Transient operation parameter prediction method and device and electronic equipment

Abstract

The invention relates to the technical field of data processing, in particular to a transient operation parameter prediction method and device and electronic equipment. The method comprises the following steps: acquiring actual acquisition data acquired by at least one data acquisition device in the nuclear power station at the current moment; inputting all the actually acquired data into a prediction model for prediction to obtain a predicted value of the transient operation parameter within a preset duration after the current moment; wherein the prediction model is obtained through training of historical collection data collected by all the data collection devices, and the transient operation parameters comprise one or more of the steam generator outlet steam mass flow, the voltage stabilizer fluctuation pipe temperature and the coolant flow; and generating a prediction result of the transient operation parameter based on the prediction value of the transient operation parameter.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and in particular to methods, apparatus and electronic devices for predicting transient operating parameters. Background Technology

[0002] Transient operating parameters (such as temperature and flow rate) of nuclear power plants are crucial to their safety. Currently, the prediction of transient operating parameters mainly relies on two methods: one is a calculation model based on physical principles, but the calculation is too complex and difficult to meet the rapid response requirements of real-time control; the other is to rely on the operator's experience and judgment, but this method is highly subjective, lacks quantitative standards, and is prone to errors in emergency situations.

[0003] Therefore, how to quickly and accurately predict transient operating parameters has become an urgent problem to be solved. Summary of the Invention

[0004] To address the aforementioned technical problems, this disclosure provides a method, apparatus, and electronic device for predicting transient operating parameters.

[0005] In a first aspect, this disclosure provides a method for predicting transient operating parameters, comprising: acquiring actual data collected at the current moment by at least one data acquisition device in a nuclear power plant; inputting all the actual data collected into a prediction model for prediction, thereby obtaining predicted values ​​of transient operating parameters within a preset time period after the current moment; wherein the prediction model is trained using historical data collected by all data acquisition devices, and the transient operating parameters include one or more of the following: steam mass flow rate at the steam generator outlet, pressurizer oscillation tube temperature, and coolant flow rate; and generating prediction results of the transient operating parameters based on the predicted values ​​of the transient operating parameters.

[0006] Secondly, this disclosure provides a device for predicting transient operating parameters, comprising: an acquisition unit for acquiring actual data collected at least once by a data acquisition device in a nuclear power plant at the current moment; a processing unit for inputting all the actual data acquired by the acquisition unit into a prediction model for prediction, thereby obtaining predicted values ​​of transient operating parameters within a preset time period after the current moment; wherein the prediction model is trained using historical data collected by all data acquisition devices, and the transient operating parameters include one or more of the following: steam mass flow rate at the steam generator outlet, pressure regulator oscillation tube temperature, and coolant flow rate; the processing unit is further configured to generate prediction results of transient operating parameters based on the predicted values ​​of the transient operating parameters.

[0007] Thirdly, this disclosure provides an electronic device, including: a communication interface, a processor, a memory, and a bus; the memory stores computer-executed instructions, and the processor is connected to the memory via the bus. When the electronic device is running, the processor executes the computer-executed instructions stored in the memory to cause the electronic device to perform the transient operating parameter prediction method provided in the first aspect above.

[0008] Fourthly, this disclosure provides a computer-readable storage medium including instructions. When executed on a computer, the instructions cause the computer to perform the method for predicting transient operating parameters as described in the first aspect above.

[0009] Fifthly, this disclosure provides a computer program product that, when run on a computer, causes the computer to perform a method for predicting transient operating parameters as described in the first aspect.

[0010] It should be noted that the aforementioned computer instructions may be stored, in whole or in part, on the first computer-readable storage medium. The first computer-readable storage medium may be packaged together with the processor of the transient operating parameter prediction device, or it may be packaged separately from the processor of the transient operating parameter prediction device; this disclosure does not limit this.

[0011] The descriptions of the second, third, fourth, and fifth aspects in this disclosure can be referenced to the detailed description of the first aspect; and the beneficial effects of the descriptions of the second, third, fourth, and fifth aspects can be referenced to the analysis of the beneficial effects of the first aspect, which will not be repeated here.

[0012] In this disclosure, the name of the aforementioned transient operating parameter prediction device does not limit the device or functional module itself. In actual implementation, these devices or functional modules may appear under other names. As long as the function of each device or functional module is similar to that of this disclosure, it falls within the scope of the claims of this disclosure and its equivalents.

[0013] These or other aspects of this disclosure will become more readily apparent in the following description.

[0014] The technical solution provided in this disclosure has the following advantages compared with the prior art:

[0015] The transient operating parameter prediction method disclosed herein pre-trains a model using historical data collected by all data acquisition devices to obtain the corresponding transient operating parameter prediction model. This allows for the acquisition of actual data collected at least once by at least one data acquisition device in the nuclear power plant at the current moment; all actual data is then input into the prediction model for prediction, yielding predicted values ​​of transient operating parameters within a preset timeframe after the current moment. Since the prediction model is trained based on historical data collected by all data acquisition devices, rather than on physical principles, the computational difficulty is significantly reduced, and the prediction time is shortened. Furthermore, because the prediction model does not rely on human experience, subjectivity is avoided, improving prediction accuracy. Subsequently, based on the predicted values ​​of the transient operating parameters, prediction results for the transient operating parameters are generated. Due to the higher timeliness and accuracy of the prediction results, maintenance personnel can better utilize these results to provide intelligent support for the operation management, status monitoring, and fault prevention of the nuclear power plant, significantly improving the safety and operational efficiency of the nuclear power plant and solving the problem of how to quickly and accurately predict transient operating parameters. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0017] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is one of the flowcharts illustrating the transient operating parameter prediction method described in this embodiment of the present disclosure;

[0019] Figure 2 This is one of the application scenarios of the transient operating parameter prediction method described in the embodiments of this disclosure;

[0020] Figure 3 This is a second schematic diagram illustrating the application scenario of the transient operating parameter prediction method described in this embodiment of the present disclosure.

