An lstm-based loss of coolant accident diagnosis method and system for pressurized water reactor nuclear power plants

Abstract

The application discloses a loss of coolant accident diagnosis method and system for a pressurized water reactor nuclear power plant based on an LSTM, constructs an LSTM model for loss of coolant accident diagnosis of the nuclear power plant, trains the LSTM model by using preprocessed and standardized training set data, verifies diagnosis accuracy and generalization ability of the model based on test set data, inputs operation data of the pressurized water reactor nuclear power plant into the model, on-line detects whether a loss of coolant accident occurs, and diagnoses a position and severity of a break. The application can realize on-line and accurate diagnosis of loss of coolant accidents of different positions and different severities of a primary coolant pipeline of the pressurized water reactor nuclear power plant, and provides valuable information for accident mitigation.

Description

Technical Field

[0001] This invention belongs to the field of intelligent accident diagnosis technology for nuclear power plants, specifically relating to a method and system for diagnosing loss-of-water accidents in pressurized water reactor nuclear power plants based on LSTM (Long Short-Term Memory) networks. Background Technology

[0002] Nuclear safety is the lifeline of nuclear energy development and utilization. In the event of a nuclear power plant accident, the ability to quickly and accurately diagnose the location and severity of the accident is crucial for emergency response and mitigation. However, nuclear reactor systems are highly complex, and the various observation parameters during operation are strongly coupled. In the event of a sudden accident, hundreds of alarm instruments often issue warnings simultaneously, making it difficult for operators to pinpoint the accident's location and determine its severity immediately. This can easily lead to misoperation or missed opportunities for optimal remediation, resulting in serious consequences. Therefore, researching and developing a nuclear power plant accident monitoring and diagnostic system, and automatically identifying accident information through this system, has significant application value.

[0003] Accident diagnosis can be achieved through three approaches: traditional analytical models, expert systems, and data-driven methods. Due to the high complexity and time-varying nature of nuclear power plants, accurate modeling is difficult, making traditional analytical model-based accident diagnosis methods unsatisfactory. When applying expert system-based diagnostic methods to nuclear power plants, acquiring knowledge from the knowledge base is challenging, and the accuracy of the diagnosis heavily depends on the richness of the expert's experience and knowledge level. With numerous rules, matching conflicts and combinatorial explosions occur during reasoning, further hindering satisfactory results. Data-driven accident diagnosis methods do not require knowledge of the system's precise analytical model, are very convenient to apply, and offer high diagnostic accuracy. Therefore, this invention employs this method for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants.

[0004] Deep learning-based accident diagnosis methods extract signal features and learn optimal network weights and thresholds for different accident features. Based on these weights and thresholds, the signal features are rationally classified, thereby achieving efficient and accurate end-to-end accident diagnosis for complex systems. This intelligent diagnostic method can quickly and effectively process massive amounts of highly coupled data, significantly improving the accuracy of accident type diagnosis and enhancing the operational safety of pressurized water reactor nuclear power plants. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants based on LSTM, which addresses the shortcomings of the prior art and solves the technical problems of difficult diagnosis, low diagnostic efficiency and poor diagnostic effect of traditional diagnostic methods.

[0006] The present invention adopts the following technical solution:

[0007] A method for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants based on LSTM includes the following steps:

[0008] S1. Collect characteristic data of pressurized water reactor nuclear power plants under different severities of loss-of-coolant accidents at different locations in the primary coolant pipeline and under normal operating conditions, and divide the characteristic data into training set and test set;

[0009] S2. Preprocess and standardize the training and test sets obtained in step S1;

[0010] S3. Construct a LSTM-based diagnostic model for loss-of-coolant accidents in pressurized water reactor nuclear power plants;

[0011] S4. Use the training set obtained in step S2 to train the pressurized water reactor nuclear power plant loss-of-coolant accident diagnosis model constructed in step S3, and input the test set data obtained in step S2 into the trained pressurized water reactor nuclear power plant loss-of-coolant accident diagnosis model to realize the diagnosis of loss-of-coolant accidents of different locations and severity in pressurized water reactor nuclear power plants.

