CO2 conveying pipeline leakage prediction method and device

Abstract

The invention discloses a CO2 conveying pipeline leakage prediction method and device, relates to the technical field of CO2 conveying pipeline safety risk assessment, and aims to solve the problem that a reliable CO2 conveying pipeline leakage prediction method is lacked in related technologies. The method comprises the steps that pipeline parameters, leakage parameters and environment parameters are obtained; inputting the pipeline parameters, the leakage parameters and the environment parameters into a prediction model to obtain prediction data; the prediction data comprises at least one of the following data: a pipeline state, a leakage rate at a future moment, a leakage rate at the future moment and CO2 concentration distribution; the prediction model is a physical information neural network model; and according to the prediction data and the concentration threshold, calculating to obtain a dangerous area.

Description

Technical Field

[0001] The present application relates to the technical field of risk assessment and management of carbon dioxide (CO2) transmission pipelines, and in particular to a pipeline leakage prediction method and device. Background Art

[0002] In recent years, the emergence of Carbon Capture Utilization and Storage (CCUS) technology has alleviated global climate change and is considered the most effective way to reduce CO2 emissions. CO2 pipeline transportation is a key link in CCUS technology, connecting carbon sources and sinks. It is the most flexible and economical way to transport CO2 on an industrial scale and over long distances. However, once a CO2 pipeline leaks, the hazard type, damage distance, duration, and risk range differ significantly from those of natural gas pipelines. It can also easily induce pipe brittle fractures, exacerbating the hazard and complicating pipeline maintenance and emergency response. Therefore, a simulation method that can realistically simulate the entire process of a CO2 pipeline leak, from the leak outlet to diffusion and determining the hazard range, is urgently needed. Summary of the Invention

[0003] The purpose of this application is to provide a CO2 transmission pipeline leakage prediction method and device, aiming to solve the problem of the lack of a reliable CO2 transmission pipeline leakage prediction method in the related art.

[0004] To achieve the above objectives, this application adopts the following technical solutions:

[0005] This application provides a CO2 transmission pipeline leakage prediction method, comprising:

[0006] Obtain pipeline parameters, leakage parameters and environmental parameters;

[0007] Inputting pipeline parameters, leakage parameters, and environmental parameters into a prediction model to obtain prediction data; the prediction data includes at least one of the following: pipeline status, leakage rate at a future time, leakage volume at a future time, and CO2 concentration distribution; the prediction model is a physical information neural network model;

[0008] The danger zone is calculated based on the predicted data and concentration threshold.

[0009] In this application, the prediction model based on the physical information neural network model architecture can combine the neural network with the physical conservation equation, and can meet the requirements of physical laws while realizing data fitting in the neural network, thereby realizing accurate prediction of pipeline leakage.

[0010] In some embodiments, the prediction model is a dual neural network architecture, and the prediction model includes:

[0011] Concentration field prediction network: input is spatial coordinates (x, y, z) and time t, and output is CO2 concentration value C(x, y, z, t);

[0012] Pipeline state prediction network: input is time t, output is normalized pressure P(t), normalized temperature T(t) and liquid fraction

[0013] In some embodiments, the prediction model is trained by:

[0014] Obtaining neural network parameters; neural network parameters include at least one of the following: number of network layers, number of nodes in each network layer, activation function, and learning rate;

[0015] Initialize the computational domain;

[0016] Generate training data based on the computational domain;

[0017] Determine the loss function;

[0018] Build an initial prediction model based on the neural network parameters;

[0019] The initial prediction model is trained based on the training data and the loss function to obtain a prediction model.

[0020] In some embodiments, generating training data based on the computational domain includes:

[0021] Based on random sampling in the computational domain, physical constraint points are obtained to determine the internal points of the physical domain; the internal points of the physical domain satisfy X internal ∈[x min ,x max ]×[y min ,y max ]×[z min ,z max ]×[0,t max ], where X internal is a point inside the physical domain, x max ,y max ,z max and x min ,y min ,z min is the physical constraint point;

[0022] Generate computational domain boundary points and initial condition constraint points;

[0023] Determine the source constraint point in the leakage source area; the source constraint point satisfies (xx i ) 2 +(yy j ) 2 +(zz l )2 ≤r s 2 , where x, y, z are the coordinates of the source constraint point, x i ,y i ,z i is the leakage source coordinate, r s is the radius of the leakage source;

[0024] Internal points of the physical domain, boundary points of the computational domain, initial condition constraint points, and source term constraint points are determined as training data.

