Nuclear power system valve life prediction method, device and equipment and storage medium

Abstract

The invention provides a nuclear power system valve life prediction method and device, equipment and a storage medium, and relates to the technical field of data processing, the method comprises the steps that a local end carries out feature extraction on full life cycle data to obtain multiple to-be-screened features, a server end determines intersection features in the to-be-screened features, and the local end carries out prediction on the valve life of a nuclear power system on the basis of the intersection features; and training the initial model to obtain a plurality of target models, and obtaining a global fusion model by the server according to each target model. The consistency of valve life prediction can be improved by performing life prediction on the target valve based on the same global fusion model, the target model is trained at each local end, the data of the local ends are only reserved locally, the problem of data management disorder can be solved, and the global fusion model is obtained based on fusion of the target models, so that the service life of the valve is prolonged. And the global fusion model can integrate related data of each valve, so that the problem of poor data circulation is solved, and the effect of improving the generalization ability of the model can also be achieved.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly relates to a method, device, equipment, and storage medium for predicting the service life of valves in a nuclear power system. Background Art

[0002] A nuclear power system refers to a system that uses the energy generated by nuclear energy to drive the operation of machines or equipment, and has a complex structure and diverse components. This system has various types and functions of valve devices, and their safety and stability directly affect the reliability and healthy operation of the entire system.

[0003] Currently, the life prediction of key valves in a nuclear power system is evaluated by means of regular tests. Regular tests are carried out to evaluate whether there are abnormalities and to carry out targeted maintenance. This method is relatively passive, and the evaluation means are mainly based on expert experience, resulting in poor evaluation consistency and potential safety hazards for the valves. In addition, each type of valve device is responsible for by specialized personnel or departments, and the data and knowledge have poor circulation, making it difficult to efficiently evaluate the target valve device. Summary of the Invention

[0004] This application provides a method, device, equipment, and storage medium for predicting the service life of valves in a nuclear power system, which can improve the consistency of valve life prediction. The target model is trained at each local end, and the data at the local end is only retained locally, which can solve the problem of chaotic data management. And a global fusion model is obtained based on the fusion of each target model, and the global fusion model can integrate the relevant data of each valve, solving the problem of poor data circulation.

[0005] A method for predicting the service life of valves in a nuclear power system provided by this application includes: Each local end obtains the full life cycle data of the corresponding target valve, and each local end corresponds to a target valve, and the target valve is a valve in the nuclear power system; The local end extracts features from the full life cycle data to obtain a plurality of features to be screened, and sends the plurality of features to be screened to the server; The server determines the intersection features among the features to be screened, and sends the intersection features to each local end; The local end trains the initial model at the local end based on the intersection features to obtain a target model, and sends the target model to the server; The server obtains a global fusion model based on each target model, and sends the global fusion model to the prediction end; The prediction end obtains the current data of the valve to be predicted in the nuclear power system, and obtains the life prediction result of the valve to be predicted according to the current data of the valve to be predicted and the global fusion model.

[0006] In one embodiment of the present application, the local end extracts features from the full life cycle data to obtain multiple features to be screened, including: For each feature parameter group of the full life cycle data: determine multiple feature arrays corresponding to the valve opening action and multiple feature arrays corresponding to the valve closing action; the full life cycle data includes multiple feature parameter groups, each feature parameter group corresponds to a feature parameter, and each feature parameter group includes data of the feature parameter during the valve opening action and data during the valve closing action, and the feature arrays include multiple feature values of the same type; Determine the correlation between two-by-two feature arrays, and determine the features to be screened from the flag features of the two feature arrays according to the correlation. Each feature array corresponds to a flag feature, and the flag feature characterizes the feature parameter corresponding to the feature array, the valve action stage, and the type of the feature value. The valve action stage includes the valve opening action and the valve closing action.

[0007] In one embodiment of the present application, the types of feature values include average value, variance, maximum value, minimum value, and volatility.

