient of determination R 2 The two efficiency evaluation indicators of RMSE and RMS were used to analyze the response surface regression models of different forms, and the best response surface regression model was selected; the best response surface regression model was the determination coefficient R 2 The model with the maximum and minimum root mean square error RMSE;

[0047] Step 3.3, constructing a polynomial response surface regression model of gas injection parameters under different reservoir conditions;

[0048] Based on the response surface model after form optimization, basic numerical simulation models for different types of reservoirs are established, and polynomial response surface regression models of gas injection parameters under different reservoir conditions are constructed according to the reservoir parameters of different types of reservoirs themselves.

[0049] Furthermore, the specific process of step 4 is:

[0050] Step 4.1, establishing a mathematical model for optimizing gas injection parameters;

[0051] The established optimization mathematical model includes three factors, namely, objective function, decision variables and constraints;

[0052] A multi-objective optimization mathematical model is established with the maximum cumulative oil production as the objective function. The objective function is shown in formula (10):

[0053] Max(cumulative oil production) (10);

[0054] In the optimization problem of flue gas reinjection strategy, the decision variables are the gas injection parameters, including four decision variables: flue gas composition, gas injection rate, gas injection volume and injection pressure;

[0055] X i =[x i1 ,x i2 ,…,x in ] (11);

[0056] In the formula, X is each decision variable, x in is the nth value of the ith decision variable;

[0057] The constraints are the range of flue gas component selection and the upper and lower limits of the gas injection working system change;

[0058] G∈[3:1,2:1,1:1,1:2,1:3](12);

[0059] Q jmin ≤Q j ≤Q jmax (13);

[0060] Where G represents the range constraint of flue gas component selection, specifically the ratio of N2 to CO2 in flue gas; Q j Indicates the upper and lower limits of the gas injection working system. The gas injection working system is specifically the various gas injection parameters, including gas injection speed, gas injection volume and injection pressure. Q jmax Indicates the maximum value of gas injection parameters, Q jmin Indicates the minimum value of gas injection parameters;

[0061] Step 4.2: Optimize the flue gas reinjection strategy parameters based on the differential evolution algorithm; the specific process is as follows:

[0062] First, the parameters are initialized and the initial values ​​of the four input parameters of flue gas composition, injection rate, injection volume and injection pressure are assigned. Then, the differential evolution algorithm is applied to update the variables. The values ​​are randomly updated within the constraints according to the constraints, and the fitness of each individual in the population is calculated. Then, an iterative loop is entered. When the termination condition is not met, mutation, crossover and selection operations are performed in sequence. Then the number of iterations is increased by one and the next iteration is entered. The fitness of each individual in the population is recalculated until the number of iterations is reached to obtain the optimal result. The optimal result at this time is the optimal solution obtained by automatic optimization, and finally the flue gas composition, injection rate, injection volume and injection pressure corresponding to the optimal solution are output.

[0063] Beneficial technical effects brought by the present invention:

[0064] By combining machine learning algorithms to optimize flue gas reinjection parameters, the ratio of CO2 and N2 in the parameters is optimized and adjusted, so that CO2 and N2 in the flue gas can fully play their role in the displacement stage, while ensuring high oil production and improving gas utilization, and realizing precise regulation of flue gas reinjection strategies. The present invention optimizes the injection speed, injection volume and injection pressure simultaneously when optimizing the components. The present invention can realize rapid optimization of flue gas reinjection strategies for different types of reservoirs, greatly improving the optimization efficiency and the accuracy of the optimization results; compared with the existing gas injection parameter optimization method, the present invention can coordinately optimize the reinjection components of flue gas, and can provide the most optimized flue gas reinjection strategy for different oil reservoirs, which is the key to the later development and adjustment of oil reservoirs. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 This is a flow chart of the flue gas reinjection parameter optimization method based on the differential evolution algorithm of the present invention.

[0066] Figure 2 This is an optimization flow chart of the differential evolution algorithm in the present invention. DETAILED DESCRIPTION

[0067] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0068] like Figure 1 As shown, a flue gas reinjection parameter optimization method based on differential evolution algorithm adopts a machine learning algorithm of differential evolution, which specifically includes the following steps:

[0069] Step 1: Determine the gas injection parameters for the flue gas reinjection process.