[0025] In some embodiments, determining the loss function includes:

[0026] Determine the physical constraint loss; the physical constraint loss satisfies Where u, v, w are velocity components, D is the diffusion coefficient, g is the acceleration due to gravity, is the Laplace operator, c is the concentration;

[0027] Determine the boundary condition loss; the boundary condition loss satisfies where ||c boundary || 2 is the L value of the concentration field c on the boundary 2 Norm squared, is the L of the partial derivative of c with respect to z on the z = 0 plane 2 norm squared;

[0028] Determine the initial condition loss; the initial condition loss satisfies L initial =||c(x,y,z,0)|| 2 , which is used to constrain the deviation between the predicted value of the concentration field c and the actual initial distribution at the initial moment of the model, so as to make the model meet the initial conditions of the physical process;

[0029] Determine the source term constraint loss; the source term constraint loss satisfies: Where Q is the leakage calculated dynamically, r s is the radius of the leakage source, and the scalar field c at the source term in the constraint model source , so that it approaches the theoretical analytical solution By L 2 The squared norm measures the deviation, allowing the model to fit the physical laws of the source term;

[0030] Based on the physical constraint loss, boundary condition loss, initial condition loss and source term constraint loss, a loss function is determined; the loss function satisfies L total =w p L physics +w b L boundary +wi L initial +w s L source ; Among them, w p 、w b 、w i 、w s is the weight.

[0031] In some embodiments, the initial prediction model is trained based on the training data and the loss function to obtain the prediction model, including:

[0032] Input the training data at different time points into the initial prediction model respectively;

[0033] For each time point, perform forward propagation to obtain the output of the initial prediction model and calculate the loss value based on the loss function and the output;

[0034] Based on the loss value, back propagation is performed to update the parameters of the initial prediction model to obtain the prediction model.

[0035] In some embodiments, the output of the initial prediction model or the leakage amount in the output of the prediction model satisfies:

[0036] The dynamic calculation method of leakage Q is:

[0037] When the CO2 in the pipeline is liquid:

[0038] When the CO2 in the pipeline is in gaseous critical flow:

[0039] Among them, C d is the flow coefficient, A is the leakage hole area, ρ l is the density of liquid CO2, ρ g is the density of gaseous CO2, γ is the specific heat ratio, P is the pipeline pressure, P a Environmental pressure;

[0040] The phase change process of CO2 satisfies:

[0041] Vapor pressure equation:

[0042] Phase transition energy balance: Q vap =m vap L;

[0043] Among them, P vap is the saturated vapor pressure of the liquid, A, B, C are Antoine constants, m vap is the mass of evaporated substance, L is the latent heat, and T is the thermodynamic temperature.

[0044] In some embodiments, the leakage amount is obtained based on at least one of the following formulas:

[0045] Conservation of mass:

[0046] Conservation of Energy:

[0047] Where m is the mass of CO2 in the pipeline, t is the time, Q is the leakage volume, h is the specific enthalpy, Q loss For heat loss.

[0048] In some embodiments, the CO2 concentration profile is determined based on the following formula:

[0049] Diffusion-convection equation:

[0050] Gravity Settling Correction:

[0051] Where c is the CO2 concentration, is the wind speed, D is the diffusion coefficient, ρ co2 is the CO2 density, ρ air is the air density, g is the acceleration due to gravity, μ is the dynamic viscosity of the air, d p is the particle diameter.

[0052] In some embodiments, based on the predicted data and the concentration threshold, the danger zone is calculated to satisfy the following formula: danger ={(x,y,z)|c(x,y,z,t)>c threshold}; where R danger Indicates the range of the danger zone, c threshold is the concentration threshold, and c(x,y,z,t) represents the CO2 concentration at the position (x,y,z) at time t.

[0053] In some embodiments, the method further comprises:

[0054] Based on the hazard zone, determine the impact range of the hazard zone and the exposure time of the hazard zone;

[0055] The impact range of the hazardous area meets R max =max{r|c(r,t)>c min};

[0056] Exposure time in hazardous areas meets

[0057] Among them, R max is the maximum influence radius of the concentration field distribution, c min is the minimum concentration threshold, T exposure(x,y,z) is the exposure time at point (x,y,z), H is the Heaviside step function, (c(x,y,z,τ) is the concentration field value at point (x,y,z) at time τ, c threshold is the exposure threshold.

[0058] In some embodiments, the present application provides a CO2 transmission pipeline leakage prediction device, comprising: a processor and a memory configured to store processor-executable instructions; wherein the processor is configured to execute the instructions to implement any of the above optional methods.

