Method and device for predicting removal factor of aerosol in containment vessel of nuclear power plant

Abstract

The invention relates to a method and a device for predicting an aerosol removal factor in a containment of a nuclear power plant. The method comprises the following steps: acquiring containment characteristic parameter information of a to-be-predicted containment, and inputting the containment characteristic parameter information into a pre-constructed aerosol removal factor prediction model to predict an aerosol removal factor in the to-be-predicted containment within a prediction time period, the containment characteristic parameter information represents multi-feature coupling information, and the aerosol removal factor prediction model comprises a multilayer physical constraint type bidirectional long-short term memory neural network; the physical constraint type representation introduces the historical spraying efficiency in the to-be-predicted containment into the bidirectional long-short term memory neural network. According to the method, the prior spraying efficiency in the containment vessel to be predicted is introduced into the bidirectional long-short term memory neural network as a physical constraint condition, so that the prediction precision of the aerosol removal factor prediction model can be improved.

Description

Technical Field

[0001] This application relates to the field of nuclear power technology, and in particular to a method and apparatus for predicting aerosol removal factors inside the containment of a nuclear power plant. Background Technology

[0002] With the development of nuclear safety technology, the containment vessel of a nuclear power plant has gradually become the last line of defense against the leakage of radioactive materials. Radioactive materials within the containment vessel are typically removed through various mechanisms, such as deposition, coagulation, and filtration. Furthermore, the radioactive material removal factor directly affects the prediction of environmental release levels and the assessment of accident consequences. Therefore, predicting the radioactive material removal factor has become particularly important.

[0003] Taking aerosols as an example, the relevant technologies mainly use empirical formulas and mechanism models based on thermal-hydraulic processes to predict aerosol removal factors.

[0004] However, related technologies suffer from inaccurate prediction results when forecasting aerosol removal factors. Summary of the Invention

[0005] Therefore, it is necessary to provide a method and apparatus for predicting aerosol removal factors inside the containment of nuclear power plants to address the aforementioned technical problems.

[0006] In a first aspect, embodiments of this application provide a method for predicting aerosol removal factors within the containment of a nuclear power plant, the method comprising:

[0007] Obtain containment characteristic parameter information of the containment to be predicted; the containment characteristic parameter information represents the coupling information corresponding to the aerosol characteristics, system characteristics, model characteristics and containment structural characteristics within the containment to be predicted;

[0008] The containment characteristic parameter information is input into the pre-built aerosol removal factor prediction model to predict the aerosol removal factor in the containment to be predicted within the prediction time period, and the predicted value of the aerosol removal factor in the containment to be predicted is obtained.

[0009] Among them, the aerosol removal factor prediction model includes a multi-layer physical constraint bidirectional long short-term memory neural network. The physical constraint means that the historical spray efficiency inside the containment to be predicted is introduced into the bidirectional long short-term memory neural network as a physical constraint condition of the bidirectional long short-term memory neural network.

[0010] In one embodiment, the process of constructing the aerosol removal factor prediction model includes:

[0011] Obtain the sample datasets corresponding to different sample safe shells;

[0012] According to the preset partitioning rules, the sample dataset corresponding to each sample safe shell is partitioned to obtain the training set, test set and validation set;

[0013] The initial aerosol removal factor prediction model is trained using the training set, test set, validation set, and loss function to generate the aerosol removal factor prediction model; the loss function satisfies the dual constraint mechanism of time sensitivity and physical process rationality.

[0014] In one embodiment, obtaining the sample dataset corresponding to different sample safe houses includes:

[0015] Based on the physical field models of different sample containment structures, the sample characteristic parameter information of each sample containment structure is obtained;

[0016] The thermal-hydraulic algorithm is used to obtain the sample prediction results of the aerosol removal factor inside the containment of each sample based on the sample characteristic parameter information of each sample containment.

[0017] Based on the sample characteristic parameters of each containment and the sample prediction results of the aerosol removal factor inside each containment, the sample dataset corresponding to each containment is obtained.

[0018] In one embodiment, obtaining the sample dataset corresponding to different sample safe houses further includes:

[0019] Based on the mass conservation equation, momentum equation, and aerosol transport and diffusion equation of each sample containment, a physical field model of each sample containment is constructed.

[0020] In one embodiment, based on the sample characteristic parameter information of each sample containment and the sample prediction results of the aerosol removal factor inside each sample containment, a sample dataset corresponding to each sample containment is obtained, including:

[0021] Outlier removal was performed on the sample prediction results of the aerosol removal factor inside the containment for each sample to obtain the reference prediction results of the aerosol removal factor inside the containment for each sample.

[0022] The sample datasets corresponding to each sample containment are obtained by combining the sample characteristic parameters of each containment and the reference prediction results of the aerosol removal factor inside each containment.

[0023] In one embodiment, the partitioning rules include the system extreme operating conditions within each sample containment and the geometric conditions of each sample containment; according to the preset partitioning rules, the sample dataset corresponding to each sample containment is partitioned to obtain a training set, a test set, and a validation set, including:

[0024] The sample datasets that satisfy the system's extreme operating conditions are designated as the test set; and the sample datasets that satisfy the geometric structure conditions are designated as the validation set.

[0025] The sample datasets excluding the test and validation sets are designated as the training set.

[0026] In one embodiment, the method further includes:

[0027] Based on the predicted value of the aerosol removal factor within the containment to be predicted during the prediction period, a safety warning message is triggered within the containment to be predicted.

[0028] In one embodiment, based on the predicted value of the aerosol removal factor within the containment to be predicted during the prediction time period, a safety warning message for the containment to be predicted is triggered, including:

[0029] If all predicted aerosol removal factors within a first duration are less than a first preset threshold, then an early warning message will be output.

[0030] If all predicted aerosol removal factors within a second duration are less than a second preset threshold, an early warning command is triggered; the second preset threshold is less than the first preset threshold.

[0031] Secondly, embodiments of this application provide an aerosol removal factor prediction device for the containment of a nuclear power plant, the device comprising:

[0032] The acquisition module is used to acquire containment characteristic parameter information of the containment to be predicted; the containment characteristic parameter information represents the coupling information corresponding to the aerosol characteristics, system characteristics, model characteristics and containment structural characteristics within the containment to be predicted.

[0033] The prediction module is used to input the containment characteristic parameter information into the pre-built aerosol removal factor prediction model, predict the aerosol removal factor in the containment to be predicted within the prediction time period, and obtain the predicted value of the aerosol removal factor in the containment to be predicted.

[0034] Among them, the aerosol removal factor prediction model includes a multi-layer physical constraint bidirectional long short-term memory neural network. The physical constraint means that the historical spray efficiency inside the containment to be predicted is introduced into the bidirectional long short-term memory neural network as a physical constraint condition of the bidirectional long short-term memory neural network.

[0035] Thirdly, embodiments of this application also provide a computer device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in the first aspect.

[0036] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0037] Fifthly, embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.

