A method and apparatus for determining reservoir pressure based on neural networks

Abstract

The application discloses a method and device for determining reservoir pressure based on a neural network, and comprises the following steps: obtaining position information and time information of a point to be interpreted in a reservoir; generating input features based on the position information and the time information; inputting the input features into a reservoir pressure prediction model to predict the reservoir pressure, wherein the reservoir pressure prediction model is a neural network model trained based on mass conservation, control equations and boundary conditions; and performing encoding processing on the predicted reservoir pressure to obtain a target reservoir pressure of the point to be interpreted. The application realizes high-precision solution of the reservoir pressure without any label data, and improves the accuracy of determining the reservoir pressure.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and apparatus for determining reservoir pressure based on neural networks. Background Technology

[0002] Reservoir pressure is a crucial parameter in oil and gas reservoir development. Although methods using neural networks to determine reservoir pressure exist, current methods require measured bottom-hole pressure data or numerical solutions obtained through numerical computation as labeled data to train the model. Without labeled data, existing methods suffer from low accuracy and limited application when solving flow equations with source-sink terms and cumulative terms. Furthermore, obtaining measured data as labels is quite challenging. Summary of the Invention

[0003] To address the aforementioned problems, this invention provides a method and apparatus for determining reservoir pressure based on neural networks, which achieves high-precision solution of reservoir pressure without requiring any tag data, thereby improving the accuracy of reservoir pressure determination.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for determining reservoir pressure based on a neural network includes:

[0006] Obtain the location and time information of the points to be interpreted in the reservoir;

[0007] Based on the location and time information, input features are generated;

[0008] The input features are input into the reservoir pressure prediction model to predict the reservoir pressure. The reservoir pressure prediction model is a neural network model trained based on mass conservation, governing equations, and boundary conditions.

[0009] The predicted reservoir pressure is encoded to obtain the target reservoir pressure at the point to be interpreted.

[0010] Optionally, the method further includes:

[0011] Obtain training data, which includes the location and time information of any point in the reservoir;

[0012] The training data is input into the main neural network to obtain the main network output;

[0013] The main network output is encoded to obtain the encoded network output;

[0014] The encoded network output is fed into a pre-constructed target loss function to train the neural network by minimizing the loss until the loss reaches the target value, thus obtaining the reservoir pressure prediction model.

[0015] The target loss function is a function determined based on mass conservation, governing equations, and boundary conditions.

[0016] Optionally, the method further includes:

[0017] The boundary condition loss function is determined based on the initial reservoir pressure, the radius of the circle centered at the well center, the radius of the production well, the formation volume coefficient, the fluid viscosity, the reservoir thickness, and the reservoir permeability.

[0018] Based on porosity, fluid compressibility, bottom layer volume factor under reference pressure, porosity under reference pressure, rock compressibility, formation volume factor, initial reservoir pressure, radius of the circle centered at the well center, and reservoir permeability, the loss function of the governing equation is determined.

[0019] The mass conservation loss function is determined based on flow rate, porosity, current reservoir pressure, formation volume factor, and total reservoir volume.

[0020] The target loss function is determined based on the boundary condition loss function, the governing equation loss function, and the mass conservation loss function.

[0021] Optionally, the step of encoding the main network output to obtain the encoded network output includes:

[0022] Obtain the location information of production wells in the reservoir;

[0023] Based on the location information and acquisition time, an asymptotic function is constructed;

[0024] The main network output is encoded based on the asymptotic function to obtain the encoded network output.

[0025] Optionally, the method further includes:

[0026] The reservoir pressure prediction model is adjusted based on the predicted reservoir pressure to obtain the adjusted reservoir pressure prediction model.

[0027] An apparatus for determining reservoir pressure based on a neural network, comprising:

[0028] The acquisition unit is used to obtain the location and time information of the points to be interpreted in the reservoir.

[0029] A generation unit is used to generate input features based on the location information and time information;

[0030] The prediction unit is used to input the input features into the reservoir pressure prediction model to predict the reservoir pressure. The reservoir pressure prediction model is a neural network model trained based on mass conservation, governing equations, and boundary conditions.