[0059] In some embodiments, the present application provides a computer-readable storage medium having instructions stored thereon. When the instructions in the computer-readable storage medium are executed by a CO2 transmission pipeline leakage prediction device, the CO2 transmission pipeline leakage prediction device is enabled to execute any of the above-mentioned optional methods.

[0060] In some embodiments, the present application provides a computer program product, which includes computer program instructions, and when the computer program instructions are executed by a processor, implements any of the above optional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0062] Figure 1 A flow chart of a pipeline leakage prediction method provided in this application;

[0063] Figure 2 A flow chart of another pipeline leakage prediction method provided in this application;

[0064] Figure 3 This is a schematic diagram of the structure of a pipeline leakage prediction device provided in this application. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0066] In the description of this application, it should be understood that the terms "upper," "lower," "left," "right," "front," "back," "inner," "outer," and the like, indicating directions or positional relationships, are based on the directions or relative positional relationships shown in the accompanying drawings and are intended solely to facilitate the description of this application and simplify the description. They do not indicate or imply that the devices or components referred to must have a specific direction, be constructed, or operate in a specific direction. Therefore, they should not be construed as limitations on this application. Unless otherwise specified, the above-mentioned directionality descriptions may be flexibly set in actual application, provided that the relative positional relationships shown in the accompanying drawings are met.

[0067] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include one or more of such features. Throughout this application, unless otherwise specified, "plurality" means two or more.

[0068] In the description of this application, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," "connected," and "connected" should be understood broadly. For example, they may refer to fixed connections, detachable connections, or integral connections. They may be directly connected, indirectly connected through an intermediary, or internally connected between two components. Those skilled in the art will understand the specific meanings of these terms in this application based on the specific circumstances.

[0069] In the embodiments of the present application, the terms "comprises," "comprising," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, article, or device comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of other identical elements in the process, article, or device comprising the element.

[0070] In the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0071] In the description of this specification, specific features, structures, materials or characteristics may be combined in an appropriate manner in any one or more embodiments or examples.

[0072] In recent years, the emergence of carbon capture, storage, and utilization (CCUS) technologies has alleviated global climate change concerns and is considered the most effective way to reduce CO2 emissions. CO2 pipeline transportation is a key link in connecting carbon sources and sinks within CCUS technology, making it the most flexible and economical way to transport CO2 on an industrial scale and over long distances. However, once a CO2 pipeline leaks, the hazard type, damage distance, duration, and risk range differ significantly from those of a natural gas pipeline. It can also easily induce pipe brittle fractures, exacerbating the hazards of an accident and complicating pipeline maintenance and emergency response. Therefore, a simulation method that can realistically simulate the entire process of a CO2 pipeline leak, from leak exit and diffusion to determining the scope of the hazard, is urgently needed.

[0073] To solve the above problems, the present application provides a CO2 transmission pipeline leakage prediction method.

[0074] like Figure 1 As shown, this application provides a CO2 transmission pipeline leakage prediction method, including: S101-S103:

[0075] S101. Obtain pipeline parameters, leakage parameters, and environmental parameters.

[0076] Exemplarily, the pipeline parameters include at least one of the following: pipeline diameter, length, initial pressure, initial temperature, flow rate, etc.

[0077] Exemplarily, the leakage parameter includes at least one of the following: leakage location, leakage aperture, emission coefficient, etc.

[0078] Exemplarily, the environmental parameters include at least one of the following: ambient temperature, pressure, wind speed, wind direction, terrain roughness, etc.

[0079] S102: Input pipeline parameters, leakage parameters and environmental parameters into the prediction model to obtain prediction data.

[0080] The prediction data includes at least one of the following: pipeline status, leakage rate at a future time, leakage volume at a future time, and CO2 concentration distribution.

[0081] In some embodiments, the prediction model is a physical information neural network model.

[0082] S103. Calculate the danger zone based on the predicted data and the concentration threshold.

[0083] In one possible implementation, calculating the danger zone based on the predicted data and the concentration threshold can be called a safety assessment and decision support process. Optionally, the safety assessment and decision support process can include the following steps:

[0084] Step 1: Calculate the range and duration of CO2 concentration exceeding the standard based on the simulation results.

[0085] Step 2: Assess the risk impact range under different meteorological conditions and leakage parameters.

[0086] Step 3: Provide emergency response advice and safe distance recommendations.

[0087] Step 4: Provide technical optimization support for pipeline design and operation.