[0038] The method and apparatus for predicting aerosol removal factors within the containment of a nuclear power plant provided in this application embodiment include: acquiring containment characteristic parameter information of the containment to be predicted, inputting the containment characteristic parameter information into a pre-constructed aerosol removal factor prediction model, predicting the aerosol removal factor within the containment to be predicted within a prediction time period, and obtaining the predicted value of the aerosol removal factor within the containment to be predicted. The containment characteristic parameter information represents the coupling information corresponding to aerosol characteristics, system characteristics, model characteristics, and containment structural characteristics within the containment to be predicted. The aerosol removal factor prediction model includes a multi-layer physically constrained bidirectional long short-term memory neural network. The physically constrained nature indicates that the historical spraying efficiency within the containment to be predicted is introduced as a bidirectional long short-term memory parameter in the bidirectional long short-term memory neural network. The physical constraints of the short-term memory neural network: The aerosol removal factor prediction model in the above method consists of a multi-layered physically constrained bidirectional long short-term memory neural network. Specifically, the prior spray efficiency within the containment to be predicted is introduced as a physical constraint in the bidirectional long short-term memory neural network, thereby improving the prediction accuracy of the aerosol removal factor prediction model. Furthermore, the above method can utilize containment characteristic parameter information of the containment within the nuclear power plant with higher quality and integrity, i.e., multi-parameter coupling information, to predict the aerosol removal factor, which can greatly improve the accuracy of the predicted aerosol removal factor. In addition, the above method does not require human intervention, which not only reduces the prediction error of the aerosol removal factor but also accelerates the prediction speed. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating a method for predicting aerosol removal factors within the containment of a nuclear power plant in one embodiment.

[0040] Figure 2 This is a flowchart illustrating a method for predicting aerosol removal factors within the containment of a nuclear power plant, as described in another embodiment.

[0041] Figure 3 This is a flowchart illustrating a method for predicting aerosol removal factors within the containment of a nuclear power plant, as described in another embodiment.

[0042] Figure 4This is a flowchart illustrating a method for predicting aerosol removal factors within the containment of a nuclear power plant, as described in another embodiment.

[0043] Figure 5 This is a flowchart illustrating a method for predicting aerosol removal factors within the containment of a nuclear power plant, as described in another embodiment.

[0044] Figure 6 This is a flowchart illustrating a method for predicting aerosol removal factors within the containment of a nuclear power plant, as described in another embodiment.

[0045] Figure 7 This is a structural block diagram of an aerosol removal factor prediction device inside the containment of a nuclear power plant in one embodiment.

[0046] Figure 8 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0048] In the field of nuclear safety, the containment vessel of a nuclear power plant has gradually become the last line of defense against the leakage of radioactive materials. Typically, in the event of a nuclear power plant accident, a large amount of radioactive material is released from the containment vessel. This radioactive material is removed through various mechanisms, such as deposition, coagulation, and filtration, to prevent its release into the environment as much as possible. The radioactive material removal factor is used to quantify the efficiency of these removal mechanisms. Obtaining the radioactive material removal factor is of great significance for containment integrity analysis, radioactive material release assessment, and emergency decision-making. Therefore, predicting the radioactive material removal factor is particularly important. Taking aerosols as an example, relevant technologies mainly use empirical formulas and mechanistic models based on thermo-hydraulic procedures to predict the aerosol removal factor.

[0049] However, related technologies have significant limitations when dealing with complex accident conditions and large amounts of time-series data. For example, under different accident scenarios, parameters such as temperature, pressure, and ventilation volume inside the containment undergo complex changes over time, and empirical formulas often fail to accurately describe the dynamic processes of the complex environment inside the actual containment (i.e., they fail to describe the dynamic evolution of aerosols). Furthermore, mechanistic models cannot cover all possible accident scenarios, leading to inaccurate predictions of aerosol removal factors. Based on this, embodiments of this application provide a method for predicting aerosol removal factors inside the containment of a nuclear power plant, which can improve the accuracy of aerosol removal factor prediction results.

[0050] The aerosol removal factor prediction method for the containment of a nuclear power plant provided in this application embodiment can be applied to an aerosol removal factor prediction system, which includes at least one safety cabin and computer equipment within the nuclear power plant. The computer equipment can be, but is not limited to, various personal computers, laptops, smartphones, and tablets; this embodiment does not limit the specific form of the computer equipment. The following embodiments will specifically describe the process of the aerosol removal factor prediction method for the containment of a nuclear power plant, using the computer equipment as the executing entity to illustrate the specific process.

[0051] like Figure 1 The diagram shown is a flowchart illustrating the method for predicting aerosol removal factors within the containment of a nuclear power plant according to an embodiment of this application. This method may include the following steps:

[0052] S100. Obtain the containment characteristic parameter information of the containment to be predicted. The containment characteristic parameter information represents the coupling information corresponding to the aerosol characteristics, system characteristics, model characteristics, and containment structural characteristics within the containment to be predicted.

[0053] In practical applications, the aforementioned containment to be predicted can be any containment within a nuclear power plant. Optionally, the containment characteristic parameter information of the aforementioned containment to be predicted can be understood as a kind of coupling information. In the embodiments of this application, the containment characteristic parameter information of the aforementioned containment to be predicted is the coupling information corresponding to the aerosol characteristics, system characteristics, model characteristics, and containment structural characteristics within the containment to be predicted.

[0054] Specifically, computer equipment can obtain real-time stored containment characteristic parameter information of the containment to be predicted from the cloud, local storage, disk, hard drive, etc., or directly receive containment characteristic parameter information of the containment to be predicted collected in real time by sensors.

[0055] It should be noted here that the containment characteristic parameter information for any computational domain within the containment to be predicted can be expressed as: ,in, This indicates the aerosol characteristics within the containment to be predicted. , Indicates aerosol density (kg / m³) 3 ), Indicates the thermal conductivity of aerosols (W / (m) K)), The specific heat of aerosols (J / (kg)) K)), and These represent the minimum and maximum particle sizes for aerosol analysis, respectively. m), This indicates the number of aerosol groups to be predicted within the containment. This represents the model feature information of the containment structure to be predicted. , Represents the shape factor of aerosol dynamics. This represents the aerosol aggregation shape factor. Indicates the slip correction factor. This indicates the thermodynamic heat adjustment factor.

[0056] and, This represents the system characteristic information within the containment to be predicted. , This represents the spray flow rate (kg / s) of the containment sprinkler system to be predicted. Indicates the droplet size of the spray system ( m), This indicates the spray system's spray start time (s). This indicates the thermal power (W) of the Passive Core Cooling System (PCCS). Indicates the PCCS start time (s); This represents the structural features of the containment structure to be predicted. , This represents the compartmental structure information of the containment to be predicted (including the volume (m³) of multiple compartments within the containment to be predicted). 3 Information such as elevation (m), dome radius, and dome height. This indicates the structural information of the flow channels between different compartments within the containment to be predicted (including the flow area (m²)). 2 (Information on upstream and downstream compartments, etc.) This indicates the structural information of the thermal components within the containment to be predicted (including area (m²)). 2 Information such as thickness (m), angle (°), material, and compartment location.

[0057] In practical applications, the aforementioned containment characteristic parameter information of the containment to be predicted can be the containment characteristic parameter information of the containment to be predicted within a historical time period. Optionally, the historical time period can be a period of time before the start-up of the sprinkler system or PCCS in the containment to be predicted. In this embodiment of the application, 20 minutes before the start-up of the sprinkler system or PCCS in the containment to be predicted is taken as an example.