[0031] The encoding unit is used to encode the predicted reservoir pressure to obtain the target reservoir pressure at the point to be interpreted.

[0032] Optionally, the device further includes:

[0033] A data acquisition unit is used to acquire training data, which includes location and time information of any point in the reservoir.

[0034] The first input unit is used to input the training data into the main neural network to obtain the main network output;

[0035] An encoding processing unit is used to encode the main network output to obtain an encoded network output.

[0036] The training unit is used to feed the encoded network output into a pre-constructed target loss function to train the neural network by minimizing the loss until the loss reaches the target value, thereby obtaining a reservoir pressure prediction model.

[0037] The target loss function is a function determined based on mass conservation, governing equations, and boundary conditions.

[0038] Optionally, the apparatus further includes: a loss function determination unit for:

[0039] The boundary condition loss function is determined based on the initial reservoir pressure, the radius of the circle centered at the well center, the radius of the production well, the formation volume coefficient, the fluid viscosity, the reservoir thickness, and the reservoir permeability.

[0040] Based on porosity, fluid compressibility, bottom layer volume factor under reference pressure, porosity under reference pressure, rock compressibility, formation volume factor, initial reservoir pressure, radius of the circle centered at the well center, and reservoir permeability, the loss function of the governing equation is determined.

[0041] The mass conservation loss function is determined based on flow rate, porosity, current reservoir pressure, formation volume factor, and total reservoir volume.

[0042] The target loss function is determined based on the boundary condition loss function, the governing equation loss function, and the mass conservation loss function.

[0043] Optionally, the encoding processing unit is specifically used for:

[0044] Obtain the location information of production wells in the reservoir;

[0045] Based on the location information and acquisition time, an asymptotic function is constructed;

[0046] The main network output is encoded based on the asymptotic function to obtain the encoded network output.

[0047] Optionally, the device further includes:

[0048] The model adjustment unit is used to adjust the reservoir pressure prediction model based on the predicted reservoir pressure to obtain the adjusted reservoir pressure prediction model.

[0049] Compared to existing technologies, this invention provides a method and apparatus for determining reservoir pressure based on a neural network, comprising: obtaining location and time information of a point to be interpreted in the reservoir; generating input features based on the location and time information; inputting the input features into a reservoir pressure prediction model to predict the reservoir pressure, wherein the reservoir pressure prediction model is a neural network model trained based on mass conservation, governing equations, and boundary conditions; and encoding the predicted reservoir pressure to obtain the target reservoir pressure of the point to be interpreted. This invention achieves high-precision solution for reservoir pressure without requiring any labeled data, thus improving the accuracy of reservoir pressure determination. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0051] Figure 1 A schematic flowchart illustrating a method for determining reservoir pressure based on a neural network, provided in an embodiment of the present invention;

[0052] Figure 2 A schematic diagram illustrating the training of a reservoir pressure prediction model provided in an embodiment of the present invention;

[0053] Figure 3 This is a schematic diagram of the structure of a device for determining reservoir pressure based on a neural network, provided in an embodiment of the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] The terms "first" and "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units may include steps or units not listed, but rather steps or units not listed.

[0056] This invention discloses a method for determining reservoir pressure based on a neural network. After obtaining the location and time information of the point to be interpreted input by the user, the location and time information of the point to be interpreted are input into a reservoir pressure prediction model. Without requiring measured bottomhole pressure data as label data, the final reservoir pressure is obtained by encoding the output of the reservoir pressure prediction model. The reservoir pressure prediction model is obtained by minimizing the loss function based on mass conservation, governing equations (also known as seepage equations), and boundary conditions. Using the reservoir pressure output by the prediction model as the interpretation result for the point to be interpreted can significantly improve the interpretation accuracy of the reservoir pressure at the point to be interpreted within the reservoir area. For the reservoir pressure (i.e., bottomhole pressure) in the area where the well is located, it can effectively solve and predict the reservoir pressure at the point to be interpreted within that area, and improve the interpretation accuracy.

