Generation method for accident prediction model and diagnosis method for nuclear power plant accident

Abstract

The present disclosure provides a generation method for an accident prediction model and a diagnosis method for a nuclear power plant accident. The generation method for an accident prediction model comprises: acquiring a plurality of sets of sample training data, the sample training data comprising sample operation data of a nuclear power plant in different historical time periods and a sample accident type corresponding to the sample operation data; training a plurality of different teacher models on the basis of the sample training data, so as to obtain a plurality of intermediate sample models for predicting a target accident type corresponding to target operation data of the nuclear power plant in an arbitrary time period; performing knowledge distillation on the basis of the plurality of different intermediate sample models, so as to obtain a student model, the student model serving as an accident prediction model. Thus, the accident prediction model enables rapid identification of a nuclear power plant accident type under the condition of limited computing resources, thereby improving the accuracy and computational efficiency of accident diagnosis.

Description

Methods for generating accident prediction models and methods for diagnosing nuclear power plant accidents

[0001] This disclosure claims priority to Chinese Patent Application No. 202510175974.1, filed on February 18, 2025. The entire contents of the aforementioned Chinese Patent Application are incorporated herein by reference. Technical Field

[0002] This disclosure relates to the field of nuclear power plant data processing, and in particular to a method for generating an accident prediction model and a method for diagnosing nuclear power plant accidents. Background Technology

[0003] In recent years, artificial intelligence (AI) technology has been increasingly widely applied across various industries, achieving remarkable results, particularly in improving efficiency, productivity, and decision-making capabilities. The nuclear power industry is also increasingly focusing on using AI technology to enhance the safety and reliability of nuclear power plants.

[0004] However, due to the highly complex operating environment of nuclear power plants and the need to comply with strict safety regulations, obtaining sufficient real-world data for training AI models has become a major challenge. At the same time, nuclear power plant accidents are relatively rare, making data collection difficult, which means that most AI research relies mainly on simulated data. Furthermore, there is a certain gap between simulated data and actual operating conditions, which means that the performance of trained AI models in real-world scenarios may not fully guarantee the safety and reliability of nuclear power plants. Summary of the Invention

[0005] The technical problem to be solved by this disclosure is to overcome the shortcomings of existing technologies in accurately diagnosing nuclear power plants, and to provide a method for generating accident prediction models and a method for diagnosing nuclear power plant accidents.

[0006] This disclosure solves the above-mentioned technical problems through the following technical solution:

[0007] According to a first aspect of this disclosure, a method for generating an accident prediction model is provided, wherein the accident prediction model is used to predict accident prediction information corresponding to the actual operating data of a nuclear power plant within a preset time period.

[0008] The generation method includes:

[0009] Acquire several sets of sample training data, including sample operation data of the nuclear power plant in different historical periods and the corresponding sample accident types under the sample operation data;

[0010] Based on the sample training data, several different teacher models are trained to obtain several intermediate sample models for predicting the target accident type of a nuclear power plant under the target operating data at any time period.

[0011] Different teacher models are based on processing the sample training data using different processing methods to obtain different intermediate sample models;

[0012] Knowledge distillation is performed based on several different intermediate sample models to obtain a student model, which serves as the accident prediction model.

[0013] Optionally, the teacher model includes at least one of a basic teacher model, a noise processing teacher model, and a time trend teacher model;

[0014] When the teacher model is the basic teacher model, the step of obtaining the intermediate sample model includes:

[0015] The preset training model is trained based on the sample training data to obtain the basic teacher model used to predict the type of target accident under any target running data.

[0016] When the teacher model is the noise-processing teacher model, the step of obtaining the intermediate sample model includes:

[0017] Different standards of Gaussian noise are added to the sample training data to obtain the first sample training data;

[0018] The preset training model is trained based on the first sample training data to obtain the noise processing teacher model used to predict the corresponding noise processing teacher model under any target running data.

[0019] When the teacher model is the time-trend teacher model, the step of obtaining the intermediate sample model includes:

[0020] Based on different time window requirements, the sample training data is subjected to time window smoothing processing to obtain the second sample training data.

[0021] The preset training model is trained based on the second sample training data to obtain the teacher model for predicting the time trend under any target running data.

[0022] Optionally, the step of performing knowledge distillation based on several different intermediate sample models to obtain the student model specifically includes:

[0023] Based on a preset function, generate soft-label losses corresponding to different teacher models;

[0024] The acquired sample training data is preprocessed to generate hard-labeled losses corresponding to different teacher models;

[0025] Based on the soft label loss and the hard label loss, a distillation loss function is generated;

[0026] Based on the distillation loss function, knowledge distillation is performed on different intermediate sample models to obtain the student model;

[0027] And / or,

[0028] After the step of acquiring several sets of sample training data, and before the step of training several different teacher models based on the sample training data to obtain several intermediate sample models for predicting the target accident type of a nuclear power plant under target operating data at any time period, the generation method further includes:

[0029] The sample training data is processed using a preset data processing method;

[0030] The preset data processing methods include normalization methods and / or sliding window techniques.

[0031] According to a second aspect of this disclosure, a method for detecting nuclear power plant accidents is provided, the method comprising:

[0032] Obtain actual operating data of nuclear power plants within a preset time period;

[0033] The actual operating data is input into the accident prediction model generated by the accident prediction model generation method according to the first aspect of this disclosure, so as to obtain the accident prediction information corresponding to the actual operating data of the nuclear power plant within a preset time period.

[0034] Optionally, after the step of obtaining the accident prediction information corresponding to the actual operating data of the nuclear power plant within a preset time period, the detection method further includes:

[0035] The accident prediction model is updated based on the actual operating data and the corresponding accident prediction information.

[0036] And / or,

[0037] The accident prediction information includes at least accident type information.

[0038] According to a third aspect of this disclosure, a system for generating an accident prediction model is provided, the system comprising:

[0039] The first acquisition module is used to acquire several sets of sample training data, including sample operation data of the nuclear power plant in different historical periods and the sample accident types corresponding to the sample operation data.

[0040] The first training module is used to train several different teacher models based on the sample training data to obtain several intermediate sample models for predicting the target accident type of the nuclear power plant under the target operating data at any time period.

