A method and device for detecting abnormality of a pump in a nuclear power plant

Abstract

The present application relates to nuclear power safety protection technical field, especially a kind of nuclear power plant pump body abnormality detection method and device.The method is: first, obtain pump body characteristic curve, based on the pump body characteristic curve, instrument arrangement and business knowledge determine the influence factor of the pump body;Then obtain the historical working data of pump body, establish training set by data analysis method;Based on training set, complete the training of regression model, reconstruction model and statistical model and obtain early warning threshold based on residual or statistical distribution;Finally, the current relevant data are input into the state reconstruction algorithm and statistical model or regression model, the statistical quantity or predicted value of pump in current state is obtained, whether there is abnormality is judged by comparing with early warning threshold.The present application selectively uses suitable model to judge the abnormality of pump body, simple and convenient.

Description

Technical Field

[0001] This invention relates to the field of nuclear power safety protection technology, and in particular to a method and device for detecting abnormalities in nuclear power plant pumps. Background Technology

[0002] During normal operation, nuclear power plants may experience unexpected protective actions, such as load shedding, shutdown, or reactor shutdown, due to reasons such as faulty or malfunctioning process equipment, instrumentation and control system failure, power grid failure, fire, insufficient personnel skills, or human error.

[0003] Pumps are crucial power units in nuclear power plants, serving as driving mechanisms to circulate liquids. Pump malfunctions can lead to unexpected system shutdowns and severe economic losses. Therefore, timely detection and troubleshooting of pump failures are of paramount importance.

[0004] The detection of abnormal conditions during pump operation currently relies mainly on alarm information from the DCS system. Alarms are only triggered when a single measurement or a combination of multiple measurements reaches a certain threshold. This lacks trend analysis and results in poor early warning capabilities for abnormal conditions. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a method and device for detecting anomalies in nuclear power plant pumps. The method uses regression and reconstruction algorithms to calculate the predicted operating parameters of the pump and compares them with the actual operating parameters, thereby timely and accurately identifying anomalies in the pump's operating parameters. At the same time, it also combines statistical models to realize the detection of anomalies in the overall pump status.

[0006] This invention provides a method for detecting abnormalities in a nuclear power plant pump, comprising the following steps:

[0007] Step S100: Obtain the pump body characteristic curve, and preliminarily determine the influencing factors of the pump body based on the pump body characteristic curve, business knowledge, and instrument layout;

[0008] Step S200: Obtain several segments of historical working data of the pump body, classify the historical working data according to the type of historical working data, and use data analysis method to analyze each type of historical working data to determine the final influencing factor of the pump body, and then generate steady-state training set and change state training set.

[0009] Step S300: Establish a first model and train the first model using the steady-state training set to obtain a first prediction model of pump-related parameters and a first early warning threshold that are fully trained. The first model includes at least one of a reconstruction model and a statistical model.

[0010] Step S400: Establish a second model by training the regression model using the changed state training set to obtain a fully trained second prediction model and a second early warning threshold, wherein the second model is a regression model;

[0011] Step S500: Obtain the current working data and influencing factor data of the pump body, determine the current state of the pump body, and input the corresponding data of the current pump body into the first prediction model or the second prediction model under the current state to obtain the prediction data of the current pump body under the current state. Then, based on the first warning threshold or the second warning threshold, determine whether the pump body is abnormal.

[0012] In one specific embodiment of the present invention, the reconstruction model employs principal component analysis (PCA) reconstruction, autoencoder (AE), and variational autoencoder (VAE).

[0013] In one specific embodiment of the present invention, the regression model is one of the following: multinomial regression model, linear regression model, decision tree regression model, support vector machine regression model, K-nearest neighbor regression model, random forest regression model, Adaboost regression model, gradient boosting random forest regression model, bagging regression model, and Extra TREE regression model.

[0014] In one specific embodiment of the present invention, the statistical model employs principal component analysis algorithm.

[0015] In one specific embodiment of the present invention, it further includes:

[0016] When the cause of the pump body abnormal alarm is determined to be a false alarm or a missed alarm, the latest historical data is obtained and the model under this state is updated using the latest historical data.

[0017] In one specific embodiment of the present invention, the working data includes at least temperature parameters, current parameters, and vibration parameters, and the influencing factor data includes at least pump fluid temperature, volumetric flow rate, pump sealing working fluid temperature and flow rate, and cooling motor working fluid temperature and flow rate.

[0018] In one specific embodiment of the present invention, the temperature parameters include motor bearing temperature, coil temperature, and pump bearing temperature; the current parameters include motor current and other parameters; and the vibration parameters include bearing vibration.

