A steam distribution method and device based on machine learning

Abstract

This application discloses a steam distribution method and device based on machine learning. The method includes: establishing a downhole comprehensive reservoir model using a sequential Gaussian simulation method based on the actual downhole formation points of a selected target well; establishing a heavy oil thermal recovery data prediction model using a linear model based on the simulation parameter values ​​of the simulated formation points in the downhole comprehensive reservoir model; determining the total amount of steam to be allocated to the target well in real time based on the heavy oil thermal recovery data prediction model and a net present value formula, and determining the steam utilization rate based on the total steam injection volume. This application establishes a heavy oil thermal recovery data prediction model through machine learning, which can achieve real-time optimization of steam distribution, rationally allocate gas injection parameters for each oil production well, avoid resource waste, and effectively improve the production efficiency of heavy oil reservoirs.

Description

Technical Field

[0001] The present application relates to the technical field of heavy oil extraction, and in particular to a steam distribution method and device based on machine learning. Background Art

[0002] Heavy oil is difficult to extract due to its high viscosity. Compared to regular crude oil, heavy oil has higher viscosity, higher density, and lower fluidity. During the heavy oil extraction process, saturated steam is injected into the underground oil to raise its temperature, reduce its viscosity, and increase its fluidity, ultimately improving the oil's recovery rate.

[0003] The steam injection distributor in heavy oil production often adopts a one-branch-multiple-branch structure, that is, a main line inputs wet saturated steam into the steam distributor, the wet steam flows in the distributor and flows out through multiple branch pipelines after the gas and liquid are remixed, and is injected into each wellhead for thermal oil production.

[0004] The wet steam distributors commonly used at present are multi-branch vertical distributors and multi-branch spherical distributors. However, due to the lack of precise steam metering and reasonable distribution, when injecting steam into the well, the steam injection volume of a single well is unclear, resulting in insufficient steam injection in some wells and excessive steam injection in others. The oil layer heating effect of under-injected steam wells is poor, and steam over-injected wells experience steam crossflow, heat energy loss, and low steam thermal energy utilization. The main reason for the difficulty in measuring wet saturated steam is the simultaneous presence of both vapor and liquid phases in the fluid. The presence of the liquid phase causes large changes in the differential pressure flowing through the throttling device. Since the droplets are not evenly distributed during the steam injection process, the pressure difference varies irregularly, resulting in large measurement errors, and traditional flow measurement devices cannot be used. V-cone flowmeters are not wear-resistant and have limited applications. Precise wet steam measurement at home and abroad usually adopts phase metering, equal dryness distribution and other technologies to separate steam and liquid and measure them separately. The equipment is complex and costly. The steam injection time of thin-layer heavy oil is short and the frequency is high. The combination of well groups in each round of steam injection is not fixed. The equipment installation and commissioning time is long and the cost is high, which is not suitable for thin-layer heavy oil development.

[0005] This section is intended to provide a background or context to the embodiments of the present application that are recited in the claims. No admission is made that the description herein is prior art by virtue of its inclusion in this section. Summary of the Invention

[0006] This application provides a steam distribution method and device based on machine learning to at least solve the problem of poor control of steam injection volume during heavy oil production in the prior art.

[0007] According to one aspect of the present application, a steam distribution method based on machine learning is provided, comprising:

[0008] Based on the actual downhole formation points of the selected target well, a downhole comprehensive reservoir model is established through the sequential Gaussian simulation method;

[0009] According to the simulated parameter values ​​of the simulated formation points in the downhole comprehensive reservoir model, a heavy oil thermal recovery data prediction model is established through a linear model;

[0010] The total amount of steam allocated to the target well in real time is determined based on the heavy oil thermal recovery data prediction model and the net present value formula, and the steam utilization rate is determined based on the total steam injection volume.

[0011] In one embodiment, the number of target wells selected is at least 3 pairs.

[0012] In one embodiment, a downhole comprehensive reservoir model is established by a sequential Gaussian simulation method, including:

[0013] Establish the multivariate normal distribution function of actual stratigraphic points;

[0014] Calculate the mathematical expectation of the conditional distribution of actual stratigraphic points according to the multivariate normal distribution conditional theorem;

[0015] The simulation parameter value at the simulation formation point is obtained according to the mathematical expectation value.