[0041] Different teacher models are based on processing the sample training data using different processing methods to obtain different intermediate sample models;

[0042] The knowledge distillation module is used to perform knowledge distillation based on several different intermediate sample models to obtain a student model, which serves as the accident prediction model.

[0043] Optionally, the teacher model includes at least one of a basic teacher model, a noise processing teacher model, and a time trend teacher model;

[0044] When the teacher model is the basic teacher model, the first training module is used to train the preset training model based on the sample training data to obtain the basic teacher model used to predict the target accident type under any target running data.

[0045] When the teacher model is the noise processing teacher model, the first training module is used to add Gaussian noise of different standards to the sample training data to obtain the first sample training data.

[0046] The preset training model is trained based on the first sample training data to obtain the noise processing teacher model used to predict the corresponding noise processing teacher model under any target running data.

[0047] When the teacher model is the time trend teacher model, the first training module is used to perform time window smoothing on the sample training data based on different time window requirements to obtain the second sample training data.

[0048] The preset training model is trained based on the second sample training data to obtain the teacher model for predicting the time trend under any target running data.

[0049] Optionally, the knowledge distillation module is used to generate soft label losses corresponding to different teacher models based on a preset function;

[0050] The acquired sample training data is preprocessed to generate hard-labeled losses corresponding to different teacher models;

[0051] Based on the soft label loss and the hard label loss, a distillation loss function is generated;

[0052] Based on the distillation loss function, knowledge distillation is performed on different intermediate sample models to obtain the student model;

[0053] And / or,

[0054] The accident prediction model generation system also includes a data preprocessing module. The data preprocessing module is used to process the sample training data using a preset data processing method after the step of obtaining several sets of sample training data and before the step of training several different teacher models based on the sample training data to obtain several intermediate sample models for predicting the target accident type of the nuclear power plant under the target operating data at any time period.

[0055] The preset data processing methods include normalization methods and / or sliding window techniques.

[0056] According to a fourth aspect of this disclosure, a detection system for nuclear power plant accidents is provided, the detection system comprising:

[0057] The second acquisition module is used to acquire the actual operating data of the nuclear power plant within a preset time period;

[0058] The detection result acquisition module is used to input the actual operating data into the accident prediction model generated by the accident prediction model generation method based on the first aspect of this disclosure, so as to obtain the accident prediction information corresponding to the actual operating data of the nuclear power plant within a preset time period.

[0059] Optionally, the nuclear power plant accident detection system includes a model update module, which is used to update the accident prediction model based on the actual operating data and the corresponding accident prediction information after the step of obtaining the accident prediction information corresponding to the actual operating data of the nuclear power plant within a preset time period.

[0060] And / or,

[0061] The accident prediction information includes at least accident type information.

[0062] According to a fifth aspect of this disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and for running on the processor, wherein the processor executes the computer program to implement the method for generating an accident prediction model according to a first aspect of this disclosure, and / or the method for detecting nuclear power plant accidents according to a second aspect of this disclosure.

[0063] According to a sixth aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method for generating an accident prediction model as described in the first aspect of this disclosure, and / or the method for detecting nuclear power plant accidents as described in the second aspect of this disclosure.

[0064] According to a seventh aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method for generating an accident prediction model as described in the first aspect of this disclosure, and / or the method for detecting nuclear power plant accidents as described in the second aspect of this disclosure.

[0065] Based on common knowledge in the field, the above-mentioned preferred conditions can be combined arbitrarily to obtain various preferred embodiments of this disclosure.

[0066] The positive and progressive effects of this disclosure are as follows:

[0067] The accident prediction model generation method provided in this disclosure is based on knowledge distillation technology, which transfers knowledge from multiple teacher models to student models, thereby obtaining an accident prediction model that can quickly identify the types of nuclear power plant accidents. By transferring the knowledge of large and complex models (teacher models) to smaller and lighter student models through knowledge distillation, the computational efficiency and adaptability of student models are improved without significant performance loss. In addition, the accident prediction model generation method provided in this disclosure can not only bridge the gap between simulated data and real-world scenario data, but also improve the response speed and stability of the model in an environment with limited computing resources, thereby facilitating the actual deployment and online updates of nuclear power plant accident diagnosis systems.

[0068] The nuclear power plant accident diagnosis method provided in this disclosure achieves real-time and efficient diagnosis of nuclear power plant accidents based on an accident prediction model. Attached Figure Description

[0069] Figure 1 is a flowchart illustrating the method for generating an accident prediction model provided in Embodiment 1 of this disclosure;

[0070] Figure 2 is a schematic diagram of the knowledge distillation process provided in Embodiment 1 of this disclosure;

[0071] Figure 3 is a flowchart illustrating the real-time process of the nuclear power plant accident diagnosis method provided in Embodiment 1 of this disclosure.

[0072] Figure 4 is a schematic diagram of the process for constructing a teacher model provided in Embodiment 1 of this disclosure;

[0073] Figure 5 is a schematic flowchart of the knowledge distillation process provided in Embodiment 1 of this disclosure;

[0074] Figure 6 is a flowchart illustrating the method for diagnosing nuclear power plant accidents provided in Embodiment 2 of this disclosure;

[0075] Figure 7 is a schematic diagram of the structure of the accident prediction model generation system provided in Embodiment 3 of this disclosure;

[0076] Figure 8 is a schematic diagram of the structure of the nuclear power plant accident diagnosis system provided in Embodiment 4 of this disclosure;

[0077] Figure 9 is a schematic diagram of the structure of the electronic device provided in Embodiment 5 of this disclosure. Detailed Implementation

[0078] The present disclosure is further illustrated below by way of embodiments, but the present disclosure is not limited to the scope of the embodiments described herein.

[0079] The prefixes such as "first" and "second" used in this disclosure are merely for distinguishing different descriptive objects and do not limit the position, order, priority, quantity, or content of the described objects. The use of ordinal numbers and other prefixes used to distinguish descriptive objects in this disclosure does not constitute a limitation on the described objects. The description of the described objects is given in the claims or the context of the embodiments, and should not be construed as an unnecessary limitation. Furthermore, in the description of this embodiment, unless otherwise stated, "multiple" means two or more.

