Water-feeding control system for evaporator of nuclear power unit of reactor

Abstract

Disclosed in the present application is a water-feeding control system for an evaporator of a nuclear power unit of a reactor. First, pressure values of a reactor at a plurality of pre-determined time points within a pre-determined period of time are collected by means of a pressure sensor, and water-injection speed values at the plurality of pre-determined time points are collected; then, the pressure values at the plurality of pre-determined time points and the water-injection speed values at the plurality of pre-determined time points are respectively arranged according to a time dimension to form a pressure time-series input vector and a water-injection-speed time-series input vector; next, time-series interactive association analysis is performed on the pressure time-series input vector and the water-injection-speed time-series input vector, so as to obtain a pressure-water injection speed time-series interaction feature; and finally, on the basis of the pressure-water injection speed time-series interaction feature, whether a water-injection speed value at a current time point should be increased or decreased is determined. In this way, the performance of a water-feeding control system can be improved, thereby maintaining the stable operation of a reactor.

Description

A water supply control system for the evaporator of a reactor nuclear power unit Technical Field

[0001] The present application relates to the field of water supply control, and more specifically, to a water supply control system for an evaporator of a reactor nuclear power unit. Background Art

[0002] A nuclear reactor is a vital energy generation device, the core of which is the reactor, including the evaporator feedwater control system. The evaporator feedwater control system is a crucial component of a nuclear power plant, responsible for replenishing feedwater to the evaporator to maintain stable water level and pressure within the evaporator. The performance of the feedwater control system directly impacts the safe and economical operation of a nuclear power plant. Traditional evaporator feedwater control systems typically employ proportional-integral-derivative (PID) controllers. However, PID controllers have limitations for nonlinear, time-varying, and strongly coupled reactor feedwater control systems, such as an inability to adapt to parameter changes and poor disturbance rejection.

[0003] Therefore, an optimized evaporator feedwater control system for a nuclear power reactor unit is desired. Technical issues

[0004] To address the aforementioned technical issues, the present application is proposed. Embodiments of the present application provide a reactor nuclear power unit evaporator feedwater control system. This system can implement intelligent regulation of the reactor feedwater control system, thereby improving its performance and maintaining stable reactor operation. Technical Solutions

[0005] According to one aspect of the present application, a reactor nuclear power unit evaporator feed water control system is provided, comprising:

[0006] A data acquisition module, configured to acquire, through a pressure sensor, the pressure values ​​of the reactor at a plurality of predetermined time points in a predetermined time period, and the water injection rate values ​​at the plurality of predetermined time points;

[0007] a data parameter time series arrangement module, configured to arrange the pressure values ​​at the plurality of predetermined time points and the water injection rate values ​​at the plurality of predetermined time points into a pressure time series input vector and a water injection rate time series input vector according to a time dimension;

[0008] a data parameter time series feature interaction module, configured to perform time series interaction correlation analysis on the pressure time series input vector and the water injection rate time series input vector to obtain a pressure-water injection rate time series interaction feature; and

[0009] The water injection rate control module is used to determine whether the water injection rate value at the current time point should be increased or decreased based on the pressure-water injection rate time series interaction characteristics. Beneficial effects

[0010] Compared to the prior art, the present application provides a reactor nuclear power unit evaporator feedwater control system. It first uses a pressure sensor to collect the reactor's pressure values ​​and water injection rate values ​​at multiple predetermined time points during a predetermined time period. Then, the pressure values ​​and water injection rate values ​​at these multiple predetermined time points are arranged according to the time dimension into a pressure time series input vector and a water injection rate time series input vector. Then, a time series interaction correlation analysis is performed on the pressure time series input vector and the water injection rate time series input vector to obtain a pressure-water injection rate time series interaction feature. Finally, based on the pressure-water injection rate time series interaction feature, it is determined whether the water injection rate value at the current time point should be increased or decreased. This improves the performance of the feedwater control system to maintain stable reactor operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for the description of the embodiments. The following drawings are not intentionally scaled to the actual size, and the focus is on illustrating the main purpose of the present application.

[0012] FIG1 is a block diagram of a water feed control system for an evaporator of a reactor nuclear power unit according to an embodiment of the present application.

[0013] FIG2 is a block diagram of the data parameter timing characteristic interaction module in the evaporator feed water control system of the reactor nuclear power unit according to an embodiment of the present application.

