An LSTM prediction method and system for heat recovery performance of enhanced geothermal system

Abstract

The present invention belongs to the technical field of geothermal capacity prediction, and specifically relates to an LSTM prediction method and system for the heat recovery performance of an enhanced geothermal system, comprising the following steps: S1: using a dual-well system to establish a three-dimensional geological model and obtain data of the enhanced geothermal system; S2: preprocessing the acquired data and using the preprocessed data as input data of a prediction model; S3: calculating a variable slope and normalizing the obtained slope; S4: constructing an LSTM prediction model to predict the outlet water temperature of the geothermal system; S5: training the LSTM model and performing prediction, and combining evaluation indicators to determine the accuracy of the prediction results.

Description

Technical Field

[0001] The present invention belongs to the technical field of geothermal capacity prediction, and specifically relates to an LSTM prediction method and system for the heat recovery performance of an enhanced geothermal system. Background Art

[0002] In the process of geothermal energy extraction, the enhanced geothermal system first needs to convert low-porosity and low-permeability rocks into high-permeability geothermal reservoirs through hydraulic fracturing and other methods, and then circulate the working fluid in the reservoir through injection wells and extraction wells to achieve heat energy extraction. From the extraction process of the enhanced geothermal system, it can be found that many characteristics of the reservoir and the working fluid will affect the extraction of heat energy, such as reservoir temperature, fracture distribution, fracture surface heat exchange capacity, permeability, porosity, working fluid injection temperature, etc. Many factors make the problem of heat energy extraction in the enhanced geothermal system more complicated.

[0003] There are two main problems in the existing research literature on thermal performance prediction: first, the data source is single, and the input values ​​used are self-set with strong regularity. In the subsequent prediction process, this will lead to a single input, which cannot play a good training role for the model and reduce the credibility of subsequent research; second, the selection of input combinations is relatively fixed, mostly a fixed set of input variables, and then the impact of specific numerical changes of the variables on the results is studied. Summary of the invention

[0004] The present invention provides an LSTM prediction method and system for the heat recovery performance of an enhanced geothermal system. By constructing a three-dimensional geological model and preprocessing the acquired data, it is no longer necessary to collect detailed stratum data, which simplifies the data acquisition process. At the same time, variable slope calculation is used to construct an LSTM prediction model, which can make the model prediction range wide, the fitting effect good, and the prediction result accurate, which is of great significance to the energy development and utilization of the enhanced geothermal system.

[0005] An LSTM prediction method for the heat recovery performance of an enhanced geothermal system includes the following steps:

[0006] S1: Use a dual well system to build a 3D geological model and obtain data for the enhanced geothermal system;

[0007] S2: preprocessing the acquired data and using the preprocessed data as input data of the prediction model;

[0008] S3: Calculate the variable slope and normalize the obtained slope;

[0009] S4: Build an LSTM prediction model to predict the outlet water temperature of the geothermal system;

[0010] S5: Train the LSTM model and make predictions, and use evaluation indicators to determine the accuracy of the prediction results.

[0011] By constructing a three-dimensional geological model and preprocessing the acquired data, it is no longer necessary to collect detailed stratigraphic data, which simplifies the data acquisition process. At the same time, variable slope calculation is used to construct an LSTM prediction model, which can make the model prediction range wide, the fitting effect good, and the prediction results accurate, which is of great significance to the energy development and utilization of enhanced geothermal systems.

[0012] Furthermore, in S1, the process of establishing a three-dimensional geological model using a dual well system specifically includes:

[0013] S11: Define the calculation area of ​​the 3D geological model and construct a 3D geological conceptual model by combining geological and engineering data;

[0014] S12: Setting mathematical assumptions and required physical parameters of the 3D geological model;

[0015] S13: Based on the 3D geological conceptual model and parameter settings, a 3D geological model is established using numerical simulation software;

[0016] S14: Based on the three-dimensional geological model, the boundary conditions and initial state are initialized, and the mathematical equations of the matrix rock block seepage field, the fracture seepage field, the matrix rock block temperature field, and the fracture water temperature field are defined.