[0080] Example 1

[0081] In this embodiment, a method for generating an accident prediction model is provided. The accident prediction model is used to predict accident prediction information corresponding to the actual operating data of a nuclear power plant within a preset time period.

[0082] As shown in Figure 1, the generation method includes:

[0083] S11: Obtain several sets of sample training data, including sample operation data of nuclear power plants in different historical periods and the corresponding sample accident types under the sample operation data;

[0084] S12: Train several different teacher models based on sample training data to obtain several intermediate sample models for predicting the target accident type of nuclear power plant under target operating data at any time period;

[0085] Among them, different teacher models are based on processing the sample training data in different ways to obtain different intermediate sample models;

[0086] S13: Knowledge distillation is performed based on several different intermediate sample models to obtain a student model, which serves as the accident prediction model.

[0087] The accident prediction model in this embodiment also generates corresponding probability distributions for common accident types in nuclear power plants (such as Loss of Coolant Accident (LOCA), Steam Generator Tube Rupture (SGTR), and Main Steam Line Break (MSLB). During the diagnosis process, if the probability value of Loss of Coolant Accident (LOCA) is the highest, the model classifies the accident as LOCA; similarly, if the probability of Steam Generator Tube Rupture (SGTR) is the highest, the current situation is classified as SGTR.

[0088] In one implementation, during the accident diagnosis process, the probability distribution generated by the accident prediction model is used to determine the specific type of accident. The accident prediction model calculates the probability of each accident category and selects the category with the highest probability as the final prediction result. The formula for determining the accident type is as follows:

[0089] Among them, y pred The predicted accident type is the accident category that the model ultimately outputs. Let S be the probability distribution or subvector output by the "teacher model," representing the predicted probability of the student model for accident category c. In other words, S is a vector composed of the teacher model's predicted values ​​for each category, indicating the model's predicted probability for each category. argmax c To find the category with the highest probability, we select the accident category that the student model considers most likely to occur.

[0090] In a specific example, suppose the output corresponding to running data A has accident categories a1, a2, and a3, where the prediction probability for accident category a1 is 20%, the prediction probability for accident category a2 is 30%, and the prediction probability for accident category a3 is 50%. Then the accident category corresponding to running data A is a3.

[0091] In this embodiment, after identifying the accident type corresponding to the nuclear power plant through the accident prediction model, specific diagnostic results will be generated and output.

[0092] The output includes the predicted accident type, the probability distribution of each type of accident, and possible accident development trends. The results are transmitted in real time to the nuclear power plant's control center and displayed to operators through a visualization system so that they can quickly understand the current accident situation.

[0093] In addition, outputting the specific accident type can help operators quickly identify the nature of the accident, such as whether it is a coolant loss accident or a main steam pipe rupture.

[0094] This embodiment also displays the probability distribution values ​​of the accident prediction model for different accident types, allowing operators to assess the reliability and severity of the accident type. High probability values ​​typically indicate a high level of confidence in the model for that accident type. Furthermore, if the accident exhibits continuous changes or shows a certain trend, the diagnostic system will provide trend predictions to help operators develop preventative measures in advance. For example, if pressure data rises rapidly within a short period, the system will alert operators to a potential risk of coolant loss.

[0095] Decision support information will provide corresponding response recommendations based on the type and severity of the accident, enabling operators to make quick decisions and reduce the impact of the accident on the nuclear power plant. Simultaneously, the diagnostic process and results will be recorded for subsequent analysis and optimization of the diagnostic model.

[0096] Furthermore, since data in the early stages of a nuclear power plant accident may be incomplete or highly variable, the accident type determination process incorporates a dynamic update mechanism. This involves sampling the input data multiple times over time and making real-time judgments at each time step. If the predicted probability of a certain type of accident is consistently at its highest, the diagnostic system adopts this result as the final determination, thereby reducing the impact of instantaneous data fluctuations on the diagnostic outcome.

[0097] The accident prediction model generation method provided in this disclosure is based on knowledge distillation technology, which transfers knowledge from multiple teacher models to student models, thereby obtaining an accident prediction model that can quickly identify the types of nuclear power plant accidents. By transferring the knowledge of large and complex models (teacher models) to smaller and lighter student models through knowledge distillation, the computational efficiency and adaptability of student models are improved without significant performance loss. In addition, the accident prediction model generation method provided in this disclosure can not only bridge the gap between simulated data and real-world scenario data, but also improve the response speed and stability of the model in an environment with limited computing resources, thereby facilitating the actual deployment and online updates of nuclear power plant accident diagnosis systems.

[0098] The teacher model in this embodiment includes at least one of the following: a basic teacher model, a noise processing teacher model, and a time trend teacher model.

[0099] When the teacher model is the base teacher model, the steps to obtain the intermediate sample model include:

[0100] The preset training model is trained based on the sample training data to obtain a basic teacher model for predicting the type of target accident under any target running data.

[0101] When the teacher model is a noisy teacher model, the steps to obtain the intermediate sample model include:

[0102] Different standards of Gaussian noise were added to the sample training data to obtain the first sample training data;

[0103] The preset training model is trained based on the first sample training data to obtain a teacher model for predicting noise processing under any target running data.

[0104] When the teacher model is a time-trend teacher model, the steps to obtain the intermediate sample model include:

[0105] Based on different time window requirements, the sample training data is smoothed by time window processing to obtain the second sample training data.

[0106] The preset training model is trained based on the second sample training data to obtain a teacher model for predicting the corresponding time trend under any target running data.

[0107] In one specific implementation, the teacher model is used to learn different characteristics from full-range simulator data of a nuclear power plant. This disclosure captures data characteristics under different operating conditions of a nuclear power plant by constructing multiple teacher models, wherein the teacher models include a basic teacher model, a noise-processing teacher model, and a time-trend teacher model.

[0108] (1) Basic Teacher Model (Teacher 1): A bidirectional long short-term memory network (Bi-LSTM) is used to learn the basic features of nuclear power plant simulation data. The training objective of this model is to minimize the classification loss function.

[0109] Among them, L base This represents the loss function of the base teacher model, used to evaluate the model's classification accuracy on the training dataset.

[0110] N represents the number of samples, which is the total number of samples used during training.