[0057] To facilitate the explanation of the method for determining reservoir pressure based on neural networks provided by this invention, the relevant terminology of this invention will now be explained.

[0058] Fully connected neural networks: consisting of fully connected layers, feedforward network models using the backpropagation learning algorithm, which can realize arbitrary nonlinear mapping from input to output.

[0059] Labeled data: Given an input variable (x) and a known output variable (Y), an algorithm is used to learn the mapping function Y = f(X) from the input to the output, where (Y) is the labeled data.

[0060] Automatic differentiation: The derivative of a function can be automatically calculated by a program.

[0061] The loss function based on physical equations is composed of the residuals constructed from the equations and boundary conditions, and is used to optimize the parameters of the neural network. For example, the seepage equation with source-sink terms and reservoir boundary conditions is shown below:

[0062] D(x,t,u)=0, seepage equation;

[0063] B1(x b1 ,t b1 ,u b1 ) = 0, source and sink terms;

[0064] B2(x b2 ,t2,u b2 ) = 0, reservoir boundary;

[0065] The loss function is then:

[0066]

[0067] See Figure 1 The above is a flowchart illustrating a method for determining reservoir pressure based on a neural network, as provided in an embodiment of the present invention. The method may include the following steps:

[0068] S101. Obtain the location and time information of the points to be interpreted in the reservoir.

[0069] S102. Based on the location information and time information, generate input features.

[0070] The point to be interpreted in the reservoir refers to the point where the reservoir pressure needs to be obtained. Its corresponding location information can be determined based on coordinate information, such as (x, y). The time information can be the acquisition time, i.e., the time when the location information was obtained, denoted by t. Based on this location information and time information, input features can be generated, denoted by (x, y, t).

[0071] S103. Input the input features into the reservoir pressure prediction model to predict the reservoir pressure.

[0072] S104. Encode the predicted reservoir pressure to obtain the target reservoir pressure at the point to be interpreted.

[0073] The reservoir pressure prediction model is a neural network model trained based on mass conservation, governing equations, and boundary conditions.

[0074] This invention also provides a method for creating a reservoir pressure prediction model, which may include:

[0075] Obtain training data, which includes the location and time information of any point in the reservoir;

[0076] The training data is input into the main neural network to obtain the main network output;

[0077] The main network output is encoded to obtain the encoded network output;

[0078] The encoded network output is fed into a pre-constructed target loss function to train the neural network by minimizing the loss until the loss reaches the target value, thus obtaining the reservoir pressure prediction model.

[0079] The target loss function is a function determined based on mass conservation, governing equations, and boundary conditions.

[0080] See Figure 2 This diagram illustrates the training of a reservoir pressure prediction model according to an embodiment of the present invention. First, training data is acquired by randomly selecting an appropriate number of training data points (x, y, t) from the reservoir, including spatial and temporal dimensions, and inputting them into the main neural network to obtain the main network output. The main network output is then encoded, where an asymptotic function D(X, t) is constructed, containing a sub-neural network, resulting in the encoded network output. The encoded network output is then fed into a constructed loss function, and the neural network is trained by minimizing the loss until the loss no longer decreases and tends to stabilize. At this point, the reservoir pressure prediction model training is complete. Inputting any time and spatial point (x, y, t) into the trained prediction model yields the final reservoir pressure.

[0081] Furthermore, in this embodiment of the invention, the loss function is a function determined based on mass conservation, governing equations, and boundary conditions, specifically:

[0082] The boundary condition loss function is determined based on the initial reservoir pressure, the radius of the circle centered at the well center, the radius of the production well, the formation volume coefficient, the fluid viscosity, the reservoir thickness, and the reservoir permeability.

[0083] Based on porosity, fluid compressibility, bottom layer volume factor under reference pressure, porosity under reference pressure, rock compressibility, formation volume factor, initial reservoir pressure, radius of the circle centered at the well center, and reservoir permeability, the loss function of the governing equation is determined.

[0084] The mass conservation loss function is determined based on flow rate, porosity, current reservoir pressure, formation volume factor, and total reservoir volume.