[0058] S200. Input the containment characteristic parameter information into the pre-built aerosol removal factor prediction model, predict the aerosol removal factor in the containment to be predicted within the prediction time period, and obtain the predicted value of the aerosol removal factor in the containment to be predicted.

[0059] Among them, the aerosol removal factor prediction model includes a multi-layer physical constraint bidirectional long short-term memory neural network. The physical constraint means that the historical spray efficiency of the containment sprinkler system to be predicted is introduced into the bidirectional long short-term memory neural network as a physical constraint condition of the bidirectional long short-term memory neural network.

[0060] In practical applications, the aforementioned aerosol removal factor prediction model can be composed of at least one of the following: convolutional neural network model, fully connected neural network model, recurrent recurrent neural network model, and long short-term memory neural network model. In this embodiment, the aforementioned aerosol removal factor prediction model may include a multi-layered physically constrained bidirectional long short-term memory neural network, wherein the term "physically constrained" indicates that the historical spray efficiency within the containment to be predicted is introduced as a physical constraint condition into the bidirectional long short-term memory neural network.

[0061] Specifically, the computer equipment can input the containment characteristic parameter information into a pre-built aerosol removal factor prediction model, predict the aerosol removal factor in the containment to be predicted within the prediction time period, and obtain the predicted value of the aerosol removal factor in the containment to be predicted within the prediction time period.

[0062] In addition, to improve prediction accuracy, the computer equipment can first preprocess the containment characteristic parameter information of the containment to be predicted, and then input the preprocessed characteristic parameter information into a pre-constructed aerosol removal factor prediction model to predict the aerosol removal factor within the containment to be predicted within the prediction time period, thereby obtaining the predicted value of the aerosol removal factor within the containment to be predicted within the prediction time period. Optionally, the above preprocessing can be outlier removal processing and / or data augmentation processing, etc.

[0063] In this embodiment, the computer device can respond to a user-triggered command to display the predicted aerosol removal factor value, and output the predicted aerosol removal factor value within the containment to be predicted within a preset time period, or the aerosol removal factor change curve within the containment to be predicted within a preset time period. Simultaneously, it can also match historical accident types based on the predicted aerosol removal factor value within the containment to be predicted within the prediction time period, and output and display the matched historical accident types to remind the user to take appropriate measures to resolve nuclear power plant accidents.

[0064] The technical solution in this application embodiment obtains the containment characteristic parameter information of the containment to be predicted, and inputs the containment characteristic parameter information into a pre-constructed aerosol removal factor prediction model to predict the aerosol removal factor in the containment within the prediction time period, thereby obtaining the predicted value of the aerosol removal factor in the containment. The containment characteristic parameter information represents the coupling information corresponding to the aerosol characteristics, system characteristics, model characteristics, and containment structural characteristics in the containment to be predicted. The aerosol removal factor prediction model includes a multi-layer physically constrained bidirectional long short-term memory neural network. The physically constrained nature means that the historical spraying efficiency in the containment to be predicted is introduced into the bidirectional long short-term memory neural network as the physical constraint of the network. Constraints: The aerosol removal factor prediction model in the above method consists of a multi-layered physically constrained bidirectional long short-term memory neural network. Specifically, the prior spray efficiency within the containment to be predicted is introduced as a physical constraint into the bidirectional long short-term memory neural network, thereby improving the prediction accuracy of the aerosol removal factor prediction model. Furthermore, the above method can utilize containment characteristic parameter information of the containment within the nuclear power plant with higher quality and integrity, i.e., multi-parameter coupling information, to predict the aerosol removal factor, which can greatly improve the accuracy of the predicted aerosol removal factor. In addition, the above method does not require human intervention, which not only reduces the prediction error of the aerosol removal factor but also accelerates the prediction speed.

[0065] The construction process of the above aerosol removal factor prediction model is described below. Before performing the steps in S200 above, as follows: Figure 2 As shown, the above method also includes the following steps:

[0066] S300: Obtain the sample dataset corresponding to different sample containment structures.

[0067] The computer equipment can read sample datasets corresponding to different sample containments from local storage, disks, hard drives, the cloud, etc. Optionally, the aforementioned sample containments may be, but are not limited to, pressurized water reactor containments, boiling water reactor containments, heavy water reactor containments, or fast reactor containments. Optionally, each sample containment may include multiple sample datasets. In this embodiment, the environmental complexity within different sample containments is not limited.

[0068] S400. According to the preset partitioning rules, the sample dataset corresponding to each sample safe shell is partitioned to obtain the training set, test set and validation set.

[0069] In practical applications, the above partitioning rules can include a preset ratio for the training set, test set, and validation set. Alternatively, the partitioning rules can also be based on the chronological order in which the sample datasets were acquired.

[0070] Specifically, the computer equipment can divide the sample dataset corresponding to each sample safe shell according to the preset division rules to obtain the training set, test set and validation set.

[0071] S500 uses the training set, test set, validation set and loss function to train the initial aerosol removal factor prediction model and generate an aerosol removal factor prediction model; the loss function satisfies the dual constraint mechanism of time sensitivity and physical process rationality.

[0072] The computer equipment can input the training set, test set, validation set, initial aerosol removal factor prediction model and loss function into the training model. The training model uses the training set, test set, validation set and loss function to train the initial aerosol removal factor prediction model to obtain the trained aerosol removal factor prediction model.

[0073] Optionally, the aforementioned initial aerosol removal factor prediction model may include a multi-layer physically constrained initial bidirectional long short-term memory neural network (i.e., BiLSTM). In this embodiment, the initial aerosol removal factor prediction model includes six serially connected physically constrained initial bidirectional long short-term memory neural networks. The first three layers of the physically constrained initial bidirectional long short-term memory neural network can be referred to as the hidden layers in the initial aerosol removal factor prediction model, and the last three layers can be referred to as the output layers in the initial aerosol removal factor prediction model. In practical applications, the physically constrained initial bidirectional long short-term memory neural network is a recurrent neural network (RNN) architecture based on an improvement of the long short-term memory network (LSTM).

[0074] Recurrent Neural Networks (RNNs) are a type of neural network capable of processing time-series data. They possess memory capabilities, making them suitable for handling time-series data related to aerosol removal within the containment of nuclear power plants. RNNs can capture the temporal dependencies within time-series data, thus enabling more accurate modeling of the aerosol removal process. In particular, the improved RNN architecture based on Long Short-Term Memory (LSTM) networks can better handle long-term dependencies, avoiding the gradient vanishing problem inherent in traditional RNNs for long-series modeling. Therefore, employing an aerosol removal factor prediction model can overcome the limitations of processing time-series data, providing a more accurate and reliable method for predicting aerosol removal factors within the containment of nuclear power plants.

[0075] It should be noted that the forget gate in the hidden layer incorporates the historical spray efficiency of the containment spray system for each sample. Furthermore, in the early stages of a nuclear power plant accident, aerosols may exhibit uneven distribution due to the complex thermal-hydraulic environment within the containment. Theoretical calculations assume uniform aerosol distribution within the containment, which overestimates aerosol removal mechanisms such as spraying, resulting in an actual aerosol removal factor lower than the theoretically calculated one. Therefore, the output layer of the initial aerosol removal factor prediction model considers compensation for aerosol mixing effects within the containment. In this embodiment, the physically constrained initial bidirectional long short-term memory neural network can process time-series data and capture long-term dependencies. By incorporating historical spray efficiencies within the containment, the physical information contained in this historical data can constrain the training and prediction process of the initial aerosol removal factor prediction model, making the output of the trained aerosol removal factor prediction model more consistent with actual physical phenomena and laws, thereby improving the prediction accuracy of the aerosol removal factor prediction model.