[0088] In some embodiments, the prediction model is a dual neural network architecture, and the prediction model includes:

[0089] Concentration field prediction network: input is spatial coordinates (x, y, z) and time t, and output is CO2 concentration value C(x, y, z, t);

[0090] Pipeline state prediction network: input is time t, output is normalized pressure P(t), normalized temperature T(t) and liquid fraction

[0091] In one possible implementation, the prediction data obtained based on the dual neural network architecture prediction model includes:

[0092] Step 1: Use the trained network to predict the pipeline state (pressure, temperature, phase fraction) at different time points;

[0093] Step 2: Calculate the leakage rate and cumulative leakage at each time point;

[0094] Step 3: Predict the CO2 concentration distribution at different locations and times in three-dimensional space;

[0095] Step 4. Generate visualization results. The visualization results may include one or more of pipeline pressure and temperature time curves, three-dimensional concentration field distribution, and two-dimensional slices.

[0096] In some embodiments, the prediction model is trained by:

[0097] Obtaining neural network parameters; neural network parameters include at least one of the following: number of network layers, number of nodes in each network layer, activation function, and learning rate;

[0098] Initialize the computational domain;

[0099] Generate training data based on the computational domain;

[0100] Determine the loss function;

[0101] Build an initial prediction model based on the neural network parameters;

[0102] The initial prediction model is trained based on the training data and the loss function to obtain a prediction model.

[0103] It should be understood that the initialization calculation domain is used to determine the size of the three-dimensional space calculation domain and perform grid division.

[0104] In one possible implementation, building an initial prediction model based on neural network parameters includes:

[0105] Constructing a concentration field prediction network:

[0106] Input layer: spatial coordinates (x, y, z) and time t, a total of 4 input nodes; (x, y, z, t) ∈ R 4 ;

[0107] Hidden layer: multi-layer fully connected network, using tanh activation function;

[0108] Output layer: CO2 concentration value C(x,y,z,t), 1 output node; C(x,y,z,t)∈R 1 .

[0109] Build the pipeline state prediction network:

[0110] Input layer: time t, 1 input node; t∈R 1 ;

[0111] Hidden layer: multi-layer fully connected network, using tanh activation function;

[0112] Output layer: normalized pressure P(t), normalized temperature T(t), liquid fraction φ(t), a total of 3 output nodes;

[0113] In one possible implementation, state normalization satisfies:

[0114]

[0115] Among them, P is the actual pressure, T is the actual temperature, P min 、P max Minimum and maximum pressure, T min 、T max Minimum and maximum temperature values.

[0116] In some embodiments, generating training data based on the computational domain includes:

[0117] Based on random sampling in the computational domain, physical constraint points are obtained to determine the internal points of the physical domain; the internal points of the physical domain satisfy X internal ∈[x min ,x max ]×[y min ,y max ]×[z min ,z max ]×[0,t max], where X internal is a point inside the physical domain, x max ,y max ,z max and x min ,y min ,z min is the physical constraint point;

[0118] Generate computational domain boundary points and initial condition constraint points;

[0119] Determine the source constraint point in the leakage source area; the source constraint point satisfies (xx i ) 2 +(yy j ) 2 +(zz l ) 2 ≤r s 2 , where x, y, z are the coordinates of the source constraint point, x i ,y i ,z i is the leakage source coordinate, r s is the radius of the leakage source;

[0120] Internal points of the physical domain, boundary points of the computational domain, initial condition constraint points, and source term constraint points are determined as training data.

[0121] In one possible implementation, the physical constraint loss can be solved by calculating the partial derivatives of the concentration field with respect to time and space through automatic differentiation, substituting them into the diffusion-convection-sedimentation equation, and calculating the sum of squares of the residuals of the equation.

[0122] In one possible implementation, the automatic differentiation calculation includes:

[0123] Step 1: Set the input coordinate requires_grad = True to activate automatic differentiation tracking;

[0124] Step 2: Calculate the first-order derivative;

[0125] Step 3: Calculation of second-order derivatives (Method 1: nested automatic differentiation; Method 2: central difference approximation).

[0126] Boundary condition loss is used to ensure that the far-field boundary concentration approaches zero and the ground boundary meets the reflection condition.

[0127] Initial condition loss is used to ensure that the concentration in the entire computational domain is zero at the initial moment.

[0128] The source term constraint loss is used to ensure that the concentration in the leakage source area is consistent with the leakage amount calculated from the pipeline state (pressure, temperature, phase fraction).

[0129] In one possible implementation, the composite loss function = physical weight × physical constraint loss + boundary weight × boundary condition loss + initial condition weight × initial condition constraint loss + source term weight × source term constraint loss.