[0021] Figure 4 This is a second schematic flowchart of the transient operating parameter prediction method described in the embodiments of this disclosure;

[0022] Figure 5 This is the third flowchart illustrating the transient operating parameter prediction method described in this embodiment of the present disclosure;

[0023] Figure 6 This is a fourth flowchart illustrating the transient operating parameter prediction method described in the embodiments of this disclosure;

[0024] Figure 7 This is the fifth flowchart illustrating the transient operating parameter prediction method described in the embodiments of this disclosure;

[0025] Figure 8 This is a schematic diagram of the structure of the transient operating parameter prediction device described in the embodiments of this disclosure;

[0026] Figure 9 This is a schematic diagram of the server structure described in an embodiment of this disclosure;

[0027] Figure 10 A schematic diagram of the structure of a computer program product for the transient operating parameter prediction method provided in an embodiment of the present invention. Detailed Implementation

[0028] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the solutions disclosed herein will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0029] Numerous specific details are set forth in the following description in order to provide a full understanding of this disclosure, but this disclosure may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some, and not all, of the embodiments of this disclosure.

[0030] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0031] It should be noted that when the electronic device executes the transient operating parameter prediction method provided in the embodiments of this disclosure, the executing entity is the processor of the electronic device. For example, when the electronic device is a server, the executing entity for executing the transient operating parameter prediction method provided in the embodiments of this disclosure is the processor of the server.

[0032] For example, taking an electronic device as a server, the method for predicting transient operating parameters provided in this disclosure will be described. Figure 1 The method for predicting transient operating parameters shown includes the following steps S11-S13:

[0033] S11. Obtain the actual data collected at the current moment by at least one data acquisition device in the nuclear power plant.

[0034] In some examples, the data acquisition device can be a sensor.

[0035] In some examples, for ease of management, data acquisition devices are typically managed through the nuclear power plant's Distributed Control System (DCS). For instance, when the data acquisition device is a sensor, maintenance personnel manage the sensor through the DCS. Thus, the DCS can acquire the actual data collected at any given moment by at least one data acquisition device in the nuclear power plant, such as the steam mass flow rate at the steam generator outlet, the pressurizer sloshing tube temperature, and the coolant flow rate.

[0036] In some examples, the transient operating parameter prediction method provided in embodiments of this disclosure is applied to, for example... Figure 2 The schematic diagram shown includes an industrial control area and a management area. The industrial control area includes sensor nodes, a nuclear power plant's distributed control system (DCS), a data acquisition module, and a one-way gateway. The management area includes a data processing module, a core algorithm module, a deep feedforward neural network (DNN), data cleaning and formatting processing, a topology optimization module (quantum genetic algorithm (QGA)), an initial parameter optimization module (model-independent meta-learning algorithm (MAML)), a prediction output module, an engineering database, and a visualization interface.

[0037] Figure 2 The data acquisition module in the system collects transient thermal hydraulic operation data (including key parameters such as steam mass flow rate at the steam generator outlet, temperature of the pressure regulator fluctuating tube, and coolant flow rate) from sensor nodes (each sensor node has at least one sensor deployed under it) through the DCS. It then amplifies and filters the raw electrical signals corresponding to the transient thermal hydraulic operation data, converts them into digital signals, and transmits them securely to the management area through a one-way network gateway to ensure network isolation between the industrial control area and the management area and prevent external interference.

[0038] Figure 2The data processing module is used to receive raw data (such as digital signals corresponding to transient thermal hydraulic operation data) from the one-way gate, and to perform data cleaning, deduplication and formatting on the digital signals to generate a structured dataset D = {(X, Y)}; where X is the transient operation parameter at any time, and Y is the parameter label. If X is the raw data at any time, Y is the raw data within a preset time period that is continuous with X.

[0039] In some examples, the data cleaning process includes outlier detection and correction (such as noise removal through mean-variance analysis) to ensure the data quality of the structured dataset D. For instance, with a preset duration of 5 seconds and a sensor acquisition cycle of 1 second, if X includes 1, 2, 3, 4, 5, 6, 7, 8, and 9, when x is 1, y is 2, 3, 4, 5, 6; when x is 2, y is 3, 4, 5, 6, 7; when x is 3, y is 4, 5, 6, 7, 8; and when x is 4, y is 5, 6, 7, 8, 9.

[0040] Figure 2 The core algorithm module employs a Deep Neural Network (DNN) as its core prediction algorithm. Based on a structured dataset D, it predicts the transient operating parameters within a preset duration, such as 1, 2, 3, 4, or 5 seconds. Based on these predicted values, the algorithm can then derive the trend of transient parameter changes. The DNN network structure is as follows: Figure 3 As shown, it includes an input layer, a hidden layer, and an output layer.

