Fault early warning method and system for nuclear power instrument control system

Abstract

The invention relates to the field of system fault detection, in particular to a fault early warning method and system for a nuclear power instrument control system, and the system comprises a plurality of data collection devices which are used for collecting the real-time data of a PCB of the nuclear power instrument control system connected with the data collection devices, and the real-time data comprise one-dimensional circuit parameter data, one-dimensional circuit parameter data and one-dimensional circuit parameter data; and two-dimensional temperature data including location and temperature; the fault early warning device is used for predicting whether the nuclear power instrument control system has a fault, a fault type and a fault position according to the circuit parameter data and the temperature data through a preset time sequence deep learning model and a preset time-space sequence deep learning model, and generating fault early warning information when the nuclear power instrument control system has the fault; and the early warning device is used for generating and sending an early warning notice and displaying the fault early warning information. According to the invention, high-precision and real-time fault early warning is realized, and powerful support is provided for ensuring safe and stable operation of industrial production.

Description

Technical Field

[0001] The present invention relates to the field of system fault detection, and in particular to a fault early warning method and system for a nuclear power instrumentation and control system. Background Art

[0002] The nuclear power instrumentation and control system is an automated control system widely used in the nuclear power industry. It ensures the stability of the production process and the consistency of product quality by real-time monitoring and control of various parameters in the production process. However, various faults may occur in the operation of the nuclear power instrumentation and control system, such as component aging, environmental interference, software vulnerabilities, etc., which may lead to system performance degradation or even complete failure. In order to avoid the serious consequences caused by failures, such as production stagnation, equipment damage, safety accidents, etc., it is necessary to monitor the nuclear power instrumentation and control system in real time and predict faults, so as to detect potential problems in time and take corresponding measures.

[0003] At present, the main methods for predicting faults in nuclear power instrumentation and control systems are based on the analysis of data such as voltage, current and single-point temperature. These methods monitor key circuit parameters and temperature changes at a single location and use statistical models or traditional machine learning algorithms, such as support vector machines and decision trees, to evaluate and predict the health status of the system. However, the prediction accuracy of these methods is not high in practical applications. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a fault warning method and system for a nuclear power instrumentation and control system.

[0005] The first aspect of the present invention discloses a fault early warning system for a nuclear power instrumentation and control system, comprising:

[0006] A plurality of data acquisition devices, used to collect real-time data of the PCB board of the nuclear power instrumentation and control system connected thereto, the real-time data including one-dimensional circuit parameter data and two-dimensional temperature data including position and temperature;

[0007] A fault warning device is used to predict whether the nuclear power instrumentation and control system has a fault, the fault type and the fault location according to the circuit parameter data and the temperature data through a preset time series deep learning model and a preset space-time series deep learning model, and generate fault warning information when there is a fault;

[0008] The early warning device is used to generate and send early warning notifications and display the fault early warning information.

[0009] Furthermore, the data acquisition device comprises:

[0010] A plurality of temperature sensors are arranged at a plurality of preset positions of the PCB board, and are used to measure the temperature at the positions and generate two-dimensional temperature data;

[0011] The thermometer is arranged above the PCB board and is used to measure the temperature of the surface of the PCB board and generate two-dimensional temperature data;

[0012] The data acquisition and transmission module is used to collect the one-dimensional circuit parameter data of the PCB board and integrate the two-dimensional temperature data and the one-dimensional circuit parameter data into real-time data.

[0013] Furthermore, the fault warning device includes:

[0014] A plurality of timing analysis modules, each corresponding to a PCB board; used to predict temperature anomaly according to real-time data of the corresponding PCB board through a preset time series deep learning model and a preset spatiotemporal series deep learning model, to obtain temperature anomaly data, wherein the temperature anomaly data includes an outlier temperature value and a temperature anomaly position;

[0015] The system health analysis module uses the preset first classifier model and the Mahalanobis distance method to predict whether a fault occurs in the nuclear power instrumentation and control system, as well as the fault type and location when a fault occurs, based on the temperature abnormality data of all PCB boards. When a fault occurs, the temperature abnormality data, fault type and fault location are used to generate fault warning information.

[0016] Furthermore, the timing analysis module includes:

[0017] A one-dimensional data prediction module, used to calculate a first predicted temperature according to the time series deep learning model based on the one-dimensional circuit parameter data of the corresponding PCB board;

[0018] A multidimensional data prediction module, used to calculate the second predicted temperature of different preset positions of the corresponding PCB board according to the spatiotemporal sequence deep learning model based on the two-dimensional temperature data of the corresponding PCB board;

[0019] The anomaly detection module is used to predict whether the temperature is abnormal based on the predicted temperature and according to a preset second classifier model. When a temperature anomaly is predicted, temperature anomaly data including an outlier temperature value and a temperature anomaly position is generated.