[0017] By constructing a three-dimensional geological conceptual model and presetting the model-related assumptions and physical parameters to establish a three-dimensional geological model, and defining the relevant mathematical equations of matrix rocks and fractures, it is easy to obtain sufficient data required for the enhanced geothermal system, thereby eliminating the need for manual acquisition and simplifying the data acquisition process.

[0018] Furthermore, in S11, the computational domain of the three-dimensional geological model is a cuboid of 1000×1000×400 m, and the well length is 80 m; the fractures include 80 randomly generated small fractures and 2 main fractures, and are located in the 200 m area in the middle of the three-dimensional geological model.

[0019] Furthermore, in said S12,

[0020] The mathematical assumptions of the 3D geological model are:

[0021] (1) The rock matrix is ​​isotropic and uniform;

[0022] (2) The flow is considered to be unidirectional;

[0023] (3) Fluid flow obeys Darcy's law and capillary forces are not considered;

[0024] (4) The heat exchange between fluid and rock follows Fourier's law and is characterized by local thermal equilibrium;

[0025] The physical parameters required for the 3D geological model include:

[0026] Bottom hole production pressure P, fracture permeability K f , reservoir density ρ s , reservoir thermal conductivity λ s , reservoir permeability K s , reservoir porosity Φ s , reservoir specific heat capacity C s , fracture porosity Φ f , working fluid injection temperature T i , injection-production well spacing D, fluid injection flow F, fracture thickness H, reservoir temperature T f .

[0027] Furthermore, in S14,

[0028] The mathematical equation of the seepage field of matrix rock is:

[0029]

[0030] In the formula, S represents the rock mass water storage coefficient; t represents time; u represents the rock mass water flow rate, and its calculation expression is: K s represents rock permeability, η represents the dynamic viscosity of water, p represents water pressure, ρ f represents the water density in the crack, g represents the acceleration of gravity, z represents the elevation in the z direction; e represents the volume strain of the rock block; Q represents the source and sink term of the seepage;

[0031] The mathematical equation of the seepage field in the fracture is:

[0032]

[0033] Where, d f Indicates the crack width; S f represents the water storage coefficient of the fracture; represents the tangential derivative along the crack; K f represents the permeability of the fracture; e f represents the volume strain of the crack surface; Q f It represents the flow exchange between rock block and fracture surface, and its calculation expression is:

[0034] The mathematical equation of the temperature field of the matrix rock is:

[0035]

[0036] In the formula, C srepresents the specific heat capacity of rock; ρ s Indicates the density of rock mass; T s represents the rock temperature; s represents the thermal conductivity of the rock; W represents the heat source;

[0037] The mathematical equation of the fracture water temperature field is:

[0038]

[0039] In the formula, C p represents effective specific heat capacity; T f represents the temperature of water in the crack; ρ f Indicates water density; C f represents the specific heat of water; u f represents the water flow velocity in the crack; f Indicates the water thermal conductivity; W f It represents the heat absorbed by the water on the fracture surface from the bedrock. Its calculation expression is: W f =h(T s -T f ), h represents the convective heat transfer coefficient.

[0040] Furthermore, in S2, the method of preprocessing the acquired enhanced geothermal system data includes removing outliers, normalizing values, and binning operations.

[0041] Furthermore, in S2, before using the preprocessed data as input data of the prediction model, the following steps are further included:

[0042] S21: grouping the processed data according to the physical parameter type;

[0043] S22: Based on the physical parameter type data, the random forest algorithm is used to rank their importance;

[0044] S23: Select the top five physical type parameters and combine them, and use them as input data for the LSTM prediction model.

[0045] By grouping the acquired data according to the data type, sorting it using the random forest algorithm, and constructing different combinations as the input of the LSTM prediction model, the prediction change rules are discovered, so as to repeatedly compare the prediction effects of different input combinations, and then obtain the optimal combination as the input of the LSTM prediction model to improve the prediction accuracy of the LSTM prediction model.