[0019] This invention provides a nuclear power plant pump anomaly detection device, comprising:

[0020] The module for preliminary determination of influencing factors acquires the pump body characteristic curve and, based on the pump body characteristic curve, business knowledge, and instrument layout, preliminarily determines the influencing factors of the pump body.

[0021] The training set generation module is used to acquire several segments of historical working data of the pump body, classify the historical working data according to the type of historical working data, and use data analysis methods to analyze each type of historical working data to determine the final influencing factor of the pump body, and then generate steady-state training set and change-state training set.

[0022] The first model building module is used to build a first model and train the first model using the steady-state training set to obtain a first prediction model of pump-related parameters and a first early warning threshold that are fully trained. The first model includes at least one of a reconstruction model and a statistical model.

[0023] The second model building module is used to build a second model. The regression model is trained using the changed state training set to obtain a fully trained second prediction model and a second early warning threshold. The second model is a regression model.

[0024] An anomaly detection module is used to acquire the current working data and influencing factor data of the pump body, determine the current state of the pump body, and input the corresponding data of the current pump body into the first prediction model or the second prediction model under the current state to obtain the prediction data of the current pump body under the current state. Based on the first warning threshold or the second warning threshold, it is determined whether there is an anomaly in the pump body.

[0025] This invention provides an electronic device, comprising: a processor and a memory;

[0026] The memory stores computer programs that can be executed by the processor;

[0027] When the processor executes the computer program, it implements the steps in the nuclear power plant pump anomaly detection method.

[0028] The present invention provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps in the nuclear power plant pump anomaly detection method.

[0029] Compared with the prior art, the nuclear power plant pump body anomaly detection method and device of the present invention obtains the influencing factors of pump body anomalies, and constructs three anomaly detection models of pump body by using statistical algorithm, reconstruction model and regression algorithm respectively, and then selectively uses the appropriate model to judge the pump body anomalies, which is simple and convenient. Attached Figure Description

[0030] Figure 1 A flowchart illustrating a method for detecting abnormalities in nuclear power plant pumps;

[0031] Figure 2A schematic diagram showing the functional modules of a nuclear power plant pump anomaly detection device;

[0032] Figure 3 A schematic diagram showing the hardware structure of an electronic device;

[0033] In the diagram, 10 is an electronic device, 110 is a processor, and 120 is a memory. Detailed Implementation

[0034] To further understand the present invention, embodiments of the present invention are described below in conjunction with examples. However, it should be understood that these descriptions are only for further illustrating the features and advantages of the present invention, and not for limiting the present invention.

[0035] An embodiment of the present invention provides a method for detecting abnormalities in a nuclear power plant pump, such as... Figure 1 As shown, it includes the following steps:

[0036] S100: Obtain the pump body characteristic curve, and preliminarily determine the influencing factors of the pump body based on the pump body characteristic curve, business knowledge, and instrument layout;

[0037] S200: Acquire several segments of historical working data of the pump body, classify the historical working data according to the type of historical working data, and use data analysis method to analyze each type of historical working data to determine the final influencing factor of the pump body, and then generate steady-state training set and change state training set.

[0038] S300: Establish a first model, and train the first model using the steady-state training set to obtain a first prediction model of pump-related parameters and a first early warning threshold that are fully trained, wherein the first model includes at least one of a reconstruction model and a statistical model;

[0039] S400: Establish a second model by training the regression model using the changed state training set to obtain a fully trained second prediction model and a second early warning threshold, wherein the second model is a regression model;

[0040] S500: Obtain the current working data and influencing factor data of the pump body, determine the current state of the pump body, and input the corresponding data of the current pump body into the first prediction model or the second prediction model under the current state to obtain the prediction data of the current pump body under the current state. Based on the first warning threshold or the second warning threshold, determine whether the pump body is abnormal.

[0041] In this embodiment, firstly, the pump body characteristic curve is obtained. Based on the pump body characteristic curve, instrument layout, and business knowledge, the influencing factors of the pump body are determined. Then, several segments of historical operating data of the pump body are obtained, and a training set is established through data analysis methods. Based on the training set, a regression model, a reconstruction model, and a statistical model are trained, and a warning threshold is obtained based on the residual or statistical distribution. Finally, the current relevant data is input into the reconstruction algorithm and the statistical model or regression model under this state to obtain the statistical or predicted value of the pump in the current state. The presence of an anomaly is determined by comparing it with the warning threshold. This invention obtains the influencing factors of pump body anomalies and uses statistical algorithms, reconstruction models, and regression algorithms to construct three anomaly detection models for the pump body, respectively. Then, the appropriate model is selectively used to determine the anomaly of the pump body, which is simple and convenient.