[0076] Optionally, the output of the output layer in the above initial aerosol removal factor prediction model can be expressed by the following formula (1):

[0077] (1)

[0078] in, Indicates at time step The output value, Indicates at time step The sample prediction results Represents the time constant. This indicates the maximum downward adjustment percentage.

[0079] Furthermore, when the spray system within each sample containment is activated, the aerosol removal factor does not reach its peak instantaneously due to the time required for the spray droplets to distribute; instead, it exhibits an exponential growth trend. In this embodiment, prior information about the spray efficiency (i.e., the spray droplets from the spray system) is incorporated into the forget gate, allowing the forget gate in the BiLSTM to adjust before and after the spray system is activated, thus predicting the impact on the aerosol removal factor more quickly. The output value of the forget gate at time step t can be expressed by the following formula:

[0080] (2)

[0081] in, This represents the activation function (used to map input to...). (interval) The weight matrix represents the forget gate. Indicates at time step ( The hidden state of ) Indicates at time step Input, The bias term representing the forget gate. Indicates the spray bias term. This represents the spray bias constant.

[0082] Optionally, the loss function described above can be, but is not limited to, the squared absolute error function, the mean square error function, the root mean square error function, and / or the mean absolute percentage error function. In the embodiments of this application, in order to improve the prediction accuracy of the final aerosol removal factor prediction model, the loss function can be designed as a function that simultaneously satisfies the dual constraint mechanism of time sensitivity and physical process rationality. The time sensitivity is reflected in the higher prediction accuracy requirements in the early stages of a nuclear accident; the actual physical process of aerosol removal factor is continuous and constrained by inertial factors such as droplet settling rate and diffusion process, and should not exhibit drastic jumps. Correspondingly, the physical process rationality can represent the smoothness of the change in aerosol removal factor conforming to the actual physical process.

[0083] In this embodiment, the loss function Loss can be expressed by the following formula:

[0084] (3)

[0085] in, The first part represents the time-weighted mean square error (MSE) (used to measure the predicted aerosol removal factor output by the initial aerosol removal factor prediction model). and sample prediction results The differences between them, in order to meet time sensitivity requirements. Indicates weight, The latter part represents the total variation regularization (to satisfy the smoothness of the predicted aerosol removal factor over time). It is the regularization coefficient. Indicates the number of training sets. This represents the total number of time steps.

[0086] In the embodiments of this application, the initial aerosol removal factor prediction model can be trained using traditional model training methods, which will not be elaborated further in the embodiments of this application.

[0087] The technical solution in this application embodiment obtains sample datasets corresponding to different sample containment structures, divides the sample datasets corresponding to each sample containment structure according to a preset partitioning rule to obtain a training set, a test set, and a validation set, and uses the training set, test set, validation set, and loss function to train the initial aerosol removal factor prediction model to generate an aerosol removal factor prediction model. The loss function satisfies the dual constraint mechanism of time sensitivity and physical process rationality. The above method uses a loss function that satisfies the dual constraint mechanism of time sensitivity and physical process rationality to train the initial aerosol removal factor prediction model, which can improve the prediction accuracy of the finally trained aerosol removal factor prediction model.

[0088] The process of obtaining sample datasets corresponding to different sample safe houses is described below. In one embodiment, as... Figure 3 As shown, the steps in S300 above can be implemented in the following ways:

[0089] S310. Based on the physical field model of different sample containment structures, obtain the sample characteristic parameter information of each sample containment structure.

[0090] The physical field model of the aforementioned sample containment can include a thermo-hydraulic model, a structural mechanics model, a fluid dynamics model, etc., which are mathematical models used to describe various physical phenomena and processes within the sample containment. In this embodiment, the physical field model of the sample containment can reflect the thermodynamic and fluid dynamic changes within the sample containment, that is, describe the flow of gas and the transport of aerosols within the sample containment. Based on this, the temporal characteristics of the aerosol removal process within the sample containment can be captured based on the physical field model of the sample containment in the nuclear power plant, thereby obtaining the containment characteristic parameter information corresponding to the sample containment, i.e., the sample characteristic parameter information of the sample containment.

[0091] Specifically, the computer equipment can first obtain the mass conservation equation, momentum equation, and aerosol transport and diffusion equation for each sample containment using a thermal-hydraulic integrated analysis module. Then, based on the mass conservation equation, momentum equation, and aerosol transport and diffusion equation for each sample containment, it can construct physical field models of different sample containments in real time. Alternatively, it can obtain pre-constructed physical field models of different sample containments and then obtain the containment characteristic parameter information corresponding to each sample containment based on the physical field models of different sample containments, i.e., the sample characteristic parameter information of each sample containment.

[0092] The aforementioned integrated thermal-hydraulic analysis module can be understood as a numerical method for analyzing and simulating thermal-hydraulic phenomena such as fluid flow, heat transfer, and phase change. Optionally, the integrated thermal-hydraulic analysis module may be, but is not limited to, a reactor coolant system analysis program, a thermal-hydraulic system analysis program, or a containment analysis system. In this embodiment, the integrated thermal-hydraulic analysis module is illustrated using the Multiphysics Emergency Logic Simulator (MELCOR) or the Accident Source Term Evaluation Code (ASTEC). In this embodiment, the complexity of the environment within the containment to be predicted is not limited.

[0093] It should be noted that, for any sample containment, in this embodiment, the sample characteristic parameter information of the sample containment can be the sample characteristic parameter information of the sample containment within the sampling time period. Specifically, the sample characteristic parameter information corresponding to each sample computational domain within the sample containment is, in other words, the sample characteristic parameter information of that sample containment. , , , and Some information remains constant over time, while other information changes over time. Based on this, the Monte Carlo method can be used to sample the sample characteristic parameters of the containment vessel to obtain multiple sets of different sample characteristic parameters corresponding to the containment vessel.

[0094] In the embodiments of this application, the mass conservation equation, momentum equation, and aerosol transport and diffusion equation of the sample containment may include the mass conservation equation, momentum equation, and aerosol transport and diffusion equation of each computational domain within the sample containment.

[0095] Due to the complex geometry of the containment vessel and the intricate phenomena of internal fluid flow and aerosol transport, the physical field model of the containment vessel must consider not only its actual layout but also various mechanisms such as gas flow, aerosol migration, and aerosol settling within it. Specifically, based on the design drawings of the containment vessel, multiple computational domains (i.e., regions) within it can be precisely divided. Then, for each computational domain, the mass conservation equation, momentum equation, and aerosol transport and diffusion equation can be obtained. This allows the construction of the physical field model of each computational domain within the containment vessel, i.e., the physical field model of the containment vessel itself, preparing for the subsequent refined acquisition of containment characteristic parameters within each computational domain.