[0046] Furthermore, in S4, the LSTM prediction model includes a forget gate, an input gate, and an output gate;

[0047] Forget Gate:

[0048] f t =σ(w fh h t-1 +w fh x t +b f );

[0049] Input Gate:

[0050] i t =σ(w ih h t-1 +w ih x t +b i );

[0051] Cell status update:

[0052]

[0053] New cell state:

[0054]

[0055] Output Gate:

[0056] o t =σ(w oh h t-1 +w ox x t +b o );

[0057] In the formula, i t represents the input of the input gate; σ() represents the sigmoid activation function, w i x represents the weight matrix of the input gate; h t-1 represents the actual value at time t-1; x t represents the input at the current time t; b i represents the bias term of the input gate; represents the cell state at the candidate time t; tanh() represents the tanh activation function; w ch Represents the weight matrix of cell state update; b c Represents the bias term for cell state update; C t represents the cell state at the current time t; f t represents the output of the forget gate; C t-1 represents the cell state at time t-1; w fh represents the weight matrix of the forget gate; b f Represents the bias term of the forget gate; o t represents the output of the output gate; w oh represents the weight matrix corresponding to the actual value at time t-1 in the output gate; wox Represents the weight matrix corresponding to the input at the current time t in the output gate; b o Represents the bias term of the output gate.

[0058] Furthermore, in S5, the evaluation indicators include:

[0059] The MSE evaluation indicator is expressed as:

[0060]

[0061] Where N is the number of samples; y i represents the i-th actual observation value; represents the i-th predicted value;

[0062] The MAPE evaluation indicator is expressed as:

[0063]

[0064] The R-squared evaluation indicator is expressed as:

[0065]

[0066] In the formula, R 2 is the coefficient of determination, which is used to measure the degree to which the model explains the variability of the data. 2 The value range is 0 to 1; Represents the average of the actual observations.

[0067] A system for an LSTM prediction method for enhanced geothermal system heat recovery performance, comprising:

[0068] A three-dimensional geological model building module, which is used to build a three-dimensional geological model;

[0069] A data acquisition module, which is used to acquire data of the enhanced geothermal system;

[0070] A preprocessing module, which is used to preprocess the acquired enhanced geothermal system data;

[0071] Model building and training module, which is used to build LSTM prediction model and perform training prediction;

[0072] The evaluation and verification module is used to evaluate and verify the prediction model.

[0073] The beneficial effects of the present invention are:

[0074] The present invention constructs a three-dimensional geological model and pre-processes the acquired data, so that there is no need to collect detailed stratum data, which simplifies the data acquisition process. At the same time, variable slope calculation is used to construct an LSTM prediction model, which can make the model prediction range wide, the fitting effect good, and the prediction result accurate, which is of great significance to the energy development and utilization of enhanced geothermal systems; the acquired data are grouped according to the data type, sorted by a random forest algorithm, and different combinations are constructed as the input of the LSTM prediction model to explore the prediction change law, so as to repeatedly compare the prediction effects of different input combinations, and then obtain the optimal combination as the input of the LSTM prediction model to improve the prediction accuracy of the LSTM prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 is a flow chart of the present invention;

[0076] Figure 2 It is a schematic diagram of the three-dimensional geological model;

[0077] Figure 3 This is a schematic diagram of the structure of the LSTM prediction model algorithm;

[0078] Figure 4 Schematic diagram of importance ranking for random forest;

[0079] Figure 5 Schematic diagram of simulation results for the best prediction input combination;

[0080] Figure 6 It is a schematic diagram of simulation results preprocessed under the best prediction input combination;

[0081] Figure 7 Schematic diagram of the structure of the LSTM prediction system for the heat recovery performance of the enhanced geothermal system. DETAILED DESCRIPTION

[0082] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0083] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present disclosure, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0084] In addition, in the following description, specific details are provided to facilitate a thorough understanding of the examples, and a person of ordinary skill in the art can understand the specific meanings of the above terms in this application under specific circumstances.