[0041] In some examples, the transient operating parameter prediction method provided by the embodiments of the present invention captures time series features through multi-layer nonlinear mapping, thereby ensuring the prediction accuracy of transient operating parameters. For example, the prediction error of steam mass flow rate at the steam generator outlet is controlled within 0 to 4%, and the prediction error of temperature fluctuation tube of the pressure regulator is controlled within 0.2 to 0.3%.

[0042] In some examples, the server executing the transient operating parameter prediction method provided in this disclosure connects to a Supervisory Control and Data Acquisition (SCADA) system, a Supervisory Information System at Plant Level (SIS), and an Enterprise Resource Planning (ERP) system via a RESTful API interface, and uses OAuth 2.0 authentication to ensure secure data exchange, such as... Figure 2The bidirectional arrows in the engineering database indicate this. The server obtains real-time operational data from the SCADA system, acquires equipment status information from the SIS system, and feeds back the prediction results to the ERP system for operation management and maintenance decisions. It can be seen that the transient operational parameter prediction method provided in this disclosure supports functional expansion through modular design, such as integrating predictive maintenance algorithms and analyzing historical data and fault records to predict potential equipment failures.

[0043] Furthermore, topology analysis of DNNs provides guidance for the optimization of nuclear power systems. By analyzing the number of layers and node distribution, key factors affecting thermal-hydraulic parameters can be identified, providing a reference direction for the optimization of physical calculation programs. This feedback mechanism enables the system not only to achieve forward prediction but also to support the design optimization of nuclear power plant systems.

[0044] S12. Input all the actual collected data into the prediction model for prediction, and obtain the predicted values ​​of the transient operating parameters within a preset time period after the current moment. The prediction model is trained using historical data collected by all data acquisition devices. The transient operating parameters include one or more of the following: steam mass flow rate at the steam generator outlet, pressure regulator oscillation tube temperature, and coolant flow rate.

[0045] S13. Based on the predicted values ​​of transient operating parameters, generate the prediction results of transient operating parameters.

[0046] In some examples, Figure 2 The prediction output module generates prediction results based on the optimized DNN and connects to the visualization interface. For example, it displays a comparison between the predicted curve and the actual value (such as "prediction of steam mass flow rate in 1 second"). The prediction time is approximately 0.7 milliseconds, which meets the real-time requirements.

[0047] In some examples, Figure 2 The engineering database in the system employs a Distributed Database Management System (DDBMS) to store transient operating data, transient operating parameters, DNN model parameters, and prediction results for thermal hydraulics. The engineering database can interact with the data processing module, prediction output module, and visualization interface, supporting time-series data management and real-time updates.

[0048] As described above, the transient operating parameter prediction method provided in this embodiment of the present disclosure trains the model in advance using historical acquisition data collected by all data acquisition devices, thereby obtaining the corresponding transient operating parameter prediction model. In this way, by acquiring the actual acquisition data collected by at least one data acquisition device in the nuclear power plant at the current moment, and inputting all the actual acquisition data into the prediction model for prediction, the predicted values ​​of transient operating parameters within a preset time period after the current moment can be obtained. Since the prediction model is trained based on historical acquisition data collected by all data acquisition devices, rather than being calculated based on physical principles, the computational difficulty is greatly reduced, and the prediction time is shortened. Simultaneously, since the prediction model does not require prediction based on human experience, subjectivity is avoided, and the prediction accuracy is improved. Then, based on the predicted values ​​of the transient operating parameters, the prediction results of the transient operating parameters are generated. Because the prediction results are more timely and accurate, operation and maintenance personnel can better provide intelligent support for the operation management, status monitoring, and fault prevention of the nuclear power plant based on the prediction results, significantly improving the safety and operational efficiency of the nuclear power plant.

[0049] In some feasible examples, combining Figure 1 ,like Figure 4 As shown, the transient operating parameter prediction method provided in this embodiment of the present disclosure can also execute S14-S16 before executing S11.

[0050] S14. Obtain training sample data. The training sample data includes historical data continuously collected by the data acquisition device.

[0051] S15. Perform preprocessing operations on the training sample data to obtain preprocessed data. The preprocessing operations include one or more of data cleaning and formatting. The preprocessed data includes training data and labeling results. The training data includes any item of historical data, and the labeling results include historical data within a preset time period continuous with the training data.

[0052] In some examples, the preprocessing operation also includes data deduplication.

[0053] S16. Train the target depth feedforward neural network based on the preprocessed data until the target depth feedforward neural network converges to obtain the prediction model.

[0054] In some examples, to reduce the cost and time of model development, the transient operating parameter prediction method provided in this embodiment of the invention, through the collaborative optimization of Quantum Genetic Algorithm (QGA) and / or Model-Agnostic Meta-Learning (MAML), allows the system to autonomously complete the topology design, parameter initialization, and training of the DNN model without manual intervention, significantly reducing the cost and time of model development. For example, by obtaining an initial deep feedforward neural network and optimizing its topology parameters based on the quantum genetic algorithm and preprocessed data, the target structure parameters are obtained; the initial deep feedforward neural network is then set according to the target structure parameters to obtain the target deep feedforward neural network.