[0096] Among them, gas flow can determine the direction and speed of aerosol transport, aerosol migration can describe the transport process of aerosols in the environmental medium inside the sample containment, and aerosol sedimentation can reflect the law of aerosols settling from the gas phase under the action of gravity, inertia and other factors.

[0097] Optionally, the containment characteristic parameter information of the aforementioned sample containment may include the containment characteristic parameter information of each computational domain in the sample containment. Optionally, the spatial volumes of different computational domains may be equal or unequal, and this embodiment of the application does not limit this.

[0098] Optionally, the sampling time period can be a period of time before the start of the sample containment spray system or PCCS. In this embodiment, 20 minutes before the start of the sample containment spray system or PCCS is taken as an example.

[0099] S320. Using a thermal-hydraulic algorithm, based on the sample characteristic parameters of each sample containment, the sample prediction results of the aerosol removal factor inside each sample containment are obtained.

[0100] Specifically, the computer equipment can employ thermal-hydraulic algorithms to process the sample characteristic parameter information of each containment sample, obtaining the sample prediction results of the aerosol removal factor within each containment sample. Optionally, the aforementioned thermal-hydraulic algorithms can be, but are not limited to, single-channel analysis, sub-channel analysis, computational fluid dynamics analysis, thermal-hydraulic multi-scale coupled calculation, or data assimilation-based calculation.

[0101] For any sample containment, the sample prediction result of the aerosol removal factor within the sample containment can include time series data of the aerosol removal factor changing over time within the sample containment during the prediction period.

[0102] S330. Based on the sample characteristic parameter information of each sample containment and the sample prediction results of the aerosol removal factor inside each sample containment, obtain the sample dataset corresponding to each sample containment.

[0103] In practical applications, for any sample containment, computer equipment can simply combine the sample characteristic parameter information of the sample containment and the sample prediction results of the aerosol removal factor inside the sample containment to obtain the sample dataset corresponding to the sample containment.

[0104] The technical solution in this application embodiment obtains the sample characteristic parameter information of each sample containment based on the physical field model of different sample containment. Using a thermal-hydraulic algorithm, the sample prediction results of the aerosol removal factor inside each sample containment are obtained based on the sample characteristic parameter information of each sample containment. Based on the sample characteristic parameter information and the sample prediction results of the aerosol removal factor inside each sample containment, the sample dataset corresponding to each sample containment is obtained. The above method does not require the participation of complex algorithms, reduces the complexity of obtaining the sample dataset corresponding to the sample containment, and can accelerate the acquisition speed of the sample dataset corresponding to the sample containment.

[0105] It should be noted that in some embodiments, such as Figure 4 As shown, before step S310 above, step S300, obtaining the sample dataset corresponding to different sample safe houses, may further include:

[0106] S340. Based on the mass conservation equation, momentum equation, and aerosol transport and diffusion equation of each sample containment, construct the physical field model of each sample containment.

[0107] In practical applications, computer equipment can use the thermal-hydraulic integrated analysis module to simulate the complex physical phenomena and processes inside each sample containment, obtain the mass conservation equation, momentum equation and aerosol transport and diffusion equation for each sample containment, and construct the physical field model of each sample containment based on the mass conservation equation, momentum equation and aerosol transport and diffusion equation for each sample containment.

[0108] In this embodiment, a physical field model of each sample containment can be constructed based on the mass conservation equation, momentum equation, and aerosol transport and diffusion equation of each sample containment. This ensures that the temporal characteristics of the aerosol removal process within each sample containment can be captured from the physical field model. Of course, in other embodiments, other methods can be used to obtain the physical field models of different sample containments, which are not limited here.

[0109] The following describes the process of obtaining the sample dataset corresponding to each containment unit based on the sample characteristic parameter information of each containment unit and the sample prediction results of the aerosol removal factor within each containment unit. In one embodiment, as... Figure 5 As shown, the steps in S330 above can be implemented in the following ways:

[0110] S331. Perform outlier removal processing on the sample prediction results of aerosol removal factors inside the containment for each sample to obtain reference prediction results of aerosol removal factors inside the containment for each sample.

[0111] In practical applications, computer equipment can employ outlier removal methods to remove outliers from the sample prediction results of aerosol removal factors within the containment for each sample, thus obtaining reference prediction results for the aerosol removal factors within the containment for each sample. These outlier removal methods can include standard deviation, modified Z-score, cluster analysis, nearest neighbor analysis, or regression analysis, among others.

[0112] In this embodiment of the application, the computer device may adopt an improved 3 Based on the criteria and removal factor elimination rules, outlier removal processing was performed on the sample prediction results of aerosol removal factors inside the containment for each sample to obtain reference prediction results of aerosol removal factors inside the containment for each sample.

[0113] It should be noted here that the improved version 3 The criteria are constructed based on nuclear safety analysis requirements and may include the rejection of sample prediction results determined by sample characteristic parameter information where the droplet size of the sprayed liquid is not within three standard deviations. In the embodiments of this application, the improved 3 The criterion can be expressed by the following formula (4):

[0114] (4)

[0115] in, This represents the reference predicted result for the aerosol removal factor within the containment for any given sample. Indicates standard deviation, The droplet size in the sample prediction results for the aerosol removal factor within the sample containment is indicated by the spray droplet size. This represents the mean droplet size in the sample prediction results of the aerosol removal factor within the containment for each sample. This represents the standard deviation of the spray droplet size in the sample prediction results of the aerosol removal factor within the containment for each sample.

[0116] Meanwhile, the aforementioned removal factor elimination rule may include a sample prediction result for aerosol removal factor within the sample containment being greater than a first preset factor threshold and less than a second preset factor threshold, wherein the second preset factor threshold is greater than the first preset factor threshold. In this embodiment, the first preset factor threshold and the second preset factor threshold may be opposites. Furthermore, for different sample containments, the corresponding first preset factor thresholds may be equal or unequal, and the corresponding second preset factor thresholds may be equal or unequal.

[0117] In practical applications, the first and second preset factor thresholds will change as the pressure inside the sample containment changes. For example, when the pressure inside the sample containment exceeds 0.5 MPa, the deviation range between the first and second preset factor thresholds can be widened to ±7%.

[0118] S332. Based on the sample characteristic parameter information of each sample containment and the reference prediction results of the aerosol removal factor in each sample containment, the sample dataset corresponding to each sample containment is obtained by combining the data.

[0119] Furthermore, for any sample containment, the computer device can combine the sample characteristic parameter information of the sample containment with the reference prediction results of the aerosol removal factor inside the sample containment to obtain the sample dataset corresponding to the sample containment.

[0120] The technical solution in this application embodiment can first perform outlier removal processing on the sample prediction results of aerosol removal factors in each sample containment, and then combine the sample characteristic parameter information of each sample containment with the obtained reference prediction results of aerosol removal factors in each sample containment to obtain the sample dataset corresponding to each sample containment. This can improve the quality of the obtained sample dataset based on the reference prediction results of aerosol removal factors in the sample containment with higher quality, so as to provide reliable data for subsequent model training and improve the prediction accuracy of the finally obtained aerosol removal factor prediction model.