[0042] In some embodiments, the anomaly detection method of the present invention will reflect: 1) whether the pump parameters are consistent with historical values ​​under the same conditions; 2) whether the pump current changes reasonably under non-starting transient conditions; and 3) whether the pump as a whole may be abnormal.

[0043] Specifically, the influencing factors of the pump's operating parameters are first determined, and then a pump anomaly detection model can be built based on these factors. Optionally, in this embodiment of the invention, the influencing factors are determined based on the characteristic curves of the pump, which reflect the pump's operating characteristics. Optionally, the operating data includes at least pump fluid temperature, volumetric flow rate, pump sealing medium temperature and flow rate, and cooling motor medium temperature and flow rate.

[0044] In this embodiment, the pump body's temperature parameters include motor bearing temperature, coil temperature, and pump bearing temperature; the pump body's current parameters include motor current; and the pump body's vibration parameters include bearing vibration, which includes vibration in the X and Y directions. The working fluid parameters include working fluid temperature, flow rate, and pressure. The sealing parameters include sealing flow rate and sealing temperature. For example, analysis of the characteristic curves reveals a significant and continuous increase in pump current during cooling; the opposite process is observed during heating. Therefore, the influence of temperature on pump current must be considered during modeling.

[0045] In some embodiments, in step S200, the anomaly detection model is constructed based on knowledge and experience. According to the operating experience of nuclear power plants, historical operating data of the pump body is obtained. This historical operating data includes at least historical operating parameters of the pump body and corresponding influencing factor data. Then, the model uses this historical operating data to predict whether the pump body operating parameters are abnormal. Specifically, the historical operating data needs to be classified according to data type, such as temperature sensors, vibration sensors, flow sensors, etc. If there are many pump-related parameters, clustering algorithms are needed for classification. For each type of data, data analysis methods are used to determine the final influencing factors and complete the required feature processing, ultimately forming the required steady-state training set and variable-state training set, such as pump start-up and pump shutdown processes. Furthermore, for more complex pumps, the steady-state training set needs to be divided into multiple data subsets, such as the continuous rise or fall of current under steady-state operation.

[0046] In some embodiments, during step S300, since power plants may undergo maintenance, the data may change after the maintenance, leading to false alarms in model prediction. Therefore, to avoid this situation, the influencing factor data is reconstructed in this embodiment of the invention. The reconstructed data can best represent the characteristics of the influencing factors and thus avoid false alarms. It should be noted that the present invention sets up two training datasets, which are used to establish different models. The first detection model is suitable for non-variable frequency pumps (except APA pumps) because the pump parameters do not change much after the pump is started. Therefore, the pump anomaly can be determined by the reconstruction algorithm and the statistical algorithm. The second detection model is suitable for pumps that are significantly affected by relevant parameters or are variable frequency pumps. Therefore, the pump anomaly is determined by the regression algorithm.

[0047] Optionally, the reconstruction model employs reconstruction algorithms such as principal component analysis (PCA), autoencoder (AE), and variational autoencoder (VAE).

[0048] In this embodiment, three reconstruction algorithms can be selected to build the reconstruction model according to actual needs. Specifically:

[0049] a. Autoencoder (AE): This is an unsupervised learning model. In machine learning and deep learning, its main goal is to learn how to represent data from the input space as a compact representation in the latent variable space and recover it back to the original input space as accurately as possible.

[0050] b. PCA Reconstruction Algorithm: Principal Component Analysis (PCA) is a dimensionality reduction method widely used in statistics, machine learning, and data analysis;

[0051] c. Variational Autoencoders (VAEs): These are unsupervised learning methods used to generate and understand the latent structure of data. They combine the concepts of autoencoders (AEs) and Bayesian inference, enabling the compression and reconstruction of input data, as well as the generation of new data samples. VAEs excel at discovering low-dimensional representations from high-dimensional data, making them widely applicable in many fields such as image generation, semantic embedding, and anomaly detection.

[0052] In some embodiments, in step S400, the regression model is a predictive modeling technique that studies the relationship between the dependent variable (target) and the independent variable (predictor) for predictive analysis, time series modeling, and discovering causal relationships between variables. Optionally, the regression model can be one of the following: linear regression model, decision tree regression model, support vector machine regression model, K-nearest neighbor regression model, random forest regression model, Adaboost regression model, gradient boosting random forest regression model, bagging regression model, and Extra TREE regression model.