[0111] y i This represents the true label (hard label) of the i-th sample, usually represented by 0 or 1.

[0112] Let represent the model's predicted probability for the i-th sample, and let represent the probability that the model believes the sample belongs to a certain category.

[0113] (2) Noise-handling Teacher Model (Teacher 2): Based on the basic teacher model, Gaussian noise with different standard deviations is added to the training data to enhance the model's robustness to noise. Its training objective is to minimize the following loss function:

[0114] Among them, Lnoise This represents the loss function of the noise-handling teacher model, used to improve the model's robustness in noisy environments.

[0115] λ represents the regularization parameter, which controls the weight of the noise term in the loss function.

[0116] ε j This represents Gaussian noise added to the training data to simulate uncertainty in the real environment.

[0117] M represents the number of noise samples, the number of samples with added noise during training.

[0118] (3) Time-Trend Teacher Model (Teacher 3): This model utilizes moving averages preprocessed with different time window sizes (e.g., 60s, 120s, 180s) to better capture time-dependent characteristics. The loss function for this model is:

[0119] Among them, L trend is the loss function for the time-trend teacher model, used to capture trends and dependencies in time series data. μ is the time smoothing coefficient, used to adjust the impact of the time trend term on the loss function. T is the total number of steps in the time series, representing the span of the data in the time dimension. Let y be the model prediction value at time t, representing the model's prediction result for the current time. t-1 The true value at time t-1 represents the data from the previous time.

[0120] This disclosure demonstrates strong robustness to noisy and time-dependent data by processing input data and combining knowledge distilled from multiple teacher models. It can identify early characteristics of accident types and generate corresponding probability distributions. By learning from nuclear power plant simulation data through multiple teacher models and transferring knowledge from complex models to lightweight student models, the student models can still effectively diagnose accidents even under computationally limited conditions.

[0121] As shown in Figure 2, the steps for knowledge distillation based on several different intermediate sample models to obtain the student model include:

[0122] S21: Based on a preset function, generate soft-label loss corresponding to different teacher models;

[0123] S22: Preprocess the acquired sample training data to generate hard-label loss corresponding to different teacher models;

[0124] S23: Generate a distillation loss function based on soft label loss and hard label loss;

[0125] S24: Based on the distillation loss function, knowledge distillation is performed on different intermediate sample models to obtain the student model.

[0126] In one specific implementation, the teacher model's output to the input data is typically a probability distribution. Smoother, softer labels are generated by increasing the distillation temperature T.

[0127] Where, σ(z) i ,T) is a temperature-adjusted softmax function (a function that converts a real vector into a probability distribution) used to generate soft tags.

[0128] z i Here, represents the logit value (logistic regression value) predicted by the model for category i, indicating the score before softmax transformation. T is the distillation temperature, used to control the smoothness of the generated probability distribution. The higher the temperature, the smoother the generated probability distribution.

[0129] The student model is trained by combining two loss functions: hard loss and soft loss. These two loss functions are derived from the extracted knowledge. The hard loss is calculated using real labels consistent with the base teacher model. The hard loss ensures that the student model directly learns to predict the correct category labels.

[0130] The soft labels from the teacher model are used as the learning objectives of the student model, and combined with the hard labels to form the distillation loss function:

[0131] Among them, L distill This is represented as the distillation loss function, used to guide the student model in learning the prediction distribution of the teacher model. α is a weighting coefficient used to balance the effects of soft-label loss and hard-label loss. Kullback-Leibler divergence (also known as relative entropy or information divergence, is an asymmetric measure of the difference between two probability distributions) is used to measure the difference between two probability distributions. The soft-label probability distribution generated for the teacher model. The probability distribution of soft labels generated for the student model. Let be the predicted probability of the student model for the i-th sample.

[0132] In this embodiment, the student model is trained using a distillation loss function, enabling it to learn the rich knowledge from the teacher model and improve its performance in real-world environments.

[0133] Furthermore, the student model achieves a lightweight design through knowledge distillation, enabling rapid response in nuclear power plant accident diagnosis. Specific optimization steps include architecture simplification, temperature adjustment, and model deployment.

[0134] (1) Simplified architecture

[0135] The student model employs a three-layer bidirectional long short-term memory network (Bi-LSTM), reducing the number of layers and hidden units by 50% compared to the teacher model. Its loss function is: L student =βL distill +(1-β)L student,task ;

[0136] Among them, L student is the total loss function for the student model, used to optimize its performance during training. β is the loss balancing coefficient, used to adjust the weights of the distillation loss and the task loss. student，task The task loss for the student model, also known as cross-entropy loss, is used to ensure that the student model can classify correctly.

[0137] (2) Adjust the distillation temperature

[0138] To improve the student model's ability to mimic the teacher model's predictions, the distillation temperature T was optimized to achieve an optimal balance between the smoothness of the soft label and the student model's learning ability.

[0139] (3) Model Deployment

[0140] Once the trained student model is deployed to the nuclear power plant accident diagnosis system, the student model can quickly identify accident types and update online with limited computing resources.

[0141] Knowledge distillation simplifies the architecture of the student model, significantly reducing its parameter count and training and inference time compared to the teacher model. This lightweight design allows the student model to operate stably in computationally limited environments such as nuclear power plants, supporting online updates and rapid deployment. Through knowledge distillation, the student model not only learns the output information of the teacher model but also extracts the decision-making basis from it.

[0142] This disclosure enables the student model to better adapt to the data distribution of real-world scenarios through knowledge distillation, thereby improving diagnostic accuracy under realistic conditions. Furthermore, through distillation, the knowledge of the teacher model is successfully compressed into a lightweight student model, significantly reducing the computational resource requirements while maintaining high performance. In addition, different types of teacher models are used for training in the teacher model to address noise and temporal trends in the data, and knowledge distillation enables the student model to cope with these changes, thus improving the robustness and stability of the model.

[0143] In this embodiment, after the step of acquiring several sets of sample training data, and before the step of training several different teacher models based on the sample training data to obtain several intermediate sample models for predicting the target accident type of a nuclear power plant under the target operating data at any time period, the generation method further includes:

[0144] The sample training data is processed using a preset data processing method;

[0145] The preset data processing methods include normalization methods and / or sliding window techniques.