[0130] In a possible implementation, computational domain boundary points may be generated and used as boundary condition constraint points.

[0131] In some embodiments, determining the loss function includes:

[0132] Determine the physical constraint loss; the physical constraint loss satisfies Where u is the wind speed vector, D is the diffusion coefficient, w is the gravity settling velocity, g is the acceleration of gravity, is the Laplace operator, c is the concentration; in this case, in one possible implementation, the second-order derivative calculation of the central difference approximation satisfies

[0133] Determine the boundary condition loss; the boundary condition loss satisfies where ||c boundary || 2 is the L value of the concentration field c on the boundary 2 Norm squared, is the L of the partial derivative of c with respect to z on the z = 0 plane 2 norm squared;

[0134] Determine the initial condition loss; the initial condition loss satisfies L initial =||c(x,y,z,0)|| 2 ; used to constrain the deviation of the predicted value of the concentration field c from the actual initial distribution at the initial time (exemplarily, t=0) of the model, so that the model satisfies the initial conditions of the physical process;

[0135] Determine the source term constraint loss; the source term constraint loss satisfies: Where Q is the leakage calculated dynamically, r s is the radius of the leakage source, and the scalar field c at the source term in the constraint model source , so that it approaches the theoretical analytical solution ), through L 2 The squared norm measures the deviation, allowing the model to fit the physical laws of the source term;

[0136] Based on the physical constraint loss, boundary condition loss, initial condition loss and source term constraint loss, a loss function is determined; the loss function satisfies L total =w p L physics +w b L boundary +wi L initial +w s L source ; Among them, w p 、w b 、w i 、w s is the weight.

[0137] In some embodiments, the initial prediction model is trained based on the training data and the loss function to obtain the prediction model, including:

[0138] Input the training data at different time points into the initial prediction model respectively;

[0139] For each time point, perform forward propagation to obtain the output of the initial prediction model and calculate the loss value based on the loss function and the output;

[0140] Based on the loss value, back propagation is performed to update the parameters of the initial prediction model to obtain the prediction model.

[0141] In some embodiments, the initial prediction model is trained based on the training data and the loss function to obtain the prediction model, including:

[0142] Step 1: Load training data in batches;

[0143] Step 2: For each time point, (1) perform forward propagation to calculate the network output, (2) calculate various loss functions, and (3) perform backpropagation to update the network parameters;

[0144] Optionally, step 3: applying a learning rate scheduling strategy to optimize the training process;

[0145] Optionally, step 4: implement early stopping mechanism to avoid overfitting;

[0146] Optionally, step 5: regularly record training history and intermediate results.

[0147] In some embodiments, the output of the initial prediction model or the leakage amount in the output of the prediction model satisfies:

[0148] The dynamic calculation method of leakage Q is:

[0149] When the CO2 in the pipeline is liquid:

[0150] When the CO2 in the pipeline is in gaseous critical flow:

[0151] Among them, C d is the flow coefficient, A is the leakage hole area, ρ l is the density of liquid CO2, ρ gis the density of gaseous CO2, γ is the specific heat ratio, P is the pipeline pressure, P a Environmental pressure;

[0152] The phase change process of CO2 satisfies:

[0153] Vapor pressure equation:

[0154] Phase transition energy balance: Q vap =m vap L;

[0155] Among them, P vap is the saturated vapor pressure of the liquid, A, B, C are Antoine constants, m vap is the mass of evaporated substance, L is the latent heat, and T is the thermodynamic temperature.

[0156] In some embodiments, the leakage amount is obtained based on at least one of the following formulas:

[0157] Conservation of mass:

[0158] Conservation of Energy:

[0159] Where m is the mass of CO2 in the pipeline, t is the time, Q is the leakage volume, h is the specific enthalpy, Q loss For heat loss.

[0160] In some embodiments, the CO2 concentration profile is determined based on the following formula:

[0161] Diffusion-convection equation:

[0162] Gravity Settling Correction:

[0163] Where c is the CO2 concentration, is the wind speed, D is the diffusion coefficient, ρ co2 is the CO2 density, ρ air is the air density, g is the acceleration due to gravity, μ is the dynamic viscosity of the air, d p is the particle diameter.

[0164] In some embodiments, based on the predicted data and the concentration threshold, the danger zone is calculated to satisfy the following formula: danger ={(x,y,z)|c(x,y,z,t)>c threshold}; where R danger Indicates the range of the danger zone, c threshold is the concentration threshold, and c(x,y,z,t) represents the CO2 concentration at the position (x,y,z) at time t.