[0085] The target loss function is determined based on the boundary condition loss function, the governing equation loss function, and the mass conservation loss function.

[0086] The target loss function (denoted by L) in this embodiment of the invention consists of three parts, namely the control equation (denoted by L) PDE (represented by L), boundary conditions (in L) BC (represented by L), the mass conservation condition (in L) mass express).

[0087] Specifically, L = L PDE +L BC +L mass .

[0088]

[0089]

[0090]

[0091] Where, φ ref To the reference pressure p ref Porosity below; C f C is the fluid compressibility coefficient. r B is the rock compressibility coefficient; ref To the reference pressure p ref The formation volume factor is below; k is the reservoir permeability; μ is the fluid viscosity; Q is the flow rate; r w r is the radius of the production well; h is the reservoir thickness; r is the radius of the circle centered at the well center; u r The pressure at a point on a circle with radius r from the center of the well; u w f V is the bottom hole flowing pressure; V is the total volume of the reservoir; t is time. L represents the gradient symbol. PDE The loss function is constructed for the governing equation, i.e., the seepage equation, so that the neural network output has physical meaning and satisfies the flow laws. L BC The loss function constructed for the boundary conditions, where L practical The boundary conditions are set according to the reservoir boundary conditions in the actual problem, such as closed boundary, isobaric boundary, etc. L mass The law of conservation of mass is intended to better align with the laws of physics.

[0092] The loss function here does not use labeled data, meaning it does not include a loss term. (i.e., the error between the network output and the label data). Simultaneously train the main neural network and the sub-neural network by minimizing the loss function until the loss no longer decreases and tends to stabilize. Then the training is complete, and the final reservoir pressure prediction model is obtained.

[0093] In one embodiment of the present invention, the step of encoding the main network output to obtain the encoded network output includes: obtaining the location information of the production well in the reservoir; constructing an asymptotic function based on the location information and the acquisition time; and encoding the main network output based on the asymptotic function to obtain the encoded network output.

[0094] First, the data (x, y, t) representing space and time are input into the main neural network (i.e., a fully connected neural network), and the output p of the main network is obtained. N(·) .

[0095] For output p N(·) The encoding process is as follows:

[0096] p = p inital +D(X,t)·p N(·)

[0097] Where p is the final reservoir pressure, i.e. the target reservoir pressure.

[0098] p inital This represents the initial reservoir pressure. D(X,t) is the constructed asymptotic function, as shown below:

[0099]

[0100] By encoding the output of the main network, the pressure can be automatically satisfied with the initial conditions, and the final reservoir pressure obtained can be more accurate.

[0101] It should be noted that the asymptotic function is a function model for simulating reservoir pressure changes constructed in this invention. It is constructed based on the asymptotic solution in the numerical solution method. Here, X is a function of (x,y,t), which is the intrinsic mechanism function used to describe the reservoir pressure with respect to time and space information, and t is the time after the well starts production.

[0102] In one implementation of this invention, the method further includes:

[0103] The reservoir pressure prediction model is adjusted based on the predicted reservoir pressure to obtain the adjusted reservoir pressure prediction model.

[0104] In other words, the reservoir pressure prediction model in this embodiment of the invention can be an iterative update, that is, the model is adjusted according to the predicted reservoir pressure until the preset convergence condition is met, thereby improving the accuracy of the model prediction.

[0105] This invention provides a method for determining reservoir pressure based on a neural network. This method achieves high-precision reservoir pressure calculation without requiring any labeled data, and is particularly effective for determining the bottom hole pressure of production wells within reservoirs. Encoding the output of the main neural network improves the model's solution performance and accuracy. Incorporating a mass conservation term into the loss function makes the calculated reservoir pressure more consistent with physical laws and actual conditions.