[0055] In some examples, when optimizing the topology parameters of an initial deep feedforward neural network using a quantum genetic algorithm and preprocessed data, multiple optimization parameters (number of layers L and number of nodes H) can be obtained based on the quantum genetic algorithm. Then, different topology parameters for the initial deep feedforward neural network are set according to each optimization parameter, resulting in the corresponding optimized deep feedforward neural network. Next, for each transient operating parameter, the training data from the preprocessed data corresponding to that transient operating parameter is input into each optimized deep feedforward neural network for iterative processing until the number of iterations equals the iteration threshold M. The loss value between the prediction result and the labeled result for that transient operating parameter in each optimized deep feedforward neural network at the Mth iteration is calculated. The optimized parameters of the optimized deep feedforward neural network corresponding to the minimum loss value are taken as the preset structural parameters for that transient operating parameter. (If multiple minimum loss values ​​exist, the optimized parameters of the DNN model corresponding to any one of these loss values ​​are selected as the preset structural parameters). Next, the target structure parameters are obtained by summarizing the preset structure parameters corresponding to all transient operating parameters. For example, the average number of layers L and the average number of nodes H corresponding to all transient operating parameters are calculated, and the average number of layers L and the average number of nodes H are used as the target structure parameters. Alternatively, different initial deep feedforward neural network topology parameters are set according to each preset structure parameter to obtain the candidate deep feedforward neural networks. Then, for each transient operating parameter, the training data in the preprocessed data corresponding to each transient operating parameter is input into each candidate deep feedforward neural network for iterative iteration until the number of iterations equals the iteration threshold M. The loss value of the prediction result and the label result of each transient operating parameter in the Mth iteration of each candidate deep feedforward neural network is calculated, and the average value of the loss values ​​of all transient operating parameters in the Mth iteration of each candidate deep feedforward neural network is calculated. The preset structure parameters of the candidate deep feedforward neural network corresponding to the smallest average value are used as the target structure parameters. Here, M is an integer greater than or equal to 1.

[0056] In some examples, the target structure parameters include the number of optimized input layers, the number of hidden layers, the number of output layers, and the number of nodes contained in each hidden layer.

[0057] In some examples, the number of input layers is typically 1, and the number of output layers is typically 1.

[0058] In some examples, Figure 2 The topology optimization module in the DNN uses QGA to search for the optimal topology parameters (number of layers L and number of nodes H). Figure 2The QGA swimlane framework comprises four sub-steps: initializing the qubit population, quantum encoding, fitness evaluation, and quantum gate update. Optimization is achieved via dashed cyclic arrows (labeled "iteration to convergence"). Initializing the qubit population yields a set of quantum chromosomes, each composed of multiple qubits. The qubit population includes at least one qubit. Unlike traditional bits that can only be 0 or 1, a qubit simultaneously represents both 0 and 1 with a certain probability; this state is called a superposition. Next, each quantum chromosome is analyzed according to a predefined quantum encoding to obtain the corresponding quantum collapse, thus completing the quantum encoding. Then, a predefined fitness function is used to evaluate each quantum collapse, obtaining a score for each quantum collapse to complete the fitness evaluation. Finally, based on the scores, the topological parameters of the DNN are adjusted until the globally optimal or near-optimal solution is obtained as the target structural parameters.

[0059] In some examples, to shorten the model construction time, it can be assumed that when the number of iterations for optimizing the topology parameters of the initial deep feedforward neural network based on quantum genetic algorithms and preprocessed data is equal to N, the topology parameters obtained on the Nth iteration are considered to be the target topology parameters. For example, N∈[20, 80].

[0060] In some examples, assuming N equals 80, and the DNN contains one input layer and one output layer, when predicting the mass flow rate of steam at the steam generator outlet, after 80 iterations, the QGA determines L = 38 and H = 24, with the loss function value decreasing to 3.106. Since the DNN contains one input layer and one output layer, it contains 36 hidden layers (38-1-1), with 24 nodes in each hidden layer. When predicting the temperature of the regulator's fluctuating tube, after 80 iterations, the QGA determines L = 68 and H = 43, with the loss function value decreasing to 0.0302. Therefore, the DNN contains 66 hidden layers, with 43 nodes in each hidden layer. The quantum encoding characteristic of the QGA makes its search efficiency higher than that of traditional genetic algorithms, avoiding getting trapped in local optima. Similarly, if L = 68 and H = 43 (fluctuating tube temperature), since the DNN contains one input layer and one output layer... Since QGA expands the search range by 2 times compared to traditional genetic algorithms, i.e., QGA's quantum encoding expands the search range, the number of iterations is reduced to 20-80, thereby shortening the construction time of DNN.

[0061] Alternatively, by obtaining an initial deep feedforward neural network and optimizing the initial weight matrix W and bias matrix B of the initial deep feedforward neural network based on MAML and preprocessed data, the optimized initial weight matrix W and bias matrix B are obtained; the initial deep feedforward neural network is set according to the optimized initial weight matrix W and bias matrix B to obtain the target deep feedforward neural network.