[0012] Specifically, in step S1, the feature data are the main physical and thermal data of a pressurized water reactor nuclear power plant.

[0013] Specifically, in step S1, the feature data is divided into a training set and a test set in a 6:2 ratio.

[0014] Specifically, in step S3, the LSTM-based pressurized water reactor nuclear power plant loss-of-coolant accident diagnosis model includes an input layer, a sequence layer composed of an LSTM network, an activation function layer, a normalization layer, a forgetting layer, a fully connected layer, a softmax layer, and an output layer.

[0015] Furthermore, the input layer receives standardized time-series training samples and sets the dimension of the feature values ​​to 36.

[0016] Sequence layer: Contains an LSTM module, with the number of neurons in the hidden layer set to 160;

[0017] Activation function layer: Introduces nonlinear factors to enhance the model's expressive power;

[0018] Normalization layer: speeds up network training;

[0019] Forget layer: Randomly discards neuron information in the neural network during training to ensure the robustness of model training;

[0020] Fully connected layer: It takes the data features obtained from the forgetting layer as input values, maps the distributed feature representation learned by the network to the category code, and outputs a single-dimensional vector that can represent the accident type features;

[0021] Softmax layer: The Softmax function is used to compress all values ​​of a one-dimensional vector of any size, and the size of the vector remains unchanged after processing;

[0022] Output layer: Outputs the final prediction results of the model.

[0023] Furthermore, the LSTM hidden layer in the sequence layer includes 160 neurons.

[0024] Furthermore, the forgetting rate of the forgetting layer is set to 50%.

[0025] Furthermore, the values ​​of all elements in the Softmax layer are compressed to the range [0, 1], and the sum of all elements is 1.

[0026] Specifically, in step S4, the maximum number of iterations in the training process is 2000, and the learning rate decays to 40% of its original value after every 500 iterations.

[0027] Secondly, embodiments of the present invention provide an LSTM-based system for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants, comprising:

[0028] The data module collects characteristic data of pressurized water reactor nuclear power plants under different conditions of loss-of-coolant accidents of varying severity at different locations in the primary coolant pipeline, as well as under normal operating conditions, and divides the characteristic data into training and testing sets.

[0029] The preprocessing module preprocesses and standardizes the training and test sets obtained from the data module.

[0030] The module is used to build a LSTM-based diagnostic model for loss-of-coolant accidents in pressurized water reactor nuclear power plants.

[0031] The diagnostic module uses the training set obtained from the preprocessing module to train the pressurized water reactor nuclear power plant loss-of-coolant accident diagnostic model constructed by the construction module. The test set data obtained from the preprocessing module is input into the trained pressurized water reactor nuclear power plant loss-of-coolant accident diagnostic model to realize the diagnosis of loss-of-coolant accidents of different locations and severities in pressurized water reactor nuclear power plants.

[0032] Compared with the prior art, the present invention has at least the following beneficial effects:

[0033] A method for diagnosing loss-of-coolant (LOD) accidents in pressurized water reactor nuclear power plants based on LSTM is proposed. The acquired sample data is preprocessed using the Z-score method. Standardized data eliminates the inconvenience caused by magnitude in data analysis, making the results analysis unaffected by magnitude. Data with large order-of-magnitude differences can independently reveal their hidden features, greatly improving the comparability between data and accelerating the network training and convergence speed. The constructed LSTM-based LOD accident diagnosis model can accurately identify the location and severity of the breach.

[0034] Furthermore, characteristic parameters of pressurized water reactor nuclear power plants during loss-of-coolant accidents are obtained through simulation methods. By analyzing the different change trajectories of these physical and thermal parameters after the accident, loss-of-coolant accidents of different locations and severity can be identified.

[0035] Furthermore, the data recording module in Simulink is used to collect and save accident simulation data, with 75% of the data serving as the training set. This includes operational data of breach accidents of varying degrees occurring at different locations and times, ensuring the diversity of the training set.