[0165] In some embodiments, the method further comprises:

[0166] Based on the hazard zone, determine the impact range of the hazard zone and the exposure time of the hazard zone;

[0167] The maximum diffusion distance in the influence range of the dangerous area meets R max =max{r|c(r,t)>c min};

[0168] Exposure time in hazardous areas meets

[0169] Among them, R max is the maximum influence radius of the concentration field distribution, c min is the minimum concentration threshold, T exposure (x,y,z) is the exposure time at point (x,y,z), H is the Heaviside step function, (c(x,y,z,τ) is the concentration field value at point (x,y,z) at time τ, c threshold is the exposure threshold.

[0170] In some embodiments, the cumulative leakage mass of the hazardous area satisfies:

[0171] The following is an exemplary description provided in the examples of this application, which is used to briefly describe the method as a whole:

[0172] For example, Figure 2 As shown, the CO2 transmission pipeline leakage prediction method includes the following steps:

[0173] Step 1: System initialization and parameter configuration;

[0174] Step 2: Build a dual neural network architecture;

[0175] Step 3: Generate training data;

[0176] Step 4: Construct a composite loss function;

[0177] Step 5: Automatic differentiation realizes high-order derivative calculation;

[0178] Step 6: Model training process;

[0179] Step 7: Generate simulation results;

[0180] Step 8: Safety assessment and decision support.

[0181] In one possible implementation, step 1 includes the following steps:

[0182] Step 1.1: Configure pipeline parameters: pipeline diameter, length, initial pressure, initial temperature, flow rate, etc.

[0183] Step 1.2: Configure leakage parameters: leakage location, leakage aperture, emission coefficient, etc.

[0184] Step 1.3: Configure environmental parameters: ambient temperature, pressure, wind speed, wind direction, terrain roughness, etc.

[0185] Step 1.4: Configure the neural network parameters: number of network layers, number of nodes per layer, activation function, learning rate, etc.

[0186] Step 1.5: Initialize the computational domain: Determine the size of the three-dimensional computational domain and the grid division.

[0187] Step 2 includes the following steps:

[0188] Step 2.1: Construct a concentration field prediction network:

[0189] Input layer: spatial coordinates (x, y, z) and time t, a total of 4 input nodes;

[0190] Hidden layer: multi-layer fully connected network, using tanh activation function;

[0191] Output layer: CO2 concentration value C(x,y,z,t), 1 output node.

[0192] Step 2.2: Build pipeline state prediction network:

[0193] Input layer: time t, 1 input node;

[0194] Hidden layer: multi-layer fully connected network, using tanh activation function;

[0195] Output layer: normalized pressure P(t), normalized temperature T(t), liquid fraction φ(t), a total of 3 output nodes.

[0196] Step 3 includes the following steps:

[0197] Step 3.1: Randomly sample points in the computational domain as physical constraint points;

[0198] Step 3.2: Generate computational domain boundary points as boundary condition constraint points;

[0199] Step 3.3: Generate the point at the initial time (t=0) as the initial condition constraint point;

[0200] Step 3.4: Generate points in the leakage source area as source constraint points.

[0201] Step 4 includes the following steps:

[0202] Step 4.1: Physical constraint loss, calculate the partial derivatives of the concentration field with respect to time and space through automatic differentiation, substitute them into the diffusion-convection-sedimentation equation, and calculate the residual sum of squares of the equation;

[0203] Step 4.2: Boundary condition loss, ensure that the far-field boundary concentration approaches zero and the ground boundary meets the reflection condition;

[0204] Step 4.3: Initial condition loss, ensuring that the concentration in the entire computational domain is zero at the initial moment;

[0205] Step 4.4: Constrain the source term loss to ensure that the concentration in the leakage source area is consistent with the leakage rate calculated from the pipeline state (pressure, temperature, phase fraction);

[0206] Step 4.5: Composite loss function = physical weight × physical constraint loss + boundary weight × boundary condition loss + initial condition weight × initial condition constraint loss + source term weight × source term constraint loss.

[0207] Step 5 includes the following steps:

[0208] Step 5.1: Set the input coordinate requires_grad = True to activate automatic differentiation tracking;

[0209] Step 5.2: Calculate the first-order derivative;

[0210] Step 5.3: Calculation of second-order derivatives (Method 1: Nested automatic differentiation);

[0211] Step 5.4: Calculation of second-order derivatives (Method 2: Central difference approximation, practical application).