[0053] In some embodiments, the input factors of the regression model are influencing factor data, and the output of the regression model is the pump body operating parameters.

[0054] In this embodiment, after obtaining a fully trained pump anomaly detection model, the current pump's influence factor data is acquired. After inputting the current pump's influence factor data into the pump anomaly detection model, the predicted pump operating data can be quickly obtained.

[0055] In some embodiments, the statistical model employs one of the following algorithms: clustering analysis, correlation analysis, and principal component analysis.

[0056] Cluster analysis algorithms classify individual samples or indicator variables according to their characteristics and find reasonable statistical measures of similarity between things. Common examples include Q-type cluster analysis and R-type cluster analysis. Correlation analysis algorithms study whether there is a certain dependency relationship between phenomena and explore the direction and degree of correlation for specific dependent phenomena. Principal component analysis algorithms transform a set of related indicators into a new set of independent indicator variables and use a few of the new indicator variables to comprehensively reflect the main information contained in the original multiple indicator variables.

[0057] In this embodiment, when the pump body is a non-variable frequency pump, regression model and statistical model are used to determine the pump body abnormality. The output results of the statistical model, such as variance, can be directly used to determine whether the output results are within the normal range to determine whether the pump body is abnormal.

[0058] When the pump body is a variable frequency pump or a pump body that is greatly affected by relevant parameters, a preset alarm threshold for abnormal working data is established. This alarm threshold is the maximum value of the error between the actual value and the predicted value. That is, within this error range, it means that the pump body has no abnormality. Therefore, after obtaining the working data through the model, the predicted working data is compared with the actual working data to obtain the difference between the two. Then, this difference is compared with the preset alarm threshold for abnormal working data. If it exceeds the alarm threshold for abnormal working data, it means that the pump body has an abnormality; otherwise, it means that the pump body has no abnormality.

[0059] In some embodiments, the method further includes:

[0060] When an abnormality is detected in the pump body, a corresponding alarm message will be issued.

[0061] In this embodiment, when an abnormality is detected in the pump body, an alarm message is issued. The alarm message includes the specific details of the pump body abnormality, the cause of the abnormality, and other factors, so that staff can respond to the alarm message quickly.

[0062] In some embodiments, the method further includes:

[0063] When it is determined that the cause of the pump body abnormal alarm is a false alarm or a missed alarm, the latest historical data is obtained and the first prediction model or the second prediction model is updated using the latest historical data.

[0064] In this embodiment, the above models are both first and second prediction models built based on historical operating data, and the accuracy of the models is constrained by the data. Since historical operating data of the unit cannot cover all normal or abnormal states, the anomaly detection model developed based on historical operating data will produce false alarms or missed alarms when encountering new normal or abnormal states. If the previous model is used for detection at this time, it will lead to an increase in the false alarm rate or missed alarm rate. Therefore, this invention adds an online update process, uses historical data for verification, and makes adjustments based on new data.

[0065] Specifically, the alarm information is first analyzed manually. If it is determined to be a false alarm, it indicates that there may be a new abnormal pattern that does not exist in the historical data. In this case, such samples need to be added to the training sample, and the model is retrained and updated based on these samples. If it is determined to be a missed alarm, it indicates that there are cases that should have been identified by the existing nuclear power important parameter anomaly identification system, but were not actually identified. In this case, the features of the missed alarm are extracted and an anomaly feature library is established. Subsequently, new training is carried out based on the abnormal sample data in the anomaly feature library to obtain the updated first and second prediction models.

[0066] Another embodiment of the present invention provides a nuclear power plant pump body anomaly detection device, see reference. Figure 2 The nuclear power plant's pump anomaly detection device includes:

[0067] The module for preliminary determination of influencing factors acquires the pump body characteristic curve and, based on the pump body characteristic curve, business knowledge, and instrument layout, preliminarily determines the influencing factors of the pump body.

[0068] The training set generation module is used to acquire several segments of historical working data of the pump body, classify the historical working data according to the type of historical working data, and use data analysis methods to analyze each type of historical working data to determine the final influencing factor of the pump body, and then generate steady-state training set and change-state training set.

[0069] The first model building module is used to build a first model and train the first model using the steady-state training set to obtain a first prediction model of pump-related parameters and a first early warning threshold that are fully trained. The first model includes at least one of a reconstruction model and a statistical model.

