Oil reservoir inversion method based on differential evolution particle swarm fusion algorithm

Abstract

The invention discloses an oil reservoir inversion method based on a differential evolution particle swarm fusion algorithm, and belongs to the technical field of oil reservoir inversion, and the method comprises the steps: carrying out the oil reservoir history fitting, and determining an objective function of the oil reservoir history fitting; establishing a particle swarm optimization algorithm; establishing a differential evolution algorithm; and based on a differential evolution particle swarm fusion algorithm, solving the oil reservoir history fitting objective function to obtain an oil reservoir inversion optimal result. By adopting the above method, through fusion of the two algorithms, the algorithms can maintain the exploration capability for the global optimal solution in the search process, and can rapidly converge to the vicinity of the optimal solution, thereby improving the precision and reliability of oil reservoir inversion.

Description

Technical Field

[0001] The present invention relates to the technical field of oil reservoir inversion, and in particular to an oil reservoir inversion method based on a differential evolution particle swarm fusion algorithm. Background Art

[0002] As oil and gas exploration and development extend into complex geological structures and deep areas, geophysical inversion technology faces unprecedented challenges. Reservoir inversion serves as a critical bridge between seismic data and subsurface reservoir parameters, and its accuracy and efficiency directly impact exploration and development benefits. Traditional reservoir inversion algorithms, such as particle swarm optimization (PSO) or differential evolution (DE) algorithms, often run the risk of falling into local optimal solutions when used alone. Although the PSO algorithm converges quickly, it is prone to premature convergence during the search process, resulting in failure to find the global optimal solution. When dealing with complex multi-peak problems, the DE algorithm may also miss the global optimal region due to the limitations of its mutation and crossover strategies. Summary of the Invention

[0003] The purpose of the present invention is to provide a reservoir inversion method based on differential evolution particle swarm fusion algorithm to solve the technical problems mentioned in the background technology.

[0004] To achieve the above object, the present invention provides a reservoir inversion method based on differential evolution particle swarm fusion algorithm, the steps comprising:

[0005] S1. Perform reservoir history fitting and determine the objective function of reservoir history fitting based on Bayesian theory;

[0006] S2, establish particle swarm optimization algorithm;

[0007] S3, establish differential evolution algorithm;

[0008] S4. Based on the differential evolution particle swarm fusion algorithm, the objective function of reservoir history fitting is solved to obtain the optimal result of reservoir inversion.

[0009] Preferably, the objective function in step S1 is:

[0010]

[0011] Where f(m) is the history matching objective function, m is the vector consisting of well connectivity parameters, and d obs is the true value, G(m) is the predicted value of the proxy model, and C d is the dynamic covariance matrix, is the conductivity value between well i and well j at the initial moment, is the connected volume value between well i and well j at the initial moment.

[0012] Preferably, the inter-well communication parameter constraint condition is:

[0013] m low ≤m≤m up ;

[0014]

[0015] Where m low and m up are vectors consisting of the upper and lower limits of the connectivity parameters, V R is the total connected volume of the reservoir, is the connected volume between well i and well j on the kth layer at the initial moment, N W is the number of all nodes, i is the i-well, and j is the j-well.

[0016] Preferably, the updating formula of particle velocity and position in the particle swarm optimization algorithm in step S2 is:

[0017] v i =ωv i +c1×rand()×(p best (i)-x i )+c2×rand()×(g best -x i );

[0018] x i =x i +v i ;

[0019] Where, v i is the current velocity of the i-th particle; ω is the inertia parameter; c1 and c2 are the individual acceleration constant and the group acceleration constant respectively; rand() is a random number between (0,1); p best (i) is the optimal value searched for the i-th particle; g best To search for the optimal value of the entire cluster; x i is the current position of the i-th particle.