[0016] In one embodiment, a heavy oil thermal recovery data prediction model is established using a linear model, including:

[0017] Get the actual oil production and disturbance value at time t;

[0018] The simulation parameter values, actual oil production at time t and disturbance value are input into the linear model to generate a heavy oil thermal recovery data prediction model.

[0019] In one embodiment, the linear model is an OE model.

[0020] In one embodiment, the total amount of steam allocated to a target well in real time is determined based on a heavy oil thermal recovery data prediction model and a net present value formula, including:

[0021] Obtain the predicted oil production of the target well at time t based on the heavy oil thermal recovery data prediction model;

[0022] The predicted oil production at time t is input into the net present value formula to obtain the total steam injection volume.

[0023] According to another aspect of the present application, a steam distribution device based on machine learning is also provided, comprising:

[0024] A comprehensive reservoir model building unit is used to build a downhole comprehensive reservoir model by sequential Gaussian simulation method based on the actual downhole formation points of the selected target well;

[0025] A heavy oil thermal recovery data prediction model establishment unit is used to establish a heavy oil thermal recovery data prediction model through a linear model based on the simulation parameter values ​​of the simulation formation points in the downhole comprehensive reservoir model;

[0026] The steam distribution unit is used to determine the total amount of steam allocated to the target well in real time based on the heavy oil thermal recovery data prediction model and the net present value formula, and to determine the steam utilization rate based on the total steam injection volume.

[0027] In one embodiment, the number of target wells selected is at least 3 pairs.

[0028] In one embodiment, the integrated reservoir model building unit includes:

[0029] Multivariate normal distribution function building module, used to build the multivariate normal distribution function of actual formation points;

[0030] A mathematical expectation value calculation module is used to calculate the mathematical expectation value of the conditional distribution of actual formation points according to the multivariate normal distribution condition theorem;

[0031] The simulation parameter module is used to obtain the simulation parameter value at the simulation formation point position according to the mathematical expectation value.

[0032] In one embodiment, the heavy oil thermal recovery data prediction model establishment unit includes:

[0033] Real-time data acquisition module, used to obtain the actual oil production and disturbance value at time t;

[0034] The model building module is used to input the simulation parameter values, the actual oil production at time t and the disturbance value into the linear model to generate a heavy oil thermal recovery data prediction model.

[0035] In one embodiment, the linear model is an OE model.

[0036] In one embodiment, the steam distribution unit comprises:

[0037] The oil production prediction module is used to obtain the predicted oil production of the target well at time t based on the heavy oil thermal recovery data prediction model;

[0038] The total steam injection volume calculation module is used to input the predicted oil production at time t into the net present value formula to obtain the total steam injection volume.

[0039] In one embodiment of the present application, a computer device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, a steam distribution method based on machine learning is implemented.

[0040] In one embodiment of the present application, a computer-readable storage medium is further provided. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, a steam distribution method based on machine learning is implemented.

[0041] In one embodiment of the present application, a computer program product is further provided. The computer program product includes a computer program. When the computer program is executed by a processor, a steam distribution method based on machine learning is implemented.

[0042] Compared with the existing heavy oil steam injection technology solutions, this method uses machine learning to establish a heavy oil thermal recovery data prediction model, optimize steam distribution in real time, and rationally allocate gas injection parameters for each oil well, thus avoiding resource waste and effectively improving the efficiency of heavy oil reservoir recovery. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0044] Figure 1 A steam distribution method based on machine learning is provided in this application;

[0045] Figure 2 This is a flow chart of establishing a downhole comprehensive reservoir model by sequential Gaussian simulation method in an embodiment of the present application;

[0046] Figure 3 This is a flow chart of establishing a heavy oil thermal recovery data prediction model using a linear model in an embodiment of the present application;

[0047] Figure 4 This is a flow chart for determining the total amount of steam allocated to a target well in real time based on a heavy oil thermal recovery data prediction model and a net present value formula in an embodiment of the present application.

[0048] Figure 5 A structural block diagram of a steam distribution device based on machine learning provided in this application;

[0049] Figure 6 This is a structural block diagram of a comprehensive reservoir model building unit in an embodiment of the present application;

[0050] Figure 7 This is a structural block diagram of a unit for establishing a heavy oil thermal recovery data prediction model in an embodiment of the present application;

[0051] Figure 8 This is a structural block diagram of the steam distribution unit in an embodiment of the present application;

[0052] Figure 9 This is a specific implementation of an electronic device in the embodiment of this application. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the embodiments of the present application are further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present application and their descriptions are used to explain the present application, but are not intended to limit the present application.