[0121] In one embodiment, the above-mentioned division rules include the system limit condition conditions within each sample containment and the geometric conditions of each sample containment; such as Figure 6 As shown, the step in S400 above, which divides the sample dataset corresponding to each sample safe house according to a preset partitioning rule to obtain a training set, a test set, and a validation set, may include:

[0122] S410. The sample datasets that satisfy the system's extreme operating conditions in each sample dataset are determined as the test set; and the sample datasets that satisfy the geometric structure conditions in each sample dataset are determined as the validation set.

[0123] In this embodiment, the computer device can select sample datasets that satisfy the system's extreme operating conditions from all sample datasets as the test set, and select sample datasets that satisfy the geometric structure conditions from all sample datasets as the validation set. The order in which the test set and validation set are selected can be interchanged, and this embodiment does not limit this process.

[0124] It should be noted that if the same sample dataset satisfies both the system limit condition and the geometric condition, then this sample dataset can be used as a test set or a validation set. Optionally, the system limit condition can be understood as a sample dataset with a spray flow rate Qs greater than 40 kg / s and both the aerosol dynamics shape factor and the aerosol condensation shape factor greater than 1.2; the aforementioned geometric condition can include sample datasets where the ratio of the dome radius to the dome height of the sample containment is in the range of [0.8, 1.2].

[0125] S420. Select the sample datasets other than the test set and validation set from each sample dataset as the training set.

[0126] Furthermore, all sample datasets except for the test and validation sets can be used as training sets. In this embodiment, the training, test, and validation sets are selected based on system extreme operating conditions and geometric conditions. This ensures that the finally trained aerosol removal factor prediction model has good robustness to the geometric scaling of the containment.

[0127] The technical solution in this application embodiment has a relatively simple process for selecting test sets, validation sets and training sets from all sample datasets. It does not require complex algorithms, thereby accelerating the acquisition of test sets, validation sets and training sets.

[0128] In some scenarios, to ensure the safety of the containment vessel, when the predicted aerosol removal factor within the containment vessel is low, corresponding safety measures need to be taken to protect the containment vessel. The following describes this approach. In one embodiment, after performing the steps in S200 above, the method may further include: triggering a safety warning message for the containment vessel based on the predicted aerosol removal factor value within the predicted time period.

[0129] In practical applications, computer equipment can pre-train an early warning model. Then, the predicted aerosol removal factor value within the containment structure for the predicted time period is input into the early warning model for processing. The model then outputs a safety warning message for the containment structure and triggers the warning. Simultaneously, the computer equipment can also display this safety warning message to remind users to take appropriate safety measures to protect the containment structure.

[0130] Optionally, the aforementioned early warning model can be implemented by at least one of the following: convolutional neural network model, fully connected neural network model, recurrent recurrent neural network model, residual neural network model, and recurrent recurrent neural network model.

[0131] In one embodiment, the step of triggering a safety warning information for the containment based on the predicted value of the aerosol removal factor within the containment within the predicted time period may include:

[0132] If all predicted aerosol removal factors within a first duration are less than a first preset threshold, an early warning message is output; if all predicted aerosol removal factors within a second duration are less than a second preset threshold, an early warning command is triggered. The second preset threshold is less than the first preset threshold.

[0133] The predicted aerosol removal factor within the containment to be predicted during the aforementioned prediction period may include the predicted aerosol removal factor within the containment to be predicted at each moment within the preset time period, that is, the predicted aerosol removal factor within the containment to be predicted that changes over time within the preset time period.

[0134] In practical applications, computer equipment can determine whether any of the predicted aerosol removal factor values ​​within the containment structure during the prediction period are all less than a first preset threshold for a first duration. If so, a warning message is output. In this embodiment, the warning message may be related to prompting the inspection of the sprinkler system or PCCS system.

[0135] In addition, the computer equipment can determine whether there are any predicted aerosol removal factors within the containment under prediction within the prediction time period that are all less than the second preset threshold during the second duration. If so, it indicates that the aerosol removal factor within the containment under prediction is small, and an early warning command can be triggered.

[0136] The first and second preset thresholds can be user-defined or determined based on historical experience. However, in this embodiment, the second preset threshold can be less than the first preset threshold and can be equal to 0.5 times the first preset threshold.

[0137] Meanwhile, the first duration and the second duration may be equal or unequal, and this application embodiment does not limit this. In this application embodiment, the aforementioned warning command may be an automatic start command for the containment sprinkler system or PCCS to be predicted.

[0138] This embodiment can trigger different levels of safety warning information according to different actual situations to ensure the safety of the containment to be predicted, thereby improving the safety of the nuclear power plant.

[0139] The technical solution in this application embodiment triggers a safety warning information for the containment to be predicted based on the predicted value of the aerosol removal factor within the predicted containment period, and then takes corresponding safety measures to protect the containment based on the safety warning information, thereby improving the safety of the nuclear power plant.

[0140] In one embodiment, this application also provides a method for predicting aerosol removal factors within the containment of a nuclear power plant, applied to computer equipment. The method includes the following steps:

[0141] (1) Based on the mass conservation equation, momentum equation and aerosol transport and diffusion equation of each sample containment, construct the physical field model of each sample containment;

[0142] (2) Based on the physical field model of different sample containment structures, obtain the sample characteristic parameter information of each sample containment structure;

[0143] (3) Using the thermal-hydraulic algorithm, based on the sample characteristic parameter information of each sample containment, the sample prediction results of the aerosol removal factor in each sample containment are obtained;

[0144] (4) The outlier removal process was performed on the sample prediction results of the aerosol removal factor inside the containment of each sample to obtain the reference prediction results of the aerosol removal factor inside the containment of each sample.

[0145] (5) Combine the sample characteristic parameter information of each sample containment and the reference prediction results of the aerosol removal factor in each sample containment to obtain the sample dataset corresponding to each sample containment.

[0146] (6) The sample datasets that satisfy the system limit condition in each sample dataset are determined as the test set; and the sample datasets that satisfy the geometric condition in each sample dataset are determined as the validation set;

[0147] (7) The sample datasets other than the test set and validation set in each sample dataset are determined as the training set;

[0148] (8) The initial aerosol removal factor prediction model is trained using the training set, test set, validation set and loss function to generate the aerosol removal factor prediction model; the loss function satisfies the dual constraint mechanism of time sensitivity and physical process rationality.

[0149] (9) Obtain containment characteristic parameter information of the containment to be predicted; the containment characteristic parameter information represents the coupling information corresponding to the aerosol characteristics, system characteristics, model characteristics and containment structural characteristics in the containment to be predicted;

[0150] (10) Input the containment characteristic parameter information into the pre-built aerosol removal factor prediction model, predict the aerosol removal factor in the containment to be predicted within the prediction time period, and obtain the predicted value of the aerosol removal factor in the containment to be predicted.

[0151] Among them, the aerosol removal factor prediction model includes a multi-layer physical constraint bidirectional long short-term memory neural network. The physical constraint means that the historical spray efficiency inside the containment to be predicted is introduced into the bidirectional long short-term memory neural network as a physical constraint condition of the bidirectional long short-term memory neural network.

[0152] (11) If all the predicted values ​​of aerosol removal factors within a first duration are less than the first preset threshold, then an early warning message is output; if all the predicted values ​​of aerosol removal factors within a second duration are less than the second preset threshold, then an early warning command is triggered; the second preset threshold is less than the first preset threshold.