[0020] Preferably, in step S3, the individual variation in the differential evolution algorithm is calculated by a differential strategy, and the formula is:

[0021]

[0022] Where r1, r2, r3 are randomly selected vectors that are different from each other; α is the mutation operator; v i is the mutant individual generated by the differential mutation strategy;

[0023] The newly generated mutant individuals are cross-recombined with individuals in the original population with a certain probability to obtain temporary individuals. The formula is:

[0024]

[0025] Where, i = 1, 2, ..., N; j = 1, 2, ..., M; j rand ∈{1,2,…,M} is a random number; rand ij is a random number between [0,1], C r is the crossover probability; u i is the i-th temporary individual (new individual after crossover); u ij is the value of the temporary individual in the jth dimension; x ij is the value of the i-th individual in the original population in the j-th dimension;

[0026] For each individual of this generation and Make a judgment and select the one with the best fitness as the individual in the same position in the next generation. The competitive selection strategy formula is:

[0027]

[0028] Where F() is the fitness function.

[0029] Preferably, step S4 specifically includes:

[0030] S41. Randomly generate a vector of preset dimension within a preset parameter range and assign an initial position and velocity to each particle;

[0031] S42. Based on a preset probability threshold, select and execute a particle swarm optimization algorithm update strategy or a differential evolution algorithm mutation update strategy;

[0032] S43, applying random perturbations to the particles that have completed the update;

[0033] S44, calculating the fitness evaluation value of each particle based on the objective function;

[0034] S45. Compare the fitness of the particle's current position with the historical optimal position, and record the current optimal position of the population;

[0035] S46, repeat step S45 at the group level, and update the global optimal solution;

[0036] S47: Determine the termination condition. If the convergence criterion is not met, repeat steps S42-S46 until the convergence criterion is met; otherwise, output the final optimization result.

[0037] Therefore, the present invention adopts the above-mentioned reservoir inversion method based on the differential evolution particle swarm fusion algorithm, which combines the advantages of the particle swarm optimization algorithm and the differential evolution algorithm. The differential evolution algorithm has a strong global search capability and can generate diverse individuals through differential mutation operations, thereby increasing the diversity of the population; the particle swarm optimization algorithm has a faster convergence speed and can guide the population to search for a better area; by fusing the two algorithms, the algorithm can maintain the ability to explore the global optimal solution during the search process and can quickly converge to the vicinity of the optimal solution, thereby improving the accuracy and reliability of reservoir inversion.

[0038] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a flow chart of the method of the present invention;

[0040] Figure 2 This is a flow chart of the differential evolution particle swarm fusion algorithm of the present invention;

[0041] Figure 3 A schematic diagram of the well points and connectivity relationships of the conceptual reservoir model of the present invention;

[0042] Figure 4 Schematic diagram showing the conductivity comparison before and after fitting of the conceptual reservoir model of the present invention;

[0043] Figure 5 Schematic diagram showing the comparison of connected volumes before and after fitting of the conceptual reservoir model of the present invention;

[0044] Figure 6 Schematic diagram of the comparison of WWCT curves of connectivity parameters fitted to the conceptual reservoir model of the present invention over the past 900 days;

[0045] Figure 7 The initial reservoir model of the complex reservoir calculation example of the present invention;

[0046] Figure 8 This is the result of history fitting of the water cut curves of 6 production wells using the DEPSO method for the complex reservoir calculation example of the present invention;

[0047] Figure 9 This is a comparison diagram of conductivity before and after fitting for a complex reservoir calculation example of the present invention;

[0048] Figure 10 This is a comparison diagram of the connected volumes before and after fitting for a complex oil reservoir calculation example of the present invention;

[0049] Figure 11 A comparison chart of history matching using three optimization methods, DEPSO, PSO, and DE, for a complex reservoir example of the present invention;

[0050] Figure 12 This is a diagram showing the change in the function rate curve before and after reservoir history fitting for a complex reservoir example of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings herein can be arranged and designed in various different configurations. In the description of the present invention, it should be noted that the orientation or position relationship indicated by the terms "upper", "lower", "inside", "outside", etc. is based on the orientation or position relationship shown in the drawings, or is the orientation or position relationship in which the product of the invention is usually placed when in use. It is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0052] Example

[0053] Reference Figure 1 The present invention provides a reservoir inversion method based on differential evolution particle swarm fusion algorithm, the steps comprising:

[0054] S1. Perform reservoir history fitting. Based on Bayesian theory, determine the objective function for reservoir history fitting. This objective function is used to reasonably quantify the error between the predicted results and the true value. As shown in the following formula, a set of well connectivity parameters that conform to the actual reservoir conditions is used to minimize the objective function.