[0036] Furthermore, the LSTM hidden layer includes multiple neurons. Too many neurons can lead to overfitting, reduce the model's generalization ability, complicate the network structure, and increase training time. Too few neurons can lead to underfitting, reduce the model's prediction accuracy, and fail to meet practical requirements. Based on experience, choosing an appropriate number of neurons can balance the model's generalization ability and prediction accuracy, resulting in the best final diagnostic effect.

[0037] Furthermore, an activation function layer is employed to introduce nonlinear factors and enhance the model's expressive power; a normalization layer and a forgetting layer are used, with the normalization layer accelerating the network's training speed and the forgetting layer randomly discarding neuron information in the neural network during training; the network model structure is reasonably set, and the optimal hyperparameters are selected to improve prediction accuracy; the fully connected layer can accurately map the extracted features to accident types, thereby achieving efficient and accurate diagnosis of multiple accident types.

[0038] Furthermore, by employing an LSTM network layer with 160 neurons, the feature extraction of operational data from water loss accidents can be effectively performed, greatly improving the accuracy of the prediction model.

[0039] Furthermore, a forgetting rate of 50% is set to ensure the robustness of model training.

[0040] Furthermore, in the Softmax layer, the values ​​of all elements are compressed to the range [0, 1], and the sum of all elements is 1. This layer normalizes a numerical vector into a probability distribution vector, mapping the outputs of multiple neurons to the [0, 1] interval. This avoids negative numbers or values ​​greater than 1, facilitating probabilistic interpretation and comparison. Simultaneously, the sum of all elements being 1 satisfies the properties of a probability distribution and avoids numerical overflow or underflow issues. The Softmax layer enhances the value of the largest element in the output vector while reducing the values ​​of other elements, amplifying the differences in classification results and improving classification accuracy.

[0041] Furthermore, the maximum number of iterations during training is set to 2000. Setting a sufficient number of iterations ensures that the loss function converges during model training. A variable learning rate is used. An excessively large learning rate can lead to overfitting, while an excessively small learning rate slows down network convergence, hindering training. This invention sets the initial learning rate to 0.01, and every 500 iterations, the learning rate increases to 40% of its original value. This ensures that the model finds the optimal value more quickly and smoothly during backpropagation to update the weight parameters, thus avoiding excessive oscillations in the loss function during training.

[0042] It is understandable that the beneficial effects of the second aspect mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0043] In summary, this invention establishes an LSTM-based neural network architecture to accurately diagnose loss-of-coolant accidents in pressurized water reactor nuclear power plants. When faced with some unknown accidents, the predicted result is the existing accident with the closest characteristics, thereby achieving the goal of real-time and effective diagnosis of loss-of-coolant accidents in pressurized water reactor nuclear power plants.

[0044] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0045] Figure 1 This is a logic flowchart of the present invention;

[0046] Figure 2 This is a diagram of the simulation platform;

[0047] Figure 3 Create a diagram for the breach;

[0048] Figure 4 Diagram of LSTM cell unit structure;

[0049] Figure 5 Here is a diagram of the LSTM network structure;

[0050] Figure 6 Set up an infographic for the LOCA operating condition type;

[0051] Figure 7 This is a diagram showing the comparison between the training set and the test set during the training process of the diagnostic model, where (a) is the model training accuracy curve and (b) is the model training loss curve.

[0052] Figure 8 A diagram showing the diagnostic results of a specific test condition over time.

[0053] Figure 9 Diagnostic accuracy graphs for different test sets. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0056] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0057] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this invention generally indicates that the preceding and following objects have an "or" relationship.

[0058] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0059] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0060] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0061] This invention provides an LSTM-based method for diagnosing loss-of-coolant (LOD) accidents in pressurized water reactor (PWR) nuclear power plants. It categorizes LODs based on the cross-sectional area of ​​the main coolant piping, classifying them as small-sized LODs (less than 1 / 50 of the main piping cross-sectional area) and medium-sized LODs (between 1 / 50 and 1 / 10 of the main piping cross-sectional area). Since small-sized LODs are more likely to occur and harder to identify, and medium-sized LODs pose a greater risk, this invention selects these two sizes of LODs as the diagnostic targets. The LOD locations are selected from three areas: the primary coolant cooling pipe section, the primary coolant cooling pipe section, and the steam generator heat transfer pipes. Because real PWR accident data is difficult to obtain, this invention simulates LOD accidents using a simulation platform to obtain sample data. From this sample data, 36 characteristic data points, including primary coolant pressure and average coolant temperature, are selected to establish an LSTM-based PWR accident diagnosis model for accurately diagnosing the location and severity of LOD accidents.

