An equivalent numerical simulation method for ion water flooding based on artificial neural network

Abstract

This invention relates to the field of oil and gas field development technology, and particularly to an equivalent numerical simulation method for ion water flooding based on artificial neural networks. Using the artificial neural network method described in this application, given the injected water ion concentration and initial relative permeability curve, the relative permeability curve corresponding to the IMEX simulator can be directly calculated, making the ion water flooding process equivalent to an IMEX simulator simulation, without needing to simulate the exchange process between ions in the injected water and rock clay minerals. By substituting the calculated relative permeability curve into the IMEX simulator, the same simulation effect can be quickly obtained. Compared to using the STARS simulator, this method reduces the number of components to be simulated and avoids complex ion exchange reaction processes, thereby significantly reducing computation time.

Description

Technical Field

[0001] The invention relates to the technical field of oil and gas field development, and in particular to an ion water drive equivalent numerical simulation method based on artificial neural network. Background Art

[0002] Ion water flooding is an important technology for improving oil recovery in low permeability reservoirs. The ions in the injected water exchange with the ions adsorbed on the surface of the formation rock, which manifests as a change in the wettability of the rock, thereby affecting the relative permeability curve. The current numerical simulation research method of ion water flooding is also to achieve the change of the relative permeability curve by ion concentration through the ion exchange reaction process and use the ion concentration as the interpolation parameter of the relative permeability curve. The above process can be achieved through the STARS simulator in the CMG software, but because the STARS simulator involves phase equilibrium, the simulation process needs to consider multiple factors such as ion exchange reaction rate, different ion concentrations, wettability, interfacial tension, temperature, etc., which makes the simulation process of ion water flooding take a lot of time.

[0003] In addition, publication number CN117057138A, invention name A numerical simulation method for ion water drive, publication number CN104632151B, invention name A method for improving oil recovery by ion matching water drive, and publication number CN108876055A, invention name A method for predicting low-salinity water drive results in ultra-low permeability reservoirs all disclose numerical simulation methods for ion water drive with low salinity water drive, but they mainly use numerical simulation methods and laboratory methods to analyze the microscopic mechanism of ion water drive and the factors affecting oil recovery, and the simulation process still consumes a lot of time. Summary of the invention

[0004] Based on the above problems, the present invention proposes an equivalent numerical simulation method for ion water flooding based on artificial neural network. The initial relative permeability curve is adjusted by artificial neural network, and the equivalent relative permeability curve after the ion concentration is changed is obtained and brought into the IMEX simulator for simulation, thereby equivalently simulating the STARS ion water flooding process and saving simulation time.

[0005] Based on the above purpose, the present application discloses an equivalent numerical simulation method of ion water flooding based on artificial neural network, comprising the following steps:

[0006] S1. Construct a neural network training database, and determine that the database structure consists of the initial relative permeability curve parameters of the target model, the ion concentration composition of the injected water, and the equivalent relative permeability curve parameters calculated by the IMEX simulator; wherein, the equivalent relative permeability curve parameter composition calculated by the IMEX simulator is constructed by: using the STARS simulator to perform ion water flooding simulation, using the initial relative permeability curve parameters and the ion concentration composition of the injected water as model parameters; while using the IMEX simulator to adjust the initial relative permeability curve for simulation, when the cumulative oil production reaches the same as the STARS simulation result, stop the simulation. At this time, the relative permeability curve of the IMEX model is used as the equivalent relative permeability curve.

[0007] S2. The relative permeability curve is converted into the Corey equation, and the Corey equation parameters (Krocw, Krwor, Sw, Sor, Scw, Swcrit,

[0008] No, Nw) and the ion concentration composition of the injected water (Na + 、SO4 2- Mg 2+ , Ca 2+ ) as input, the Corey equation parameters corresponding to the equivalent relative permeability curve parameters (Krocw, Krwor,

[0009] Sw, Sor, Scw, Swcrit, No, Nw) are used as output and imported into the artificial neural network for training;

[0010] S3. The initial relative permeability curve parameters of the target model and the ion concentration composition of the injected water are brought into the artificial neural network, and the equivalent relative permeability curve of the IMEX simulator is used as the output result. It is brought into IMEX for simulation, and the simulation results are compared with the results of the STARS simulator. When the cumulative oil production error is ≤10%, it proves that the simulation results meet the accuracy requirements, and the equivalent numerical simulation results of ion water flooding of the artificial neural network are obtained.