[0146] In one specific implementation, real-time operational data is acquired from the nuclear power plant, including various state variables such as temperature, pressure, flow rate, and radiation levels. This data is collected in real-time by a sensor system, reflecting the characteristics of the nuclear power plant under different operating conditions and forming the basis for accurate analysis by the diagnostic system.

[0147] Because the different sensors used to collect data from nuclear power plants have varying dimensions and data ranges, direct input may cause the model to become overly sensitive to features with particularly large values. Therefore, data preprocessing ensures that the input data is consistent and standardized, further improving the accuracy of the model.

[0148] This disclosure employs a 0-to-1 normalization method to map the data to a standardized range, thereby eliminating scale differences between feature values ​​and making the model's influence on each feature more balanced. The normalization formula is as follows:

[0149] In the preprocessing stage, the time series data can be further processed using the sliding window technique. The sliding window technique enables the model to capture the temporal dependencies of the data during the diagnostic process, such as monitoring pressure change trends or temperature rise rates over a period of time to infer the potential occurrence of accidents. For time series data, a sliding window of size ω is used for preprocessing to generate input data containing multiple time steps, in order to better capture the dynamic characteristics of accident evolution.

[0150] By preprocessing the acquired data, the accuracy of nuclear power plant accident assessments has been further improved.

[0151] The preprocessed data is input into the optimized student model, and the probability distribution of various accidents is output through the forward propagation process of the neural network. The model inference formula is as follows:

[0152] Let x be the prediction result of the student model for the input data, and let y represent the classification probability distribution of the input sample by the student model. Let θ be the function mapping for the student model, representing the prediction process of the student model. s This is the set of parameters for the student model, representing the weights and biases learned during training.

[0153] The following example illustrates the implementation principle of the accident prediction model generation method in this embodiment:

[0154] As shown in Figure 3, this disclosure proposes a method for generating an accident prediction model, which achieves efficient diagnosis of nuclear power plant accidents by combining the design of teacher and student models.

[0155] The architecture of the accident prediction model provided in this disclosure includes the following modules: data processing module, teacher model module, knowledge distillation module, student model module, and result output module.

[0156] The data processing module is responsible for receiving real-time status data from sensors at the nuclear power plant, including temperature, pressure, flow rate, and radiation levels. This real-time status data can be acquired through the nuclear power plant's sensors.

[0157] The data processing module preprocesses the input data, primarily through normalization and sliding window processing, to ensure the standardization and consistency of the input data. Normalization maps data of different dimensions to the range of 0 to 1, eliminating scale differences between data. The sliding window is used to capture time-series characteristics, enabling the model to better understand how state data changes over time.

[0158] The teacher model module is used to learn from nuclear power plant simulation data using different teacher models, including a basic teacher model, a noise treatment teacher model, and a time trend teacher model. Each teacher model captures different characteristics of the data: the basic teacher model learns the main accident features, the noise treatment model enhances the system's robustness to noise data, and the time trend model is used to capture the time dependencies of the data.

[0159] In this embodiment, the teacher module includes at least a basic teacher model, a noise processing teacher model, and a time trend teacher model; and the different models in the teacher module are trained using data simulated by the nuclear power plant full-range simulator and / or data obtained by the data processing module to obtain the target model.

[0160] The knowledge distillation module is used to distill the knowledge learned in the teacher model into a lightweight student model. The process includes generating soft labels and calculating the distillation loss. The soft labels, output by the teacher model, are generated using a temperature-adjusted softmax function and serve as the learning objectives for the student model. The distillation loss combines the soft label loss and the hard label loss to achieve knowledge transfer.

[0161] The student model is a lightweight bidirectional long short-term memory network (Bi-LSTM) that learns knowledge from the teacher model through knowledge distillation, enabling efficient real-time diagnostic capabilities. The student model receives processed data and quickly identifies the type of incident.

[0162] The diagnostic results from the student model are transmitted to the results output module. The results output module is used to feed back the identified accident type and its probability distribution value to the nuclear power plant control center, and to display the diagnostic results and development trends through a visual interface, providing decision support for operators.

[0163] The construction and training process of the teacher model is shown in Figure 4. The teacher model captures the diverse characteristics of nuclear power plant data through different structures. The specific training steps for each model are as follows:

[0164] 1. Basic Teacher Model Training

[0165] The training process for the basic teacher model is as follows: after obtaining the sample training data, the sample training data is not processed in any way. Simulation data is input, feature learning is performed, the model is optimized, and the model is output.

[0166] The basic teacher model, based on a bidirectional long short-term memory (Bi-LSTM) network structure, is used to learn key accident features from nuclear power plant simulation data. The model's training data comes from nuclear power plant simulation platforms (such as the CNS (Central Nervous System) dataset), covering various accident types and their corresponding operating parameters. The training objective of the basic teacher model is to minimize the cross-entropy loss, with the loss function being:

[0167] 2. Noise Processing Teacher Model Training

[0168] The training process for the noise processing teacher model is as follows: after obtaining the sample training data, Gaussian noise is added to the sample training data, simulation data is input, feature learning is performed, the model is optimized, and the model is output.

[0169] To enhance the stability of the diagnostic system with noisy data, the noise processing teacher model incorporates Gaussian noise of varying standard deviations (e.g., 0.1, 0.2, 0.3) into the training data to simulate fluctuations and anomalies in actual operating data. The loss function of the noise processing model includes a regularization term to limit the impact of noise on the model.

[0170] 3. Time Trend Teacher Model Training

[0171] The training process of the time trend teacher model is as follows: after obtaining the sample training data, the sample training data is smoothed by adding a time window, the simulation data is input, feature learning is performed, the model is optimized, and the model is output.