[0014] FIG3 is a flow chart of a method for controlling water supply to an evaporator of a reactor nuclear power unit according to an embodiment of the present application.

[0015] FIG4 is a schematic diagram of the system architecture of the evaporator feed water control method for a reactor nuclear power unit according to an embodiment of the present application.

[0016] FIG5 is a diagram showing an application scenario of an evaporator feed water control system for a reactor nuclear power unit according to an embodiment of the present application. Modes for Carrying Out the Invention

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts also fall within the scope of protection of this application.

[0018] As used in this application and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0019] Although the present application makes various references to certain modules in the system according to embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are illustrative only, and different aspects of the system and method can use different modules.

[0020] Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the various steps may be processed in reverse order or simultaneously, as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0021] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0022] In response to the above technical problems, the technical concept of this application is to collect the reactor pressure value and water injection rate value in real time through a pressure sensor, and introduce data processing and analysis algorithms at the back end to perform time-series interactive collaborative analysis of the reactor pressure value and water injection rate value, so as to perform real-time control of the water injection rate. In this way, intelligent adjustment of the reactor water supply control system can be achieved, thereby improving the performance of the water supply control system to maintain stable operation of the reactor.

[0023] FIG1 is a block diagram of a reactor nuclear power unit evaporator feedwater control system according to an embodiment of the present application. As shown in FIG1 , the reactor nuclear power unit evaporator feedwater control system 100 according to an embodiment of the present application includes: a data acquisition module 110 for collecting pressure values ​​of the reactor at multiple predetermined time points in a predetermined time period and water injection rate values ​​at the multiple predetermined time points via a pressure sensor; a data parameter time series arrangement module 120 for arranging the pressure values ​​at the multiple predetermined time points and the water injection rate values ​​at the multiple predetermined time points into a pressure time series input vector and a water injection rate time series input vector, respectively, according to a time dimension; a data parameter time series feature interaction module 130 for performing time series interaction correlation analysis on the pressure time series input vector and the water injection rate time series input vector to obtain a pressure-water injection rate time series interaction feature; and a water injection rate control module 140 for determining whether the water injection rate value at the current time point should be increased or decreased based on the pressure-water injection rate time series interaction feature.

[0024] Specifically, in the technical solution of the present application, first, the pressure values ​​of the reactor at multiple predetermined time points in a predetermined time period, as well as the water injection rate values ​​at the multiple predetermined time points, collected by the pressure sensor, are obtained. Next, considering that the pressure value of the reactor and the water injection rate value each have their own dynamic change laws in the time dimension, and there is also a synergistic correlation between the two in time series, which jointly affect the stable operation of the reactor. Therefore, in the technical solution of the present application, before extracting the time series features of the two parameters, it is necessary to further arrange the pressure values ​​at the multiple predetermined time points and the water injection rate values ​​at the multiple predetermined time points into pressure time series input vectors and water injection rate time series input vectors according to the time dimension, so as to respectively integrate the time series distribution information of the pressure value of the reactor and the water injection rate value in the time dimension.

[0025] It should be understood that in the evaporator water supply control system, both pressure and injection rate are time-varying time-series data. To obtain more parameter time-series information and subsequently capture more time-series features, the time-series distribution information of these parameter data can be represented as a time-series image. Specifically, in the technical solution of the present application, the pressure time-series input vector and the injection rate time-series input vector are passed through a format converter comprising an upsampling layer and a vector-to-image conversion unit to obtain pressure time-series images and injection rate time-series images. In particular, the upsampling layer serves to expand the input vectors to a size suitable for image processing. This preserves the temporal relationships of the original data and enables better capture of the temporal details and patterns of the data parameters in subsequent processing. Furthermore, the vector-to-image conversion unit can convert the reactor data parameter time-series input vector into an image, converting the temporal relationships of the reactor pressure data and injection rate data into spatial relationships of the image. Because images contain rich spatial information, image processing methods and techniques can be used to process time-series data. In addition, since image data has higher dimensions and richer information, compared with simple vector forms, images can better express the correlation and change trends between data, so as to better reflect the temporal dynamic changes of the reactor pressure and the water injection rate.