[0054] Steam injection distributors used in heavy oil recovery often employ a single-branch, multi-branch structure. This involves a single main line feeding wet, saturated steam into the distributor. The wet steam flows through the distributor and, after remixing, exits through multiple branch lines, injecting it into various wellheads for thermal recovery. Currently, the most commonly used wet steam distributors are single-branch, multi-branch vertical distributors and single-branch, multi-branch spherical distributors. However, due to a lack of precise steam metering and proper distribution, the amount of steam injected into individual wells is unclear, resulting in under-injection and over-injection in some wells. Under-injected wells heat the oil layer poorly, while over-injected wells experience steam crossflow, resulting in heat loss and low steam thermal energy utilization. The primary reason for the difficulty in measuring wet, saturated steam is the coexistence of both vapor and liquid phases in the fluid. The presence of the liquid phase results in large variations in the differential pressure across the throttling device. Because the liquid droplets are unevenly distributed during steam injection, the pressure differential fluctuates erratically, leading to significant measurement errors and making traditional flow measurement devices ineffective. V-cone flowmeters are not wear-resistant and have limited applications. Precise wet steam measurement at home and abroad usually adopts phase metering, equal dryness distribution and other technologies to separate steam and liquid and measure them separately. The equipment is complex and costly. The steam injection time of thin-layer heavy oil is short and the frequency is high. The combination of well groups in each round of steam injection is not fixed. The equipment installation and commissioning time is long and the cost is high, which is not suitable for thin-layer heavy oil development.

[0055] In order to solve the above problems, this application provides a steam distribution method based on machine learning, such as Figure 1 Shown, including:

[0056] S101: Based on the actual downhole formation points of the selected target well, a downhole comprehensive reservoir model is established by sequential Gaussian simulation method.

[0057] S102: Establishing a heavy oil thermal recovery data prediction model through a linear model based on the simulation parameter values ​​of the simulation formation points in the downhole comprehensive reservoir model.

[0058] S103: Determine the total amount of steam allocated to the target well in real time based on the heavy oil thermal recovery data prediction model and the net present value formula, and determine the steam utilization rate based on the total steam injection volume.

[0059] In one specific embodiment, a machine learning-based steam allocation method includes three steps. First, actual formation points (several) of the target well are selected and a set vector of these points is established. A comprehensive reservoir model for the well is then constructed using the Sequential Gaussian Simulation method. Once the comprehensive reservoir model is obtained, the simulated formation points and their parameters are determined. Next, an existing linear model is selected and a SAGD model (heavy oil thermal recovery data prediction model) is established based on the parameters of the simulated formation points and the linear model. This SAGD model can predict oil production at a specific moment. The predicted oil production, combined with the NPV formula (net present value), yields the optimal steam allocation solution.

[0060] Figure 1 The executors of the provided machine learning-based steam distribution method can be computers, servers, etc. By establishing a heavy oil thermal recovery data prediction model through machine learning, real-time optimization of steam distribution is achieved, and the gas injection parameters of each oil well are reasonably allocated, thus avoiding resource waste and effectively improving the efficiency of heavy oil reservoir recovery.

[0061] In one embodiment, the number of target wells selected is at least 3 pairs.

[0062] In one specific embodiment, the number of horizontal wells selected is 3 pairs. In the description of the subsequent embodiments, 3 pairs of horizontal wells are selected as the specific implementation method. In practice, the number of horizontal wells selected can be more than 3 pairs, and this application is not limited to this.

[0063] In one embodiment, a downhole comprehensive reservoir model is established by sequential Gaussian simulation method, such as Figure 2 Shown, including:

[0064] S201: Establishing the multivariate normal distribution function of actual stratigraphic points;

[0065] S202: Calculate the mathematical expectation of the conditional distribution of the actual stratigraphic points according to the multivariate normal distribution condition theorem;

[0066] S203: Obtain simulation parameter values ​​at the simulated formation point locations according to the mathematical expectation values.