[0153] For details of the execution process of (1) to (11) above, please refer to the description of the above embodiments. The implementation principle and technical effect are similar, and will not be repeated here.

[0154] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0155] Based on the same inventive concept, this application also provides an apparatus for predicting aerosol removal factors inside the containment of a nuclear power plant to implement the aforementioned method for predicting aerosol removal factors inside the containment of a nuclear power plant. The solution provided by this apparatus is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the apparatus for predicting aerosol removal factors inside the containment of a nuclear power plant provided below can be found in the limitations of the method for predicting aerosol removal factors inside the containment of a nuclear power plant described above, and will not be repeated here.

[0156] In one embodiment, Figure 7 This is a schematic diagram of the structure of an aerosol removal factor prediction device inside the containment of a nuclear power plant according to one embodiment of this application. The aerosol removal factor prediction device inside the containment of a nuclear power plant provided in this embodiment can be applied to computer equipment. Figure 7 As shown, the aerosol removal factor prediction device inside the containment of a nuclear power plant according to an embodiment of this application may include: an acquisition module 11 and a prediction module 12, wherein:

[0157] The acquisition module 11 is used to acquire containment characteristic parameter information of the containment to be predicted; the containment characteristic parameter information represents the coupling information corresponding to the aerosol characteristics, system characteristics, model characteristics and containment structural characteristics within the containment to be predicted;

[0158] Prediction module 12 is used to input containment characteristic parameter information into a pre-built aerosol removal factor prediction model, predict the aerosol removal factor in the containment to be predicted within the prediction time period, and obtain the predicted value of the aerosol removal factor in the containment to be predicted.

[0159] Among them, the aerosol removal factor prediction model includes a multi-layer physical constraint bidirectional long short-term memory neural network. The physical constraint means that the historical spray efficiency inside the containment to be predicted is introduced into the bidirectional long short-term memory neural network as a physical constraint condition of the bidirectional long short-term memory neural network.

[0160] The aerosol removal factor prediction device for nuclear power plant containment provided in this application can be used to execute the technical solutions in the above-described embodiments of the aerosol removal factor prediction method for nuclear power plant containment. Its implementation principle and technical effects are similar, and will not be repeated here.

[0161] In one embodiment, the aerosol removal factor prediction device inside the containment of a nuclear power plant further includes: a dataset acquisition module, a dataset partitioning module, and a model training module, wherein:

[0162] The dataset acquisition module is used to acquire sample datasets corresponding to different sample safe shells;

[0163] The dataset partitioning module is used to partition the sample dataset corresponding to each sample safe shell according to a preset partitioning rule, so as to obtain the training set, test set and validation set.

[0164] The model training module is used to train the initial aerosol removal factor prediction model using the training set, test set, validation set, and loss function, and generate the aerosol removal factor prediction model; the loss function satisfies the dual constraint mechanism of time sensitivity and physical process rationality.

[0165] The aerosol removal factor prediction device for nuclear power plant containment provided in this application can be used to execute the technical solutions in the above-described embodiments of the aerosol removal factor prediction method for nuclear power plant containment. Its implementation principle and technical effects are similar, and will not be repeated here.

[0166] In one embodiment, the dataset acquisition module includes: a first acquisition unit, a second acquisition unit, and a dataset acquisition unit, wherein:

[0167] The first acquisition unit is used to acquire sample characteristic parameter information of each sample safe house based on the physical field model of different sample safe houses;

[0168] The second acquisition unit is used to obtain the sample prediction results of the aerosol removal factor inside the containment of each sample by using a thermal-hydraulic algorithm based on the sample characteristic parameter information of each containment sample.

[0169] The dataset acquisition unit is used to obtain the sample dataset corresponding to each sample containment based on the sample characteristic parameter information of each sample containment and the sample prediction results of the aerosol removal factor inside each sample containment.

[0170] The aerosol removal factor prediction device for nuclear power plant containment provided in this application can be used to execute the technical solutions in the above-described embodiments of the aerosol removal factor prediction method for nuclear power plant containment. Its implementation principle and technical effects are similar, and will not be repeated here.

[0171] In one embodiment, the dataset acquisition module further includes: a model building unit, wherein:

[0172] The model building unit is used to construct the physical field model of each sample containment based on the mass conservation equation, momentum equation, and aerosol transport and diffusion equation of each sample containment.

[0173] The aerosol removal factor prediction device for nuclear power plant containment provided in this application can be used to execute the technical solutions in the above-described embodiments of the aerosol removal factor prediction method for nuclear power plant containment. Its implementation principle and technical effects are similar, and will not be repeated here.

[0174] In one embodiment, the dataset acquisition unit is specifically used for:

[0175] Outlier removal was performed on the sample prediction results of the aerosol removal factor inside the containment for each sample to obtain the reference prediction results of the aerosol removal factor inside the containment for each sample.

[0176] The sample datasets corresponding to each sample containment are obtained by combining the sample characteristic parameters of each containment and the reference prediction results of the aerosol removal factor inside each containment.

[0177] The aerosol removal factor prediction device for nuclear power plant containment provided in this application can be used to execute the technical solutions in the above-described embodiments of the aerosol removal factor prediction method for nuclear power plant containment. Its implementation principle and technical effects are similar, and will not be repeated here.

[0178] In one embodiment, the partitioning rules include the system extreme operating conditions within each sample containment and the geometric conditions of each sample containment; the dataset partitioning module is specifically used for:

[0179] The sample datasets that satisfy the system's extreme operating conditions are designated as the test set; and the sample datasets that satisfy the geometric structure conditions are designated as the validation set.

[0180] The sample datasets excluding the test and validation sets are designated as the training set.

[0181] The aerosol removal factor prediction device for nuclear power plant containment provided in this application can be used to execute the technical solutions in the above-described embodiments of the aerosol removal factor prediction method for nuclear power plant containment. Its implementation principle and technical effects are similar, and will not be repeated here.

[0182] In one embodiment, the aerosol removal factor prediction device inside the containment of a nuclear power plant further includes: a warning information determination module, wherein:

[0183] The early warning information determination module is used to trigger safety early warning information for the containment to be predicted based on the predicted value of the aerosol removal factor within the containment to be predicted within the predicted time period.

[0184] The aerosol removal factor prediction device for nuclear power plant containment provided in this application can be used to execute the technical solutions in the above-described embodiments of the aerosol removal factor prediction method for nuclear power plant containment. Its implementation principle and technical effects are similar, and will not be repeated here.

[0185] In one embodiment, the warning information determination module is specifically used for:

[0186] If all predicted aerosol removal factors within a first duration are less than a first preset threshold, then an early warning message will be output.

[0187] If all predicted aerosol removal factors within a second duration are less than a second preset threshold, an early warning command is triggered; the second preset threshold is less than the first preset threshold.

[0188] The aerosol removal factor prediction device for nuclear power plant containment provided in this application can be used to execute the technical solutions in the above-described embodiments of the aerosol removal factor prediction method for nuclear power plant containment. Its implementation principle and technical effects are similar, and will not be repeated here.