[0026] Then, a convolutional neural network model, which excels in extracting implicit features from images, is used to extract temporal features from the pressure time-series image and the water injection rate time-series image, respectively. In particular, considering that the reactor pressure and water injection rate values ​​fluctuate in the time dimension and their temporal variations may be relatively subtle, it is difficult to effectively capture and characterize their temporal features using traditional feature extraction methods. Therefore, in order to extract more representative and useful features from the time-series images, the technical solution of the present application further performs feature mining on the pressure time-series image and the water injection rate time-series image, respectively, using a temporal feature extractor based on a multi-scale neighborhood feature extraction module. This extracts multi-scale temporal feature information of the reactor pressure and water injection rate values ​​at different time spans in the temporal dimension, thereby obtaining a pressure time-series feature vector and a water injection rate time-series feature vector. In other words, since the data parameter time-series image contains information about the temporal variations of the reactor pressure and water injection rate, feature extraction can transform this information into a higher-level abstract representation for subsequent classification and decision-making. In particular, the time series feature extractor based on the multi-scale neighborhood feature extraction module can analyze and process the time series images of data parameters at different scales. This can capture features and patterns at different time scales, thereby more comprehensively describing the changes in pressure and water injection rate.

[0027] Furthermore, a feature interaction layer is used to interactively fuse the pressure time-series feature vector and the water injection rate time-series feature vector to obtain a pressure-water injection rate time-series interaction feature vector. This vector captures the correlation and mutual influence between the multi-scale temporal features of the reactor pressure value and the multi-scale temporal features of the water injection rate value. It should be understood that the goal of the traditional attention mechanism is to learn an attention weight matrix that assigns greater weights to important features and smaller weights to less important features, thereby selecting information that is more critical to the current task objective. This approach focuses on weighting the importance of each feature while ignoring the dependencies between features. However, the feature interaction layer, through feature interaction based on the attention mechanism, can capture the correlation and mutual influence between the multi-scale temporal features of the reactor pressure value and the multi-scale temporal features of the water injection rate value. It can learn the dependencies between different features and interact and integrate the features based on these dependencies to obtain the pressure-water injection rate time-series interaction feature vector.

[0028] Correspondingly, as shown in Figure 2, the data parameter time series feature interaction module 130 includes: a format conversion unit 131, which is used to convert the pressure time series input vector and the water injection rate time series input vector into a format converter including an upsampling layer and a vector-to-image conversion unit to obtain a pressure time series image and a water injection rate time series image; a parameter time series multi-scale feature extraction unit 132, which is used to perform multi-scale feature extraction on the pressure time series image and the water injection rate time series image respectively through a time series feature extractor based on a deep neural network model to obtain a pressure time series feature vector and a water injection rate time series feature vector; and a pressure-water injection rate interaction correlation coding unit 133, which is used to perform feature interaction correlation coding on the pressure time series feature vector and the water injection rate time series feature vector to obtain a pressure-water injection rate time series interaction feature vector as the pressure-water injection rate time series interaction feature. It should be understood that the data parameter time series feature interaction module 130 consists of three main units: a format conversion unit 131, a parameter time series multi-scale feature extraction unit 132, and a pressure-injection rate interactive correlation encoding unit 133. The format conversion unit 131 converts the input pressure time series vector and water injection rate time series vector into a pressure time series image and a water injection rate time series image. To achieve this conversion, the unit utilizes an upsampling layer and a vector-to-image conversion unit. The upsampling layer expands the vector data into an image form, while the vector-to-image conversion unit converts the vector data into a corresponding image representation. The parameter time series multi-scale feature extraction unit 132 aims to extract multi-scale features from the pressure time series image and the water injection rate time series image. To achieve this goal, the unit utilizes a time series feature extractor based on a deep neural network model. This feature extractor analyzes and processes the input pressure time series image and water injection rate time series image, extracts their multi-scale features, and obtains a pressure time series feature vector and a water injection rate time series feature vector. The function of the pressure-injection rate interaction coding unit 133 is to perform feature interaction coding on the pressure time series feature vector and the injection rate time series feature vector. Through this interaction coding, the temporal interaction characteristics between pressure and injection rate can be captured. Ultimately, this unit generates a pressure-injection rate temporal interaction feature vector as the required pressure-injection rate temporal interaction feature. In summary, the various units of the data parameter temporal feature interaction module are responsible for formatting the input data, extracting multi-scale features, and performing feature interaction coding, ultimately obtaining the pressure-injection rate temporal interaction features. These features can be used for further analysis and applications, such as predicting, controlling, or optimizing related systems or processes.