[0067] In a specific embodiment, a comprehensive reservoir model of three pairs of horizontal wells is established according to the sequential Gaussian simulation method, where X is the actual formation point:

[0068] Assume that the random variable X=(X1,X2,...,X I )′ is an I×1 random vector that obeys a multivariate normal distribution, and its distribution function is X~N I (μ,∑), where μ=E(X), ∑=Cov(X i ,X J ),i,j=1,2,..,I.

[0069] in, C is the covariance, ∑ is the covariance matrix.

[0070] Divide X into two parts, X=(X1,X2,...,X m ,X m+1 ,X m+2 ,...,X m+n＝I ), the corresponding μ=(μ m ,μ n ),∑=(Σ mm ,∑ mn ,∑ nm ,∑ nm ).

[0071] in,

[0072] According to the multivariate normal distribution condition calculation theorem, when X is known n (X m+1 ,X m+2 ,...,X m+n ) condition, X m The conditional distribution of

[0073]

[0074] in, For X m The variance of the conditional distribution of X m The mathematical expectation of the conditional distribution of is

[0075] Let X m The simulation parameter matrix Z′(x m ) is equal to the expected value of the conditional probability distribution function at that location plus the random number l m The product of the standard deviation of the conditional probability distribution function. m ~N(0,1) is a standard normal distribution with an expected value of 0 and a variance of 1.

[0076]

[0077] For regionalized variables, μ m Equal to x m The expected value m(x m ),∑ mn is the point x to be simulated m and the known point x n The autocovariance matrix of m ,x j) where m is the position to be simulated and can only take one value, and the value of n increases as the simulated value is continuously added to the known condition matrix.

[0078] And nm is the autocovariance matrix between known points, written as C(x i ,x j ), so there is

[0079]

[0080] Among them, i,j=1,2,...,n,n+1,n+2,...,n+m-1.

[0081] As more simulated formation points generate simulated parameter values, the capacity of known points will increase from n to n+1, n+2, ..., n+m-1. Based on these simulated formation points, a comprehensive reservoir model of the well can be obtained.

[0082] In one embodiment, a heavy oil thermal recovery data prediction model is established using a linear model, such as Figure 3 Shown, including:

[0083] S301: Obtaining the actual oil production and disturbance value at time t;

[0084] S302: Inputting simulation parameter values, actual oil production at time t, and disturbance values ​​into a linear model to generate a heavy oil thermal recovery data prediction model.

[0085] In one embodiment, the linear model is an OE model.

[0086] In a specific embodiment, the OE model expression is:

[0087]

[0088] in,

[0089] B(z -1 )=b1z -1 +...+b nb z -nb

[0090] F(z -1 )=1+f1z -1 +...+f nf z -nf

[0091] Where u t is the input sequence of the system at time t, that is, the actual oil production at time t; y t is the output value of the system at time t, that is, the predicted oil production at time t; ε tis the disturbance value of the system at time t, that is, the system error; B(z -1 )、F(z -1 ) is the unit shift operator z -1 The purpose is to better fit the predicted value with the actual value, where b and f are constants, z -1 Represents the unit backshift operator. In order to simplify the expression of the difference equation and facilitate the solution, an operator that performs a "shift" operation on the time variable is introduced. The commonly used backshift operator is z -1 Used to define an operation that moves back one step in time.

[0092] The constants b and f are determined by comparing the predicted values ​​with the actual values ​​and using the degree of fit.

[0093] In one embodiment, the total amount of steam allocated to the target well in real time is determined based on the heavy oil thermal recovery data prediction model and the net present value formula, such as Figure 4 Shown, including:

[0094] S401: Obtaining the predicted oil production of the target well at time t based on the heavy oil thermal recovery data prediction model;

[0095] S402: Input the predicted oil production at time t into the net present value formula to obtain the total steam injection amount.

[0096] In a specific embodiment, the prediction data analysis model for three pairs of target wells (horizontal wells) established according to S301-S302 is as follows:

[0097]

[0098]

[0099]

[0100] Real-time nonlinear optimization of steam distribution based on NPV formula:

[0101]

[0102] in,

[0103]

[0104]

[0105]

[0106] Where N w is the total number of oil wells, dimensionless; p o is the oil price, in $ / bbl, To predict oil production rate, dimensionless; p wis the steam treatment cost, in $ / bbl; u is the allocated steam injection rate, dimensionless; D is the discount factor, dimensionless; is the total steam injection volume, unit: m 3 .