[0189] Specific limitations regarding the aerosol removal factor prediction device within the containment of a nuclear power plant can be found in the limitations on the aerosol removal factor prediction method within the containment of a nuclear power plant mentioned above, and will not be repeated here. Each module in the aforementioned aerosol removal factor prediction device within the containment of a nuclear power plant can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0190] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 8 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements a method for predicting aerosol removal factors within the containment of a nuclear power plant. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0191] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0192] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0193] Obtain containment characteristic parameter information of the containment to be predicted; the containment characteristic parameter information represents the coupling information corresponding to the aerosol characteristics, system characteristics, model characteristics and containment structural characteristics within the containment to be predicted;

[0194] The containment characteristic parameter information is input into the pre-built aerosol removal factor prediction model to predict the aerosol removal factor in the containment to be predicted within the prediction time period, and the predicted value of the aerosol removal factor in the containment to be predicted is obtained.

[0195] Among them, the aerosol removal factor prediction model includes a multi-layer physical constraint bidirectional long short-term memory neural network. The physical constraint means that the historical spray efficiency inside the containment to be predicted is introduced into the bidirectional long short-term memory neural network as a physical constraint condition of the bidirectional long short-term memory neural network.

[0196] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0197] Obtain containment characteristic parameter information of the containment to be predicted; the containment characteristic parameter information represents the coupling information corresponding to the aerosol characteristics, system characteristics, model characteristics and containment structural characteristics within the containment to be predicted;

[0198] The containment characteristic parameter information is input into the pre-built aerosol removal factor prediction model to predict the aerosol removal factor in the containment to be predicted within the prediction time period, and the predicted value of the aerosol removal factor in the containment to be predicted is obtained.

[0199] Among them, the aerosol removal factor prediction model includes a multi-layer physical constraint bidirectional long short-term memory neural network. The physical constraint means that the historical spray efficiency inside the containment to be predicted is introduced into the bidirectional long short-term memory neural network as a physical constraint condition of the bidirectional long short-term memory neural network.

[0200] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0201] Obtain containment characteristic parameter information of the containment to be predicted; the containment characteristic parameter information represents the coupling information corresponding to the aerosol characteristics, system characteristics, model characteristics and containment structural characteristics within the containment to be predicted;

[0202] The containment characteristic parameter information is input into the pre-built aerosol removal factor prediction model to predict the aerosol removal factor in the containment to be predicted within the prediction time period, and the predicted value of the aerosol removal factor in the containment to be predicted is obtained.

[0203] Among them, the aerosol removal factor prediction model includes a multi-layer physical constraint bidirectional long short-term memory neural network. The physical constraint means that the historical spray efficiency inside the containment to be predicted is introduced into the bidirectional long short-term memory neural network as a physical constraint condition of the bidirectional long short-term memory neural network.

[0204] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0205] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0206] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for predicting aerosol removal factors within the containment of a nuclear power plant, characterized in that, The method includes: Obtain containment characteristic parameter information of the containment to be predicted; the containment characteristic parameter information represents the coupling information corresponding to the aerosol characteristics, system characteristics, model characteristics and containment structural characteristics within the containment to be predicted; The containment characteristic parameter information is input into a pre-built aerosol removal factor prediction model to predict the aerosol removal factor in the containment to be predicted within the prediction time period, and the predicted value of the aerosol removal factor in the containment to be predicted is obtained. The aerosol removal factor prediction model includes a multi-layered physically constrained bidirectional long short-term memory neural network. The physically constrained nature means that the historical spray efficiency within the containment to be predicted is introduced into the bidirectional long short-term memory neural network as a physical constraint condition for the network.

2. The method according to claim 1, characterized in that, The process of constructing the aerosol removal factor prediction model includes: Obtain the sample datasets corresponding to different sample safe shells; According to the preset partitioning rules, the sample datasets corresponding to each sample safety shell are partitioned to obtain training set, test set and validation set; The initial aerosol removal factor prediction model is trained using the training set, the test set, the validation set, and the loss function to generate the aerosol removal factor prediction model.

3. The method according to claim 2, characterized in that, The loss function satisfies the dual constraints of time sensitivity and physical process rationality.

4. The method according to claim 2, characterized in that, The process of obtaining sample datasets corresponding to different sample safe houses includes: Based on the physical field models of different sample safe houses, the sample characteristic parameter information of each sample safe house is obtained; The thermal-hydraulic algorithm is used to obtain the sample prediction results of the aerosol removal factor inside the containment of each sample based on the sample characteristic parameter information of each sample containment. Based on the sample characteristic parameter information of each sample containment and the sample prediction results of the aerosol removal factor inside each sample containment, the sample dataset corresponding to each sample containment is obtained.

5. The method according to claim 4, characterized in that, The process of obtaining sample datasets corresponding to different sample safe houses also includes: Based on the mass conservation equation, momentum equation, and aerosol transport and diffusion equation of each sample containment vessel, a physical field model of each sample containment vessel is constructed.

6. The method according to claim 4, characterized in that, The step of obtaining the sample dataset corresponding to each sample containment based on the sample characteristic parameter information of each sample containment and the sample prediction results of the aerosol removal factor inside each sample containment includes: Outlier removal processing was performed on the sample prediction results of the aerosol removal factor inside the containment of each sample to obtain the reference prediction results of the aerosol removal factor inside the containment of each sample. The sample datasets corresponding to each sample containment are obtained by combining the sample characteristic parameter information of each sample containment and the reference prediction results of the aerosol removal factor inside each sample containment.

7. The method according to claim 2, characterized in that, The partitioning rules include the system extreme operating conditions within each sample containment and the geometric structural conditions of each sample containment; the partitioning of the sample dataset corresponding to each sample containment according to the preset partitioning rules to obtain a training set, a test set, and a validation set includes: The sample datasets that satisfy the system's extreme operating condition from each of the aforementioned sample datasets are determined as the test set; and the sample datasets that satisfy the geometric structure condition from each of the aforementioned sample datasets are determined as the verification set; The sample datasets other than the test set and the validation set in each of the aforementioned sample datasets are determined as the training set.

8. The method according to claim 1, characterized in that, The method further includes: Based on the predicted value of the aerosol removal factor within the containment to be predicted during the predicted time period, a safety warning message is triggered within the containment to be predicted.

9. The method according to claim 8, characterized in that, The step of triggering a safety warning message for the containment to be predicted based on the predicted aerosol removal factor value within the predicted time period includes: If all the predicted aerosol removal factors within a first duration are less than a first preset threshold, then an early warning message is output. If all predicted aerosol removal factors within a second duration are less than a second preset threshold, an early warning command is triggered; the second preset threshold is less than the first preset threshold.

10. A device for predicting aerosol removal factors inside the containment of a nuclear power plant, characterized in that, The device includes: The acquisition module is used to acquire containment characteristic parameter information of the containment to be predicted; the containment characteristic parameter information represents the coupling information corresponding to the aerosol characteristics, system characteristics, model characteristics and containment structural characteristics within the containment to be predicted. The prediction module is used to input the containment characteristic parameter information into the pre-built aerosol removal factor prediction model, predict the aerosol removal factor in the containment to be predicted within the prediction time period, and obtain the predicted value of the aerosol removal factor in the containment to be predicted. The aerosol removal factor prediction model includes a multi-layered physically constrained bidirectional long short-term memory neural network. The physically constrained nature means that the historical spray efficiency within the containment to be predicted is introduced into the bidirectional long short-term memory neural network as a physical constraint condition for the network.

Images