[0172] The time-trend teacher model is used to capture the time dependence of data. Moving averages are applied through different time windows (e.g., 60 seconds, 120 seconds, and 180 seconds) to enable the model to perceive time trends. The model's loss function includes a time smoothing term:

[0173] As shown in Figure 5, the knowledge distillation process transfers knowledge from the teacher model to the student model through a distillation loss function. The specific steps are as follows:

[0174] 1. Generating soft tags

[0175] The soft tags in the knowledge distillation process are generated by the output of the teacher model (including the basic teacher model, the noisy teacher model, and the time-trend teacher model) and a softmax function adjusted for temperature parameters. The specific formula is as follows:

[0176] 2. Calculation of distillation loss

[0177] The distillation loss function includes soft-label loss and hard-label loss, which are defined as follows:

[0178] IV. Diagnostic Process for Student Models

[0179] After being trained through knowledge distillation, the student model is able to perform real-time accident diagnosis in nuclear power plants. The specific diagnostic steps are as follows:

[0180] 1. Data Acquisition and Preprocessing

[0181] Real-time data is acquired from sensors at nuclear power plants and normalized to eliminate differences in feature scale. Furthermore, a sliding window approach is used to convert the data into time-series input, enhancing the student model's understanding of time-dependent characteristics.

[0182] 2. Model Inference

[0183] The preprocessed data is input into the student model. Through forward propagation, the student model makes probability predictions for different accident types, generating probability distributions for each type of accident.

[0184] 3. Accident type determination

[0185] The student model determines the accident type with the highest probability based on the probability distribution as the prediction result:

[0186] 4. Results Output and Decision Support

[0187] The diagnostic results from the student model are output to the nuclear power plant control center and displayed on a visual interface via the results output module, enabling operators to promptly grasp the accident situation. The results output includes the following information:

[0188] Accident type: Displays the accident type predicted by the model, such as Loss of Coolant Accident (LOCA), Steam Generator Tube Rupture (SGTR), or Main Steam Tube Rupture (MSLB).

[0189] Probability distribution: Displays the model's predicted probability values ​​for various types of accidents, helping operators understand the severity of the accident and the system's diagnostic confidence.

[0190] Time trend analysis: During the diagnosis process, the system analyzes the time trend information of the accident type, such as the rate of change of key parameters, to determine whether the accident has the potential to develop further.

[0191] This information is transmitted to the nuclear power plant control center via the results output module to help operators respond quickly to accident situations, take appropriate protective measures, and ensure the safety of the nuclear power plant.

[0192] To address changes in the operating environment of nuclear power plants and equipment aging, the student model needs to be updated and optimized regularly. This disclosure also employs an incremental learning strategy to ensure the diagnostic system maintains accuracy and adaptability during long-term operation.

[0193] 1. Knowledge Update

[0194] During operation, the system periodically collects the latest simulation data and historical accident data from nuclear power plants to retrain the teacher model. The new teacher model then transfers knowledge to the student model through a knowledge distillation process, ensuring that the student model can make judgments based on the latest accident characteristics.

[0195] 2. Incremental learning

[0196] To minimize the impact of model updates on system operation, this disclosure employs an incremental learning approach, requiring only fine-tuning with a small amount of new data. The advantage of incremental learning is that it eliminates the need for complete model retraining; adjustments are made only to the features of the new data, ensuring real-time model performance.

[0197] 3. Performance Evaluation and Optimization

[0198] After updating the model, the system will use an offline dataset to evaluate its performance, verifying whether the student model meets the accuracy and robustness standards. If the expected diagnostic results are achieved, the optimized model will be redeployed to the nuclear power plant's accident diagnosis system, achieving a seamless switchover.

[0199] Through the aforementioned update and optimization mechanisms, this disclosure can adapt to changes in the dynamic operating environment of nuclear power plants, further enhancing the long-term stability and reliability of the system.

[0200] This disclosure utilizes knowledge distillation technology to transfer knowledge from multiple teacher models to student models. Combined with a bidirectional long short-term memory (Bi-LSTM) network structure, it enables real-time and efficient diagnosis of nuclear power plant accidents.

[0201] The specific technical advantages of this disclosure are as follows:

[0202] 1. High diagnostic accuracy

[0203] By introducing multiple teacher models, the student model demonstrates high accuracy in different accident scenarios, especially when dealing with noisy data and time series data, it can fully understand the characteristics of accident data and improve diagnostic accuracy.

[0204] 2. High computational efficiency

[0205] The student model achieves a lightweight design after knowledge distillation, reducing the number of network layers and parameters. It can run efficiently in environments with limited computing resources, making it suitable for the real-time accident diagnosis needs of nuclear power plants.

[0206] 3. Strong adaptability and robustness

[0207] The student model disclosed herein continuously adapts to environmental changes through incremental learning, and with the support of a noise-processing teacher model and a time-trend teacher model, it exhibits strong robustness and can maintain stable performance in complex environments.

[0208] 4. Visualized decision support

[0209] The diagnostic results are presented in a user-friendly and intuitive manner through the visualization of the results output module, which helps operators make quick and reasonable response decisions in the event of an accident, thereby improving the safety assurance capabilities of nuclear power plants.

[0210] The implementation of this disclosure has yielded significant results, providing a practical and reliable diagnostic method for nuclear power plant accidents. It can effectively bridge the gap between simulated data and real data, and provide strong technical support for the safe operation of nuclear power plants.

[0211] Example 2

[0212] As shown in Figure 6, this embodiment provides a method for detecting nuclear power plant accidents, characterized in that the detection method includes:

[0213] S61: Obtain actual operating data of the nuclear power plant within a preset time period;

[0214] S62: Input the actual operating data into the accident prediction model generated by the accident prediction model generation method to obtain the accident prediction information corresponding to the actual operating data of the nuclear power plant within a preset time period.

[0215] The nuclear power plant accident diagnosis method provided in this disclosure achieves real-time and efficient diagnosis of nuclear power plant accidents based on an accident prediction model.

[0216] In this embodiment, after obtaining the accident prediction information corresponding to the actual operating data of the nuclear power plant within a preset time period, the detection method further includes:

[0217] The accident prediction model is updated based on actual operational data and corresponding accident prediction information; the accident prediction information in this embodiment includes at least accident type information.

[0218] In this embodiment, after the step of inputting real-time operational data into the target prediction model to output real-time accident information of the nuclear power plant, the diagnostic method further includes:

[0219] The target prediction model is updated based on real-time operational data and real-time accident information.