[0020] Furthermore, the early warning device includes:

[0021] A monitoring module, used to analyze temperature anomaly data and fault type according to the fault warning information, display a temperature distribution map according to the temperature anomaly data, and identify the fault location and fault type in the temperature distribution map;

[0022] The early warning module is used to generate and send an early warning notification according to the fault early warning information.

[0023] Furthermore, the fault warning system also includes:

[0024] The data management device is used to manage the storage and reading of the real-time data.

[0025] The second aspect of the present invention discloses a fault warning method for a nuclear power instrumentation and control system, using the fault warning system for a nuclear power instrumentation and control system disclosed in any one of the first aspects of the present invention, the fault warning method comprises:

[0026] Collecting real-time data of each PCB board in the nuclear power instrumentation and control system, the real-time data includes one-dimensional circuit parameter data and two-dimensional temperature data including position and temperature;

[0027] Based on a preset time series deep learning model and a preset space-time series deep learning model, according to the circuit parameter data and temperature data, predict whether there is a fault in the nuclear power instrumentation and control system, the fault type and the fault location;

[0028] When the prediction result is a fault, the fault warning information is displayed, and a warning notification is generated and sent.

[0029] Furthermore, based on the preset time series deep learning model and the preset space-time series deep learning model, the steps of predicting whether there is a fault, the fault type and the fault location in the nuclear power instrumentation and control system according to the circuit parameter data and the temperature data include:

[0030] For the real-time data of each PCB board, based on a preset time series deep learning model and a preset spatiotemporal series deep learning model, a temperature anomaly prediction is performed on the PCB board to obtain temperature anomaly data, wherein the temperature anomaly data includes an outlier temperature value and a temperature anomaly position;

[0031] Based on the preset first classifier model, the Mahalanobis distance method is used to predict whether there is a fault in the nuclear power instrumentation and control system, the fault type and the fault location according to the temperature abnormality data of all PCB boards.

[0032] Furthermore, for the real-time data of each PCB board, based on a preset time series deep learning model and a preset spatiotemporal series deep learning model, the step of performing temperature anomaly prediction on the PCB board to obtain temperature anomaly data includes:

[0033] Based on a preset time series deep learning model, according to the one-dimensional circuit parameter data of each PCB board, the temperature of the PCB board is predicted to obtain a first predicted temperature;

[0034] Based on a preset spatiotemporal sequence deep learning model, and according to the two-dimensional temperature data of each PCB board, predict a plurality of second predicted temperatures of the PCB board at a plurality of preset positions;

[0035] Based on the first predicted temperature and the plurality of second predicted temperatures, according to a preset second classifier model, it is predicted whether the temperature is abnormal, and when a temperature anomaly is predicted, temperature anomaly data including an outlier temperature value and a temperature anomaly position is generated.

[0036] Furthermore, the pre-training method of the first classifier model includes:

[0037] Test the PCB board under predefined conditions to obtain experimental data;

[0038] Collecting operation data of nuclear power instrumentation and control systems;

[0039] The selected first classifier model is trained in combination with the experimental data and the operating data to obtain a trained first classifier model.

[0040] The present invention makes full use of the value of multi-source heterogeneous data in nuclear power instrumentation and control systems, combines the powerful characterization capabilities of deep learning technology, and achieves high-precision and real-time fault warning, providing strong support for ensuring the safe and stable operation of industrial production. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0042] Figure 1 It is a structural schematic diagram of a fault early warning system of a nuclear power instrumentation and control system disclosed in an embodiment of the present invention;

[0043] Figure 2 It is a structural schematic diagram of a data acquisition device disclosed in an embodiment of the present invention;

[0044] Figure 3 The present invention discloses a flow chart of a fault warning method for a nuclear power instrumentation and control system. DETAILED DESCRIPTION

[0045] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, or product end including a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0047] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0048] See also Figure 1 As shown, Figure 1 Schematic diagram of a fault warning system for a nuclear power instrumentation and control system disclosed in an embodiment of the present invention. Figure 1 As shown, the fault warning system of the nuclear power instrumentation and control system may include the following modules:

[0049] A plurality of data acquisition devices 101, used to collect real-time data of the PCB board of the nuclear power instrumentation and control system connected thereto, wherein the real-time data includes one-dimensional circuit parameter data and two-dimensional temperature data including position and temperature;

[0050] In an optional embodiment, the structure of the data acquisition device is as follows: Figure 2 As shown, including:

[0051] A plurality of temperature sensors are arranged at a plurality of preset positions of the PCB board, and are used to measure the temperature at the positions and generate two-dimensional temperature data;

[0052] The thermometer is arranged above the PCB board and is used to measure the temperature of the surface of the PCB board and generate two-dimensional temperature data;

[0053] The data acquisition and transmission module is used to collect the one-dimensional circuit parameter data of the PCB board and integrate the two-dimensional temperature data and the one-dimensional circuit parameter data into real-time data.