[0085] Example 1

[0086] Figure 1 The figure shows an LSTM prediction method for the heat recovery performance of an enhanced geothermal system. By constructing a three-dimensional geological model and preprocessing the acquired data, it is no longer necessary to collect detailed stratum data, which simplifies the data acquisition process. At the same time, variable slope calculation is used to construct an LSTM prediction model, which can make the model prediction range wide, the fitting effect good, and the prediction result accurate, which is of great significance to the energy development and utilization of the enhanced geothermal system. It includes the following steps:

[0087] S1: Use a dual well system to build a 3D geological model and obtain data for the enhanced geothermal system;

[0088] Among them, Figure 2 As shown in FIG. 1 , the process of establishing a three-dimensional geological model using a dual-well system specifically includes:

[0089] S11: Define the calculation area of ​​the 3D geological model and construct a 3D geological conceptual model by combining geological and engineering data;

[0090] In this embodiment, the computational domain of the 3D geological model is a cuboid of 1000×1000×400 m, and the well length is 80 m; the fractures include 80 randomly generated small fractures and 2 main fractures, and are located in the 200 m area in the middle of the 3D geological model.

[0091] In this embodiment, the geological and engineering data include rock parameters and fracture parameters; wherein the rock parameters include rock density, rock specific heat capacity, rock thermal conductivity, rock porosity, and rock permeability; and the fracture parameters include fracture density, fracture specific heat capacity, fracture thermal conductivity, fracture porosity, fracture permeability, and fracture width.

[0092] S12: Setting mathematical assumptions and required physical parameters of the 3D geological model;

[0093] Among them, the mathematical assumptions of the three-dimensional geological model are:

[0094] (1) The rock matrix is ​​isotropic and uniform;

[0095] (2) The flow is considered to be unidirectional;

[0096] (3) Fluid flow obeys Darcy's law and capillary forces are not considered;

[0097] (4) The heat exchange between fluid and rock follows Fourier's law and is characterized by local thermal equilibrium.

[0098] Among them, the physical parameters required for the three-dimensional geological model are shown in Table 1;

[0099] Table 1 Physical parameters required for 3D geological model

[0100] variable character unit Variable meaning Bottom hole production pressure P Pa Bottom well pressure Fracture permeability <![CDATA[K f ]]> <![CDATA[m 2 ]]> Average permeability of fracture surface Reservoir density <![CDATA[ρ s ]]> <![CDATA[kg / m 3 ]]> Average density of the rock in which the reservoir is located Reservoir thermal conductivity <![CDATA[λ s ]]> W / (m·K) Average thermal conductivity of the rock in which the reservoir is located Reservoir permeability <![CDATA[K s ]]> <![CDATA[m 2 ]]> Average permeability of the rock in which the reservoir is located Reservoir porosity <![CDATA[Φ s ]]> / Porosity of the rock where the reservoir is located Reservoir specific heat capacity <![CDATA[C s ]]> J / (kg·K) Average specific heat capacity of the reservoir rock Fracture porosity <![CDATA[Φ f ]]> / Porosity of the rock where the reservoir is located Working fluid injection temperature <![CDATA[T i ]]> K The working fluid reaches the injection wellhead temperature Spacing between injection and production wells D m Horizontal distance between injection well and heat production well Working fluid injection flow F kg / s Working fluid flow rate in the injection well Crack thickness H m Average width of hydraulic fractures Reservoir temperature <![CDATA[T f ]]> K The temperature at the bottom of the wellbore in the reservoir

[0101] S13: Based on the 3D geological conceptual model and parameter settings, a 3D geological model is established using numerical simulation software;

[0102] S14: Based on the three-dimensional geological model, the boundary conditions and initial state are initialized, and the mathematical equations of the matrix rock block seepage field, the fracture seepage field, the matrix rock block temperature field, and the fracture water temperature field are defined.

[0103] Among them, the mathematical equation of the seepage field of the matrix rock is:

[0104]

[0105] In the formula, S represents the rock mass water storage coefficient; t represents time; u represents the rock mass water flow rate, and its calculation expression is: K s represents rock permeability, η represents the dynamic viscosity of water, p represents water pressure, ρ f represents the water density in the crack, g represents the acceleration of gravity, z represents the elevation in the z direction; e represents the volume strain of the rock block; Q represents the source and sink term of the seepage;

[0106] Among them, the mathematical equation of the seepage field in the fracture is:

[0107]