[0062] Please see Figure 1 This invention discloses a method for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants based on LSTM, comprising the following steps:

[0063] S1, please refer to Figure 2(a) Figure shows the structure of the simulation platform, and (b) Figure shows the internal details of the FEEDWATER SYSTEM MODEL. The simulation platform was used to collect characteristic data of pressurized water reactor nuclear power plants under different severity loss-of-coolant accident conditions and normal operating conditions at different locations in the coolant pipelines. Training and testing sets were then obtained. The simulation platform was built using a coupled modeling approach with RELAP5 and Simulink. The main equipment and auxiliary systems of the primary and secondary loops of the pressurized water reactor were modeled using RELAP5, while the control logic for reactor start-up / shutdown and the activation of various auxiliary systems during accidents was modeled using Simulink.

[0064] Different locations are selected for the location of the rupture, namely the primary loop hot pipe section, the primary loop cold pipe section, and the steam generator heat transfer pipe. Small ruptures and medium-sized ruptures are further categorized. In this embodiment, the main pressurized water reactor coolant pipe size is 0.4536m. 2 The break size selection range is 20cm. 2 Up to 200cm 2 20cm intervals 2 Set up a working condition that meets the requirements for small and medium-sized small breaks. For details on how to set up the break, please refer to [link / reference needed]. Figure 3 Connect 240 and 993 with a card of type VALVE. The valve is a trigger valve (trpvlv).

[0065] Based on the above accident types, 36-dimensional feature data were selected for each type, with a time series length of 200s. Data was recorded every 0.1s. The breach opening time was selected as 10 / 20 / 30 / 40 / 50 / 60 / 70 / 80s after the start of the simulation, and the data were divided into training set and test set in a 6:2 ratio.

[0066] S2. Preprocess and standardize the training set and test set collected in step S1. Preprocessing is to process the multidimensional signal into a format that the training framework can accept.

[0067] The standardization method is as follows:

[0068]

[0069] Where, μ i It is the average value of the i-th data sample; σ i It is the standard deviation of the i-th data sample.

[0070] Standardization helps to accelerate the convergence of network training and also allows data with large differences in magnitude to independently reveal their hidden features.

[0071] S3. Construct a LSTM-based diagnostic model for loss-of-coolant accidents in pressurized water reactor nuclear power plants;

[0072] Please see Figure 4 A single LSTM cell unit is composed of a forgetting gate f t Input gate i t and output gate o t It consists of three threshold units. Each LSTM cell unit outputs h from the cell structure at the previous time step through these three threshold units. t-1 The cell state C at the previous time step t-1 Input x at the current time t The data is processed, and then the output of the current LSTM cell is obtained through the tanh function. The key calculation process is as follows:

[0073] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0074] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0075]

[0076]

[0077] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0078] h t =o t ·tanh(C t )

[0079] Where σ represents the Sigmoid activation function, W f W i W c and W o The weight matrix represents the forget gate, input gate, cell state calculation gate, and output gate; b f b i b c and b o The bias terms represent the forget gate, input gate, computation cell state, and output gate.

[0080] Please see Figure 5The LSTM network consists of a general network structure: sequenceInputLayer, lstmLayer, fullyConnectedLayer, softmaxLayer, and classificationLayer. It also includes three unique layers specific to this invention: reluLayer (activation function layer), batchNormalizationLayer, and dropoutLayer, as detailed below:

[0081] Input layer: Receives standardized 36×2001 time series training samples;

[0082] LSTM layer: Set the number of hidden layer neurons to 160;

[0083] Activation function layer: The ReLU function is selected as the activation function. The addition of the activation function introduces nonlinear factors and enhances the expressive power of the model.