[0029] More specifically, in the parameter time series multi-scale feature extraction unit 132, the deep neural network model is a multi-scale neighborhood feature extraction module, which includes a first and second parallel convolutional layers, and a multi-scale feature fusion layer connected to the first and second convolutional layers, wherein the first and second convolutional layers use one-dimensional convolution kernels of different scales. It is worth noting that the multi-scale neighborhood feature extraction module is a component of the parameter time series multi-scale feature extraction unit 132, which is used to extract multi-scale features from the input pressure time series images and water injection rate time series images. This module consists of a first and second parallel convolutional layers and a multi-scale feature fusion layer connected thereto. The first and second convolutional layers use one-dimensional convolution kernels of different scales, which can effectively capture features of different scales in time series data. By using convolution kernels of different scales, the first and second convolutional layers can respectively extract features of different scales. For example, smaller convolution kernels can capture local detail features, while larger convolution kernels can capture a wider range of contextual information. The multi-scale feature fusion layer is used to fuse the multi-scale features extracted by the first and second convolution layers. Fusion can be performed in various ways, such as concatenating features, summing them, or using an attention mechanism. The purpose of the multi-scale feature fusion layer is to comprehensively utilize feature information at different scales to improve the diversity and expressiveness of feature representations. The multi-scale neighborhood feature extraction module enhances the feature extraction capabilities of pressure time series images and water injection rate time series images within the parameter time series multi-scale feature extraction unit. By using convolution kernels of different scales and multi-scale feature fusion, the module can capture feature information at different levels and scales, thereby improving the understanding and modeling capabilities of time series data. This helps to extract richer and more accurate feature representations, providing more valuable input for subsequent feature interaction encoding.

[0030] More specifically, the pressure-injection rate interactive correlation encoding unit 133 is configured to interactively fuse the pressure time series feature vector and the injection rate time series feature vector using a feature interaction layer to obtain the pressure-injection rate time series interaction feature vector. It is worth noting that the feature interaction layer is a component within the pressure-injection rate interactive correlation encoding unit 133 and is configured to interactively fuse the pressure time series feature vector and the injection rate time series feature vector to obtain the pressure-injection rate time series interaction feature vector. The feature interaction layer interactively fuses the pressure time series feature vector and the injection rate time series feature vector, fusing and interleaving their information. This interactive fusion can be achieved in various ways, such as using element-by-element multiplication, concatenation, and weighted summation. By fusing the pressure and injection rate feature information, the correlation and interaction between them can be captured, resulting in a more expressive time series interaction feature. The purpose of the feature interaction layer is to enhance the representational capabilities of features. Through interactive fusion, the pressure and injection rate features can be mutually influenced and complemented, providing a more comprehensive and richer feature representation. This helps capture the complex relationship between pressure and injection rate in the system, improving the ability to model and predict system behavior. The feature interaction layer enables the interaction and fusion of pressure and injection rate features within the pressure-injection rate interaction encoding unit. This interactive fusion yields more informative and expressive pressure-injection rate time series interaction feature vectors. These feature vectors can be used in subsequent analysis, prediction, or control tasks to better understand and apply the relationship between pressure and injection rate.

[0031] The pressure-injection rate time series interaction feature vector is then passed through a classifier to obtain a classification result, which indicates whether the injection rate value at the current time point should be increased or decreased. In other words, the interactive correlation feature information between the reactor pressure time series and the injection rate time series is used for classification processing, thereby achieving real-time control of the injection rate. This approach enables intelligent regulation of the reactor water supply control system, thereby improving its performance and maintaining stable reactor operation.

[0032] Accordingly, the water injection rate control module 140 is used to pass the pressure-water injection rate time series interaction feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the water injection rate value at the current time point should be increased or decreased.

[0033] More specifically, the water injection rate control module 140 is further used to: use the fully connected layer of the classifier to fully connect the pressure-water injection rate time series interaction feature vector to obtain a coded classification feature vector; and input the coded classification feature vector into the Softmax classification function of the classifier to obtain the classification result.

[0034] That is, in the technical solution disclosed in the present invention, the labels of the classifier include the value of the water injection rate at the current time point should increase (first label), and the value of the water injection rate at the current time point should decrease (second label), wherein the classifier determines to which classification label the pressure-water injection rate time series interaction feature vector belongs through the soft maximum function. It is worth noting that the first label p1 and the second label p2 here do not contain artificially set concepts. In fact, during the training process, the computer model does not have the concept of "the value of the water injection rate at the current time point should increase or decrease". It only has two classification labels and the probability of the output feature under these two classification labels, that is, the sum of p1 and p2 is one. Therefore, the classification result of whether the value of the water injection rate at the current time point should increase or decrease is actually converted from the classification label to a binary classification class probability distribution that conforms to natural laws. In essence, what is used is the physical meaning of the natural probability distribution of the label, rather than the linguistic text meaning of "the value of the water injection rate at the current time point should increase or decrease".