[0070] The flue gas reinjection process can be approximated as a combined displacement process of CO2 and N2. The present invention starts from the displacement parameters directly related to oil production and gas production, and determines the injection parameters of the flue gas reinjection process as flue gas composition, injection speed, injection volume and injection pressure. The evaluation index of the reinjection development effect is the cumulative oil production.

[0071] Step 2: Design a numerical simulation scheme based on the flue gas reinjection parameters, and establish a flue gas reinjection strategy optimization sample set by changing different gas ratios. The specific process is:

[0072] The ratio of CO2 and N2 in the flue gas is regulated to make the content of N2 in the flue gas higher than that of CO2. The ratio of N2:CO2 is designed to be reduced from the initial ratio of 3:1 to 1:3. The parameter ranges of gas injection rate, injection volume and injection pressure are set according to the actual reservoir conditions. A large number of numerical simulation schemes are generated within the value range using the Latin hypercube sampling method to establish a sample set for optimizing the flue gas reinjection strategy.

[0073] Step 3: Construct a polynomial response surface regression model to establish the correlation between gas injection parameters.

[0074] With the help of the flue gas reinjection strategy optimization sample set generated in step 2, the response surface analysis method is used to construct the response surface regression model of gas injection parameters, and the correlation between gas injection parameters is established to achieve accurate prediction of the cumulative oil production under different gas injection parameters. The specific process is as follows:

[0075] Step 3.1, construct a polynomial response surface regression model;

[0076] The present invention designs and establishes a polynomial function response surface equation based on Taylor series expansion. The response surface equation is as follows. It is concise and clear, and easy to understand. It replaces complex implicit relationships with intuitive explicit functions, and has obvious advantages in solving the correlation of gas injection parameters.

[0077] f(d)=f(d1,d2,…,d i ,…,d m ) (1);

[0078] Where f(d) represents the cumulative oil production of the reservoir, d i represents the i-th flue gas reinjection parameter, and m represents the total number of flue gas reinjection parameters.

[0079] Based on the polynomial function of Taylor series expansion, various forms of polynomial response surface regression models are constructed by the progression of orders. Formula (1) is expanded into seven different forms of polynomial functions according to Taylor series, and seven different forms of response surface regression models are constructed as follows.

[0080] Second-order model without interaction terms:

[0081]

[0082] Where f1(d) represents the cumulative oil production of the reservoir in the second-order model without cross terms; α0 represents the constant term to be determined; α i represents the undetermined coefficient of the i-th flue gas reinjection parameter; α ii represents the undetermined coefficient of the square term of the i-th flue gas reinjection parameter;

[0083] Full second-order model:

[0084]

[0085] Where f2(d) represents the cumulative oil production of the reservoir in the complete second-order model; α ij represents the undetermined coefficient of the cross term between the ith and jth flue gas reinjection parameters;

[0086] The third-order model without interaction terms:

[0087]

[0088] Where f3(d) represents the cumulative oil production of the reservoir in the third-order model without cross terms; α iii represents the undetermined coefficient of the cubic term of the i-th flue gas reinjection parameter;

[0089] The third-order model with cubic interactions:

[0090]

[0091] Where f4(d) represents the cumulative oil production of the reservoir in the third-order model with cubic cross terms; α ijk represents the unknown coefficients of the cross terms of the i-th, j-th and k-th flue gas reinjection parameters;

[0092] Third-order model with squared cross terms:

[0093]

[0094] Where f5(d) represents the cumulative oil production of the reservoir in the third-order model with square cross terms; α iij represents the undetermined coefficient of the cross term between the square of the i-th flue gas reinjection parameter and the j-th flue gas reinjection parameter;

[0095] Full third-order model:

[0096]

[0097] Where, f6(d) represents the cumulative oil production of the reservoir in the complete third-order model;

[0098] The fourth-order model without interaction terms:

[0099]

[0100] Where f6(d) represents the cumulative oil production of the reservoir in the fourth-order model without cross terms;

[0101] Step 3.2, optimizing the response surface model form;

[0102] First, the F test is performed on the response surface regression model of the same form to determine the rationality of the response surface regression model of different forms. The F statistic of each response surface equation is calculated using the variance analysis method:

[0103]

[0104] Among them, F represents the statistical quantity of the influence of flue gas reinjection parameters on the cumulative oil production of the reservoir, Q R represents the regression sum of squares, Q Erepresents the residual sum of squares, n represents the number of flue gas reinjection parameters, and p represents the number of non-constant terms in the response surface equation.