[0084] Drop layer: Set the drop rate to 50%;

[0085] Fully connected layer: The input value is the data features after the discard layer, and the output value is a one-dimensional vector that can represent the accident type features;

[0086] The softmax layer maps all values ​​of a one-dimensional vector of arbitrary size to real numbers between (0,1) using the softmax function, and ensures that the sum of all elements is 1.

[0087] The Softmax function is as follows:

[0088]

[0089] Where K represents the number of categories, this function implements the above mapping process, enhances the value of the largest element in the output vector, and reduces the values ​​of other elements, amplifying the differences in classification results and improving classification accuracy.

[0090] S4. Using the training set data obtained in step S2, train the LOCA diagnostic model for pressurized water reactor nuclear power plants constructed in step S3. Input the test set data obtained in step S2 into the trained diagnostic model, and the accident diagnostic model gives the final diagnostic result. During the actual operation of the pressurized water reactor, feature data of the same dimensions as the training set are collected by various measuring instruments and input into the diagnostic model to realize the diagnosis of loss-of-coolant accidents of different locations and severities in pressurized water reactor nuclear power plants.

[0091] The training set samples are input into the LSTM-based pressurized water reactor nuclear power plant loss-of-coolant accident diagnostic model for training. The LSTM network needs to continuously extract features from a large number of samples and optimize the network weight parameters to achieve the ideal classification effect. In addition, an excessively large learning rate can lead to overfitting and may also cause a large number of neurons in the network to fail to be activated and die; an excessively small learning rate will slow down the network convergence speed, which is not conducive to network training. Therefore, this invention sets a variable learning rate, with an initial learning rate of 0.01, which is changed to 40% of the original rate every 500 iterations.

[0092] In another embodiment of the present invention, an LSTM-based pressurized water reactor nuclear power plant loss-of-coolant accident diagnosis system is provided. This system can be used to implement the above-mentioned LSTM-based pressurized water reactor nuclear power plant loss-of-coolant accident diagnosis method. Specifically, the LSTM-based pressurized water reactor nuclear power plant loss-of-coolant accident diagnosis system includes a data module, a preprocessing module, a construction module, and a diagnosis module.

[0093] The data module collects characteristic data of pressurized water reactor nuclear power plants under different conditions of loss of coolant at different locations in the primary coolant pipeline and under normal operating conditions, and divides the characteristic data into training set and test set;

[0094] The preprocessing module preprocesses and standardizes the training and test sets obtained from the data module.

[0095] The module is used to build a LSTM-based diagnostic model for loss-of-coolant accidents in pressurized water reactor nuclear power plants.

[0096] The diagnostic module uses the training set obtained from the preprocessing module to train the pressurized water reactor nuclear power plant loss-of-coolant accident diagnostic model constructed by the construction module. The test set data obtained from the preprocessing module is input into the trained pressurized water reactor nuclear power plant loss-of-coolant accident diagnostic model to realize the diagnosis of loss-of-coolant accidents of different locations and severities in pressurized water reactor nuclear power plants.

[0097] In another embodiment of the present invention, a terminal device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used for the operation of an LSTM-based pressurized water reactor nuclear power plant loss-of-coolant accident diagnosis method, including:

[0098] Feature data of loss-of-coolant accidents of varying severity at different locations in the primary coolant piping of a pressurized water reactor (PWR) nuclear power plant, as well as under normal operating conditions, were collected and divided into training and testing sets. The training and testing sets were preprocessed and standardized. An LSTM-based PWR accident diagnosis model was constructed. The PWR accident diagnosis model was trained using the training set, and the test set data was input into the trained model to diagnose loss-of-coolant accidents of varying severity at different locations in the PWR nuclear power plant.

[0099] In another embodiment of the present invention, a storage medium is also provided, specifically a computer-readable storage medium (memory). This computer-readable storage medium is a memory device in a terminal device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and extended storage media supported by the terminal device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device.