[0008] In one embodiment of the present application, the valve opening action includes before valve opening stabilization, during valve opening stabilization, and after valve opening stabilization, and the valve closing action includes before valve closing stabilization, during valve closing stabilization, and after valve closing stabilization.

[0009] In one embodiment of the present application, determining the correlation between two-by-two feature arrays and determining the features to be screened from the flag features of the two feature arrays according to the correlation includes: Determine the correlation coefficient between two-by-two feature arrays; When the correlation coefficient is less than the preset value, determine the flag features of the two feature arrays as the features to be screened; When the correlation coefficient is greater than or equal to the preset value, determine the flag feature of one of the two feature arrays as the feature to be screened.

[0010] In one embodiment of the present application, the correlation coefficient is the Pearson correlation coefficient.

[0011] In one embodiment of the present application, the server obtains a global fusion model according to each target model and sends the global fusion model to the prediction end to be predicted, including: Determine the weighting coefficient according to the local verification results of each target model, or determine the weighting coefficient according to the number of training samples corresponding to each target model; Perform weighted fusion on each target model according to each target model and the weighting coefficient to obtain the global fusion model.

[0012] To achieve the above and other related purposes, the present application provides a valve life prediction device for a nuclear power system, including: A data acquisition module, configured to acquire the full life cycle data of the corresponding target valve for each local end, where each local end corresponds to a target valve, and the target valve is a valve in a nuclear power system; A feature screening module, configured to extract features from the full life cycle data at the local end to obtain multiple features to be screened, and send the multiple features to be screened to the server; A feature processing module, configured to determine the intersection features among the features to be screened at the server, and send the intersection features to each local end; A model training module, configured to train the initial model at the local end based on the intersection features to obtain a target model, and send the target model to the server; A model fusion module, configured to obtain a global fusion model according to each target model at the server, and send the global fusion model to the prediction end to be predicted; A life prediction module, configured to acquire the current data of the valve to be predicted in the nuclear power system at the prediction end to be predicted, and obtain the life prediction result of the valve to be predicted according to the current data of the valve to be predicted and the global fusion model.

[0013] To achieve the above object and other related objects, the present application further provides an electronic device, where the electronic device includes: One or more processors; A memory for storing the executable program code of the processor; Wherein, the processor is configured to execute the program code to implement the above nuclear power system valve life prediction method.

[0014] To achieve the above object and other related objects, the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by the processor of the computer, the computer is enabled to execute the foregoing one or more nuclear power system valve life prediction methods.

[0015] As described above, a nuclear power system valve life prediction method, device, equipment, and storage medium provided by the present application have the following beneficial effects: A method for predicting the service life of valves in a nuclear power system in this application. This method obtains the full life cycle data of the corresponding target valves from each local end, extracts features from the full life cycle data to obtain multiple features to be screened, and sends the multiple features to be screened to the server. The server determines the intersection features among the features to be screened and sends the intersection features to each local end. The local end trains the initial model of the local end based on the intersection features to obtain a target model and sends the target model to the server. The server obtains a global fusion model based on each target model and sends the global fusion model to the prediction end to be predicted. The prediction end to be predicted obtains the service life prediction result of the valve to be predicted according to the current data of the valve to be predicted and the global fusion model. Predicting the service life of the target valve based on the same global fusion model can improve the consistency of valve service life prediction. Training the target model at each local end and only retaining the local data at the local end can solve the problem of chaotic data management. And a global fusion model is obtained by fusing each target model. The global fusion model can integrate the relevant data of each valve, solve the problem of poor data circulation, and also achieve the effect of improving the generalization ability of the model.

[0016] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit this application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings herein are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with this application, and are used together with the specification to explain the principles of this application. Obviously, the drawings in the following description are only some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. In the drawings: Figure 1 It is a schematic diagram of the implementation environment of the method for predicting the service life of valves in a nuclear power system shown in an exemplary embodiment of this application; Figure 2 It is a flowchart of the method for predicting the service life of valves in a nuclear power system shown in an exemplary embodiment of this application; Figure 3 It is a block diagram of the structure of the device for predicting the service life of valves in a nuclear power system shown in an exemplary embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The embodiments of the present application will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be understood that the preferred embodiments are only for illustrating the present application, rather than for limiting the protection scope of the present application.