[0105] For a given significance level γ, look up the F statistic critical value table F γ (p,np-1), compare F with it, if F ≥ F γ The model is considered reasonable, otherwise it is unreasonable.

[0106] Secondly, using two efficiency evaluation indicators (determination coefficient R 2 The response surface regression models of different forms were analyzed by using the root mean square error (RMSE) and the best response surface regression model was selected. The best response surface regression model was the one with the coefficient of determination R 2 The model corresponding to the maximum and minimum root mean square error RMSE. Step 3.3, construct a polynomial response surface regression model of gas injection parameters under different reservoir conditions;

[0107] Based on the response surface model after form optimization, the basic numerical simulation model of different types of reservoirs is established, and the polynomial response surface regression model of gas injection parameters under different reservoir conditions is constructed according to the reservoir parameters of different types of reservoirs. The constructed polynomial response surface regression model of gas injection parameters under different types of reservoir conditions can reflect the correlation between gas injection parameters and provide a more accurate and efficient proxy model for the optimization of subsequent flue gas reinjection strategy.

[0108] Step 4: Establish a mathematical model for gas injection parameter optimization based on the differential evolution algorithm to optimize the flue gas reinjection parameters. The algorithm is used to automatically select the flue gas components and gas injection parameter combinations under different reservoir conditions to obtain the optimal solution. Finally, the flue gas components, injection rate, injection volume and injection pressure corresponding to the optimal solution are output.

[0109] Step 4.1, establishing a mathematical model for optimizing gas injection parameters;

[0110] The established optimization mathematical model mainly includes three factors, namely objective function, decision variables and constraints.

[0111] (1) Objective function

[0112] In order to ensure that underground crude oil can be efficiently extracted in the CO2 displacement development project and avoid extreme development plans caused by over-emphasizing the gas injection development effect, a multi-objective optimization mathematical model is established with the maximum cumulative oil production as the objective function. The objective function is shown in formula (10), and the objective is finally formulated as a maximization problem:

[0113] Max(cumulative oil production) (10);

[0114] (2) Decision variables

[0115] In the optimization problem of flue gas reinjection strategy, the decision variables are the gas injection parameters, including four decision variables: flue gas composition, gas injection rate, gas injection volume and injection pressure;

[0116] X i =[x i1 ,x i2 ,…,x in ] (11);

[0117] In the formula, X is each decision variable, x in is the nth value of the ith decision variable;

[0118] (3) Constraints

[0119] The constraints that need to be considered in this optimization model are the range constraints for flue gas component selection and the upper and lower limit constraints for changes in the gas injection working system.

[0120] G∈[3:1,2:1,1:1,1:2,1:3](12);

[0121] Q jmin ≤Q j ≤Q jmax (13);

[0122] Where G represents the range constraint of flue gas component selection, specifically the ratio of N2 to CO2 in flue gas; Q j Indicates the upper and lower limits of the gas injection working system. The gas injection working system is specifically the various gas injection parameters, including gas injection speed, gas injection volume and injection pressure. Q jmax Indicates the maximum value of gas injection parameters, Q jmin Indicates the minimum value of the gas injection parameter.