[0011] Furthermore, the specific steps of constructing the target model neural network database are:

[0012] S1. Preset the injected water ion concentration composition;

[0013] S2. Using the preset injection water ion concentration composition as a parameter, the STARS simulator is used to simulate the model, and the relative permeability curve in the IMEX simulator is adjusted according to the simulation results until the error between the production effect and the STARS simulation result is within an acceptable range, and the initial relative permeability curve, ion concentration, and equivalent relative permeability curve are obtained;

[0014] S3. Change the ion concentration composition of the injected water, repeat steps S1 and S2 many times, obtain the data required by the neural network database, and input parameters: initial relative permeability curve parameters (Krocw, Krwor, Sw, Sor, Scw, Swcrit, No, Nw), ion concentration (Na + 、SO4 2- Mg 2+ , Ca 2+ ); Output parameters: equivalent relative permeability curve (Krocw, Krwor, Sw, Sor, Scw, Swcrit, No, Nw).

[0015] Furthermore, the Corey equation is:

[0016]

[0017] Among them, equation ① is the relative permeability of the oil phase under Corey, equation ② is the relative permeability of the water phase; Krocw is the relative permeability of oil under native water conditions; Krwor is the relative permeability of water in residual oil; Sw is water saturation; Sor is residual oil saturation; Scw is native water saturation; Swcrit is critical water saturation; No is the Corey index of oil; Nw is the Corey index of water.

[0018] Furthermore, the optimization method also includes comparing and verifying the results of IMEX simulation using an equivalent relative permeability curve and STARS simulation using an initial relative permeability curve to determine its feasibility.

[0019] Beneficial effect: Through the equivalent method described in this application, under the condition of given injected water ion concentration, the corresponding relative permeability curve can be directly obtained. At this time, it is no longer necessary to use the complex physical and chemical simulation process of the STARS simulator, and the relative permeability curve can be brought into the IMEX simulator to quickly obtain the same development effect. Compared with the traditional numerical simulation method, this method avoids the complex ion exchange reaction process and greatly reduces the calculation time. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a numerical simulation model diagram of the case selected by the present invention;

[0021] Figure 2 It is a schematic diagram of the relative permeability curves before and after equivalent;

[0022] Figure 3 It is a schematic diagram of the network structure of the relative permeability curve equivalent model of the present invention;

[0023] Figure 4 is a schematic diagram of an equivalent model of the relative permeability curve of the present invention;

[0024] Figure 5 This is a comparison chart of IMEX and STARS production forecasts. DETAILED DESCRIPTION

[0025] In order to more clearly illustrate the purpose, research methods and advantages of the present invention, the present invention is further described in detail through the following examples.

[0026] To clearly describe the technical concept of the present application, the principle of the present application is first explained: The basic structure of the artificial neural network (ANN) is composed of neurons, which are divided into an input layer, a hidden layer, and an output layer. ANN learns through two stages, forward propagation and reverse propagation, to optimize the model parameters so that it can accurately predict the output. During the forward propagation process, the input information is transmitted from the input layer to the hidden layer, and then transmitted to the output layer after being processed by the neurons, and finally the predicted output of the network is generated. During the reverse propagation process, by calculating the error between the actual output and the expected output, the error signal returns to the hidden layer and the input layer in the original path, thereby adjusting the connection weights between the neurons to minimize the error. The alternating iteration of forward propagation and reverse propagation is the core process of neural network training. In the forward propagation, the network generates a prediction result based on the current weight parameter; in the reverse propagation, the weight parameter is updated by an optimization algorithm such as gradient descent according to the error between the actual output and the expected output. This process is repeated until the predetermined training stop condition is reached and the final result is output.

[0027] Example 1 takes a well group in a typical test area of ​​a certain X oil reservoir as a case (target model) to specifically illustrate an equivalent numerical simulation method for ion water flooding based on an artificial neural network.

[0028] S1. Construct a neural network training database, and determine that the database structure consists of the initial relative permeability curve parameters of ion water drive, the ion concentration of injected water, and the equivalent relative permeability curve parameters calculated by the IMEX simulator.

[0029] Taking a typical test area well group of a certain X reservoir as an example, its production history has adopted ionized water flooding. In order to minimize the simulation time, two typical well groups are selected (the numerical simulation model is shown in Figure 2). Figure 1 As shown in the figure, the model has 35280 grids and 24 layers. The well group model has 15 wells, including 13 production wells and 2 water injection wells (L86-36, L86-38). A square reverse nine-point well pattern is used for ion water injection development. The average reservoir permeability is 3mD, the average oil layer porosity is 0.12, the initial reservoir pressure is 13.2MPa, the oil-water interface depth is 250m, and the average crude oil viscosity is 1.23cp. If it is necessary to explain, the initial relative permeability curve parameters have been determined when the geological model is established, such as Figure 2 shown.