[0035] It's easy to understand that a classifier uses given categories and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistic regression and SVM are commonly used to solve binary classification problems. For multi-class classification, logistic regression or SVM can also be used, but multiple binary classifications are required to form a multi-class classification. However, this approach is error-prone and inefficient. A commonly used multi-classification method is the Softmax classification function.

[0036] Furthermore, in the technical solution of this application, the reactor nuclear power unit evaporator feedwater control system further includes a training module for training the format converter comprising an upsampling layer and a vector-to-image conversion unit, the time series feature extractor based on a multi-scale neighborhood feature extraction module, and the classifier. It should be understood that the training module plays a key role in the reactor nuclear power unit evaporator feedwater control system. It is used to train the format converter comprising an upsampling layer and a vector-to-image conversion unit, the time series feature extractor based on a multi-scale neighborhood feature extraction module, and the classifier. The following are the main functions and uses of the training module: 1. Model parameter learning: The training module learns parameters for various components in the system using labeled training data. The goal of model parameter learning is to enable the model to learn the underlying patterns and feature representations of the input data, so that it can accurately predict and infer new unlabeled data. 2. Optimizing model performance: The training module uses iterative optimization algorithms (such as gradient descent) to minimize the model's prediction error or loss function on the training data. By continuously adjusting the model parameters, the training module can improve the performance of the model, enabling it to better adapt to the training data and have better generalization capabilities. 3. Hyperparameter tuning: The training module can also be used to tune the hyperparameters of the model. Hyperparameters refer to parameters that need to be manually set before model training, such as learning rate, regularization coefficient, etc. By trying and comparing different hyperparameter values ​​during the training process, the training module can find the best hyperparameter combination to improve the performance and robustness of the model. Through the training process of the training module, the various components of the evaporator feed water control system of the reactor nuclear power unit can learn feature representation and pattern recognition capabilities suitable for specific tasks. The role of the training module is to provide an effective learning framework that enables the system to automatically learn from data and perform tasks such as prediction, inference or control.

[0037] More specifically, the training module includes: a training data acquisition unit for acquiring training data, wherein the training data includes training pressure values ​​of the reactor at multiple predetermined time points in a predetermined time period, training water injection rate values ​​at the multiple predetermined time points, and a real value of the water injection rate value at the current time point that should be increased or decreased; a training data parameter time series arrangement unit for arranging the training pressure values ​​at the multiple predetermined time points and the training water injection rate values ​​at the multiple predetermined time points into a training pressure time series input vector and a training water injection rate time series input vector according to the time dimension; a training parameter time series format conversion unit for passing the training pressure time series input vector and the training water injection rate time series input vector through the format converter including the upsampling layer and the vector-to-image conversion unit to obtain a training pressure time series image and a training water injection rate time series image; a training parameter time series feature extraction unit for respectively passing the training pressure time series image and the training water injection rate time series image through the multi-scale based time series format conversion unit. A time series feature extractor of the degree neighborhood feature extraction module is used to obtain a training pressure time series feature vector and a training water injection rate time series feature vector; a training pressure-water injection rate time series feature interaction coding unit is used to interactively fuse the training pressure time series feature vector and the training water injection rate time series feature vector using the feature interaction layer to obtain a training pressure-water injection rate time series interaction feature vector; a classification loss unit is used to pass the training pressure-water injection rate time series interaction feature vector through the classifier to obtain a classification loss function value; and a model training unit is used to train the format converter including the upsampling layer and the vector-image conversion unit, the time series feature extractor based on the multi-scale neighborhood feature extraction module and the classifier based on the classification loss function value and through gradient descent direction propagation, wherein, at each weight matrix iteration of the training, the training pressure-water injection rate time series interaction feature vector is subjected to weight space exploration constraint optimization based on class degree matrix regularization.

