Nuclear power fault diagnosis and transient prediction method and system based on generative adversarial network

Abstract

The present invention discloses a method and system for nuclear power fault diagnosis and transient prediction based on a generative adversarial network, and relates to the field of nuclear power plant safety technology. The key points of the technical solution are: randomly generating a key state parameter set according to the transient type and the nuclear power plant location where the transient type may appear; taking the key state parameter set as input and different transient types at different locations as output, a discriminant model is obtained through preliminary training; a neural network model for nuclear power plant fault diagnosis and transient prediction is constructed based on a naive generative adversarial network framework; combined with real-time monitoring of nuclear power plant parameters, a neural network model is used to perform fault diagnosis and transient consequence prediction on various transient abnormal conditions that occur during the operation of the nuclear power plant. The present invention can quickly perform fault diagnosis on abnormal operating conditions after detecting abnormalities in the operating state of the nuclear power plant, identify transient types, and determine the location where transients occur, and then quickly predict the consequences of the accident, thereby achieving the purpose of reducing transients or accidents in nuclear reactors.

Description

Technical Field

[0001] The present invention relates to the field of nuclear power plant safety technology, and more specifically, to a nuclear power fault diagnosis and transient prediction method and system based on a generative adversarial network. Background Art

[0002] Transient safety analysis and confrontation of nuclear power plants has always been an important direction for the development of nuclear energy. For traditional nuclear power plants, if an accident transient occurs during its operation, the operator cannot immediately know the type of accident and the location of the accident, let alone the consequences of the accident. The operator can only wait for the relevant protection system of the nuclear reactor to trigger the signal and then perform relevant operations to confront the accident. This passive response method has high requirements for operators, short response time after the accident, and it is difficult to carry out relevant response measures in the early stage of the accident. In order to ensure the safety of nuclear reactors after an accident, most commercial nuclear power plants have set up redundant safety. For example, Hualong One can ensure the safety of nuclear reactors without human intervention 30 minutes after the accident, but this safety setting will increase the construction cost of the power plant and reduce economic efficiency.

[0003] In order to quickly determine the type and location of anomalies after an abnormality occurs in a nuclear power plant, quickly predict the consequences, and provide guidance to operators, it is necessary to develop relevant technologies to monitor the status of the power plant in real time. At present, there are some technologies at home and abroad that use substitution models, simple feedback neural networks, etc., and use the main parameters of the nuclear power plant operation process to conduct fault diagnosis of Class 3 and 4 accidents in nuclear power plants. The main application objects of these studies are concentrated on the design basis accidents of nuclear power plants. These accidents have obvious parameter changes and the accident process is relatively fixed. Therefore, using relatively simple technologies can achieve better prediction results.

[0004] However, in the actual operation of nuclear power plants, most of the time the nuclear power plants are in normal operation and transient operation state, with a small probability of expected operation events, also called medium-frequency events, and the probability of Class III and IV design basis accidents is extremely small, so the detection and diagnosis of such accidents is of little practical value. The monitoring and diagnosis of normal operation, transient operation and expected operation events of nuclear power plants are mainly due to instrument measurement fluctuation interference, slow parameter changes after transients occur, and similar transients of accidents at different locations. Conventional methods are difficult to perform accurate fault diagnosis and transient prediction. Therefore, how to study and design a nuclear power fault diagnosis and transient prediction method and system based on generative adversarial networks that can overcome the above defects is an urgent problem that we need to solve. Summary of the invention

[0005] In order to address the deficiencies in the prior art, the purpose of the present invention is to provide a nuclear power fault diagnosis and transient prediction method and system based on a generative adversarial network, which can accurately perform fault diagnosis and transient consequence prediction when various transient abnormal conditions occur during the operation of a nuclear power plant, and provide guidance for the confrontation with accident transients.

[0006] The above technical objectives of the present invention are achieved through the following technical solutions:

[0007] In the first aspect, a method for nuclear power fault diagnosis and transient prediction based on a generative adversarial network is provided, comprising the following steps:

[0008] Randomly generate a set of key state parameters based on transient types and nuclear power plant locations where transient types may occur;

[0009] With the key state parameter set as input and different transient types at different locations as output, the discriminant model is initially trained;

[0010] Constructing a neural network model for nuclear power plant fault diagnosis and transient prediction based on the naive generative adversarial network framework;

[0011] Combined with real-time monitoring of nuclear power plant parameters, neural network models are used to perform fault diagnosis and transient consequence prediction for various transient abnormal situations that occur during the operation of nuclear power plants.

[0012] Furthermore, the generation process of the key state parameter set is specifically as follows:

[0013] Determine different transient types based on the initial event library of the nuclear power plant;

[0014] A large number of different transient types at different locations are randomly generated by sampling methods;

[0015] The calculations were performed using a thermal-hydraulic program, outputting the set of key state parameters required to build the neural network.