[0062] In some examples, Figure 2 The initial parameter optimization module in the DNN uses MAML to optimize the initial weight matrix W and bias matrix B. Figure 2 The MAML swimlane box contains four sub-steps: creating a few-sample training task set, meta-learning inner loop, outer loop, and parameter optimization. Nested optimization is achieved through dashed loop arrows (labeled "inner loop iteration" and "outer loop iteration").

[0063] For example, an initial deep feedforward neural network (DNN) can be trained by sampling task sets (each task containing a small number of samples) from multiple training datasets. This ensures that the DNN can learn common initialization parameters from different tasks, thus creating a small-sample training task set. Then, gradients are calculated using the small amount of training data within each task to adjust the model parameters (task-specific parameters) of the DNN, optimizing its performance on that task, completing the meta-learning inner loop. Afterward, by calculating the test loss of the prediction and label results of the DNN after completing the meta-learning inner loop, the initial parameters are updated using gradient descent, enabling the model to converge faster on future tasks, completing the meta-learning outer loop. Finally, through alternating iterations of the inner and outer loops, the initialization parameters of the DNN are optimized, completing parameter optimization. Using a model-independent meta-learning algorithm and preprocessing data to train the DNN allows for training with less data, thus reducing the number of training iterations. Fewer training iterations shorten the DNN's construction time.

[0064] Alternatively, an initial deep feedforward neural network can be obtained, and its topological parameters can be optimized using a quantum genetic algorithm and preprocessed data to obtain target structural parameters. The initial deep feedforward neural network can then be configured according to these target structural parameters to obtain an optimized deep feedforward neural network. Subsequently, based on MAML and preprocessed data, the initial weight matrix W and bias matrix B of the optimized deep feedforward neural network are optimized to obtain optimized initial weight matrices W and B. The optimized deep feedforward neural network is then configured according to the optimized initial weight matrix W and bias matrix B to obtain the target deep feedforward neural network.

[0065] In some examples, when training a target deep feedforward neural network (DNN) based on preprocessed data, a QGA (Quantum Gaussian Approach) can be used to search for the optimal topological parameters of the DNN. This involves initializing a qubit population to obtain a group of quantum chromosomes, each composed of multiple qubits. The qubit population includes at least one qubit. Unlike traditional bits that can only be 0 or 1, a qubit simultaneously represents both 0 and 1 with a certain probability; this state is called a superposition state. Then, each quantum chromosome is analyzed according to a predefined quantum encoding to obtain the corresponding quantum collapse, thus completing the quantum encoding. Next, a predefined fitness function is used to evaluate each quantum collapse, obtaining a score for each quantum collapse to complete the fitness assessment. Based on the scores, the topological parameters of the DNN are adjusted until a globally optimal or near-optimal solution is obtained as the target structure parameters. Finally, the initial deep feedforward neural network is set according to the target structure parameters to obtain the optimized deep feedforward neural network. Subsequently, a task set (each task containing a small number of samples) can be sampled from multiple training datasets to train the initial deep feedforward neural network. This ensures that the initial deep feedforward neural network can learn common initialization parameters from different tasks, thus creating a small-sample training task set. Next, gradients are calculated using the small amount of training data within each task to adjust the model parameters (task-specific parameters) of the initial deep feedforward neural network, optimizing its performance on that task, thus completing the meta-learning inner loop. Afterward, by calculating the test loss of the prediction and label results of the initial deep feedforward neural network after completing the meta-learning inner loop, the initial parameters are updated using gradient descent, enabling the model to converge faster on future new tasks, thus completing the meta-learning outer loop. Finally, through alternating iterations of the inner and outer loops, the initialization parameters of the initial deep feedforward neural network are optimized, completing parameter optimization. Using a model-independent meta-learning algorithm and preprocessing data to train the initial deep feedforward neural network allows for training with less data, thereby reducing the number of training iterations. Since the number of training iterations is reduced, the construction time of the DNN can be shortened.

[0066] As can be seen from the above, the transient operating parameter prediction method provided in this embodiment of the present disclosure can greatly reduce the difficulty of calculation and shorten the prediction time because the prediction model is trained based on historical data collected by all data acquisition devices, rather than being calculated based on physical principles. At the same time, since the prediction model does not need to be based on human experience, subjectivity is avoided and the accuracy of prediction is improved.

[0067] In some feasible examples, combining Figure 4 like Figure 5As shown, the transient operating parameter prediction method provided in this embodiment of the present disclosure can also execute S17-S19 before executing S14.

[0068] S17. Obtain the initial deep feedforward neural network.

[0069] S18. Based on the quantum genetic algorithm and preprocessed data, the topology parameters of the initial deep feedforward neural network are optimized to obtain the target structure parameters. The topology parameters include at least the number of input layers, the number of hidden layers, the number of output layers, and the number of nodes in each hidden layer.

[0070] S19. Set the initial deep feedforward neural network according to the target structure parameters to obtain the target deep feedforward neural network.

[0071] As described above, the transient operating parameter prediction method provided in this embodiment obtains an initial deep feedforward neural network, optimizes its topological parameters based on a quantum genetic algorithm and preprocessed data to obtain target structural parameters, and then sets the initial deep feedforward neural network according to the target structural parameters to obtain the target deep feedforward neural network. Since the search range of QGA is twice that of the traditional genetic algorithm, i.e., QGA's quantum encoding expands the search range, the number of iterations of the DNN model can be reduced, thus shortening the DNN construction time.