[0123] Step 4.2, optimizing flue gas reinjection strategy parameters based on differential evolution algorithm;

[0124] The differential evolution algorithm DE is a type of population-based adaptive global optimization algorithm. It is a type of evolutionary algorithm that simulates the process of cooperation and competition among individuals in a population. It uses computer programming to solve the optimal solution to the problem through random search and selection memory in a continuous domain space. The algorithm has a simple principle and few control parameters, only crossover probability and scaling factor. It is highly robust and easy to implement. In the differential evolution algorithm, the object of evolution is called a population, which is composed of multiple groups of individuals. The algorithm includes four genetic operations: population initialization, mutation, crossover, and selection. When all individuals in the population complete the four genetic operations once, the population completes a generation of evolution. The specific process is as follows: Figure 2As shown, the parameters are initialized first, and the initial values ​​of the four input parameters of flue gas composition, injection speed, injection volume and injection pressure are assigned. Then, the differential evolution algorithm is applied to update the variables, and the values ​​are randomly updated within the constraints according to the constraints. The fitness value of each individual in the population is calculated, and the iteration cycle is entered. When the termination condition is not met, the mutation, crossover and selection operations are performed in sequence, and then the number of iterations t is increased by one, and the next iteration is entered. The fitness value of each individual in the population is recalculated until the number of iterations is reached to obtain the optimal result. The optimal result at this time is the optimal solution obtained by automatic optimization, and the flue gas composition, injection speed, injection volume and injection pressure corresponding to the optimal solution are finally output.

[0125] The present invention couples the response surface regression model with the differential evolution algorithm to establish a method for optimizing gas injection parameters for a flue gas reinjection strategy, wherein the response surface regression model is mainly used to establish a mapping relationship between different injection-production parameter combinations and development effects, and the differential evolution algorithm is mainly used to automatically optimize flue gas components and gas injection parameter combinations under different reservoir conditions.

[0126] In order to demonstrate the feasibility and superiority of the present invention, the following examples are given. The present invention takes a block of Daqing Oilfield as an example. The reservoir of this oilfield block is dense, with an average porosity of 8%, and the average permeability of the core analysis is only 1×10 -3 μm 2 A core-scale one-injection-one-production displacement numerical simulation model was established for the oilfield block, and production wells were set up for constant pressure production.

[0127] The specific steps of optimizing the flue gas reinjection components and injection parameters using the method of the present invention are as follows:

[0128] Step 1: According to the reservoir properties of the target block, a core-scale displacement numerical simulation model is established, and the parameter range of the optimization variables of the target block is determined.

[0129] Step 2. Design the N2:CO2 ratio to be 3:1, 2:1, 1:1, 1:2, and 1:3 respectively. Set the parameter range of injection rate, injection volume, and injection pressure according to the actual reservoir conditions. Call the numerical simulator to automatically construct the optimized sample of flue gas reinjection strategy. The input parameters are the injection rate, injection volume, and injection pressure of the flue gas components. The output parameter is the cumulative oil production.

[0130] Step 3. First, based on the polynomial function of Taylor series expansion, seven different forms of polynomial response surface models are constructed by the progression of orders; secondly, these different forms of response surface models are subjected to F-test to determine the rationality of the response surface model. The F statistic and its critical value are shown in Table 1.

[0131] Table 1 F statistics and critical values ​​of seven response surface models

[0132]

[0133]

[0134] The 7 response surface equations in Table 1 correspond to seven different forms of response surface regression models represented by formula (2) to formula (8). As shown in Table 1, the statistics of each response surface equation are much larger than their critical values, indicating that these 7 response surface regression models are effective and reasonable. Then, the overall accuracy of the response surface regression model is evaluated, and the best response surface regression model is selected. The R 2 The RMSE evaluation results are shown in Table 2.

[0135] Table 2 Overall accuracy of seven response surface models

[0136]

[0137] Based on the analysis results of the frequency response surface model and the vibration response surface model of typical measuring points, f4(d) has higher model accuracy than other models. Therefore, the best response surface model is determined to be f4(d) - a third-order model with cubic cross terms.

[0138] Step 4: Coupling the response surface regression model and the differential evolution algorithm to establish a flue gas reinjection strategy gas injection parameter optimization method, wherein the response surface regression model is mainly used to establish the mapping relationship between the input parameters and the development effect, and the differential evolution algorithm is mainly used to automatically optimize the flue gas components and gas injection parameter combinations under different reservoir conditions. According to the designed mathematical model of the optimization problem, the differential evolution algorithm is called to optimize the flue gas reinjection strategy. According to the rationality judgment and comprehensive analysis of the identification results, the most suitable differential evolution algorithm parameter settings with the fastest convergence speed and the highest convergence accuracy for the present invention are shown in Table 3.