[0019] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present application in a schematic manner. Therefore, only the components related to the present application are shown in the diagrams, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and ratio of each component in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0020] In the following description, a large number of details are explored to provide a more thorough explanation of the embodiments of the present application. However, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present application difficult to understand.

[0021] Please refer to Figure 1 , which is a schematic diagram of the implementation environment of the nuclear power system valve life prediction method shown in an exemplary embodiment of the present application. The implementation environment may include a local end 110, a server end 120, and a to-be-predicted end 130. Among them, the solid line represents that the data or model of the local end is uploaded to the server end, and the dashed line represents that the data or model of the server end is sent to the to-be-predicted end.

[0022] Both the local end 110 and the to-be-predicted end 130 can be connected to the server end 120 in a wired or wireless manner. Based on the connection between the local end 110, the to-be-predicted end 130, and the server end 120, the local end 110 can perform data transmission with the server end 120, and the to-be-predicted end 130 can perform data transmission with the server end 120.

[0023] The server end 120 can send data to the local end 110 and the to-be-predicted end 130, and receive data sent by the local end 110 and the to-be-predicted end 130.

[0024] The valve life prediction method provided by the embodiments of this application can be executed by software or hardware installed on the local end 110, the server end 120, and the to-be-predicted end 130. The server end 120 includes, but is not limited to, servers, server clusters, cloud servers, or cloud server clusters, etc. The local end 110 and the to-be-predicted end 130 include, but are not limited to, any one of intelligent terminal devices such as smart phones, personal computers (PCs), laptop computers, tablet computers, e-readers, Internet TVs, and wearable devices.

[0025] Please refer to Figure 2 , Figure 2 which is a flowchart of the valve life prediction method for the nuclear power system shown in an exemplary embodiment of this application. This valve life prediction method for the nuclear power system can be applied to Figure 1 the implementation environment shown. Referring to Figure 2 it can be seen that this valve life prediction method for the nuclear power system can include: Step S210, each local end acquires the full life cycle data of the corresponding target valve.

[0026] Among them, each local end corresponds to a target valve, and the target valve is a valve in the nuclear power system.

[0027] In an embodiment of this application, each local end can acquire the full life cycle data of the corresponding target valve. The full life cycle data can include all data during the entire usage process (or test process) of the valve from the start of use to abandonment. The full life cycle data can include characteristic parameters such as current, voltage, power, temperature, and pressure. A nuclear power system is a device that uses the heat energy generated by a nuclear reactor to drive a steam turbine generator to generate electricity, and is widely used in nuclear power plants. In a nuclear power system, valves are one of the key components to ensure safe and reliable operation. Due to the particularity and safety requirements of the nuclear power system, the monitoring of specific valves is crucial. The target valves can include main steam isolation valves, pressurizer safety valves, emergency feed water isolation valves, safety injection system isolation valves, boric acid injection valves, drain valves, exhaust valves, fuel handling system valves, chemical and volume control system valves, etc.

[0028] Each local end can correspond to a target valve. The local end can acquire the full life cycle data corresponding to the target valve when the tester tests the target valve, or the local end can acquire the full cycle life data of the target valve from the historical data in the local database. The full life cycle data of the target valve can be collected by multi-dimensional sensors.

[0029] Exemplarily, tester A is responsible for testing the target valve a, which is damaged after operating n times. Each operation includes two stages: opening the valve and closing the valve, and it has multi-dimensional sensors that can obtain characteristic parameters such as current and power. Tester B is responsible for testing the target valve b, which is damaged after operating m times. Each operation has two stages: opening the valve and closing the valve, and it has multi-dimensional sensors that can obtain characteristic parameters such as current and voltage.

[0030] It should be noted that the full life cycle data of each target valve includes common characteristic parameters such as current, voltage, and power. Based on the different characteristics of each target valve, the full life cycle data of each target valve may include specific characteristic parameters such as pressure or flow rate.