[0038] In particular, in the technical solution of the present application, the training pressure time series feature vector and the training water injection speed time series feature vector respectively express the local neighborhood correlation features in the local time domain of the training pressure value and the training water injection speed value. Thus, after using the inter-feature attention layer to perform feature interaction based on the attention mechanism on the training pressure time series feature vector and the training water injection speed time series feature vector, the training pressure-water injection speed time series interaction feature vector, in addition to containing the local neighborhood correlation features in the local time domain of the training pressure value and the training water injection speed value, also includes the training pressure value and the training water injection speed value in the time series distribution direction. That is, the training pressure-water injection rate time series interaction feature vector also includes the feature representation of diversified time series correlation dimensions corresponding to the intra-sample time series correlation and the inter-sample time series correlation. In this way, while improving the time series correlation feature representation of the training pressure-water injection rate time series interaction feature vector under the cross dimension, it will also cause the training pressure-water injection rate time series interaction feature vector to be classified and regressed through the classifier. The label distribution corresponding to the diversification of the feature distribution of the cross-correlation dimension in the probability distribution domain of the classification result is enriched, thereby affecting the convergence effect of the weight matrix of the classifier during the classification process.

[0039] Based on this, when the applicant of this application classifies the training pressure-water injection rate time series interaction feature vector through a classifier, at each iteration of the weight matrix, the applicant performs weight space exploration constraints on the training pressure-water injection rate time series interaction feature vector based on class matrix regularization.

[0040] Accordingly, in a specific example, during each weight matrix iteration of the training, the training pressure-water injection rate time series interaction feature vector is subjected to weight space exploration constraint optimization based on class matrix regularization to obtain the optimized training pressure-water injection rate time series interaction feature vector; wherein, the optimization formula is: in, is the training pressure-water injection rate time series interaction feature vector, is the optimized training pressure-water injection rate time series interaction feature vector, and is a column vector, is a row vector, is a learnable domain transfer matrix, for example, it can be initially set to a diagonal matrix consisting of the diagonal elements of the weight matrix of the previous iteration, represents the weight matrix of the previous iteration, represents the weight matrix after iteration, represents matrix multiplication, Represents a transpose operation.

[0041] Here, considering the weight space domain of the weight matrix and the training pressure-water injection rate time series interaction feature vector The domain gap between the probability distribution domains of the classification results is calculated by the weight matrix Relative to the training pressure-injection speed time series interaction feature vector The regularized representation of the degree matrix acts as an inter-domain transferring agent to transfer the probability distribution of valuable label constraints into the weight space. This prevents the richly labeled probability distribution domain from over-exploiting the weight distribution within the weight space during the weight space-based classification process. This improves the convergence of the weight matrix and, consequently, the training effect of the training pressure-injection rate time-series interaction feature vector when performing classification regression through the classifier. This enables real-time adaptive control of the injection rate based on the temporal changes in reactor pressure and injection rate, thereby achieving intelligent regulation of the reactor feedwater control system and improving its performance to maintain stable reactor operation.

[0042] In summary, the evaporator feed water control system 100 of a reactor nuclear power unit based on an embodiment of the present application is explained, which can improve the performance of the feed water control system to maintain stable operation of the reactor.

[0043] As described above, the reactor nuclear power unit evaporator feedwater control system 100 according to the embodiment of the present application can be implemented in various terminal devices, such as a server equipped with a reactor nuclear power unit evaporator feedwater control algorithm based on the embodiment of the present application. In one example, the reactor nuclear power unit evaporator feedwater control system 100 according to the embodiment of the present application can be integrated into the terminal device as a software module and / or hardware module. For example, the reactor nuclear power unit evaporator feedwater control system 100 according to the embodiment of the present application can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the reactor nuclear power unit evaporator feedwater control system 100 according to the embodiment of the present application can also be one of the many hardware modules of the terminal device.

[0044] Alternatively, in another example, the reactor nuclear power unit evaporator feed water control system 100 based on the embodiment of the present application and the terminal device may also be separate devices, and the reactor nuclear power unit evaporator feed water control system 100 may be connected to the terminal device through a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0045] FIG3 is a flow chart of a method for controlling evaporator water supply for a nuclear power plant according to an embodiment of the present application. FIG4 is a schematic diagram of a system architecture of a method for controlling evaporator water supply for a nuclear power plant according to an embodiment of the present application. As shown in FIG3 and FIG4 , the method for controlling evaporator water supply for a nuclear power plant according to an embodiment of the present application includes: S110, collecting pressure values ​​of the reactor at multiple predetermined time points in a predetermined time period and water injection rate values ​​at the multiple predetermined time points via a pressure sensor; S120, arranging the pressure values ​​at the multiple predetermined time points and the water injection rate values ​​at the multiple predetermined time points into a pressure time series input vector and a water injection rate time series input vector according to a time dimension; S130, performing a time series interaction correlation analysis on the pressure time series input vector and the water injection rate time series input vector to obtain a pressure-water injection rate time series interaction feature; and S140, determining whether the water injection rate value at the current time point should be increased or decreased based on the pressure-water injection rate time series interaction feature.