[0072] In some feasible examples, combining Figure 4 ,like Figure 6 As shown, the above S16 can be implemented through the following S160 and S161.

[0073] S160. Input the training data into the target deep feedforward neural network to obtain the prediction results of the target deep feedforward neural network on the training data.

[0074] S161. Based on the prediction results and labeling results, adjust the network parameters of the target depth feedforward neural network until the target depth feedforward neural network converges to obtain the prediction model.

[0075] In some examples, the loss values ​​for the prediction and labeling results can be calculated based on the loss function. If the loss value is greater than the loss threshold, the network parameters of the target depth feedforward neural network are adjusted until the loss value is less than or equal to the loss threshold. At this point, the target depth feedforward neural network is considered to have converged, and the converged target depth feedforward neural network is used as the prediction model.

[0076] As can be seen from the above, the transient operating parameter prediction method provided in this embodiment of the present disclosure can greatly reduce the difficulty of calculation and shorten the prediction time because the prediction model is trained based on historical data collected by all data acquisition devices, rather than being calculated based on physical principles. At the same time, since the prediction model does not need to be based on human experience, subjectivity is avoided and the accuracy of prediction is improved.

[0077] In some feasible examples, combining Figure 4 ,like Figure 7 As shown, the above S16 can be specifically implemented through the following S163.

[0078] S163. Based on a model-independent meta-learning algorithm, preprocessed data is used to train the target deep feedforward neural network until the target deep feedforward neural network converges, thus obtaining the prediction model.

[0079] In some examples, Figure 2 The initial parameter optimization module in the DNN uses MAML to optimize the initial weight matrix W and bias matrix B. Figure 2 The MAML swimlane box contains four sub-steps: creating a few-sample training task set, meta-learning inner loop, outer loop, and parameter optimization. Nested optimization is achieved through dashed loop arrows (labeled "inner loop iteration" and "outer loop iteration").

[0080] For example, a target deep feedforward neural network can be trained by sampling task sets (each task containing a small number of samples) from multiple training datasets. This ensures that the target deep feedforward neural network can learn common initialization parameters from different tasks, thus creating a small-sample training task set. Then, gradients are calculated using the small amount of training data within each task to adjust the model parameters (task-specific parameters) of the target deep feedforward neural network, optimizing its performance on that task, thus completing the meta-learning inner loop. Afterward, by calculating the test loss of the prediction and label results of the target deep feedforward neural network after completing the meta-learning inner loop, the initial parameters are updated using gradient descent, enabling the model to converge faster on future new tasks, thus completing the meta-learning outer loop. Finally, through alternating iterations of the inner and outer loops, the initialization parameters of the target deep feedforward neural network are optimized, completing parameter optimization. When using a model-independent meta-learning algorithm and preprocessing data to train the target deep feedforward neural network, less data can be used for training, thereby reducing the number of training iterations. Since the number of training iterations is reduced, the training time of the target deep feedforward neural network can be shortened.

[0081] As can be seen from the above, the transient operating parameter prediction method provided in this embodiment of the present disclosure can greatly reduce the difficulty of calculation and shorten the prediction time because the prediction model is trained based on historical data collected by all data acquisition devices, rather than being calculated based on physical principles. At the same time, since the prediction model does not need to be based on human experience, subjectivity is avoided and the accuracy of prediction is improved.

[0082] The foregoing primarily describes the solutions provided by the embodiments of this disclosure from a methodological perspective. To achieve the aforementioned functions, it includes corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein, this disclosure can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.

[0083] This disclosure embodiment can divide the transient operating parameter prediction device into functional modules according to the above method example. For example, each function can be divided into its own functional module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. It should be noted that the module division in this disclosure embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0084] like Figure 8 As shown, the transient operating parameter prediction device 10 provided in this embodiment includes an acquisition unit 101 and a processing unit 102.

[0085] The acquisition unit 101 is used to acquire actual data collected by at least one data acquisition device in the nuclear power plant at the current moment; the processing unit 102 is used to input all the actual data acquired by the acquisition unit 101 into the prediction model for prediction, and obtain the predicted values ​​of transient operating parameters within a preset time period after the current moment; wherein, the prediction model is trained by all the historical data collected by the data acquisition devices, and the transient operating parameters include one or more of the following: steam mass flow rate at the steam generator outlet, pressurizer oscillation tube temperature, and coolant flow rate; the processing unit 102 is also used to generate the prediction results of the transient operating parameters based on the predicted values ​​of the transient operating parameters.

[0086] In some feasible examples, the acquisition unit 101 is further configured to acquire training sample data; wherein the training sample data includes historical acquisition data continuously acquired by the data acquisition device; the processing unit 102 is further configured to perform preprocessing operations on the training sample data acquired by the acquisition unit 101 to obtain preprocessed data; wherein the preprocessing operations include one or more of data cleaning and formatting processing, the preprocessed data includes training data and labeling results, the training data includes any item of historical acquisition data, and the labeling results include historical acquisition data within a preset time period continuous with the training data; the processing unit 102 is further configured to train the target deep feedforward neural network based on the preprocessed data until the target deep feedforward neural network converges to obtain a prediction model.