[0031] Step S220, the local end extracts features from the full life cycle data to obtain multiple features to be screened, and sends the multiple features to be screened to the server.

[0032] In an embodiment of the present application, the local end can extract features from the full life cycle data to obtain multiple features to be screened, and send the multiple features to be screened to the server. The features to be screened characterize the types of characteristic values of the characteristic parameters in different operation stages. The operation stages can include opening the valve and closing the valve.

[0033] Furthermore, the operation stages can include before the opening valve is stable, when the opening valve is stable, after the opening valve is stable, before the closing valve is stable, when the closing valve is stable, and after the closing valve is stable.

[0034] Step S230, the server determines the intersection features among the features to be screened, and sends the intersection features to each local end.

[0035] In an embodiment of the present application, the server can determine the intersection features among the features to be screened, and send the intersection features to each local end. Determining the intersection features can ensure that a unified input can be determined for each valve, enabling the global fusion model to adapt to the life prediction of any valve.

[0036] It should be noted that the intersection features can characterize the types of characteristic values of the characteristic parameters in different operation stages.

[0037] Exemplarily, the intersection features can include the average value, variance, maximum value, minimum value, and volatility of the current in different operation stages.

[0038] Step S240, each local end trains the initial model of the local end based on the intersection features to obtain a target model, and sends the target model to the server.

[0039] In an embodiment of the present application, each local end can train the initial model of the local end based on the intersection features to obtain a target model, and each local end can send the target model to the server. Since the intersection features represent the types of feature values of the feature parameters in different action stages, after determining the intersection features, the local end can determine training samples in the full life cycle data based on the intersection features, and train the initial model based on the training samples to obtain a target model at the local end. Each target valve has multiple intersection features.

[0040] Exemplarily, the initial model can include any one of machine learning methods such as Multilayer Perceptron (MLP), Long Short-Term Memory network (LSTM), Temporal Convolutional Network (TCN), Transformer, etc. An operator can select one of the above multiple initial models as the initial model of the local end according to the characteristics of the target valve, or can also train multiple initial models based on the training samples, and determine the initial model with the best prediction effect as the target model.

[0041] In a possible implementation manner, the process in step S240 where each local end trains the initial model of the local end based on the intersection features to obtain a target model may include: Step 1: Use the full-cycle test time or number of times of the target valve at the local end as a label. For example, if the total number of full-cycle test actions of the target valve is times, the label for its th time is ; if the total number of full-cycle test actions of the target valve is times, the label for its th time is .

[0042] Step 2: Use linear function normalization, and use the maximum and minimum values of each feature parameter in the full life cycle data to scale the training samples and labels proportionally to the [0, 1] interval.

[0043] ; Among them, is the dataset corresponding to the th intersection feature of the target valve in the nuclear power system, is the normalized data, is the maximum value in the dataset corresponding to the th intersection feature, is the The minimum value in the dataset corresponding to the intersection feature.

[0044] Step 3: On the basis of normalization, in order to make full use of all data, perform a sliding window operation on the original data, and divide the data of the training set and the test set into training samples by combining the sliding window and slicing. Set the width of the sliding window and the sliding step, and use the historical slice data to test the data of the valve parameter to be detected at the current moment. The expression is as follows: ; ; ; Among them, is the data input to the initial model at time is the predicted value corresponding to the sample data at time is the sample label or the value predicted by the initial model (that is, the test time or number of times), is the data of the feature parameter at time is the number of sliding window slices, that is, the sampling dimension, is the data of the feature parameter in the previous 1 timestamp after the sliding window operation, is the data of the feature parameter in the previous timestamps after the sliding window operation, and so on.

[0045] Exemplarily, the number of sliding window slices can be set to 10. The input data for the th action of valve a should be the normalized data features from the -9th to th actions. The label data should be 1 - / n, that is, it is considered that the remaining life is (1 - / n) × 100%.