[0030] In a further embodiment, in order to construct its neural network database, the ion concentration composition of the water injected into the experimental area model is preset, as shown in Table 1.

[0031] Table 1 Injection water ion concentration composition

[0032] PH <![CDATA[Cl - ]]> <![CDATA[SO4 2- ]]> <![CDATA[Ca 2+ ]]> <![CDATA[Mg 2+ ]]> <![CDATA[HCO 3- ]]> <![CDATA[Na + ]]> 7 17500 100 100 100 100 17500

[0033] It should be noted that the ion concentration of the injected water for any well group can be set arbitrarily, but the ion concentration of the formation water in a model is fixed. In this specific embodiment, the present application presets that the test area model changes the ion concentration of the injected water, and does not change the ion composition. Different combinations of injected water ion concentrations can simulate the ion water flooding effect under different salinities. The data parameters shown in Table 1 remain unchanged in all test models, and the changes in production results are only attributed to different ion concentration combinations.

[0034] As shown in Table 1, the four aqueous phase reactions in which ions interact include:

[0035]

[0036] In a further embodiment, in order to construct a database, the well group model is selected, and the model is simulated by the STARS simulator in the CMG software with the preset injection water ion concentration composition as a parameter: specifically, 150 simulations are performed with random injection water ion concentration composition within the formation water ion concentration range, and in every 50 simulations, the interpolation ion (SO4 2- Mg 2+ , Ca 2+ ) to increase the diversity of ion combinations and their corresponding effects, and adjust the relative permeability curve in the IMEX simulator to obtain the same simulation effect as STARS. Finally, a database was obtained for all 150 samples, and the data required for the next step of artificial neural network training were extracted, namely, the initial relative permeability curve parameters, the injected water ion concentration, and the equivalent relative permeability curve parameters. The example structure of the database is shown in Table 2.

[0037] Table 2 Artificial neural network database

[0038]

[0039] S2. The initial relative permeability curve and the equivalent relative permeability curve ( Figure 2 ) is converted into the Corey equation, and the Corey equation parameters (Krocw, Krwor, Sw, Sor, Scw, Swcrit, No, Nw) and ion concentration (Na+ 、SO4 2- Mg 2+ , Ca 2+ ) is used as input, and the Corey equation parameters (Krocw, Krwor, Sw, Sor, Scw, Swcrit, No, Nw) corresponding to the optimized relative permeability curve are used as output and imported into the artificial neural network for training.

[0040] In a specific embodiment, the Corey equation is:

[0041]

[0042]

[0043] Wherein equation (1) is the relative permeability of the oil phase under Corey, and equation (2) is the relative permeability of the water phase.

[0044] The parameters used for artificial neural network training are:

[0045] ●Krocw – relative permeability of oil in connate water conditions;

[0046] Krwor – relative permeability of water in residual oil;

[0047] ●Sw – water saturation;

[0048] ●Sor – residual oil saturation;

[0049] Scw – connate water saturation;

[0050] ●Swcrit – critical water saturation;

[0051] ●No – Corey index of oil;

[0052] ●Nw – Corey index of water.

[0053] Therefore, the number of input parameters is determined to be 12, including 8 parameters from the Corey equation, 3 ion concentrations, and 1 ion interpolator. The output parameters are the corresponding 8 parameters of the Corey equation, such as Figure 3 shown.

[0054] S3. The initial relative permeability of the target model and the ion concentration composition of the injected water are brought into the artificial neural network. The output result of the model is the relative permeability curve equivalent to the IMEX simulator. It is brought into IMEX for simulation, and the simulation results are compared with the results of the STARS simulator.

[0055] In a specific embodiment, the initial relative permeability curve parameters (Krocw, Krwor, Sw, Sor, Scw, Swcrit, No, Nw) and ion concentration composition are used as input, and the equivalent relative permeability ((Krocw, Krwor, Sw, Sor, Scw, Swcrit, No, Nw) curve is used as output, of which 70% is used as a training set and 30% is used as a test set, which are imported into an artificial neural network to predict the relative permeability curve.

[0056] The predicted relative permeability curve was brought into the IMEX simulator to obtain the same ion water flooding numerical simulation results as those of STARS.