[0055]

[0056] Where f(m) is the history matching objective function, m is the vector consisting of well connectivity parameters, and d obs is the true value, G(m) is the predicted value of the proxy model, and C d is the dynamic covariance matrix, is the conductivity value between well i and well j at the initial moment, is the connected volume value between well i and well j at the initial moment.

[0057] The inter-well connectivity parameters need to meet some conditions, the formula is:

[0058] m low ≤m≤m up ;

[0059]

[0060] Where mlow and m up are vectors composed of the upper and lower limits of the connectivity parameters, respectively. The lower limit is generally 0. The specific upper and lower limits need to consider the reservoir physical parameters. R is the total connected volume of the reservoir. Since the reserves of the reservoir are relatively reliable, the sum of the connected volumes of each connected unit should remain basically unchanged or change very little. is the connected volume between well i and well j on the kth layer at the initial moment, N W is the number of all nodes, i is the i-well, and j is the j-well.

[0061] S2. Establish a particle swarm optimization algorithm (PSO). The update formula of particle velocity and position is:

[0062] v i =ωv i +c1×rand()×(p best (i)-x i )+c2×rand()×(g best -x i );

[0063] x i =x i +v i ;

[0064] Where, v i is the current velocity of the i-th particle; ω is the inertia parameter; c1 and c2 are the individual acceleration constant and the group acceleration constant respectively; rand() is a random number between (0,1); p best (i) is the optimal value searched for the i-th particle; g best To search for the optimal value of the entire cluster; x i is the current position of the i-th particle.

[0065] S3. Establish a differential evolution algorithm (DE). The individual variation of the differential evolution algorithm is calculated by the differential strategy, and the formula is:

[0066]

[0067] Where r1, r2, and r3 are randomly selected vectors that are different from each other; α is the mutation operator. The larger α is, the less likely it is to fall into a local extreme point; the smaller α is, the more conducive it is to converge to a local extreme point; v i is the mutant individual generated by the differential mutation strategy.

[0068] In the continuous iteration, the newly generated mutant individuals are cross-recombined with individuals in the original population with a certain probability to increase the diversity of the population. The temporary individuals are obtained according to the following formula:

[0069]

[0070] Where, i = 1, 2, ..., N; j = 1, 2, ..., M; j rand ∈{1,2,…,M} is a random number; rand ij is a random number between [0,1], C r is the crossover probability; u i is the i-th temporary individual (new individual after crossover); u ij is the value of the temporary individual in the jth dimension; x ij is the value of the i-th individual in the original population in the j-th dimension.

[0071] Finally, for each individual of this generation and Make a judgment and select the one with the best fitness as the individual in the same position in the next generation. The competitive selection strategy formula is:

[0072]

[0073] Where F() is the fitness function.

[0074] If the individual obtained by the crossover operation is not worse than the original individual, the original individual is replaced as the new generation individual; otherwise, the new generation individual is still the original individual.

[0075] S4. Based on the differential evolution particle swarm fusion algorithm (DEPSO), the objective function of reservoir history fitting is solved to obtain the optimal result of reservoir inversion.

[0076] like Figure 2 As shown, specifically including:

[0077] S41. Algorithm parameter configuration: assign values ​​to control parameters and convergence conditions.

[0078] S42. Population state initialization: within the preset parameter range, randomly generate a vector of preset dimension and assign an initial position and velocity to each particle.

[0079] S43, iterative update of population state: Based on the preset probability threshold, choose to execute the particle swarm optimization algorithm update strategy or the differential evolution algorithm mutation update strategy. In each cycle, a random number is generated. If the number is between 0-0.5 (excluding 0.5), the differential evolution algorithm is selected. If the number is between 0.5-1 (excluding 1), the particle swarm algorithm is selected.

[0080] S44, diversity enhancement processing: applying random perturbations to the particles that have completed the update.

[0081] S45, fitness evaluation: calculate the fitness evaluation value of each particle based on the objective function.

[0082] S46. Update individual optimal solution: Compare the fitness of the particle’s current position with the historical optimal position, and record the current optimal position of the population.

[0083] S47, group optimal solution update: repeat step S46 at the group level and update the global optimal solution based on the best one.