[0046] Step 4: The local end uses the data of the corresponding target valve for model training. In the initial model training of the local end, gradient descent updates the parameters of the initial model by calculating the gradient of the loss function. Assume the loss function is , where represents the parameters of the initial model, then the gradient descent update formula is: ; Among them, is the learning rate, is the gradient of the loss function with respect to the model parameters, is the updated parameter.

[0047] Step 5: Send the model with the best training result to the server as the target model.

[0048] Exemplarily, for the target valve a, an LSTM can be selected for training on the local side; for the target valve b, a Transformer can be selected for training on its local side.

[0049] Exemplarily, for the target valve a, an LSTM and a Transformer can be respectively selected for training on the local side, and the model with the best prediction effect is selected as the target model.

[0050] Step S250: The server obtains a global fusion model based on each target model and sends the global fusion model to the prediction end.

[0051] In an embodiment of the present application, the server can obtain a global fusion model based on each target model and send the global fusion model to the prediction end. The prediction end can be any target end in the nuclear power system, and the prediction end can also be other ends in the nuclear power system except the target end. By aggregating the target models from multiple local sides, the server can construct a more robust and generalized global fusion model. The data of different local sides may cover a wider range of situations, enabling the global fusion model to better adapt to various application environments.

[0052] Step S260: The prediction end obtains the current data of the valve to be predicted in the nuclear power system, and obtains the life prediction result of the valve to be predicted according to the current data of the valve to be predicted and the global fusion model.

[0053] In an embodiment of the present application, the prediction end can obtain the current data of the valve to be predicted in the nuclear power system, and obtain the life prediction result of the valve to be predicted according to the current data of the valve to be predicted and the global fusion model. The prediction end can be any one of the local sides, and the prediction end can also be other ends in the nuclear power system except the local side, that is, the valve to be predicted can be other valves in the nuclear power system except the target valve.

[0054] It should be noted that each prediction end can correspond to a valve to be predicted, that is, each prediction end can respectively obtain the relevant data of a valve to be predicted and predict the life of the valve to be predicted.

[0055] As an implementation manner, in the embodiment of the present application, step S220 includes: Step S221: For each feature parameter group of the full life cycle data, determine a plurality of feature arrays corresponding to the valve opening action and a plurality of feature arrays corresponding to the valve closing action.

[0056] Among them, the full life cycle data includes multiple characteristic parameter groups, each characteristic parameter group corresponds to a characteristic parameter, each characteristic parameter group includes the data of the characteristic parameter during the valve opening operation and the data of the valve closing operation, and the characteristic array includes multiple characteristic values of the same type.

[0057] In an embodiment of the present application, for each characteristic parameter group of the full life cycle data: determine the multiple characteristic arrays corresponding to the valve opening operation and the multiple characteristic arrays corresponding to the valve closing operation. Each action of the target valve may include a valve opening operation and a valve closing operation. The tester can pre-determine the time period corresponding to each valve opening operation (valve opening operation) and the time period corresponding to each valve closing operation (valve closing operation) during the valve usage process (or testing process); it can also be recognized by the local end to obtain the time period corresponding to the valve opening operation and the time period corresponding to the valve closing operation during the valve usage process (or testing process). Determine the valve opening data as the data in the full life cycle data whose time stamps are located in the time period corresponding to the valve opening operation, and determine the valve closing data as the data in the full life cycle data whose time stamps are located in the time period corresponding to the valve closing operation. Multiple characteristic arrays can be determined according to the valve opening data, and multiple characteristic arrays can be determined according to the valve closing data.

[0058] In an embodiment of the present application, the types of characteristic values include average value, variance, maximum value, minimum value, and volatility.

[0059] Since the valve has multiple actions during the usage process (or testing process), the valve opening data includes all the data corresponding to the valve opening actions, and the valve closing data includes all the data corresponding to the valve closing actions. For each characteristic parameter group, during one valve opening operation (or valve closing operation), there are values of multiple characteristic parameters, and multiple different types of characteristic values can be determined according to the values of the multiple characteristic parameters. For example, for the values of multiple characteristic parameters, the types of characteristic values that can be determined may include average value, variance, maximum value, minimum value, and volatility. The characteristic values of the same type corresponding to all valve opening operations (or valve closing operations) can be grouped into a set of characteristic arrays.