[0087] In some feasible examples, the acquisition unit 101 is further configured to acquire an initial deep feedforward neural network; the processing unit 102 is further configured to optimize the topological parameters of the initial deep feedforward neural network based on a quantum genetic algorithm and preprocessed data to obtain target structural parameters; wherein, the topological parameters include at least the number of input layers, the number of hidden layers, the number of output layers, and the number of nodes contained in each hidden layer; the processing unit 102 is further configured to set the initial deep feedforward neural network according to the target structural parameters to obtain the target deep feedforward neural network.

[0088] In some feasible examples, the processing unit 102 is specifically used to input training data into the target deep feedforward neural network to obtain the prediction result of the target deep feedforward neural network on the training data; the processing unit 102 is specifically used to adjust the network parameters of the target deep feedforward neural network based on the prediction result and the labeling result until the target deep feedforward neural network converges to obtain the prediction model.

[0089] In some feasible examples, the processing unit 102 is specifically used to train the target deep feedforward neural network based on a model-independent meta-learning algorithm and preprocessed data until the target deep feedforward neural network converges to obtain a prediction model.

[0090] All relevant content of each step involved in the above method embodiments can be referenced from the functional description of the corresponding functional module, and their functions will not be repeated here.

[0091] Of course, the transient operating parameter prediction device 10 provided in this embodiment includes, but is not limited to, the modules described above. For example, the transient operating parameter prediction device 10 may also include a storage unit 103. The storage unit 103 may be used to store the program code of the transient operating parameter prediction device 10, and may also be used to store data generated by the transient operating parameter prediction device 10 during operation, such as data in a write request.

[0092] Figure 9 This is a schematic diagram of the structure of a server provided in an embodiment of the present disclosure, such as... Figure 9 As shown, the server may include at least one processor 51, a memory 52, a communication interface 53, and a communication bus 54.

[0093] The following is combined Figure 9 A detailed introduction to each component of the server:

[0094] The processor 51 is the control center of the server and can be a single processor or a collective term for multiple processing elements. For example, the processor 51 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this disclosure, such as one or more DSPs, or one or more field-programmable gate arrays (FPGAs).

[0095] In a specific implementation, as one example, the processor 51 may include one or more CPUs, for example... Figure 9 CPU0 and CPU1 are shown in the diagram. Furthermore, as one embodiment, the server may include multiple processors, for example... Figure 9 The processors 51 and 55 are shown. Each of these processors can be a single-core processor (Single-CPU) or a multi-core processor (Multi-CPU). Here, "processor" can refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0096] The memory 52 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or it may be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. The memory 52 may exist independently and be connected to the processor 51 via a communication bus 54. The memory 52 may also be integrated with the processor 51.

[0097] In a specific implementation, memory 52 is used to store data disclosed herein and execute software programs disclosed herein. Processor 51 can perform various functions of the air conditioner by running or executing software programs stored in memory 52 and by calling data stored in memory 52.

[0098] Communication interface 53 uses any transceiver-like device for communicating with other devices or communication networks, such as Radio Access Network (RAN), Wireless Local Area Networks (WLAN), terminals, and the cloud. Communication interface 53 may include acquisition unit 101 to implement acquisition functions.

[0099] The communication bus 54 can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0100] As an example, combined Figure 8The functions implemented by the acquisition unit 101 in the server are the same as Figure 9 The communication interface 53 has the same function, and the processing unit 102 implements the same function. Figure 9 The processor 51 in the memory unit has the same function as the memory unit 103. Figure 9 The memory 52 in it has the same function.

[0101] Another embodiment of this disclosure also provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the methods shown in the above-described method embodiments.

[0102] In some embodiments, the disclosed method may be implemented as computer program instructions encoded in a machine-readable format on a computer-readable storage medium or on other non-transitory media or articles of art.

[0103] Figure 10 A conceptual partial view of a computer program product provided in an embodiment of the present disclosure is shown schematically. The computer program product includes a computer program for executing computer processes on a computing device.

[0104] In one embodiment, the computer program product is provided using signal bearer medium 410. Signal bearer medium 410 may include one or more program instructions that, when executed by one or more processors, can provide the above-mentioned... Figure 1 The described function or part of the function. Therefore, for example, refer to... Figure 1 In the embodiment shown, one or more features of S11-S13 can be fulfilled by one or more instructions associated with the signal carrying medium 410. Furthermore, Figure 10 The program instructions in the document also describe example instructions.

[0105] In some examples, the signal carrying medium 410 may include a computer-readable medium 411, such as, but not limited to, a hard disk drive, a compact disc (CD), a digital video disc (DVD), a digital magnetic tape, a memory, a read-only memory (ROM), or a random access memory (RAM), etc.

[0106] In some implementations, the signal carrying medium 410 may include a computer recordable medium 412, such as, but not limited to, a memory, a read / write (R / W) CD, a R / W DVD, and so on.

[0107] In some implementations, the signal carrying medium 410 may include a communication medium 413, such as, but not limited to, digital and / or analog communication media (e.g., fiber optic cables, waveguides, wired communication links, wireless communication links, etc.).

[0108] The signal-bearing medium 410 can be transmitted by a wireless communication medium 413 (e.g., a wireless communication medium conforming to the IEEE 802.41 standard or other transmission protocols). One or more program instructions can be, for example, computer-executable instructions or logical implementation instructions.