[0139] Table 3 Operation method and parameter setting values ​​of differential evolution algorithm

[0140]

[0141] Table 4 Comparison between basic solution and optimized solution

[0142]

[0143] Table 4 shows the comparison results between the basic scheme and the optimized scheme. The basic scheme is the data scheme designed initially, and the optimized scheme is the scheme obtained after optimization using the method of the present invention. As can be seen from Table 4, the optimization scheme of the coupling response surface regression model and the differential evolution algorithm of the present invention has a higher cumulative oil production. Therefore, by optimizing the flue gas components and gas injection parameters, the formation development potential can be maximized and the oil reservoir recovery can be improved. It is proved that this method can solve the optimization problem of flue gas components very well, and the schemes obtained by its optimization calculation have good performance in development effect, indicating that the flue gas reinjection parameter optimization method based on the differential evolution algorithm has a high degree of credibility and can well assist reservoir engineers in deployment decisions.

[0144] 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 protection scope of the present invention.

Claims

1. A flue gas reinjection parameter optimization method based on differential evolution algorithm, It is characterized in that The steps include: Step 1, determining the gas injection parameters of the flue gas reinjection process; Step 2: design a numerical simulation scheme based on the flue gas reinjection parameters, and establish a flue gas reinjection strategy optimization sample set by changing different gas ratios; Step 3: Construct a polynomial response surface regression model to establish the correlation between gas injection parameters; the specific process is as follows: Step 3.1, construct a polynomial response surface regression model; A polynomial function response surface equation based on Taylor series expansion is established. The response surface equation is as follows: f(d)=f(d 1 ,d 2 ,…,d i ,…,d m ) (1); Where f(d) represents the cumulative oil production of the reservoir, d i represents the i-th flue gas reinjection parameter, and m represents the total number of flue gas reinjection parameters; Based on the polynomial function of Taylor series expansion, various forms of polynomial response surface regression models are constructed by the progression of orders. Formula (1) is expanded into seven different forms of polynomial functions according to Taylor series, and seven different forms of response surface regression models are constructed as follows: Second-order model without interaction terms: Among them, f 1 (d) represents the cumulative oil production of the reservoir in the second-order model without cross terms; α 0 represents the unknown coefficient of the constant term; α i represents the undetermined coefficient of the i-th flue gas reinjection parameter; α ii represents the undetermined coefficient of the square term of the i-th flue gas reinjection parameter; Full second-order model: Among them, f 2 (d) represents the cumulative oil production of the reservoir in the complete second-order model; α ij represents the undetermined coefficient of the cross term between the i-th and j-th flue gas reinjection parameters; The third-order model without interaction terms: Among them, f 3 (d) represents the cumulative oil production of the reservoir in the third-order model without cross terms; α iii represents the undetermined coefficient of the cubic term of the i-th flue gas reinjection parameter; The third-order model with cubic interactions: Among them, f 4 (d) represents the cumulative oil production of the reservoir in the third-order model with cubic cross terms; α ijk represents the unknown coefficients of the cross terms of the i-th, j-th and k-th flue gas reinjection parameters; Third-order model with squared cross terms: Among them, f 5 (d) represents the cumulative oil production of the reservoir in the third-order model with square cross terms; α iij represents the undetermined coefficient of the cross term between the square of the i-th flue gas reinjection parameter and the j-th flue gas reinjection parameter; Full third-order model: Among them, f 6 (d) represents the cumulative oil production of the reservoir in the complete third-order model; The fourth-order model without interaction terms: Among them, f 6 (d) represents the cumulative oil production of the reservoir in the fourth-order model without cross terms; Step 3.2, optimizing the response surface model form; First, the F test is performed on the response surface regression model of the same form to determine the rationality of the response surface regression model of different forms; the F statistic of each response surface equation is calculated using the variance analysis method: Among them, F represents the statistical quantity of the influence of flue gas reinjection parameters on the cumulative oil production of the reservoir, Q R represents the regression sum of squares, Q E represents the residual sum of squares, n represents the number of flue gas reinjection parameters, and p represents the number of non-constant terms in the response surface equation; For a given significance level γ, look up the F statistic critical value table F γ (p,np-1), compare F with it, if F ≥ F γ The model is considered reasonable, otherwise it is unreasonable; Secondly, using the coefficient of determination R 2 The two efficiency evaluation indicators of RMSE and RMS were used to analyze the response surface regression models of different forms, and the best response surface regression model was selected; the best response surface regression model was the determination coefficient R 2 The model with the maximum and minimum root mean square error RMSE; Step 3.3, constructing a polynomial response surface regression model of gas injection parameters under different reservoir conditions; Based on the response surface model after form optimization, the basic numerical simulation model of different types of reservoirs is established, and the polynomial response surface regression model of gas injection parameters under different reservoir conditions is constructed according to the reservoir parameters of different types of reservoirs. Step 4: Establish a mathematical model for gas injection parameter optimization based on the differential evolution algorithm. The algorithm is used to automatically select the flue gas components and gas injection parameter combinations under different reservoir conditions to obtain the optimal solution. Finally, the flue gas components, gas injection rate, gas injection volume and injection pressure corresponding to the optimal solution are output.