[0108] Where, d f Indicates the crack width; S f represents the water storage coefficient of the fracture; represents the tangential derivative along the crack; K frepresents the permeability of the fracture; e f represents the volume strain of the crack surface; Q f It represents the flow exchange between rock block and fracture surface, and its calculation expression is:

[0109] Among them, the mathematical equation of the temperature field of the matrix rock is:

[0110]

[0111] In the formula, C s represents the specific heat capacity of rock; ρ s Indicates the density of rock mass; T s represents the rock temperature; s represents the thermal conductivity of the rock; W represents the heat source;

[0112] Among them, the mathematical equation of the fracture water temperature field is:

[0113]

[0114] In the formula, C p represents effective specific heat capacity; T f represents the temperature of water in the crack; ρ f Indicates water density; C f represents the specific heat of water; u f represents the water flow velocity in the crack; f Indicates the water thermal conductivity; W f It represents the heat absorbed by the water on the fracture surface from the bedrock. Its calculation expression is: W f =h(T s -T f ), h represents the convective heat transfer coefficient.

[0115] By constructing a three-dimensional geological conceptual model and presetting the model-related assumptions and physical parameters to establish a three-dimensional geological model, and defining the relevant mathematical equations of matrix rocks and fractures, it is easy to obtain sufficient data required for the enhanced geothermal system, thereby eliminating the need for manual acquisition and simplifying the data acquisition process.

[0116] S2: preprocessing the acquired data and using the preprocessed data as input data of the prediction model;

[0117] In this embodiment, the method of preprocessing the acquired enhanced geothermal system data includes removing outliers, normalizing values, and binning operations.

[0118] Before using the preprocessed data as input data for the prediction model, the following steps are also included:

[0119] S21: grouping the processed data according to the physical parameter type;

[0120] S22: Based on the physical parameter type data, the random forest algorithm is used to rank their importance;

[0121] S23: Select the top five physical type parameters and combine them, and use them as input data for the LSTM prediction model.

[0122] By grouping the acquired data according to the data type, sorting it using the random forest algorithm, and constructing different combinations as the input of the LSTM prediction model, the prediction change rules are discovered, so as to repeatedly compare the prediction effects of different input combinations, and then obtain the optimal combination as the input of the LSTM prediction model to improve the prediction accuracy of the LSTM prediction model.

[0123] S3: Calculate the variable slope and normalize the obtained slope;

[0124] S4: Build an LSTM prediction model to predict the outlet water temperature of the geothermal system;

[0125] The LSTM prediction model is a long short-term memory network prediction model, such as Figure 3 As shown in the figure, it includes a forget gate, an input gate, an output gate, and a memory cell. The memory cell, as the core of LSTM, is responsible for storing important information and passing the information to the subsequent network layer, that is, it can follow the different inputs of the input gate, save historical information and use it for current predictions. The input gate can determine whether the current input is written into the memory cell and control the impact of the input signal on the memory cell. The forget gate controls the degree of forgetting of the information stored in the memory cell. The output gate controls whether the information stored in the memory will affect the subsequent network layer, which determines the output of the information.

[0126] Forget Gate:

[0127] f t =σ(w fh h t-1 +w fh x t +b f );

[0128] Input Gate:

[0129] i t =σ(w ih h t-1 +w ih x t +b i );

[0130] Memory cell status update:

[0131]

[0132] New memory cell status:

[0133]

[0134] Output Gate:

[0135] o t =σ(w oh h t-1 +w ox x t +b o );

[0136] In the formula, i t represents the input of the input gate; σ() represents the sigmoid activation function, w ih represents the weight matrix of the input gate; h t-1 represents the actual value at time t-1; x t represents the input at the current time t; b i represents the bias term of the input gate; represents the cell state at the candidate time t; tanh() represents the tanh activation function; w ch Represents the weight matrix of cell state update; b c Represents the bias term for cell state update; C t represents the cell state at the current time t; f t represents the output of the forget gate; C t-1 represents the cell state at time t-1; w fh represents the weight matrix of the forget gate; b f Represents the bias term of the forget gate; o t represents the output of the output gate; w oh represents the weight matrix corresponding to the actual value at time t-1 in the output gate; w ox Represents the weight matrix corresponding to the input at the current time t in the output gate; b o Represents the bias term of the output gate.