[0106] This invention also provides an apparatus for determining reservoir pressure based on a neural network, see [link to relevant documentation]. Figure 3The device may include:

[0107] Acquisition unit 301 is used to obtain the location and time information of the points to be interpreted in the reservoir;

[0108] The generation unit 302 is used to generate input features based on the location information and time information;

[0109] Prediction unit 303 is used to input the input features into the reservoir pressure prediction model to predict the reservoir pressure. The reservoir pressure prediction model is a neural network model trained based on mass conservation, governing equations and boundary conditions.

[0110] The encoding unit 304 is used to encode the predicted reservoir pressure to obtain the target reservoir pressure of the point to be interpreted.

[0111] This invention provides a device for determining reservoir pressure based on a neural network, comprising: an acquisition unit obtaining location and time information of a point to be interpreted in the reservoir; a generation unit generating input features based on the location and time information; a prediction unit inputting the input features into a reservoir pressure prediction model to predict the reservoir pressure, wherein the reservoir pressure prediction model is a neural network model trained based on mass conservation, governing equations, and boundary conditions; and an encoding unit encoding the predicted reservoir pressure to obtain the target reservoir pressure of the point to be interpreted. This invention achieves high-precision solution for reservoir pressure without requiring any labeled data, thus improving the accuracy of reservoir pressure determination.

[0112] In one embodiment, the device further includes:

[0113] A data acquisition unit is used to acquire training data, which includes location and time information of any point in the reservoir.

[0114] The first input unit is used to input the training data into the main neural network to obtain the main network output;

[0115] An encoding processing unit is used to encode the main network output to obtain an encoded network output.

[0116] The training unit is used to feed the encoded network output into a pre-constructed target loss function to train the neural network by minimizing the loss until the loss reaches the target value, thereby obtaining a reservoir pressure prediction model.

[0117] The target loss function is a function determined based on mass conservation, governing equations, and boundary conditions.

[0118] Furthermore, the apparatus further includes: a loss function determination unit for:

[0119] The boundary condition loss function is determined based on the initial reservoir pressure, the radius of the circle centered at the well center, the radius of the production well, the formation volume coefficient, the fluid viscosity, the reservoir thickness, and the reservoir permeability.

[0120] Based on porosity, fluid compressibility, bottom layer volume factor under reference pressure, porosity under reference pressure, rock compressibility, formation volume factor, initial reservoir pressure, radius of the circle centered at the well center, and reservoir permeability, the loss function of the governing equation is determined.

[0121] The mass conservation loss function is determined based on flow rate, porosity, current reservoir pressure, formation volume factor, and total reservoir volume.

[0122] The target loss function is determined based on the boundary condition loss function, the governing equation loss function, and the mass conservation loss function.

[0123] Optionally, the encoding processing unit is specifically used for:

[0124] Obtain the location information of production wells in the reservoir;

[0125] Based on the location information and acquisition time, an asymptotic function is constructed;

[0126] The main network output is encoded based on the asymptotic function to obtain the encoded network output.

[0127] In one embodiment, the device further includes:

[0128] The model adjustment unit is used to adjust the reservoir pressure prediction model based on the predicted reservoir pressure to obtain the adjusted reservoir pressure prediction model.

[0129] Based on the foregoing embodiments, embodiments of this application provide a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps of the method for determining reservoir pressure based on a neural network as described in any of the above claims.

[0130] This invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the steps of a method for determining reservoir pressure based on a neural network.

[0131] It should be noted that the aforementioned processor or CPU can be at least one of the following: Application-Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), Central Processing Unit (CPU), Controller, Microcontroller, and Microprocessor. Understandably, the electronic device implementing the above processor function can also be other types, and this application embodiment does not specifically limit its capabilities.