[0060] In an embodiment of the present application, the valve opening operation includes before valve opening stabilization, during valve opening stabilization, and after valve opening stabilization, and the valve closing operation includes before valve closing stabilization, during valve closing stabilization, and after valve closing stabilization. The time period corresponding to the valve opening operation can be divided to determine before valve opening stabilization, during valve opening stabilization, and after valve opening stabilization; the time period corresponding to the valve closing operation can be divided to determine before valve closing stabilization, during valve closing stabilization, and after valve closing stabilization.

[0061] Exemplarily, the duration of the time period corresponding to the valve opening action can be 20 s. The first 4 s can be determined as the period before the valve opening stabilizes, the middle 10 s can be determined as the period when the valve opening is stable, and the last 6 s can be determined as the period after the valve opening stabilizes. The determination methods for before the valve closing stabilizes, when the valve closing is stable, and after the valve closing stabilizes are similar to the determination methods for the relevant stages of the valve opening described above.

[0062] Exemplarily, the characteristic parameter can be current, and the types of characteristic values can include average value, variance, maximum value, minimum value, and volatility. One of the characteristic arrays can be the average value of the current before the valve opening stabilizes, and another characteristic array can be the minimum value of the current when the valve closing is stable.

[0063] Step S222: Determine the correlation between pairwise characteristic arrays, and determine the features to be screened from the flag features of the two characteristic arrays according to the correlation.

[0064] Among them, each characteristic array corresponds to a flag feature, and the flag feature represents the characteristic parameter, valve action stage, and type of characteristic value corresponding to the characteristic array. The valve action stage includes the valve opening action and the valve closing action.

[0065] In an embodiment of the present application, the correlation between pairwise characteristic arrays can be determined, and the features to be screened can be determined from the flag features of the two characteristic arrays according to the correlation. When determining the correlation between pairwise characteristic arrays, when the correlation between two groups of characteristic arrays is relatively strong, the flag feature of one of the characteristic arrays can be determined as the data to be screened, and the data can be dimensionally reduced to reduce the number of characteristic data. The features to be screened can be the features that have a greater impact on the valve life at the local end.

[0066] In a possible implementation manner, the process of step S222 for determining the correlation between pairwise characteristic arrays and determining the features to be screened from the flag features of the two characteristic arrays according to the correlation can include: determining the correlation coefficient between pairwise characteristic arrays; when the correlation coefficient is less than the preset value, determining the flag features of the two characteristic arrays as the features to be screened; when the correlation coefficient is greater than or equal to the preset value, determining the flag feature of one of the two characteristic arrays as the features to be screened.

[0067] It should be noted that first, the full life cycle data is expanded in terms of data volume through step S221, and then the correlation coefficient is used to reduce the dimension of the data volume. While ensuring the data volume, the characteristic data with strong correlation can be eliminated.

[0068] In one embodiment of the present application, the correlation coefficient is the Pearson correlation coefficient. Using the Pearson Correlation Coefficient can quickly determine the degree of association between two sets of feature arrays and compare the degrees of association between different pairs of feature arrays.

[0069] Exemplarily, the process of determining the correlation coefficient may include: ; where and are two sets of feature arrays respectively, is the correlation coefficient, is the average value of the feature array and is the average value of the feature array respectively.

[0070] Exemplarily, between two feature arrays with a correlation coefficient greater than 0.9, one of the signature features of the feature array can be determined as the feature to be screened. If there is a multiple relationship between the feature parameters in the two signature features (such as current and voltage), the signature feature corresponding to the current can be determined as the feature to be screened.

[0071] As an implementation manner, in the embodiment of the present application, step S260 includes: Step S261, determining the weighting coefficient according to the local verification results of each target model, or determining the weighting coefficient according to the number of training samples corresponding to each target model.