[0137] S5: Train the LSTM model and make predictions, and use evaluation indicators to determine the accuracy of the prediction results.

[0138] Among them, the evaluation indicators include MSE evaluation indicator, MAPE evaluation indicator, and R-squared evaluation indicator, which are used to determine the accuracy of the prediction results.

[0139] The smaller the calculated value of the MSE evaluation index, the smaller the model prediction error, that is, the better the model prediction performance. The expression of the MSE evaluation index is:

[0140]

[0141] Where N is the number of samples; y i represents the i-th actual observation value; represents the i-th predicted value;

[0142] The smaller the calculated value of MAPE, the higher the prediction accuracy of the model. The expression of MAPE is:

[0143]

[0144] The closer the calculated value of R-squared is to 1, the better the model fit is, that is, the stronger the ability to explain variability is. The expression of R-squared evaluation index is:

[0145]

[0146] In the formula, R 2 is the coefficient of determination, which is used to measure the degree to which the model explains the variability of the data. 2 The value range is 0 to 1; Represents the average of the actual observations.

[0147] Example 2

[0148] This embodiment provides an LSTM prediction method for the heat recovery performance of an enhanced geothermal system, which is used to accurately predict the outlet water temperature of the geothermal system several years later, and specifically includes the following steps:

[0149] T1: Use a dual well system to build a 3D geological model and obtain data for the enhanced geothermal system;

[0150] T11: Define the calculation area of ​​the 3D geological model and construct a 3D geological conceptual model by combining geological and engineering data;

[0151] In this embodiment, the computational domain of the 3D geological model is a cuboid of 1000×1000×400 m, and the well length is 80 m; the fractures include 80 randomly generated small fractures and 2 main fractures, and are located in the 200 m area in the middle of the 3D geological model.

[0152] In this embodiment, the geological and engineering data include rock parameters and fracture parameters, as shown in Table 2.

[0153] Table 2 Geological and engineering data

[0154]

[0155]

[0156] T12: Set the mathematical assumptions and required physical parameters of the 3D geological model;

[0157] Among them, the mathematical assumptions of the three-dimensional geological model are:

[0158] (1) The rock matrix is ​​isotropic and uniform;

[0159] (2) The flow is considered to be unidirectional;

[0160] (3) Fluid flow obeys Darcy's law and capillary forces are not considered;

[0161] (4) The heat exchange between fluid and rock follows Fourier's law and is characterized by local thermal equilibrium.

[0162] Among them, the physical parameters required for the 3D geological model include 13, namely, bottom hole production pressure P, fracture permeability K f , reservoir density ρ s , reservoir thermal conductivity λ, reservoir permeability K s , reservoir porosity Φ s , reservoir specific heat capacity C s , fracture porosity Φ f , working fluid injection temperature T i , injection-production well spacing D, fluid injection flow F, fracture thickness H, reservoir temperature T f .

[0163] T13: Based on the 3D geological conceptual model and parameter settings, a 3D geological model is established using numerical simulation software;

[0164] T14: Based on the three-dimensional geological model, initialize the boundary conditions and initial state, and define the mathematical equations of matrix rock seepage field, fracture seepage field, matrix rock temperature field, and fracture water temperature field.

[0165] In this embodiment, when initializing the boundary conditions and initial state, the top temperature of the fracture reservoir is defined as 473K; the temperature gradient is 50K / km; the initial pressure of the fracture reservoir is 30MPa; the initial pressure of the production well is 20MPa; the injection mass flow rate is 10kg / s; and the injection mass temperature is 293K.

[0166] In this embodiment, a total of 323 sets of simulation data are acquired.

[0167] T2: preprocess the acquired data and use the preprocessed data as input data for the prediction model;

[0168] In this embodiment, the method for preprocessing the acquired enhanced geothermal system data includes removing outliers, normalizing values, and binning operations; and dividing 60% of the acquired 323 sets of simulation data into training sets and 40% into test sets.