[0100] One or more instructions stored in the computer-readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the LSTM-based pressurized water reactor nuclear power plant loss-of-coolant accident diagnosis method in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by the processor in the following steps:

[0101] Feature data of loss-of-coolant accidents of varying severity at different locations in the primary coolant piping of a pressurized water reactor (PWR) nuclear power plant, as well as under normal operating conditions, were collected and divided into training and testing sets. The training and testing sets were preprocessed and standardized. An LSTM-based PWR accident diagnosis model was constructed. The PWR accident diagnosis model was trained using the training set, and the test set data was input into the trained model to diagnose loss-of-coolant accidents of varying severity at different locations in the PWR nuclear power plant.

[0102] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0103] Example

[0104] Since real reactor accident data is difficult to obtain, this invention simulates the loss-of-coolant accident conditions of pressurized water reactors through a simulation platform, and collects simulation data of loss-of-coolant accident conditions at different locations and with different severities, to obtain a sample dataset containing 36 characteristic parameters such as primary loop pressure and average coolant temperature, and divides it into training set and test set.

[0105] Please see Figure 6 The data represents sample data collected in the simulation platform for this invention, including cold pipe sections, hot pipe sections, and steam generator heat transfer tubes, which exhibit different degrees of variation (20:20:200cm) at 10:10:80s. 2 The breach accidents were classified into 9 types, and the time series length was 2001.

[0106] Based on the above fault types, the sample dataset is divided into a training set and a test set in a 6:2 ratio. The training set data is used to train the established water loss accident diagnosis model, and the test set data is used to evaluate the training effect of the model.

[0107] Please see Figure 7 Figure (a) shows the change in diagnostic accuracy of the established water loss accident diagnostic model with the number of iterations, and Figure (b) shows the change in cross-entropy loss with the number of iterations. A smaller cross-entropy loss indicates a smaller error, a closer match between the predicted and actual values, and a better classification effect. It can be seen that as the number of iterations increases, the model's loss value on both the training and test sets decreases, while the diagnostic accuracy increases. Finally, the diagnostic accuracy corresponding to the minimum cross-entropy loss is selected as the final accuracy, and its corresponding network weights and thresholds are used as the final weights and thresholds of the network model.

[0108] Please see Figure 8 The horizontal axis represents the time series, and the vertical axis represents the accident type. The two curves represent the actual accident type and the accident type obtained through network diagnosis, respectively. The overlap of the two curves indicates that the obtained network diagnosis model has high accuracy.

[0109] Please see Figure 9 The horizontal axis represents accidents in different test sets, and the vertical axis represents the proportion of time points in the 2001 sequence lengths where accurate diagnosis (i.e., the actual accident type is the same as the network diagnosis type) is achieved. The figure shows that the diagnostic accuracy rate for each type of water loss accident is above 80%, and the accuracy rate for most of the diagnostics is above 95%, indicating that the trained network model has high accuracy.

[0110] In summary, this invention provides a method and system for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants based on LSTM. By simulating the data on a simulation platform to obtain the training and testing sets required for the model, an LSTM-based model for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants is constructed. Through training, the model can effectively learn and extract the operational characteristics of loss-of-coolant accident conditions, achieving accurate diagnosis of different types of loss-of-coolant accidents. This verifies that the proposed method can accurately diagnose loss-of-coolant accidents of different locations and severity.

[0111] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments 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. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0112] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0113] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software 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 implementations should not be considered beyond the scope of this invention.

[0114] In the embodiments provided by this invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal 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 system, 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 between devices or units may be electrical, mechanical, or other forms.

[0115] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0116] Furthermore, the functional units in the various embodiments of the present invention 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.