[0084] S48. Determine the termination condition: if the convergence criterion is not met, repeat steps S43-S47 until the convergence criterion is met; otherwise, output the final optimization result.

[0085] In order to verify the effectiveness of the method of the present invention, a specific case is used for illustration.

[0086] Conceptual Reservoir Example

[0087] Establish a two-dimensional heterogeneous conceptual reservoir model, Figure 3 The well points and connectivity relationships for this model are shown. The reservoir parameters are set as follows: 5 injection wells and 4 production wells, a reservoir thickness of 8.5 m, a porosity of 0.3, an average permeability of 950 mD, an initial reservoir pressure of 8 MPa, crude oil and water viscosities of 43 mPa·s and 1.1 mPa·s, respectively, and a crude oil compressibility of 2.365×10 -5 MPa -1 , the formation water compressibility coefficient is 6.368×10 -5 MPa -1 , the rock compression coefficient is 17×10 -5 MPa -1 The reservoir data were acquired during 3000 days of production, with each time step interval set at 30 days.

[0088] A sample was randomly selected from the data set, and the optimal well connectivity parameter value was determined using DEPSO. In this process, the population size was set to 50, and the algorithm was set to terminate after 100 iterations. Tables 1 and 2 list the comparison data between the connectivity parameters obtained by DEPSO inversion and the actual connectivity parameters. Figure 4 and Figure 5 The conductivity and connected volume before and after fitting are compared. To further verify the fitting results, the fitted connectivity parameters and corresponding production performance data were fed into the simulation software to obtain the water cut of each production well. Figure 6 The comparison of the WWCT curves fitted with the connectivity parameters for the past 900 days shows that the two show similar trends. Based on the above comparative data analysis, the DEPSO algorithm is an effective and reliable reservoir inversion method.

[0089] Table 1 The true conductivity and the conductivity obtained by history fitting of the DEPSO algorithm

[0090]

[0091] Table 2 The real connected volume and the connected volume obtained by history fitting of the DEPSO algorithm

[0092]

[0093]

[0094] Complex Reservoir Examples

[0095] A complex reservoir example is selected to verify the above-mentioned method. Figure 7 The initial reservoir model is shown. It features eight injection wells and 13 production wells. The reservoir model uses a 30-day time step, for a total production time of 4,500 days. A CRM model was developed based on reservoir geology data, including a 2,000-sample dataset for model training and history matching.

[0096] A test sample was randomly selected from the test set and history matching was performed using DEPSO. The algorithm parameters were set the same as those in the conceptual reservoir case. This fitting included a total of 57 parameters. Figure 8 This figure shows water cut plots for six wells after fitting using the DEPSO algorithm's history matching method. The red dotted line represents actual observed data, the green curve represents the predicted values ​​from the updated model after history matching using the DEPSO hybrid optimization algorithm, and the gray curve represents the predicted data from the initial model set. The distribution of the predicted data from the gray curve is relatively discrete. Compared to the production curve derived from the initial reservoir model, the target reservoir production curve derived using the DEPSO method is closer to the actual observed data and has a higher degree of fit with the actual historical production data.

[0097] A well was randomly selected and a comparison chart of its conductivity and connected volume before and after history fitting was drawn. The results are as follows: Figure 9 , Figure 10 As shown. Figure 11 As shown in the figure, six randomly selected production wells were historically fitted using three optimization methods: DEPSO, PSO, and DE. Each subplot contains three curves, representing the water cut changes obtained using different optimization methods. The red dotted line in the figure represents the historical water cut data, while the green curve represents the predicted data after optimization using the DEPSO algorithm. This data shows a high degree of consistency with the historical data, demonstrating extremely high fitting accuracy. The orange and pink lines correspond to the predicted data using the DE and PSO algorithms, respectively, with the fitting accuracy decreasing in descending order.

[0098] Figure 12The figure shows the water-cut changes before and after historical fitting, and the predicted water-cut curves for the past 2000 days. The dashed line indicates the dividing point before and after fitting. The red line represents the water-cut change after fitting, and the blue line represents the water-cut change before fitting. It can be seen that the fitting effect is good. Table 3 shows the water-cut fitting rate for all producing wells in this reservoir model, with the average water-cut fitting rate for each well reaching 87.8%.