[0220] In one specific implementation, due to the complex operating environment of nuclear power plants and the changing data characteristics caused by equipment aging and external environmental changes, the student model of the accident diagnosis system needs to be updated and optimized regularly to ensure its adaptability to real-time data. This disclosure updates and optimizes the model in the following aspects:

[0221] Knowledge Update: The teacher model is retrained using newly collected simulation data and historical accident data to acquire the latest knowledge. This knowledge is then re-distilled into the student model to ensure that the student model can learn the latest accident characteristics and environmental changes.

[0222] Parameter tuning: The distillation temperature parameter and loss function weights in the student model may need to be dynamically adjusted over time to adapt to new operational data characteristics. For example, in a new round of training, the distillation temperature can be experimentally adjusted to optimize the smoothness of the soft labels, thereby improving the generalization ability of the student model.

[0223] Incremental learning: To minimize the impact of model updates on the normal operation of nuclear power plants, the accident diagnosis system employs incremental learning to make small updates to the model. During incremental learning, only a small amount of new data is added to fine-tune the model, without requiring a complete retraining of the entire model, thus ensuring the real-time performance and stability of the diagnosis system.

[0224] Model performance evaluation and optimization: After the model is updated, its performance is evaluated using offline data to ensure that the accuracy and robustness of the updated model meet the set standards. If the performance meets the requirements, the updated student model is deployed to the diagnostic system to achieve continuous optimization of real-time accident monitoring and judgment at nuclear power plants.

[0225] By updating and optimizing the target prediction model, the accuracy of the target prediction model in predicting nuclear power plant accidents has been further improved.

[0226] Example 3

[0227] As shown in Figure 7, this embodiment provides a system for generating an accident prediction model. The system includes:

[0228] The first acquisition module 701 is used to acquire several sets of sample training data, including sample operation data of nuclear power plants in different historical periods and the corresponding sample accident types under the sample operation data.

[0229] The first training module 702 is used to train several different teacher models based on sample training data to obtain several intermediate sample models for predicting the target accident type of the nuclear power plant under the target operating data at any time period.

[0230] Among them, different teacher models are based on processing the sample training data in different ways to obtain different intermediate sample models;

[0231] The knowledge distillation module 703 is used to perform knowledge distillation based on several different intermediate sample models to obtain a student model, which serves as an accident prediction model.

[0232] The teacher model in this embodiment includes at least one of the following: a basic teacher model, a noise processing teacher model, and a time trend teacher model.

[0233] When the teacher model is the basic teacher model, the first training module is used to train the preset training model based on the sample training data to obtain the basic teacher model for predicting the target accident type under any target running data.

[0234] When the teacher model is a noise-processing teacher model, the first training module is used to add Gaussian noise of different standards to the sample training data to obtain the first sample training data.

[0235] The preset training model is trained based on the first sample training data to obtain a teacher model for predicting noise processing under any target running data.

[0236] When the teacher model is a time-trend teacher model, the first training module is used to perform time window smoothing on the sample training data based on different time window requirements to obtain the second sample training data.

[0237] The preset training model is trained based on the second sample training data to obtain a teacher model for predicting the corresponding time trend under any target running data.

[0238] In this embodiment, the knowledge distillation module is used to generate soft-label loss corresponding to different teacher models based on a preset function;

[0239] The acquired sample training data is preprocessed to generate hard-labeled losses for different teacher models;

[0240] A distillation loss function is generated based on soft label loss and hard label loss;

[0241] Knowledge distillation is performed on different intermediate sample models based on the distillation loss function to obtain the student model;

[0242] The accident prediction model generation system also includes a data preprocessing module 704. The data preprocessing module 704 is used to process the sample training data using a preset data processing method before the step of training several different teacher models based on the sample training data to obtain several intermediate sample models for predicting the target accident type of the nuclear power plant under the target operating data at any time period after obtaining several sets of sample training data.

[0243] The preset data processing methods include normalization methods and / or sliding window techniques.

[0244] The accident prediction model generation system provided in this disclosure uses knowledge distillation technology to transfer knowledge from multiple teacher models to student models, thereby obtaining an accident prediction model capable of quickly identifying nuclear power plant accident types. By transferring knowledge from large and complex models (teacher models) to smaller and lighter student models through knowledge distillation, the computational efficiency and adaptability of student models are improved without significant performance loss. Furthermore, the accident prediction model generation method provided in this disclosure not only bridges the gap between simulated data and real-world scenario data, but also improves the model's response speed and stability in environments with limited computing resources, thus facilitating the practical deployment and online updates of nuclear power plant accident diagnosis systems.

[0245] For the system embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The system embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs.

[0246] Example 4

[0247] As shown in Figure 8, this embodiment provides a nuclear power plant accident detection system, which includes:

[0248] The second acquisition module 801 is used to acquire the actual operating data of the nuclear power plant within a preset time period;

[0249] The detection result acquisition module 802 is used to input the actual operating data into the accident prediction model generated by the accident prediction model generation method based on the first aspect of this disclosure, so as to obtain the accident prediction information corresponding to the actual operating data of the nuclear power plant within a preset time period.

[0250] The nuclear power plant accident detection system in this embodiment includes a model update module 803. After obtaining the accident prediction information corresponding to the actual operating data of the nuclear power plant within a preset time period, the model update module 803 updates the accident prediction model based on the actual operating data and the corresponding accident prediction information.

[0251] Accident prediction information should include at least accident type information.

[0252] The nuclear power plant accident diagnosis system provided in this disclosure achieves real-time and efficient diagnosis of nuclear power plant accidents based on an accident prediction model.

[0253] For the system embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The system embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs.

[0254] Example 5

[0255] Figure 9 is a schematic diagram of the structure of an electronic device provided in Embodiment 5 of this disclosure. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the methods described in the above embodiments. The electronic device 30 shown in Figure 9 is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this disclosure.

[0256] As shown in Figure 9, the electronic device 30 can be represented in the form of a general computing device, such as a server device. The components of the electronic device 30 may include, but are not limited to: at least one processor 31, at least one memory 32, and a bus 33 connecting different system components (including memory 32 and processor 31).

[0257] Bus 33 includes a data bus, an address bus, and a control bus.

[0258] The memory 32 may include volatile memory, such as random access memory (RAM) 321 and / or cache memory 322, and may further include read-only memory (ROM) 323.