[0212] Step 6 includes the following steps:

[0213] Step 6.1: Load training data in batches;

[0214] Step 6.2: For each time point, (1) perform forward propagation to calculate the network output, (2) calculate various loss functions, and (3) perform backpropagation to update the network parameters;

[0215] Step 6.3: Apply the learning rate scheduling strategy to optimize the training process;

[0216] Step 6.4: Implement early stopping mechanism to avoid overfitting;

[0217] Step 6.5: Regularly record training history and intermediate results.

[0218] Step 7 includes the following steps:

[0219] Step 7.1: Use the trained network to predict the pipeline state (pressure, temperature, phase fraction) at different time points;

[0220] Step 7.2: Calculate the leakage rate and cumulative leakage at each time point;

[0221] Step 7.3: Predict the CO2 concentration distribution at different locations and times in three-dimensional space;

[0222] Step 7.4. Generate visualization results, including pipeline pressure and temperature time curves, three-dimensional concentration field distribution, two-dimensional slices, etc.

[0223] Step 8 includes the following steps:

[0224] Step 8.1: Calculate the range and duration of CO2 concentration exceeding the standard based on the simulation results.

[0225] Step 8.2: Assess the scope of risk impact under different meteorological conditions and leakage parameters.

[0226] Step 8.3: Provide emergency response advice and safe distance recommendations.

[0227] Step 8.4: Provide technical optimization support for pipeline design and operation.

[0228] In the embodiments of the present application, the CO2 pipeline leakage prediction device, etc., can be divided into functional modules based on the above-mentioned method examples. For example, each functional module can be divided according to each function, or two or more functions can be integrated into a single processing module. The above-mentioned integrated modules are implemented in the form of software functional modules. It should be noted that the module division in the embodiments of the present application is schematic and is only a logical functional division. In actual implementation, other division methods may be used.

[0229] In the case of dividing each functional module into corresponding functional modules, Figure 3 A possible structural diagram of the CO2 transmission pipeline leakage prediction device involved in the above embodiment is shown. Figure 3 As shown, the CO2 transmission pipeline leakage prediction device may include: an acquisition module 301 and a processing module 302.

[0230] An acquisition module 301 is used to acquire pipeline parameters, leakage parameters and environmental parameters;

[0231] Processing module 302 is used to input pipeline parameters, leakage parameters and environmental parameters into the prediction model to obtain prediction data; the prediction data includes at least one of the following: pipeline status, leakage rate at a future time, leakage volume at a future time, CO2 concentration distribution; the prediction model is a physical information neural network model

[0232] The processing module 302 is further configured to calculate the danger zone based on the predicted data and the concentration threshold.

[0233] The above are only specific embodiments of the present application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A CO2 pipeline leakage prediction method, characterized in that: The method comprises: Obtain pipeline parameters, leakage parameters and environmental parameters; Inputting the pipeline parameters, the leakage parameters, and the environmental parameters into a prediction model to obtain prediction data; the prediction data includes at least one of the following: pipeline status, leakage rate at a future time, leakage volume at a future time, and CO2 concentration distribution; the prediction model is a physical information neural network model; The danger zone is calculated based on the predicted data and the concentration threshold.

2. The method according to claim 1, characterized in that The prediction model is a dual neural network architecture, and the prediction model includes: Concentration field prediction network: input is spatial coordinates (x, y, z) and time t, and output is CO2 concentration value C(x, y, z, t); Pipeline state prediction network: input is time t, output is normalized pressure P(t), normalized temperature T(t) and liquid fraction 3. The method according to claim 2, characterized in that The prediction model is trained in the following way: Obtaining neural network parameters; the neural network parameters include at least one of the following: the number of network layers, the number of nodes in each network layer, the activation function, and the learning rate; Initialize the computational domain; generating training data based on the computational domain; Determine the loss function; constructing an initial prediction model based on the neural network parameters; The initial prediction model is trained based on the training data and the loss function to obtain the prediction model.

4. The method according to claim 3, characterized in that Generating training data based on the computing domain includes: Based on random sampling in the computational domain, physical constraint points are obtained to determine internal points of the physical domain; the internal points of the physical domain satisfy X internal ∈[x min ,x max ]×[y min ,y max ]×[z min ,z max ]×[0,t max ], where X internal is a point inside the physical domain, x max ,y max ,z max and x min ,y min ,z min is the physical constraint point; Generate computational domain boundary points and initial condition constraint points; Determine the source term constraint point in the leakage source area; the source term constraint point satisfies (xx i ) 2 +(yy j ) 2 +(zz l ) 2 ≤r s 2 , where x, y, z are the coordinates of the source constraint point, x i ,y i ,z i is the leakage source coordinate, r s is the radius of the leakage source; The internal points of the physical domain, the boundary points of the computational domain, the initial condition constraint points, and the source term constraint points are determined as the training data.