[0169] Before using the preprocessed data as input data for the prediction model, the following steps are also included:

[0170] T21: Group the processed data according to the physical parameter type;

[0171] T22: Based on the physical parameter type data, the random forest algorithm is used to rank their importance;

[0172] T23: Select the top five physical type parameters and combine them as input data for the LSTM prediction model.

[0173] In this embodiment, if Figure 4 As shown in the figure, the data of 13 physical type parameters are used as characteristic variables, and the outlet water temperature in the first, fifth, tenth, twentieth and thirtieth years is used as the dependent variable. The importance of each characteristic variable is obtained, and the first five are selected, that is, the reservoir temperature T is selected. f , fracture thickness H, fluid injection flow rate F, fluid injection temperature T i , the distance D between injection and production wells is used as the input of the LSTM prediction model.

[0174] T3: Calculate the variable slope and normalize the obtained slope;

[0175] T4: Build an LSTM prediction model to predict the outlet water temperature T of the geothermal system n ;

[0176] T5: Train the LSTM model and make predictions, and use evaluation indicators to determine the accuracy of the prediction results.

[0177] like Figure 5 As shown, the reservoir temperature T f , fracture thickness H, fluid injection flow rate F, fluid injection temperature T i , the distance between injection and production wells D is used as the input of the LSTM prediction model to obtain the outlet water temperature T in the 1st, 5th, 10th, 20th and 30th years. n By comparing the predicted results with the actual results, it can be seen that the predicted results are relatively close to the actual situation, and the error range is within 3%.

[0178] Based on the selection of the optimal input combination, we choose to perform smoothing, slope processing, and both slope processing and smoothing on the input data to compare the prediction results of the LSTM model. It is found that the prediction effect of the LSTM prediction model under slope processing is the best. The comparison between its prediction results and the actual situation is as follows: Figure 6 shown.

[0179] Example 3

[0180] like Figure 7 As shown, this embodiment provides an LSTM prediction system for the heat recovery performance of an enhanced geothermal system, including:

[0181] A three-dimensional geological model building module, which is used to build a three-dimensional geological model;

[0182] A data acquisition module, which is used to acquire data of the enhanced geothermal system;

[0183] A preprocessing module, which is used to preprocess the acquired enhanced geothermal system data;

[0184] Model building and training module, which is used to build LSTM prediction model and perform training prediction;

[0185] The evaluation and verification module is used to evaluate and verify the prediction model.