[0099] Table 3 Single well water cut fitting rate

[0100]

[0101] Therefore, the present invention adopts the above-mentioned reservoir inversion method based on differential evolution particle swarm fusion algorithm. By fusing the two algorithms, the algorithm can not only maintain the ability to explore the global optimal solution during the search process, but also quickly converge to the vicinity of the optimal solution, thereby improving the accuracy and reliability of reservoir inversion.

[0102] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A reservoir inversion method based on differential evolution particle swarm fusion algorithm, characterized in that the steps include: S1. Perform reservoir history fitting and determine the objective function of reservoir history fitting based on Bayesian theory; S2, establish particle swarm optimization algorithm; S3, establish differential evolution algorithm; S4. Based on the differential evolution particle swarm fusion algorithm, the objective function of reservoir history fitting is solved to obtain the optimal result of reservoir inversion.

2. The reservoir inversion method based on differential evolution particle swarm fusion algorithm according to claim 1, characterized in that: The objective function in step S1 is: Where f(m) is the history matching objective function, m is the vector consisting of well connectivity parameters, and d obs is the true value, G(m) is the predicted value of the proxy model, and C d is the dynamic covariance matrix, is the conductivity value between well i and well j at the initial moment, is the connected volume value between well i and well j at the initial moment.

3. The reservoir inversion method based on differential evolution particle swarm fusion algorithm according to claim 2, characterized in that: The inter-well connectivity parameter constraints are: m low ≤m≤m up ; Where m low and m up are vectors consisting of the upper and lower limits of the connectivity parameters, V R is the total connected volume of the reservoir, is the connected volume between well i and well j on the kth layer at the initial moment, N W is the number of all nodes, i is the i-well, and j is the j-well.

4. The reservoir inversion method based on differential evolution particle swarm fusion algorithm according to claim 3 is characterized by: The update formula of particle velocity and position in step S2 particle swarm optimization algorithm is: v i =ωv i +c1×rand()×(p best (i)-x i )+c2×rand()×(g best -x i ); x i =x i +v i ; Where, v i is the current velocity of the i-th particle; ω is the inertia parameter; c1 and c2 are the individual acceleration constant and the group acceleration constant respectively; rand() is a random number between (0,1); P best (i) is the optimal value searched for the i-th particle; g best To search for the optimal value of the entire cluster; x i is the current position of the i-th particle.

5. The reservoir inversion method based on differential evolution particle swarm fusion algorithm according to claim 4 is characterized in that: In step S3, the individual variation in the differential evolution algorithm is calculated using the differential strategy, and the formula is: Where r1, r2, r3 are randomly selected vectors that are different from each other; α is the mutation operator; v i is the mutant individual generated by the differential mutation strategy; The newly generated mutant individuals are cross-recombined with individuals in the original population with a certain probability to obtain temporary individuals. The formula is: Where, i = 1, 2, ..., N; j = 1, 2, ..., M; j rand ∈{1,2,…,M} is a random number; rand ij is a random number between [0,1], C r is the crossover probability; u i is the i-th temporary individual; u ij is the value of the temporary individual in the jth dimension; x ij is the value of the i-th individual in the original population in the j-th dimension; For each individual of this generation and Make a judgment and select the one with the best fitness as the individual in the same position in the next generation. The competitive selection strategy formula is: Where F() is the fitness function.

6. The reservoir inversion method based on differential evolution particle swarm fusion algorithm according to claim 5, characterized in that: Step S4 specifically includes: S41. Randomly generate a vector of preset dimension within a preset parameter range and assign an initial position and velocity to each particle; S42. Based on a preset probability threshold, select and execute a particle swarm optimization algorithm update strategy or a differential evolution algorithm mutation update strategy; S43, applying random perturbations to the particles that have completed the update; S44, calculating the fitness evaluation value of each particle based on the objective function; S45. Compare the fitness of the particle's current position with the historical optimal position, and record the current optimal position of the population; S46, repeat step S45 at the group level, and update the global optimal solution; S47: Determine the termination condition. If the convergence criterion is not met, repeat steps S42-S46 until the convergence criterion is met; otherwise, output the final optimization result.

Images