[0046] In a specific example, in the above-mentioned reactor nuclear power unit evaporator feed water control method, the pressure time series input vector and the water injection rate time series input vector are subjected to time series interaction correlation analysis to obtain pressure-water injection rate time series interaction characteristics, including: passing the pressure time series input vector and the water injection rate time series input vector through a format converter including an upsampling layer and a vector-image conversion unit to obtain a pressure time series image and a water injection rate time series image; performing multi-scale feature extraction on the pressure time series image and the water injection rate time series image respectively through a time series feature extractor based on a deep neural network model to obtain a pressure time series feature vector and a water injection rate time series feature vector; and, performing feature interaction correlation encoding on the pressure time series feature vector and the water injection rate time series feature vector to obtain a pressure-water injection rate time series interaction feature vector as the pressure-water injection rate time series interaction characteristic.

[0047] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned reactor nuclear power unit evaporator feed water control method have been introduced in detail in the description of the reactor nuclear power unit evaporator feed water control system 100 with reference to Figures 1 to 2 above, and therefore, its repeated description will be omitted.

[0048] Figure 5 illustrates an application scenario for a reactor nuclear power unit evaporator feedwater control system according to an embodiment of the present application. As shown in Figure 5 , in this application scenario, first, the reactor pressure values ​​(e.g., D1 shown in Figure 5 ) and the water injection rate values ​​(e.g., D2 shown in Figure 5 ) at multiple predetermined time points in a predetermined time period are obtained. Then, the pressure values ​​and the water injection rate values ​​at these multiple predetermined time points are input into a server (e.g., S shown in Figure 5 ) that is configured with a reactor nuclear power unit evaporator feedwater control algorithm. The server then processes the pressure values ​​and the water injection rate values ​​at these multiple predetermined time points using the reactor nuclear power unit evaporator feedwater control algorithm to obtain a classification result indicating whether the water injection rate value at the current time point should be increased or decreased.

[0049] This application uses specific terms to describe the embodiments of this application. For example, "first / second embodiment", "one embodiment", and / or "some embodiments" refer to a certain feature, structure, or characteristic related to at least one embodiment of this application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or multiple times in different places in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of this application may be appropriately combined.

[0050] In addition, it will be understood by those skilled in the art that various aspects of the present application can be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of the present application can be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may all be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of the present application may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0051] Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology and should not be interpreted in an idealized or highly formal sense, unless expressly defined as such herein.

[0052] The above is an illustration of the present invention and should not be considered as limiting thereof. Although several exemplary embodiments of the present invention have been described, it will be readily understood by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present invention. Therefore, all such modifications are intended to be included within the scope of the present invention as defined by the claims. It should be understood that the above is an illustration of the present invention and should not be considered as being limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present invention is defined by the claims and their equivalents.

Claims

1. A reactor nuclear power unit evaporator water supply control system, characterized in that: include: A data acquisition module, used to collect pressure values ​​of the reactor at multiple predetermined time points in a predetermined time period through a pressure sensor, and water injection rate values ​​at the multiple predetermined time points; A data parameter time series arrangement module, used for arranging the pressure values ​​at the plurality of predetermined time points and the water injection speed values ​​at the plurality of predetermined time points into a pressure time series input vector and a water injection speed time series input vector according to the time dimension respectively; A data parameter time series feature interaction module, used for performing time series interaction correlation analysis on the pressure time series input vector and the water injection rate time series input vector to obtain a pressure-water injection rate time series interaction feature; as well as The water injection speed control module is used to determine whether the water injection speed value at the current time point should be increased or decreased based on the pressure-water injection speed time series interaction characteristics.