[0072] In one embodiment of the present application, the weighting coefficient can be determined according to the local verification results of each target model, or the weighting coefficient can be determined according to the number of training samples corresponding to each target model. After training the target model at the local end, the Mean Squared Error (MSE) or Root Mean Squared Error (RMSE) can be used to determine the difference between the predicted value and the true value using the test set, and the difference result is determined as the local verification result.

[0073] Step S262, performing weighted fusion on each target model according to each target model and the weighting coefficient to obtain a global fusion model.

[0074] In one embodiment of the present application, a global fusion model can be obtained by performing weighted fusion on each target model according to each target model and the weighting coefficient. The global fusion model can be obtained by performing weighted fusion on the outputs of each target model.

[0075] Exemplarily, the process of determining the global fusion model may include: ; Among them, is the output of the th target model, is the weight of the th target model, is the output of the global fusion model.

[0076] Figure 3 is a block diagram of a valve life prediction device for a nuclear power system shown in an exemplary embodiment of the present application. As Figure 3 shown, the exemplary valve life prediction device 300 for a nuclear power system includes: A data acquisition module 310, configured to acquire full life cycle data of a corresponding target valve for each local end. Each local end corresponds to a target valve, and the target valve is a valve in the nuclear power system.

[0077] A feature screening module 320, configured to extract multiple features to be screened from the full life cycle data for the local end and send the multiple features to be screened to the server.

[0078] A feature processing module 330, configured to determine intersection features among the features to be screened for the server and send the intersection features to each local end.

[0079] A model training module 340, configured to train an initial model of the local end based on the intersection features to obtain a target model and send the target model to the server.

[0080] A model fusion module 350, configured to obtain a global fusion model according to each target model for the server and send the global fusion model to the end to be predicted.

[0081] A life prediction module 360, configured to obtain current data of a valve to be predicted in the nuclear power system for the end to be predicted and obtain a life prediction result of the valve to be predicted according to the current data of the valve to be predicted and the global fusion model.

[0082] It should be noted that the valve life prediction device for a nuclear power system provided in the above embodiment and the nuclear power system valve life prediction method provided in the above embodiment belong to the same concept. The specific manners in which each module and unit perform operations have been described in detail in the method embodiment, and will not be elaborated herein. In practical applications, the valve life prediction device for a nuclear power system provided in the above embodiment may, according to needs, allocate the above functions to different functional modules, that is, divide the internal structure of the device into different functional modules to complete all or part of the functions described above. This is not limited herein either.

[0083] Embodiments of the present application also provide an electronic device, including: one or more processors; a storage device for storing one or more programs, which, when executed by the one or more processors, cause the electronic device to implement the nuclear power system valve life prediction method provided in each of the above embodiments.

[0084] On the other hand, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor of a computer, the computer is caused to execute the nuclear power system valve life prediction method provided in each of the above embodiments. The computer-readable storage medium may be included in the electronic device described in the above embodiments, or may exist alone without being assembled into the electronic device.

[0085] On the other hand, the present application also provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to execute the nuclear power system valve life prediction method provided in each of the above embodiments.

[0086] In the embodiments of the present application, the terms "first" and "second" are only used for descriptive purposes and should not be construed as indicating or implying relative importance. The terms "comprising" and "including" mentioned throughout the specification and claims are open-ended terms and should be interpreted as "including but not limited to".

[0087] The above embodiments are only used to exemplarily illustrate the principles and effects of the present application, rather than to limit the present application. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present application. Therefore, all equivalent modifications or changes made by those with ordinary knowledge in the technical field without departing from the spirit and technical idea disclosed by the present application should still be covered by the claims of the present application.

Claims

1. A method for predicting the life of valves in a nuclear power system, characterized in that: include: Each local terminal obtains the full life cycle data of the corresponding target valve, each local terminal corresponds to a target valve, and the target valve is a valve in the nuclear power system; The local end extracts features from the full life cycle data to obtain multiple features to be screened, and sends the multiple features to be screened to the server end; The server determines the intersection features among the features to be screened, and sends the intersection features to each local end; The local end trains the initial model of the local end based on the intersection feature to obtain a target model, and sends the target model to the server end; The server obtains a global fusion model based on each target model, and sends the global fusion model to the end to be predicted; The end to be predicted acquires current data of the valve to be predicted in the nuclear power system, and obtains a life prediction result of the valve to be predicted based on the current data of the valve to be predicted and the global fusion model.