[0259] The memory 32 may also include a program / utility 325 having a set (at least one) of program modules 324, including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0260] The processor 31 performs various functional applications and data processing, such as the methods described in the above embodiments of this disclosure, by running computer programs stored in the memory 32.

[0261] Electronic device 30 can also communicate with one or more external devices 34 (e.g., keyboard, pointing device, etc.). This communication can be performed via input / output (I / O) interface 35. Furthermore, the model-generated device 30 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public network, such as the Internet) via network adapter 36. As shown in Figure 9, network adapter 36 communicates with other modules of the model-generated device 30 via bus 33. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with the model-generated device 30, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems.

[0262] It should be noted that although several units / modules or sub-units / modules of the electronic device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0263] Example 6

[0264] This disclosure also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method for generating the accident prediction model provided in Embodiment 1 or the method for diagnosing nuclear power plant accidents provided in Embodiment 2.

[0265] The readable storage medium may be more specifically adopted, including but not limited to: portable disk, hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.

[0266] Example 7

[0267] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the method for generating an accident prediction model provided in Embodiment 1 or the method for diagnosing nuclear power plant accidents provided in Embodiment 2.

[0268] The program code for executing the computer program product disclosed herein can be written in any combination of one or more programming languages. The program code can be executed entirely on a user device, partially on a user device, as a stand-alone software package, partially on a user device and partially on a remote device, or entirely on a remote device.

[0269] While specific embodiments of this disclosure have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of this disclosure is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of this disclosure, but all such changes and modifications fall within the scope of protection of this disclosure.

Claims

1. A method for generating an accident prediction model, characterized in that, The accident prediction model is used to predict accident information corresponding to the actual operating data of a nuclear power plant within a preset time period. The generation method includes: Acquire several sets of sample training data, including sample operation data of the nuclear power plant in different historical periods and the corresponding sample accident types under the sample operation data; Based on the sample training data, several different teacher models are trained to obtain several intermediate sample models for predicting the target accident type of a nuclear power plant under the target operating data at any time period. Different teacher models are based on processing the sample training data using different processing methods to obtain different intermediate sample models; Knowledge distillation is performed based on several different intermediate sample models to obtain a student model, which serves as the accident prediction model.

2. The method for generating an accident prediction model according to claim 1, characterized in that, The teacher model includes at least one of the following: a basic teacher model, a noise processing teacher model, and a time trend teacher model. When the teacher model is the basic teacher model, the step of obtaining the intermediate sample model includes: The preset training model is trained based on the sample training data to obtain the basic teacher model used to predict the type of target accident under any target running data. When the teacher model is the noise-processing teacher model, the step of obtaining the intermediate sample model includes: Different standards of Gaussian noise are added to the sample training data to obtain the first sample training data; The preset training model is trained based on the first sample training data to obtain the noise processing teacher model used to predict the corresponding noise processing teacher model under any target running data. When the teacher model is the time-trend teacher model, the step of obtaining the intermediate sample model includes: Based on different time window requirements, the sample training data is subjected to time window smoothing processing to obtain the second sample training data. The preset training model is trained based on the second sample training data to obtain the teacher model for predicting the time trend under any target running data.

3. The method for generating an accident prediction model according to claim 1 or 2, characterized in that, The step of performing knowledge distillation based on several different intermediate sample models to obtain the student model specifically includes: Based on a preset function, generate soft-label losses corresponding to different teacher models; The acquired sample training data is preprocessed to generate hard-labeled losses corresponding to different teacher models; Based on the soft label loss and the hard label loss, a distillation loss function is generated; Based on the distillation loss function, knowledge distillation is performed on different intermediate sample models to obtain the student model; And / or, After the step of acquiring several sets of sample training data, and before the step of training several different teacher models based on the sample training data to obtain several intermediate sample models for predicting the target accident type of a nuclear power plant under target operating data at any time period, the generation method further includes: The sample training data is processed using a preset data processing method; The preset data processing methods include normalization methods and / or sliding window techniques.

4. A method for detecting nuclear power plant accidents, characterized in that, The detection method includes: Obtain actual operating data of nuclear power plants within a preset time period; The actual operating data is input into the accident prediction model generated by the accident prediction model generation method according to any one of claims 1 to 3, so as to obtain the accident prediction information corresponding to the actual operating data of the nuclear power plant within a preset time period.

5. The method for detecting nuclear power plant accidents according to claim 4, characterized in that, After the step of obtaining the accident prediction information corresponding to the actual operating data of the nuclear power plant within a preset time period, the detection method further includes: The accident prediction model is updated based on the actual operating data and the corresponding accident prediction information. And / or, The accident prediction information includes at least accident type information.

6. A system for generating accident prediction models, characterized in that, The generation system includes: The first acquisition module is used to acquire several sets of sample training data, including sample operation data of the nuclear power plant in different historical periods and the sample accident types corresponding to the sample operation data. The first training module is used to train several different teacher models based on the sample training data to obtain several intermediate sample models for predicting the target accident type of the nuclear power plant under the target operating data at any time period. Different teacher models are based on processing the sample training data using different processing methods to obtain different intermediate sample models; The knowledge distillation module is used to perform knowledge distillation based on several different intermediate sample models to obtain a student model, which serves as the accident prediction model.

7. A detection system for nuclear power plant accidents, characterized in that, The detection system includes: The second acquisition module is used to acquire the actual operating data of the nuclear power plant within a preset time period; The detection result acquisition module is used to input the actual operating data into the accident prediction model generated by the accident prediction model generation method according to any one of claims 1 to 3, so as to obtain the accident prediction information corresponding to the actual operating data of the nuclear power plant within a preset time period.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and for running on the processor, characterized in that, When the processor executes the computer program, it implements the method for generating the accident prediction model according to any one of claims 1 to 3, and / or the method for detecting nuclear power plant accidents according to claim 4 or 5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for generating the accident prediction model according to any one of claims 1 to 3, and / or the method for detecting nuclear power plant accidents according to claim 4 or 5.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for generating the accident prediction model according to any one of claims 1 to 3, and / or the method for detecting nuclear power plant accidents according to claim 4 or 5.

Images