5. The method according to claim 3, characterized in that Determining the loss function includes: Determine a physical constraint loss; the physical constraint loss satisfies Where u, v, w are velocity components, D is the diffusion coefficient, g is the acceleration due to gravity, is the Laplace operator, c is the concentration; Determine the boundary condition loss; the boundary condition loss satisfies where ||c boundary || 2 is the L value of the concentration field c on the boundary 2 Norm squared, is the L of the partial derivative of c with respect to z on the z = 0 plane 2 norm squared; Determine the initial condition loss; the initial condition loss satisfies L initial =||c(x,y,z,0)|| 2 , which is used to constrain the deviation between the predicted value of the concentration field c and the actual initial distribution at the initial moment of the model, so as to make the model meet the initial conditions of the physical process; Determine a source term constraint loss; the source term constraint loss satisfies: Where Q is the leakage calculated dynamically, r s s is the radius of the leakage source, and the scalar field c at the source term in the constraint model is source , so that it approaches the theoretical analytical solution By L 2 The squared norm measures the deviation, allowing the model to fit the physical laws of the source term; The loss function is determined based on the physical constraint loss, the boundary condition loss, the initial condition loss and the source term constraint loss; the loss function satisfies L total =w p L physics +w b L boundary +w i L initial +w s L source ; Among them, w p 、w b 、w i 、w s is the weight.

6. The method according to claim 3, characterized in that The initial prediction model is trained based on the training data and the loss function to obtain the prediction model, including: inputting the training data at different time points into the initial prediction model respectively; For each of the time points, performing forward propagation to obtain an output of the initial prediction model and calculating a loss value based on the loss function and the output; Based on the loss value, back propagation is performed to update the parameters of the initial prediction model to obtain the prediction model.

7. The method according to claim 6, characterized in that The output of the initial prediction model or the leakage amount in the output of the prediction model satisfies: The dynamic calculation method of leakage Q is: When the CO2 in the pipeline is liquid: When the CO2 in the pipeline is in gaseous critical flow: Among them, C d is the flow coefficient, A is the leakage hole area, ρ l is the density of liquid CO2, ρ g is the density of gaseous CO2, γ is the specific heat ratio, P is the pipeline pressure, P a Environmental pressure; The phase change process of CO2 satisfies: Vapor pressure equation: Phase transition energy balance: Q vap =m vap L; Among them, P vap is the saturated vapor pressure of the liquid, A, B, V are Antoine constants, m vap is the mass of evaporated substance, L is the latent heat, and T is the thermodynamic temperature.

8. The method according to any one of claims 1 to 7, characterized in that The leakage rate is obtained based on at least one of the following formulas: Conservation of mass: Conservation of Energy: Where m is the mass of CO2 in the pipeline, t is the time, Q is the leakage volume, h is the specific enthalpy, Q loss For heat loss.

9. The method according to any one of claims 1 to 7, characterized in that The CO2 concentration distribution is determined based on the following formula: Diffusion-convection equation: Gravity Settling Correction: Where c is the CO2 concentration, is the wind speed, D is the diffusion coefficient, ρ co2 is the CO2 density, ρ air is the air density, g is the acceleration due to gravity, μ is the dynamic viscosity of the air, d p is the particle diameter.

10. The method according to any one of claims 1 to 7, characterized in that The dangerous area calculated based on the predicted data and concentration threshold satisfies the following formula: R danger ={(x,y,z)|c(x,y,z,t)>c threshold }; where R danger Indicates the scope of the dangerous area, c threshold is the concentration threshold, and c(x,y,z,t) represents the CO2 concentration at the position (x,y,z) at time t.

11. The method according to claim 10, characterized in that The method further comprises: Based on the dangerous area, determining the impact range of the dangerous area and the exposure time of the dangerous area; The influence range of the dangerous area meets R max =max{r|c(r,t)>c min }; The exposure time of the hazardous area meets Among them, R max is the maximum influence radius of the concentration field distribution, c min is the minimum concentration threshold, T exposure (x,y,z) is the exposure time at point (x,y,z), H is the Heaviside step function, (c(x,y,z,τ) is the concentration field value at point (x,y,z) at time τ, c threshold is the exposure threshold.

12. A CO2 pipeline leakage prediction device, characterized in that: The CO2 transmission pipeline leakage prediction device comprises: processor; a memory configured to store instructions executable by the processor; The processor is configured to execute the instructions to implement the method according to any one of claims 1 to 11.

Images