[0109] In some examples, such as targeting Figure 1 The described data writing device can be configured to provide various operations, functions, or actions in response to one or more program instructions in a computer-readable medium 411, a computer-recordable medium 412, and / or a communication medium 413.

[0110] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0111] In the several embodiments provided in this disclosure, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0112] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units; that is, it can be located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0113] Furthermore, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0114] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this disclosure, in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0115] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. 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 this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method of prediction of transient operating parameters, characterized in that, The method comprises the following steps: acquiring actual collection data collected by at least one data collection device in a nuclear power plant at a current time; inputting all the actual collection data into a prediction model for prediction to obtain predicted values of transient operation parameters within a preset time period after the current time; wherein the prediction model is trained by historical collection data collected by all the data collection devices, and the transient operation parameters include one or more of steam generator outlet steam mass flow, pressurizer fluctuation tube temperature and coolant flow; generating a prediction result of the transient operation parameters based on the predicted values of the transient operation parameters.

2. The method of prediction of transient operating parameters according to claim 1, characterized in that, Before the step of acquiring actual collection data collected by at least one data collection device in a nuclear power plant at a current time, the method further comprises: acquiring training sample data; wherein the training sample data includes historical collection data continuously collected by the data collection device; performing a preprocessing operation on the training sample data to obtain preprocessed data; wherein the preprocessing operation includes one or more of data cleaning and format processing, and the preprocessed data includes training data and a labeled result, the training data includes any one of the historical collection data, and the labeled result includes historical collection data within a preset time period continuous to the training data; training a target deep feedforward neural network based on the preprocessed data until the target deep feedforward neural network converges to obtain a prediction model.

3. The method of prediction of transient operating parameters according to claim 2, characterized in that, Before the step of acquiring training sample data, the method further comprises: acquiring an initial deep feedforward neural network; optimizing a topology structure parameter of the initial deep feedforward neural network based on a quantum genetic algorithm and the preprocessed data to obtain a target structure parameter; wherein the topology structure parameter at least includes the number of input layers, the number of hidden layers, the number of output layers, and the number of nodes contained in each of the hidden layers; setting an initial deep feedforward neural network according to the target structure parameter to obtain a target deep feedforward neural network.

4. The method of predicting transient operating parameters according to claim 2, wherein, The step of training a target deep feedforward neural network based on the preprocessed data until the target deep feedforward neural network converges to obtain a prediction model comprises: inputting the training data into the target deep feedforward neural network to obtain a prediction result of the target deep feedforward neural network on the training data; adjusting network parameters of the target deep feedforward neural network based on the prediction result and the labeled result until the target deep feedforward neural network converges to obtain a prediction model.

5. The method of predicting transient operating parameters of claim 2, wherein, The step of training a target deep feedforward neural network based on the preprocessed data until the target deep feedforward neural network converges to obtain a prediction model comprises: training a target deep feedforward neural network based on a model-independent meta-learning algorithm and the preprocessed data until the target deep feedforward neural network converges to obtain a prediction model.

6. A device for predicting a transient operating parameter, characterized in that The method comprises the following steps: an acquiring unit, configured to acquire actual collection data collected by at least one data collection device in a nuclear power plant at a current time; The processing unit is configured to input all the actual acquisition data acquired by the acquisition unit into a prediction model to obtain predicted values of transient operation parameters in a preset time period after the current time; the prediction model is trained by historical acquisition data collected by all the data acquisition devices, and the transient operation parameters include one or more of steam mass flow at the outlet of a steam generator, a surge line temperature of a pressurizer, and coolant flow; The processing unit is further configured to generate a prediction result of the transient operation parameters based on the predicted values of the transient operation parameters.

7. The device for predicting transient operating parameters according to claim 6, characterized in that, The acquisition unit is further configured to acquire training sample data; the training sample data includes historical acquisition data continuously collected by the data acquisition devices. The processing unit is further configured to perform a preprocessing operation on the training sample data acquired by the acquisition unit to obtain preprocessed data; the preprocessing operation includes one or more of data cleaning and format processing, and the preprocessed data includes training data and a labeling result; the training data includes any one of the historical acquisition data, and the labeling result includes historical acquisition data in a preset time period continuous to the training data. The processing unit is further configured to train a target deep feedforward neural network based on the preprocessed data until the target deep feedforward neural network converges to obtain the prediction model.

8. The device for predicting transient operating parameters according to claim 6, characterized in that, The acquisition unit is further configured to acquire an initial deep feedforward neural network. The processing unit is further configured to optimize topology structure parameters of the initial deep feedforward neural network based on a quantum genetic algorithm and the preprocessed data to obtain target structure parameters; the topology structure parameters at least include a number of input layers, a number of hidden layers, a number of output layers, and a number of nodes included in each of the hidden layers. The processing unit is further configured to set an initial deep feedforward neural network according to the target structure parameters to obtain the target deep feedforward neural network.

9. An electronic device, comprising: The electronic device comprises: a memory and a processor; the memory is configured to store a computer program; and the processor is configured to enable the electronic device to implement the method in any one of claims 1-5 when executing the computer program.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program; and when the computer program is executed by a computing device, the computing device is enabled to implement the method in any one of claims 1-5.

Images