2. According to the flue gas reinjection parameter optimization method based on differential evolution algorithm according to claim 1, It is characterized in that The injection parameters of the flue gas reinjection process are flue gas composition, injection rate, injection volume and injection pressure; the evaluation index of the reinjection development effect is the cumulative oil production.

3. According to the flue gas reinjection parameter optimization method based on differential evolution algorithm according to claim 1, It is characterized in that The specific process of step 2 is: design N 2 :CO 2 The ratio is reduced from the initial ratio of 3:1 to 1:

3. The parameter ranges of gas injection rate, injection volume and injection pressure are set according to the actual reservoir conditions. The Latin hypercube sampling method is used to generate a large number of numerical simulation schemes within the value range, and the optimized sample set of flue gas reinjection strategy is established.

4. According to the flue gas reinjection parameter optimization method based on differential evolution algorithm according to claim 1, It is characterized in that The specific process of step 4 is as follows: Step 4.1, establishing a mathematical model for optimizing gas injection parameters; The established optimization mathematical model includes three factors, namely, objective function, decision variables and constraints; A multi-objective optimization mathematical model is established with the maximum cumulative oil production as the objective function. The objective function is shown in formula (10): Max(cumulative oil production) (10); In the optimization problem of flue gas reinjection strategy, the decision variables are the gas injection parameters, including four decision variables: flue gas composition, gas injection rate, gas injection volume and injection pressure; X i =[x i1 ,x i2 ,…,x in ] (11); In the formula, X is each decision variable, x in is the nth value of the ith decision variable; The constraints are the range of flue gas component selection and the upper and lower limits of the gas injection working system change; G∈[3:1,2:1,1:1,1:2,1:3](12); Q jmin ≤Q j ≤Q jmax (13); In the formula, G represents the range constraint of flue gas component selection, specifically N 2 With CO 2 Ratio; Q j Indicates the upper and lower limits of the gas injection working system. The gas injection working system is specifically the various gas injection parameters, including gas injection speed, gas injection volume and injection pressure. Q jmax Indicates the maximum value of gas injection parameters, Q jmin Indicates the minimum value of gas injection parameters; Step 4.2: Optimize the flue gas reinjection strategy parameters based on the differential evolution algorithm; the specific process is as follows: First, the parameters are initialized and the initial values ​​of the four input parameters of flue gas composition, injection rate, injection volume and injection pressure are assigned. Then, the differential evolution algorithm is applied to update the variables. The values ​​are randomly updated within the constraints according to the constraints, and the fitness of each individual in the population is calculated. Then, an iterative loop is entered. When the termination condition is not met, mutation, crossover and selection operations are performed in sequence. Then the number of iterations is increased by one and the next iteration is entered. The fitness of each individual in the population is recalculated until the number of iterations is reached to obtain the optimal result. The optimal result at this time is the optimal solution obtained by automatic optimization, and finally the flue gas composition, injection rate, injection volume and injection pressure corresponding to the optimal solution are output.