2. The method for predicting the life of valves in a nuclear power system according to claim 1, characterized in that: The local end extracts features from the full life cycle data to obtain multiple features to be screened, including: For each characteristic parameter group of the full life cycle data: determine a plurality of characteristic arrays corresponding to the valve opening action and a plurality of characteristic arrays corresponding to the valve closing action; the full life cycle data includes a plurality of characteristic parameter groups, each characteristic parameter group corresponds to a characteristic parameter, each characteristic parameter group includes data of the characteristic parameter in the valve opening action and data of the valve closing action, and the characteristic array includes a plurality of characteristic values ​​of the same type; Determine the correlation between the feature arrays, and determine the features to be screened in the flag features of the two feature arrays based on the correlation. Each feature array corresponds to a flag feature, and the flag feature characterizes the feature parameters, valve action stages, and feature value types corresponding to the feature array. The valve action stages include valve opening action and valve closing action.

3. The method for predicting the life of valves in a nuclear power system according to claim 2, characterized in that: Types of eigenvalues ​​include mean, variance, maximum, minimum, and volatility.

4. The method for predicting the life of valves in a nuclear power system according to claim 2, characterized in that: The valve opening action includes before the valve is opened stably, when the valve is opened stably, and after the valve is opened stably. The valve closing action includes before the valve is closed stably, when the valve is closed stably, and after the valve is closed stably.

5. The method for predicting the life of valves in a nuclear power system according to claim 2, characterized in that: Determine the correlation between the two feature arrays, and determine the features to be screened from the signature features of the two feature arrays based on the correlation, including: Determine the correlation coefficient between two feature arrays; When the correlation coefficient is less than a preset value, the signature features of the two feature arrays are determined as features to be screened; When the correlation coefficient is greater than or equal to the preset value, the marker feature of one of the two feature arrays is determined as the feature to be screened.

6. The method for predicting the life of valves in a nuclear power system according to claim 5, characterized in that: The correlation coefficient is the Pearson correlation coefficient.

7. The method for predicting the life of valves in a nuclear power system according to claim 1, characterized in that: The server obtains a global fusion model according to each target model, and sends the global fusion model to the end to be predicted, including: Determine the weighting coefficient according to the local verification result of each target model, or determine the weighting coefficient according to the number of training samples corresponding to each target model; The global fusion model is obtained by weighted fusion of each target model according to each target model and the weight coefficient.

8. A valve life prediction device for a nuclear power system, characterized in that: include: A data acquisition module, used for each local terminal to acquire the full life cycle data of the corresponding target valve, each local terminal corresponds to a target valve, and the target valve is a valve in the nuclear power system; The feature screening module is used for extracting features from the whole life cycle data locally to obtain multiple features to be screened, and sending the multiple features to be screened to the server; The feature processing module is used for the server to determine the intersection feature among the features to be screened and send the intersection feature to each local end; A model training module is used for the local end to train the initial model of the local end based on the intersection feature to obtain a target model, and send the target model to the server end; A model fusion module is used for the server to obtain a global fusion model according to each target model, and send the global fusion model to the end to be predicted; The life prediction module is used for the end to be predicted to obtain the current data of the valve to be predicted in the nuclear power system, and obtain the life prediction result of the valve to be predicted based on the current data of the valve to be predicted and the global fusion model.

9. An electronic device, characterized in that: The electronic device comprises: one or more processors; a memory for storing program code executable by the processor; Wherein, the processor is configured to execute the program code to implement the nuclear power system valve life prediction method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor of a computer, the computer is enabled to execute the method for predicting the life of a valve in a nuclear power system as claimed in any one of claims 1 to 7.

Images