[0070] The second model building module is used to build a second model. The regression model is trained using the changed state training set to obtain a fully trained second prediction model and a second early warning threshold. The second model is a regression model.

[0071] An anomaly detection module is used to acquire the current working data and influencing factor data of the pump body, determine the current state of the pump body, and input the corresponding data of the current pump body into the first prediction model or the second prediction model under the current state to obtain the prediction data of the current pump body under the current state. Based on the first warning threshold or the second warning threshold, it is determined whether there is an anomaly in the pump body.

[0072] It should be noted that the module referred to in this invention refers to a series of computer program instruction segments that can perform specific functions. It is more suitable than a program for describing the execution process of nuclear power plant pump body anomaly detection. For the specific implementation of each module, please refer to the corresponding method embodiments above, which will not be repeated here.

[0073] In some embodiments, the operating data includes at least temperature parameters, current parameters, and vibration parameters, and the influencing factor data includes at least pump fluid temperature, volumetric flow rate, pump sealing working fluid temperature and flow rate, and cooling motor working fluid temperature and flow rate.

[0074] In some embodiments, the reconstruction model employs principal component analysis (PCA) reconstruction, autoencoder (AE), and variational autoencoder (VAE).

[0075] In some embodiments, the regression model is one of the following: multinomial regression model, linear regression model, decision tree regression model, support vector machine regression model, K-nearest neighbor regression model, random forest regression model, Adaboost regression model, gradient boosting random forest regression model, bagging regression model, and Extra TREE regression model.

[0076] In some embodiments, the statistical model primarily employs clustering analysis algorithms.

[0077] In some embodiments, the method further includes:

[0078] When the cause of the pump body abnormal alarm is determined to be a false alarm or a missed alarm, the latest historical data is obtained and the model under this state is updated using the latest historical data.

[0079] In some embodiments, the apparatus further includes:

[0080] The online update module is used to obtain the latest historical data and update the model under this state when the cause of the pump body abnormal alarm is determined to be a false alarm or a missed alarm.

[0081] Another embodiment of the present invention provides an electronic device, such as... Figure 3 As shown, the electronic device 10 includes:

[0082] One or more processors 110 and memory 120, Figure 3 The following description uses a processor 110 as an example. The processor 110 and the memory 120 can be connected via a bus or other means. Figure 3 Taking the example of a connection between China and Israel via a bus.

[0083] Processor 110 is used to perform various control logics of electronic device 10. It can be a general-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), microcontroller, ARM (Acorn RISC Machine) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination of these components. Furthermore, processor 110 can also be any conventional processor, microprocessor, or state machine. Processor 110 can also be implemented as a combination of computing devices, such as a combination of DSP and microprocessor, multiple microprocessors, one or more microprocessors combined with DSP and / or any other such configuration.

[0084] The memory 120, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions corresponding to the nuclear power plant pump anomaly detection method in this embodiment of the invention. The processor 110 executes various functional applications and data processing of the electronic device 10 by running the non-volatile software programs, instructions, and units stored in the memory 120, thereby implementing the nuclear power plant pump anomaly detection method in the above method embodiment.

[0085] The memory 120 may include a program storage area and a data storage area. The program storage area may store applications required for the operating platform and at least one function; the data storage area may store data created based on the use of the electronic device 10. Furthermore, the memory 120 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 120 may optionally include memory remotely located relative to the processor 110, and these remote memories may be connected to the electronic device 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0086] One or more units are stored in memory 120. When executed by one or more processors 110, they perform the nuclear power plant pump anomaly detection method in any of the above method embodiments, for example, performing the above-described... Figure 1 The method steps S100 to S500.

[0087] Another embodiment of the present invention provides a computer-readable storage medium storing computer-executable instructions that are executed by one or more processors, for example, to perform the above-described instructions. Figure 1 The method steps S100 to S500.

[0088] In summary, the present invention discloses a method, device, electronic equipment, and storage medium for detecting pump anomalies in nuclear power plants. First, it acquires the pump's characteristic curve and determines the influencing factors based on the characteristic curve, instrument layout, and operational knowledge. Then, it acquires several segments of historical operating data from the pump and establishes a training set using data analysis methods. Based on the training set, it trains a regression model, a reconstruction model, and a statistical model, and obtains a warning threshold based on the residual or statistical distribution. Finally, it inputs the current relevant data into the reconstruction algorithm and the statistical model or regression model for that state to obtain the pump's statistical or predicted value in the current state. This value is then compared with the warning threshold to determine if an anomaly exists. This invention acquires the influencing factors of pump anomalies and constructs three anomaly detection models for the pump using statistical algorithms, reconstruction models, and regression algorithms, respectively. It then selectively uses the appropriate model to determine pump anomalies, making the process simple and convenient.