[0132] It should be noted that the aforementioned computer storage media / memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM), etc.; it can also be various terminals that include one or any combination of the above-mentioned memory, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0133] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0134] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0135] Furthermore, in the various embodiments of this application, all functional units can be integrated into one processing module, or each unit can be a separate unit, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in a combination of hardware and software functional units. Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0136] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0137] The features disclosed in the several product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0138] The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0139] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0140] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0141] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method of determining reservoir pressure based on neural networks, characterized in that, include: Obtain the location and time information of the points to be interpreted in the reservoir; Based on the location and time information, input features are generated; The input features are input into the reservoir pressure prediction model to predict the reservoir pressure. The reservoir pressure prediction model is a neural network model trained based on mass conservation, governing equations, and boundary conditions. The predicted reservoir pressure is encoded to obtain the target reservoir pressure at the point to be interpreted. The steps for creating the reservoir pressure prediction model are as follows: obtaining training data, which includes the location and time information of any point in the reservoir; inputting the training data into the main neural network to obtain the main network output; encoding the main network output to obtain the encoded network output; and substituting the encoded network output into a pre-constructed target loss function to train the neural network by minimizing the loss until the loss reaches the target value, thereby obtaining the reservoir pressure prediction model. The process for determining the target loss function is as follows: The boundary condition loss function is determined based on the initial reservoir pressure, the radius of the circle centered at the well center, the radius of the production well, the formation volume coefficient, the fluid viscosity, the reservoir thickness, and the reservoir permeability. Based on porosity, fluid compressibility, formation volume factor at reference pressure, porosity at reference pressure, rock compressibility, formation volume factor, initial reservoir pressure, radius of the circle centered at the well center, and reservoir permeability, the loss function of the governing equation is determined. The mass conservation loss function is determined based on flow rate, porosity, current reservoir pressure, formation volume factor, and total reservoir volume. The target loss function is determined based on the boundary condition loss function, the governing equation loss function, and the mass conservation loss function.

2. The method of claim 1, wherein, The process of encoding the main network output to obtain the encoded network output includes: Obtain the location information of production wells in the reservoir; Based on the location information and acquisition time, an asymptotic function is constructed; The main network output is encoded based on the asymptotic function to obtain the encoded network output.

3. The method of claim 1, wherein, The method further includes: The reservoir pressure prediction model is adjusted based on the predicted reservoir pressure to obtain the adjusted reservoir pressure prediction model.

4. An apparatus for determining reservoir pressure based on neural networks, characterized by include: The acquisition unit is used to obtain the location and time information of the points to be interpreted in the reservoir. A generation unit is used to generate input features based on the location information and time information; The prediction unit is used to input the input features into the reservoir pressure prediction model to predict the reservoir pressure. The reservoir pressure prediction model is a neural network model trained based on mass conservation, governing equations, and boundary conditions. The encoding unit is used to encode the predicted reservoir pressure to obtain the target reservoir pressure at the point to be interpreted. A data acquisition unit is used to acquire training data, which includes location and time information of any point in the reservoir. The first input unit is used to input the training data into the main neural network to obtain the main network output; An encoding processing unit is used to encode the main network output to obtain an encoded network output. The training unit is used to substitute the encoded network output into a pre-constructed target loss function to train the neural network by minimizing the loss until the loss reaches the target value, thereby obtaining a reservoir pressure prediction model. The loss function determination unit is used to determine the boundary condition loss function based on the initial reservoir pressure, the radius of the circle centered at the well center, the radius of the production well, the formation volume coefficient, the fluid viscosity, the reservoir thickness, and the reservoir permeability; to determine the governing equation loss function based on porosity, fluid compressibility, formation volume coefficient at reference pressure, porosity at reference pressure, rock compressibility, formation volume coefficient, initial reservoir pressure, the radius of the circle centered at the well center, and the reservoir permeability; to determine the mass conservation loss function based on flow rate, porosity, the reservoir pressure at the current time, the formation volume coefficient, and the total reservoir volume; and to determine the target loss function based on the boundary condition loss function, the governing equation loss function, and the mass conservation loss function.

5. The apparatus according to claim 4, characterized in that, The encoding processing unit is specifically used for: Obtain the location information of production wells in the reservoir; Based on the location information and acquisition time, an asymptotic function is constructed; The main network output is encoded based on the asymptotic function to obtain the encoded network output.

6. The apparatus according to claim 4, characterized in that, The device further includes: The model adjustment unit is used to adjust the reservoir pressure prediction model based on the predicted reservoir pressure to obtain the adjusted reservoir pressure prediction model.

Images