[0107] Since produced water is recycled, equation (14) is applied to account for produced water as part of the allocated injected steam treatment cost.

[0108]

[0109] in,

[0110]

[0111]

[0112] The steam utilization rate can be set during the optimization process as follows:

[0113] Full utilization

[0114]

[0115] Partial utilization

[0116]

[0117] Utilization rate greater than 80%

[0118]

[0119] Finding the best utilization

[0120]

[0121] The machine learning-based steam distribution method provided in this application is to establish a comprehensive reservoir model of three pairs of horizontal wells based on the sequential Gaussian simulation method, then establish a SAGD prediction data analysis model based on the OE model, and finally apply the NPV formula to achieve real-time nonlinear optimization of steam distribution.

[0122] To address the challenges of accurate steam metering and proper distribution in heavy oil thermal recovery, this application proposes building a comprehensive reservoir model for three pairs of horizontal wells using the Sequential Gaussian Simulation method. This model then establishes a SAGD prediction data analysis model based on the OE model. This model then uses the NPV formula for steam allocation to achieve real-time nonlinear optimization of steam injection rates, thereby optimizing steam distribution. This application is of great significance to the research of heavy oil thermal recovery technology.

[0123] The present application also provides a machine learning-based steam distribution device, as described in the following embodiments. Since the principle of solving the problem of this device is similar to that of the machine learning-based steam distribution method, the implementation of this device can refer to the implementation of the machine learning-based steam distribution method, and the repeated parts will not be repeated.

[0124] This application provides a steam distribution device based on machine learning, such as Figure 5 Shown, including:

[0125] The comprehensive reservoir model building unit 501 is used to build a downhole comprehensive reservoir model by sequential Gaussian simulation method based on the actual downhole formation points of the selected target well;

[0126] The heavy oil thermal recovery data prediction model establishment unit 502 is used to establish a heavy oil thermal recovery data prediction model through a linear model based on the simulation parameter values ​​of the simulation formation points in the downhole comprehensive reservoir model;

[0127] The steam allocation unit 503 is used to determine the total amount of steam allocated to the target well in real time according to the heavy oil thermal recovery data prediction model and the net present value formula, and to determine the steam utilization rate according to the total steam injection amount.

[0128] In one embodiment, the number of target wells selected is at least 3 pairs.

[0129] In one embodiment, if Figure 6 As shown, the comprehensive reservoir model building unit 501 includes:

[0130] A multivariate normal distribution function establishment module 601 is used to establish a multivariate normal distribution function of actual formation points;

[0131] The mathematical expectation value calculation module 602 is used to calculate the mathematical expectation value of the conditional distribution of the actual formation point according to the multivariate normal distribution condition theorem;

[0132] The simulation parameter module 603 is used to obtain the simulation parameter value at the simulation formation point position according to the mathematical expectation value.

[0133] In one embodiment, if Figure 7 As shown, the heavy oil thermal recovery data prediction model building unit 502 includes:

[0134] Real-time data acquisition module 701, used to obtain the actual oil production and disturbance value at time t;

[0135] The model building module 702 is used to input the simulation parameter values, the actual oil production at time t and the disturbance value into the linear model to generate a heavy oil thermal recovery data prediction model.

[0136] In one embodiment, the linear model is an OE model.

[0137] In one embodiment, if Figure 8 As shown, the steam distribution unit 503 includes:

[0138] The oil production prediction module 801 is used to obtain the predicted oil production of the target well at time t based on the heavy oil thermal recovery data prediction model;

[0139] The total steam injection amount calculation module 802 is used to input the predicted oil production at time t into the net present value formula to obtain the total steam injection amount.

[0140] An embodiment of the present application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned steam distribution method based on machine learning when executing the computer program.

[0141] An embodiment of the present application further provides a computer-readable storage medium storing a computer program, which implements the above-mentioned steam distribution method based on machine learning when executed by a processor.

[0142] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the above-mentioned steam distribution method based on machine learning.

[0143] In the embodiment of the present application, compared with the technical solutions in the prior art, a heavy oil thermal recovery data prediction model is established through machine learning, steam distribution is optimized in real time, and gas injection parameters of each oil well are reasonably allocated to avoid waste of resources and effectively improve the efficiency of heavy oil reservoir recovery.