[0089] The above description of the embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. It should be noted that those skilled in the art can make several improvements and modifications to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

[0090] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting anomalies in a nuclear power plant pump body, characterized in that, Includes the following steps: Step S100: Obtain the pump body characteristic curve, and preliminarily determine the influencing factors of the pump body based on the pump body characteristic curve, business knowledge, and instrument layout; Step S200: Obtain several segments of historical working data of the pump body, classify the historical working data according to the type of historical working data, and use data analysis method to analyze each type of historical working data to determine the final influencing factor of the pump body, and then generate steady-state training set and change state training set. Step S300: Establish a first model and train the first model using the steady-state training set to obtain a first prediction model of pump-related parameters and a first early warning threshold that are fully trained. The first model includes at least one of a reconstruction model and a statistical model. Step S400: Establish a second model, and train the second model using the changed state training set to obtain a fully trained second prediction model and a second early warning threshold, wherein the second model is a regression model; Step S500: Obtain the current working data and influencing factor data of the pump body, determine the current state of the pump body, and input the corresponding data of the current pump body into the first prediction model or the second prediction model under the current state to obtain the prediction data of the current pump body under the current state. Then, based on the first warning threshold or the second warning threshold, determine whether the pump body is abnormal.

2. The method for detecting abnormalities in a nuclear power plant pump body according to claim 1, characterized in that, The reconstruction model employs principal component analysis (PCA) reconstruction, autoencoder (AE), and variational autoencoder (VAE).

3. The method for detecting abnormalities in a nuclear power plant pump body according to claim 1, characterized in that, The regression model is one of the following: multinomial regression model, linear regression model, decision tree regression model, support vector machine regression model, K-nearest neighbor regression model, random forest regression model, Adaboost regression model, gradient boosting random forest regression model, bagging regression model, and Extra TREE regression model.

4. The method for detecting abnormalities in a nuclear power plant pump body according to claim 1, characterized in that, The statistical model employs principal component analysis.

5. The method for detecting abnormalities in a nuclear power plant pump body according to claim 1, characterized in that, Also includes: When the cause of the pump body abnormal alarm is determined to be a false alarm or a missed alarm, the latest historical data is obtained and the model under this state is updated using the latest historical data.

6. The method for detecting abnormalities in a nuclear power plant pump body according to claim 1, characterized in that, The operating data includes at least temperature parameters, current parameters, and vibration parameters, and the influencing factor data includes at least pump fluid temperature, volumetric flow rate, pump sealing medium temperature and flow rate, and cooling motor medium temperature and flow rate.

7. The method for detecting abnormalities in a nuclear power plant pump body according to claim 6, characterized in that, The temperature parameters include motor bearing temperature, coil temperature, and pump bearing temperature; the current parameters include motor current; and the vibration parameters include bearing vibration.

8. A nuclear power plant pump body anomaly detection device, characterized in that, include: The module for preliminary determination of influencing factors acquires the pump body characteristic curve and, based on the pump body characteristic curve, business knowledge, and instrument layout, preliminarily determines the influencing factors of the pump body. The training set generation module is used to acquire several segments of historical working data of the pump body, classify the historical working data according to the type of historical working data, and use data analysis methods to analyze each type of historical working data to determine the final influencing factor of the pump body, and then generate steady-state training set and change-state training set. The first model building module is used to build a first model and train the first model using the steady-state training set to obtain a first prediction model of pump-related parameters and a first early warning threshold that are fully trained. The first model includes at least one of a reconstruction model and a statistical model. The second model building module is used to build a second model. The second model is trained using the changed state training set to obtain a fully trained second prediction model and a second early warning threshold. The second model is a regression model. An anomaly detection module is used to acquire the current working data and influencing factor data of the pump body, determine the current state of the pump body, and input the corresponding data of the current pump body into the first prediction model or the second prediction model under the current state to obtain the prediction data of the current pump body under the current state. Based on the first warning threshold or the second warning threshold, it is determined whether there is an anomaly in the pump body.

9. An electronic device, characterized in that, include: Processor and memory; The memory stores computer programs that can be executed by the processor; When the processor executes the computer program, it implements the steps in the nuclear power plant pump anomaly detection method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps in the nuclear power plant pump anomaly detection method according to any one of claims 1 to 7.

Images