[0016] Furthermore, the key status parameter set includes primary circuit pressure, pressurizer pressure, reactor power, hot spot cladding temperature, active zone collapse liquid level, rupture flow, cold / hot pipe section temperature of each circuit, flow of each circuit, speed of each main pump, feed water / steam flow of each steam generator and liquid level of each steam generator.

[0017] Furthermore, the discrimination model is constructed using a feedback fully connected neural network.

[0018] Furthermore, the neural network model includes:

[0019] The discriminant model uses different nuclear power plant status parameter data as input to discriminate the current status of the nuclear power plant;

[0020] Generate a model that generates prediction results of key parameters that characterize the operating status of a nuclear power plant based on different transient types at given locations.

[0021] Furthermore, the construction process of the neural network model is specifically as follows:

[0022] Randomly generate N samples of different transient types at different locations in the nuclear power plant, where N is less than the number of training samples in the key state parameter set;

[0023] Taking N randomly generated samples as input, the generative model is used to generate prediction results for a set of key state parameters of a nuclear power plant;

[0024] At the same time, N different transient types at different positions and the calculation results of the key state parameter sets corresponding to the transient types are randomly extracted from the key state parameter set by the resampling method;

[0025] The samples of 2×N key state parameter sets are used as the input of the discriminant model, and the discriminant model outputs different transient type results at different locations;

[0026] Compare the samples of different transient types at different positions that are input to the generative model and the results of different transient types at different positions that are output by the discriminative model, and construct a loss function based on the difference in the results;

[0027] Then, during the training process, back propagation is used to calculate the gradients of the generative model and the discriminative model at the same time, and the model parameters are continuously updated until the Nash equilibrium is reached to obtain the final neural network model.

[0028] Furthermore, the process of transient consequence prediction is specifically as follows:

[0029] Real-time monitoring of key status parameters of nuclear power plants;

[0030] When a single parameter is monitored to have a deviation of more than 2% relative to the preset operating value, the monitoring data of all state parameters in the previous 10 seconds are used as input, and the discriminant model is used to determine whether the current operating state of the nuclear power plant deviates from normal operation; if a transient accident condition occurs, the transient type and location are output;

[0031] The transient type and occurrence location output by the discriminant model are used as the input of the generative model, and the subsequent change curve of the key state parameters of the nuclear power plant is output to achieve rapid prediction of accident consequences.

[0032] Secondly, a nuclear power fault diagnosis and transient prediction system based on generative adversarial networks is provided, including:

[0033] A sample generation module, used for randomly generating a set of key state parameters according to the transient type and the nuclear power plant location where the transient type may appear;

[0034] The discriminant training module is used to take the key state parameter set as input and different transient types at different positions as output to obtain the discriminant model through preliminary training;

[0035] A model building module, which is used to build a neural network model for nuclear power plant fault diagnosis and transient prediction based on the naive generative adversarial network framework;

[0036] The prediction and analysis module is used to combine the real-time monitoring of nuclear power plant parameters and perform fault diagnosis and transient consequence prediction on various transient abnormal conditions that occur during the operation of the nuclear power plant through a neural network model.

[0037] In a third aspect, a computer terminal is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for nuclear power fault diagnosis and transient prediction based on a generative adversarial network as described in any one of the first aspects is implemented.

[0038] In a fourth aspect, a computer-readable medium is provided, on which a computer program is stored, and the computer program is executed by a processor to implement the nuclear power fault diagnosis and transient prediction method based on a generative adversarial network as described in any one of the first aspects.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. The nuclear power fault diagnosis and transient prediction method based on generative adversarial network proposed in the present invention can quickly diagnose the abnormal operating conditions after monitoring the abnormal operation status of the nuclear power plant, identify the transient type, and determine the location where the transient occurs, and then quickly predict the consequences of the accident, so as to achieve the purpose of reducing nuclear reactor transients or accidents;

[0041] 2. The fault diagnosis function implemented by the present invention can help nuclear power plant operators gain early response time for nuclear power plant transients or accidents, and assist operators in making correct judgments; the accident transient rapid prediction function can further assist operators in rapid human intervention in accidents, greatly reducing the possibility of human errors, thereby effectively mitigating the consequences of transients or accidents;

[0042] 3. The present invention can greatly reduce the operational burden on personnel when a nuclear power plant experiences a transient state or accident, and achieve the purpose of alleviating the consequences of transient accidents and reducing radioactive leakage. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0044] Figure 1 is a flow chart in an embodiment of the present invention;

[0045] Figure 2 It is a system block diagram in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0047] Example 1: A nuclear power fault diagnosis and transient prediction method based on a generative adversarial network is specifically implemented by the following steps.

[0048] Step 1: Generate training samples.