[0057] Test case 1: Verification of numerical simulation results of ion water flooding

[0058] like Figure 4 As shown in the relative permeability curve optimization model, the equivalent relative permeability curve obtained in Example 1 was brought into the IMEX simulator to obtain numerical simulation results, which were compared and verified with the results obtained by ion water flooding in the STARS simulator. The results are as follows Figure 5 As shown in the figure, the cumulative oil production results show that the simulation error between IMEX and STARS is 5.5%, and the error is less than 10%, which verifies the model.

[0059] The present invention uses an artificial neural network to change the relative permeability curve by given ion conditions, and uses the IMEX simulator to achieve the same prediction results as the STARS simulator. The simulation results of STARS are compared with the results of the equivalent relative permeability curve under IMEX, and the errors are analyzed to ensure the reliability of the strategy of the present invention. The simulation results of the model of the present invention are only related to the change of ion concentration, avoiding the complex ion exchange reaction process in the STARS simulation process and greatly saving simulation time.

Claims

1. An equivalent numerical simulation method for ion water flooding based on artificial neural network, characterized in that: The steps include: S1. Construct a neural network training database, and determine that the database structure consists of the initial relative permeability curve parameters of the target model, the ion concentration composition of the injected water, and the equivalent relative permeability curve parameters calculated by the IMEX simulator; wherein the equivalent relative permeability curve parameter composition calculated by the IMEX simulator is constructed by: using the STARS simulator to perform ion water flooding simulation, using the initial relative permeability curve parameters and the ion concentration composition of the injected water as model parameters; while using the IMEX simulator to perform simulation by adjusting the initial relative permeability curve, when the cumulative oil production reaches the same as the STARS simulation result, stop the simulation, and at this time the relative permeability curve of the IMEX model is used as the equivalent relative permeability curve; S2. The relative permeability curve is converted into the Corey equation, and the Corey equation parameters Krocw, Krwor, Sw, Sor, Scw, Swcrit, No, Nw corresponding to the initial relative permeability curve parameters and the ion concentration of the injected water form the Na + 、SO4 2- Mg 2 + , Ca 2+ As input, the Corey equation parameters Krocw, Krwor, Sw, Sor, Scw, Swcrit, No, and Nw corresponding to the parameters of the equivalent relative permeability curve are used as output and imported into the artificial neural network for training; S3. The initial relative permeability curve parameters of the target model and the ion concentration composition of the injected water are introduced into the artificial neural network, and the equivalent relative permeability curve of the IMEX simulator is used as the output result, which is introduced into IMEX for simulation. The simulation results are compared with the results of the STARS simulator. When the cumulative oil production error is ≤10%, it is proved that the simulation results meet the accuracy requirements, and the equivalent numerical simulation results of ion water flooding of the artificial neural network are obtained; Among the Corey equation parameters, Krocw is the relative permeability of oil under native water conditions; Krwor is the relative permeability of water in residual oil; Sw is water saturation; Sor is residual oil saturation; Scw is native water saturation; Swcrit is critical water saturation; No is the Corey index of oil; and Nw is the Corey index of water.

2. The simulation method according to claim 1, characterized in that: The specific steps of constructing the neural network training database are: S1. Preset the injected water ion concentration composition; S2. Using the preset injection water ion concentration composition as a parameter, the STARS simulator is used to simulate the model, and the relative permeability curve in the IMEX simulator is adjusted according to the simulation results until the error between the production effect and the STARS simulation result is within an acceptable range, and the initial relative permeability curve, ion concentration, and equivalent relative permeability curve are obtained; S3. Change the ion concentration composition of the injected water, repeat steps S1 and S2 many times, obtain the data required by the neural network database, and input parameters: initial relative permeability curve parameters Krocw, Krwor, Sw, Sor, Scw, Swcrit, No, Nw, ion concentration Na + 、SO4 2- Mg 2+ , Ca 2+ ; Output parameters: equivalent relative permeability curves Krocw, Krwor, Sw, Sor, Scw, Swcrit, No, Nw.

3. The simulation method according to claim 1, characterized in that: The Corey equation is: ① ② Among them, equation ① is the relative permeability of the oil phase under Corey, equation ② is the relative permeability of the water phase; Krocw is the relative permeability of oil under native water conditions; Krwor is the relative permeability of water in residual oil; Sw is water saturation; Sor is residual oil saturation; Scw is native water saturation; Swcrit is critical water saturation; No is the Corey index of oil; Nw is the Corey index of water.

4. The simulation method according to claim 1, characterized in that: The simulation method further includes comparing and verifying the results of the IMEX simulation using an equivalent relative permeability curve and the STARS simulation using an initial relative permeability curve to determine its feasibility.

Images