[0144] Figure 9 The physical structure diagram of the electronic device provided in the embodiment of the present application is as follows: Figure 9 As shown, the electronic device includes: a processor (processor) 901, a memory (memory) 902 and a bus 903.

[0145] The processor 901 and the memory 902 communicate with each other via a bus 903 .

[0146] The processor 901 is configured to call program instructions in the memory 902 to execute the methods provided by the above method embodiments.

[0147] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0148] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0149] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0150] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0151] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A steam distribution method based on machine learning, characterized in that: include: Based on the actual downhole formation points of the selected target well, a downhole comprehensive reservoir model is established through the sequential Gaussian simulation method; Establishing a heavy oil thermal recovery data prediction model through a linear model based on the simulation parameter values ​​of the simulation formation points in the downhole comprehensive reservoir model; Determining the total amount of steam allocated to the target well in real time based on the heavy oil thermal recovery data prediction model and the net present value formula, and determining the steam utilization rate based on the total steam injection volume; The method of establishing a heavy oil thermal recovery data prediction model using a linear model includes: Get the actual oil production and disturbance value at time t; Inputting the simulation parameter value, the actual oil production at time t and the disturbance value into the linear model to generate a heavy oil thermal recovery data prediction model; The method of determining the total amount of steam allocated to the target well in real time according to the heavy oil thermal recovery data prediction model and the net present value formula includes: Obtain the predicted oil production of the target well at time t based on the heavy oil thermal recovery data prediction model; The predicted oil production at time t is input into the net present value formula to obtain the total steam injection amount.

2. The steam distribution method based on machine learning according to claim 1, characterized in that: The number of target wells selected is at least 3 pairs.

3. The steam distribution method based on machine learning according to claim 1, characterized in that: The method of establishing a downhole comprehensive reservoir model by sequential Gaussian simulation includes: Establishing a multivariate normal distribution function of the actual stratigraphic points; Calculate the mathematical expectation value of the conditional distribution of the actual formation point according to the multivariate normal distribution condition theorem; The simulation parameter value at the simulation formation point position is obtained according to the mathematical expectation value.

4. The steam distribution method based on machine learning according to claim 1, characterized in that: The linear model is an OE model.

5. A steam distribution device based on machine learning, characterized in that: include: A comprehensive reservoir model building unit is used to build a downhole comprehensive reservoir model by sequential Gaussian simulation method based on the actual downhole formation points of the selected target well; a heavy oil thermal recovery data prediction model establishment unit, configured to establish a heavy oil thermal recovery data prediction model through a linear model according to simulation parameter values ​​of simulation formation points in the downhole comprehensive reservoir model; a steam distribution unit, configured to determine the total amount of steam to be distributed to the target well in real time according to the heavy oil thermal recovery data prediction model and the net present value formula, and to determine the steam utilization rate according to the total steam injection amount; The heavy oil thermal recovery data prediction model establishment unit includes: Real-time data acquisition module, used to obtain the actual oil production and disturbance value at time t; a model building module, configured to input the simulation parameter values, the actual oil production at time t, and the disturbance value into the linear model to generate a heavy oil thermal recovery data prediction model; The steam distribution unit comprises: The oil production prediction module is used to obtain the predicted oil production of the target well at time t based on the heavy oil thermal recovery data prediction model; The total steam injection amount calculation module is used to input the predicted oil production at time t into the net present value formula to obtain the total steam injection amount.

6. The machine learning-based steam distribution device according to claim 5, characterized in that: The number of target wells selected is at least 3 pairs.

7. The machine learning-based steam distribution device according to claim 5, characterized in that: The comprehensive reservoir model building unit includes: A multivariate normal distribution function establishment module, used for establishing the multivariate normal distribution function of the actual formation point; A mathematical expectation value calculation module, used for calculating the mathematical expectation value of the conditional distribution of the actual formation point according to the multivariate normal distribution condition theorem; A simulation parameter module is used to obtain the simulation parameter value at the simulation formation point position according to the mathematical expectation value.

8. The machine learning-based steam distribution device according to claim 5, characterized in that: The linear model is an OE model.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 4 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

11. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images