[0186] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0187] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. An LSTM prediction method for the heat recovery performance of an enhanced geothermal system, characterized in that: The steps include: S1: Use a dual well system to build a 3D geological model and obtain data for the enhanced geothermal system; S2: preprocessing the acquired data and using the preprocessed data as input data of the prediction model; S3: Calculate the variable slope and normalize the obtained slope; S4: Build an LSTM prediction model to predict the outlet water temperature of the geothermal system; S5: Train the LSTM model and make predictions, and use evaluation indicators to determine the accuracy of the prediction results; In S1, the process of establishing a three-dimensional geological model using a dual well system specifically includes: S11: Define the calculation area of ​​the 3D geological model and construct a 3D geological conceptual model by combining geological and engineering data; S12: Setting mathematical assumptions and required physical parameters of the 3D geological model; S13: Based on the 3D geological conceptual model and parameter settings, a 3D geological model is established using numerical simulation software; S14: Based on the three-dimensional geological model, the boundary conditions and initial state are initialized, and the mathematical equations of the matrix rock block seepage field, the fracture seepage field, the matrix rock block temperature field, and the fracture water temperature field are defined; In S11, the calculation domain of the three-dimensional geological model is The length of the well is The fractures include 80 randomly generated small fractures and 2 main fractures, and are located in the middle of the 3D geological model. in the region; In the S12, The mathematical assumptions of the 3D geological model are: (1) The rock matrix is ​​isotropic and uniform; (2) The flow is considered to be unidirectional; (3) Fluid flow obeys Darcy's law and capillary force is not considered; (4) The heat exchange between fluid and rock follows Fourier's law and is characterized by local thermal equilibrium; The physical parameters required for the 3D geological model include: Bottom hole production pressure , fracture permeability , reservoir density , reservoir thermal conductivity , reservoir permeability , reservoir porosity , Reservoir specific heat capacity , fracture porosity , working fluid injection temperature , Spacing between injection and production wells , working fluid injection flow , crack thickness , reservoir temperature ; In the S14, The mathematical equation of the seepage field of matrix rock is: ； In the formula, It represents the water storage coefficient of rock mass; Indicates time; represents the rock water flow rate, and its calculation expression is: , represents the rock permeability, is the dynamic viscosity of water, Indicates water pressure, represents the water density in the cracks, represents the acceleration due to gravity, Indicated in Elevation in direction; It represents the volume strain of the rock mass; represents the source and sink terms of seepage; The mathematical equation of the seepage field in the fracture is: ； In the formula, Indicates the crack width; represents the water storage coefficient of the fracture; represents the tangential derivative along the crack; represents the permeability of the fracture; represents the volume strain of the crack surface; It represents the flow exchange between rock block and fracture surface, and its calculation expression is: ; The mathematical equation of the temperature field of the matrix rock is: ； In the formula, It represents the specific heat capacity of rock; Indicates the density of rock mass; Indicates the rock temperature; Indicates the thermal conductivity of the rock mass; Indicates heat source; The mathematical equation of the fracture water temperature field is: ； In the formula, represents the effective specific heat capacity; Indicates the temperature of the water in the fracture; Indicates water density; represents the specific heat of water; Indicates the water flow rate in the fracture; represents the water heat conductivity; It represents the heat absorbed by the water on the fracture surface from the bedrock. Its calculation expression is: , represents the convective heat transfer coefficient; In S2, before using the pre-processed data as input data of the prediction model, the method further includes: S21: grouping the processed data according to the physical parameter type; S22: Based on the physical parameter type data, the random forest algorithm is used to rank their importance; S23: Select the top five physical type parameters and combine them, and use them as input data for the LSTM prediction model.

2. The LSTM prediction method for the heat recovery performance of an enhanced geothermal system according to claim 1 is characterized in that: In S2, the method of preprocessing the acquired enhanced geothermal system data includes removing outliers, normalizing values, and binning operations.

3. The LSTM prediction method for the heat recovery performance of an enhanced geothermal system according to claim 1 is characterized in that: In S4, the LSTM prediction model includes a forget gate, an input gate, and an output gate; Forget Gate: ； Input Gate: ； Cell status update: ； New cell state: ； Output Gate: ； In the formula, Represents the input of the input gate; represents the sigmoid activation function, represents the weight matrix of the input gate; express Actual value at the moment; Indicates the current Input of time; represents the bias term of the input gate; Indicates candidate The cell state at a given moment; represents the tanh activation function; The weight matrix representing the cell state update; Represents the bias term for cell state update; Indicates the current The cell state at a given moment; Represents the output of the forget gate; express The cell state at a given moment; Represents the weight matrix of the forget gate; Represents the bias term of the forget gate; represents the output of the output gate; Indicates the output gate The weight matrix corresponding to the actual value at the moment; Indicates the current The weight matrix corresponding to the input at the moment; Represents the bias term of the output gate.

4. The LSTM prediction method for the heat recovery performance of an enhanced geothermal system according to claim 1 is characterized in that: In S5, the evaluation indicators include: The MSE evaluation indicator is expressed as: ； In the formula, represents the number of samples; Indicates actual observations; Indicates predicted values; The MAPE evaluation indicator is expressed as: ； The R-squared evaluation indicator is expressed as: ； In the formula, is the coefficient of determination, which is used to measure the degree to which the model explains the variability of the data. The value range is ; Represents the average of the actual observations.

5. A system for the LSTM prediction method of the heat recovery performance of the enhanced geothermal system as claimed in claim 1, characterized in that: include: A three-dimensional geological model building module, which is used to build a three-dimensional geological model; A data acquisition module, which is used to acquire data of the enhanced geothermal system; A preprocessing module, which is used to preprocess the acquired enhanced geothermal system data; Model building and training module, which is used to build LSTM prediction model and perform training prediction; The evaluation and verification module is used to evaluate and verify the prediction model.

Images