2. The evaporator water supply control system for a reactor nuclear power unit according to claim 1, characterized in that: The data parameter time series feature interaction module includes: a format conversion unit, configured to obtain a pressure time series image and a water injection rate time series image by passing the pressure time series input vector and the water injection rate time series input vector through a format converter including an upsampling layer and a vector-to-image conversion unit; a parameter time series multi-scale feature extraction unit, configured to perform multi-scale feature extraction on the pressure time series image and the water injection rate time series image respectively through a time series feature extractor based on a deep neural network model to obtain a pressure time series feature vector and a water injection rate time series feature vector; and The pressure-water injection rate interaction coding unit is used to perform feature interaction coding on the pressure time series feature vector and the water injection rate time series feature vector to obtain a pressure-water injection rate time series interaction feature vector as the pressure-water injection rate time series interaction feature.

3. The evaporator feed water control system for a reactor nuclear power unit according to claim 2, characterized in that: The deep neural network model is a multi-scale neighborhood feature extraction module, which includes a first convolution layer and a second convolution layer in parallel, and a multi-scale feature fusion layer connected to the first convolution layer and the second convolution layer, wherein the first convolution layer and the second convolution layer use one-dimensional convolution kernels with different scales.

4. The evaporator feed water control system for a reactor nuclear power unit according to claim 3, characterized in that: The pressure-water injection rate interactive correlation coding unit is used to: The pressure time series feature vector and the water injection rate time series feature vector are interactively fused using a feature interaction layer to obtain the pressure-water injection rate time series interaction feature vector.

5. The evaporator feed water control system for a reactor nuclear power unit according to claim 4, characterized in that: The water injection speed control module is used to: The pressure-water injection rate time series interaction feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the water injection rate value at the current time point should be increased or decreased.

6. The evaporator feed water control system for a reactor nuclear power unit according to claim 5, characterized in that: It also includes a training module for training the format converter including the upsampling layer and the vector-to-image conversion unit, the temporal feature extractor based on the multi-scale neighborhood feature extraction module, and the classifier.

7. The evaporator feed water control system for a reactor nuclear power unit according to claim 6, characterized in that: The training module comprises: A training data acquisition unit, used to acquire training data, wherein the training data includes training pressure values ​​of the reactor at multiple predetermined time points in a predetermined time period, training water injection rate values ​​at the multiple predetermined time points, and a real value of the water injection rate value at the current time point that should be increased or decreased; A training data parameter time series arrangement unit, used to arrange the training pressure values ​​at the plurality of predetermined time points and the training water injection speed values ​​at the plurality of predetermined time points into a training pressure time series input vector and a training water injection speed time series input vector according to the time dimension; A training parameter time series format conversion unit, used for passing the training pressure time series input vector and the training water injection rate time series input vector through the format converter including the upsampling layer and the vector-image conversion unit to obtain a training pressure time series image and a training water injection rate time series image; A training parameter time series feature extraction unit, used for respectively passing the training pressure time series image and the training water injection rate time series image through the time series feature extractor based on the multi-scale neighborhood feature extraction module to obtain a training pressure time series feature vector and a training water injection rate time series feature vector; A training pressure-water injection rate time series feature interactive correlation encoding unit, used for interactively fusing the training pressure time series feature vector and the training water injection rate time series feature vector using the feature interaction layer to obtain a training pressure-water injection rate time series interactive feature vector; a classification loss unit, configured to pass the training pressure-water injection rate time series interaction feature vector through the classifier to obtain a classification loss function value; and A model training unit is used to train the format converter including the upsampling layer and the vector-to-image conversion unit, the time series feature extractor based on the multi-scale neighborhood feature extraction module, and the classifier based on the classification loss function value and through the direction propagation of gradient descent, wherein, at each weight matrix iteration of the training, the training pressure-water injection rate time series interaction feature vector is subjected to weight space exploration constraint optimization based on class degree matrix regularization.

8. The evaporator feed water control system for a reactor nuclear power unit according to claim 7, characterized in that: At each weight matrix iteration of the training, the training pressure-water injection rate time series interaction feature vector is subjected to weight space exploration constraint optimization based on class matrix regularization using the following optimization formula to obtain an optimized training pressure-water injection rate time series interaction feature vector; Wherein, the optimization formula is: in, is the training pressure-water injection rate time series interaction feature vector is the optimized training pressure-water injection rate time series interaction feature vector, and is a column vector, is a row vector, is the learnable domain transfer matrix, represents the weight matrix of the previous iteration, represents the weight matrix after iteration, Represents matrix multiplication Represents a transpose operation.

Images