[0117] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0118] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0119] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0120] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0121] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants based on LSTM, characterized in that, Includes the following steps: S1. Collect characteristic data of pressurized water reactor nuclear power plants under different severities of loss-of-coolant accidents at different locations in the primary coolant pipeline and under normal operating conditions, and divide the characteristic data into training set and test set; S2. Preprocess and standardize the training and test sets obtained in step S1; S3. Construct an LSTM-based diagnostic model for loss-of-coolant accidents in pressurized water reactor nuclear power plants. The LSTM-based diagnostic model for loss-of-coolant accidents in pressurized water reactor nuclear power plants includes an input layer, a sequence layer composed of an LSTM network, an activation function layer, a normalization layer, a forgetting layer, a fully connected layer, a softmax layer, and an output layer. Input layer: Receives standardized time series training samples, with the feature dimensions set to 36; Sequence layer: contains LSTM modules; Activation function layer: Introduces nonlinear factors to enhance the model's expressive power; Normalization layer: speeds up network training; Forget layer: Randomly discards neuron information in the neural network during training to ensure the robustness of model training; Fully connected layer: It takes the data features obtained from the forgetting layer as input values, maps the distributed feature representation learned by the network to the category code, and outputs a single-dimensional vector that can represent the accident type features; Softmax layer: The Softmax function is used to compress all values ​​of a one-dimensional vector of any size, and the size of the vector remains unchanged after processing; Output layer: Outputs the final prediction results of the model; S4. Use the training set obtained in step S2 to train the pressurized water reactor nuclear power plant loss-of-coolant accident diagnosis model constructed in step S3, and input the test set data obtained in step S2 into the trained pressurized water reactor nuclear power plant loss-of-coolant accident diagnosis model to realize the diagnosis of loss-of-coolant accidents of different locations and severity in pressurized water reactor nuclear power plants.

2. The LSTM-based method for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants according to claim 1, characterized in that, In step S1, the feature data are the main physical and thermal data of a pressurized water reactor nuclear power plant.

3. The LSTM-based method for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants according to claim 1, characterized in that, In step S1, the feature data is divided into a training set and a test set in a 6:2 ratio.

4. The LSTM-based method for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants according to claim 1, characterized in that, The LSTM hidden layer in the sequence layer contains 160 neurons.

5. The LSTM-based method for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants according to claim 1, characterized in that, The forgetting rate of the forgetting layer is set to 50%.

6. The LSTM-based method for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants according to claim 1, characterized in that, In the Softmax layer, the values ​​of all elements are compressed to the range [0, 1], and the sum of all elements is 1.

7. The LSTM-based method for diagnosing loss-of-coolant accidents in pressurized water reactor nuclear power plants according to claim 1, characterized in that, In step S4, the maximum number of iterations in the training process is 2000, and the learning rate decays to 40% of its original value every 500 iterations.

8. A LSTM-based loss-of-coolant accident diagnosis system for pressurized water reactor nuclear power plants, characterized in that, include: The data module collects characteristic data of pressurized water reactor nuclear power plants under different conditions of loss-of-coolant accidents of varying severity at different locations in the primary coolant pipeline, as well as under normal operating conditions, and divides the characteristic data into training and testing sets. The preprocessing module preprocesses and standardizes the training and test sets obtained from the data module. The module constructs a LSTM-based diagnostic model for loss-of-coolant accidents in pressurized water reactor nuclear power plants. The LSTM-based diagnostic model for loss-of-coolant accidents in pressurized water reactor nuclear power plants includes an input layer, a sequence layer composed of an LSTM network, an activation function layer, a normalization layer, a forgetting layer, a fully connected layer, a softmax layer, and an output layer. Input layer: Receives standardized time series training samples, with the feature dimensions set to 36; Sequence layer: contains LSTM modules; Activation function layer: Introduces nonlinear factors to enhance the model's expressive power; Normalization layer: speeds up network training; Forget layer: Randomly discards neuron information in the neural network during training to ensure the robustness of model training; Fully connected layer: It takes the data features obtained from the forgetting layer as input values, maps the distributed feature representation learned by the network to the category code, and outputs a single-dimensional vector that can represent the accident type features; Softmax layer: The Softmax function is used to compress all values ​​of a one-dimensional vector of any size, and the size of the vector remains unchanged after processing; Output layer: Outputs the final prediction results of the model; The diagnostic module uses the training set obtained from the preprocessing module to train the pressurized water reactor nuclear power plant loss-of-coolant accident diagnostic model constructed by the construction module. The test set data obtained from the preprocessing module is input into the trained pressurized water reactor nuclear power plant loss-of-coolant accident diagnostic model to realize the diagnosis of loss-of-coolant accidents of different locations and severities in pressurized water reactor nuclear power plants.

Images