[0049] First, different transient types are determined according to the nuclear power plant's initial event library, and the locations of nuclear power plants where different transient types may occur are sorted out. Subsequently, a large number of operating conditions of different transient types at different locations are generated by random sampling. The input files of all sampled operating conditions are constructed using thermal hydraulic programs, and program calculations are performed to obtain output calculation results. Finally, the key parameter calculation data for fault diagnosis and transient consequence prediction are extracted from the program calculation output files. The key parameters include primary circuit pressure, pressurizer pressure, reactor power, hot spot cladding temperature, active zone collapse level, breach flow, cold / hot pipe section temperature of each circuit, flow of each circuit, speed of each main pump, feedwater / steam flow of each steam generator, and liquid level of each steam generator.

[0050] Step 2: Build an initial discriminant model.

[0051] The function of the discriminant model is to use different nuclear power plant state parameter data as input to discriminate the current state of the nuclear power plant, so its output is different transient types at different locations. Using all the training samples generated in step 1, the calculated key parameter set as input, and different transient types at different locations as output, a feedback-type fully connected neural network is used to construct an initial discriminant model with an accuracy of 90% or more.

[0052] Step 3: Build and train a generative adversarial model.

[0053] There are generative models and discriminative models in the generative adversarial network. The function of the discriminative model is to use different nuclear power plant state parameter data as input to discriminate the current state of the nuclear power plant, while the main function of the generative model is to generate key parameter prediction results that characterize the operating state of the nuclear power plant based on given different transient types at different locations.

[0054] First, N samples of different transient types at different locations of the nuclear power plant are randomly generated, where N is less than the number of training samples in step 1. The N randomly sampled samples are used as input, and the prediction results of the key state parameter set of the nuclear power plant are generated using the generative model. At the same time, the calculation results of N different transient types at different locations and their corresponding key state parameter sets are randomly extracted from the training sample set obtained in step 1 using the resampling method. The (2×N) key state parameter set samples are used as the input of the discriminant model to obtain the results of different transient types at different locations output by the discriminant model. This process belongs to fault diagnosis.

[0055] Then, the samples of different transient types at different positions that are input to the generative model are compared with the results of different transient types at different positions output by the discriminative model, and the loss function is constructed based on the difference in the results. Then, during the training process, back propagation is used to calculate the gradients of both the generative model and the discriminative model, and the model parameters are continuously updated to eventually reach a Nash equilibrium, that is, the probability that the samples generated by the generative model are recognized by the discriminative model is consistent with the probability of the real training samples. At this time, the discriminative model cannot distinguish between the samples generated by the generative model and the real training samples.

[0056] Step 4: Nuclear power plant status detection, fault diagnosis and transient consequence prediction.

[0057] The generative model and discriminant model obtained after reaching Nash equilibrium are used to construct a nuclear power plant fault diagnosis and transient consequence prediction module. First, the key state parameters of the nuclear power plant are monitored in real time, including the pressurizer pressure, reactor power, the temperature of the cold / hot pipe sections of each loop, the flow of each loop, the speed of each main pump, the feed water / steam flow of each steam generator, and the liquid level of each steam generator. When a single parameter is detected to have a deviation of more than 2% relative to the nominal operating value, the monitoring data of all state parameters in the previous 10s are used as input, and the discriminant model is used to determine whether the current operating state of the nuclear power plant deviates from normal operation, that is, whether a transient accident condition occurs. If a transient accident condition occurs, the transient type and location are output. Finally, the transient type and location output by the discrimination model are used as the input of the generation model, and the subsequent change curve of the key state parameters of the nuclear power plant is output to achieve the function of rapid prediction of accident consequences. The key state parameters in this step include primary circuit pressure, pressurizer pressure, reactor power, hot spot cladding temperature, active zone collapse liquid level, breach flow, temperature of cold / hot pipe sections of each circuit, flow of each circuit, speed of each main pump, feed water / steam flow of each steam generator, and liquid level of each steam generator.

[0058] Embodiment 2: A nuclear power fault diagnosis and transient prediction system based on a generative adversarial network, the system is used to implement the nuclear power fault diagnosis and transient prediction method based on a generative adversarial network described in Embodiment 1, such as Figure 2 As shown, it includes a sample generation module, a discrimination training module, a model building module and a prediction analysis module.

[0059] Among them, the sample generation module is used to randomly generate a set of key state parameters according to the transient type and the location of the nuclear power plant where the transient type may appear; the discriminant training module is used to take the key state parameter set as input and the different transient types at different locations as output, and preliminarily train to obtain the discriminant model; the model construction module is used to build a neural network model for nuclear power plant fault diagnosis and transient prediction based on the naive generative adversarial network framework; the prediction and analysis module is used to combine the real-time monitoring of nuclear power plant parameters, and use the neural network model to perform fault diagnosis and transient consequence prediction on various transient abnormal conditions that occur during the operation of the nuclear power plant.

[0060] Working principle: The nuclear power fault diagnosis and transient prediction method based on generative adversarial network proposed in the present invention can quickly diagnose abnormal operating conditions after monitoring the abnormal operation status of the nuclear power plant, identify the transient type, and determine the location where the transient occurs, and then quickly predict the consequences of the accident, so as to achieve the purpose of reducing nuclear reactor transients or accidents.

[0061] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0062] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0063] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1A function specified in one or more boxes.

[0064] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0065] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A nuclear power fault diagnosis and transient prediction method based on generative adversarial networks, characterized by: The following steps are involved: Randomly generate a set of key state parameters based on transient types and nuclear power plant locations where transient types may occur; With the key state parameter set as input and different transient types at different locations as output, the discriminant model is initially trained; Constructing a neural network model for nuclear power plant fault diagnosis and transient prediction based on the naive generative adversarial network framework; Combined with real-time monitoring of nuclear power plant parameters, neural network models are used to perform fault diagnosis and transient consequence prediction for various transient abnormal conditions that occur during the operation of nuclear power plants; The generation process of the key state parameter set is specifically as follows: Determine different transient types based on the initial event library of the nuclear power plant; A large number of different transient types at different locations are randomly generated by sampling methods; Perform calculations using a thermal hydraulic program to output a set of key state parameters required to construct a neural network; The key state parameter set includes primary circuit pressure, pressurizer pressure, reactor power, hot spot cladding temperature, active zone collapse liquid level, breach flow, each circuit cold / hot pipe section temperature, each circuit flow, each main pump speed, each steam generator feed water / steam flow and each steam generator liquid level; The process of transient consequence prediction is specifically as follows: Real-time monitoring of key status parameters of nuclear power plants; When a single parameter is monitored to deviate by more than 2% from the preset operating value, the monitoring data of all state parameters in the previous 10 seconds are used as input, and the discriminant model is used to determine whether the current nuclear power plant operating state deviates from normal operation; If a transient accident condition occurs, the transient type and location will be output; The transient type and occurrence location output by the discriminant model are used as the input of the generative model, and the subsequent change curve of the key state parameters of the nuclear power plant is output to achieve rapid prediction of accident consequences.

2. The method for nuclear power fault diagnosis and transient prediction based on generative adversarial network according to claim 1 is characterized in that: The discrimination model is constructed using a feedback fully connected neural network.

3. The method for nuclear power fault diagnosis and transient prediction based on generative adversarial network according to claim 1 is characterized in that: The neural network model includes: The discriminant model uses different nuclear power plant status parameter data as input to discriminate the current status of the nuclear power plant; Generate a model that generates prediction results of key parameters that characterize the operating status of a nuclear power plant based on different transient types at given locations.

4. The method for nuclear power fault diagnosis and transient prediction based on generative adversarial network according to claim 1 is characterized in that: The construction process of the neural network model is specifically as follows: Randomly generate N samples of different transient types at different locations in the nuclear power plant, where N is less than the number of training samples in the key state parameter set; Taking N randomly generated samples as input, the generative model is used to generate prediction results for a set of key state parameters of a nuclear power plant; At the same time, N different transient types at different positions and the calculation results of the key state parameter sets corresponding to the transient types are randomly extracted from the key state parameter set by the resampling method; The samples of 2×N key state parameter sets are used as the input of the discriminant model, and the discriminant model outputs different transient type results at different locations; Compare the samples of different transient types at different positions that are input to the generative model and the results of different transient types at different positions that are output by the discriminative model, and construct a loss function based on the difference in the results; Then, during the training process, back propagation is used to calculate the gradients of the generative model and the discriminative model at the same time, and the model parameters are continuously updated until the Nash equilibrium is reached to obtain the final neural network model.

5. A nuclear power fault diagnosis and transient prediction system based on generative adversarial networks, characterized by: The system is used to implement the nuclear power fault diagnosis and transient prediction method based on a generative adversarial network as described in any one of claims 1 to 4, comprising: A sample generation module, used for randomly generating a set of key state parameters according to the transient type and the nuclear power plant location where the transient type may appear; The discriminant training module is used to take the key state parameter set as input and different transient types at different positions as output to obtain the discriminant model through preliminary training; A model building module, which is used to build a neural network model for nuclear power plant fault diagnosis and transient prediction based on the naive generative adversarial network framework; The prediction and analysis module is used to combine the real-time monitoring of nuclear power plant parameters and use the neural network model to perform fault diagnosis and transient consequence prediction for various transient abnormal situations that occur during the operation of the nuclear power plant.

6. A computer terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, it implements the nuclear power fault diagnosis and transient prediction method based on generative adversarial network as described in any one of claims 1-4.

7. A computer readable medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the nuclear power fault diagnosis and transient prediction method based on a generative adversarial network as described in any one of claims 1 to 4.

Images