[0054] In this optional embodiment, multiple temperature sensors are arranged at multiple preset positions of the PCB board to form a two-dimensional temperature monitoring matrix, and each temperature sensor measures the temperature at its position. The multiple preset positions may be the surface or nearby positions of key components on the PCB board, such as power modules, processors, and high-frequency components.

[0055] In this optional embodiment, the thermometer can be a miniature infrared thermometer, which is installed above the PCB board to obtain an infrared temperature image of the PCB board surface and provide overall temperature distribution information. In this case, the PCB board housing is made of a transparent material, such as a highly transparent resin or glass.

[0056] In this optional embodiment, the data acquisition and transmission module is connected to the temperature monitoring matrix and the micro infrared thermometer, is responsible for collecting temperature data, and preliminarily processing and transmitting the data. The data acquisition and transmission module can perform synchronous measurements on multiple signals.

[0057] A fault warning device 102 is used to predict whether the nuclear power instrumentation and control system has a fault, the fault type and the fault location according to the circuit parameter data and the temperature data through a preset time series deep learning model and a preset space-time series deep learning model, and generate fault warning information when a fault occurs;

[0058] In an optional embodiment, the fault warning device includes:

[0059] A plurality of timing analysis modules, each corresponding to a PCB board; used to predict temperature anomaly according to real-time data of the corresponding PCB board through a preset time series deep learning model and a preset spatiotemporal series deep learning model, to obtain temperature anomaly data, wherein the temperature anomaly data includes an outlier temperature value and a temperature anomaly position;

[0060] The system health analysis module uses the preset first classifier model and the Mahalanobis distance method to predict whether a fault occurs in the nuclear power instrumentation and control system, as well as the fault type and location when a fault occurs, based on the temperature abnormality data of all PCB boards. When a fault occurs, the temperature abnormality data, fault type and fault location are used to generate fault warning information.

[0061] In this optional embodiment, Mahalanobis distance is a distance measurement method that takes into account the correlation between variables. Unlike Euclidean distance, Mahalanobis distance takes into account the covariance matrix of data and can effectively measure the similarity between multidimensional data points. In the health status assessment, by calculating the Mahalanobis distance between the temperature abnormality data of the PCB board to be tested and the temperature data under normal state, the degree of deviation between the current state and the normal state can be quantified. Specifically, the mean vector and covariance matrix of the temperature data are first estimated based on historical data, and then the Mahalanobis distance between the temperature abnormality data and the mean vector is calculated. The larger the Mahalanobis distance, the greater the difference between the current state and the normal state, that is, the health state may be abnormal. By presetting a suitable threshold, automatic assessment and abnormality detection of the health status of the PCB board can be achieved.

[0062] In an optional embodiment, the timing analysis module includes:

[0063] A one-dimensional data prediction module, used to calculate a first predicted temperature according to the time series deep learning model based on the one-dimensional circuit parameter data of the corresponding PCB board;

[0064] A multidimensional data prediction module, used to calculate the second predicted temperature of different preset positions of the corresponding PCB board according to the spatiotemporal sequence deep learning model based on the two-dimensional temperature data of the corresponding PCB board;

[0065] The anomaly detection module is used to predict whether the temperature is abnormal based on the predicted temperature and according to a preset second classifier model. When a temperature anomaly is predicted, temperature anomaly data including an outlier temperature value and a temperature anomaly position is generated.

[0066] In this optional embodiment, the time series deep learning model in the one-dimensional data prediction module runs in parallel with the spatiotemporal series deep learning model in the multidimensional data prediction module.

[0067] The classifier model is a supervised learning model that builds a model that can map input data to discrete categories by learning the features of labeled data and the corresponding category labels. In the scenario of temperature anomaly detection, the classifier model can be used to determine whether the given temperature data is abnormal.

[0068] Specifically, the second classifier model in the anomaly detection module can be trained and applied through the following steps:

[0069] Data preparation: Collect the temperature data of the PCB board under normal working conditions, as well as the known temperature abnormality data, and mark them as normal (0) or abnormal (1);

[0070] Feature extraction: extract relevant features from the raw temperature data, such as temperature statistics (mean, variance, maximum, minimum, etc.), temperature change rate, etc. These features can help the classifier model better capture the pattern of temperature anomalies;

[0071] Model training: Use the extracted features and corresponding labels to train the second classifier model. Common classifier models include logistic regression, support vector machine decision tree, random forest, etc. Optimize model parameters by minimizing objective functions such as classification error rate or cross entropy loss;

[0072] Model evaluation: Use methods such as holdout or cross-validation to evaluate the performance of the trained classifier model on unseen data, such as accuracy, precision, recall, and other indicators;

[0073] Anomaly detection: In actual applications, the temperature predicted by the one-dimensional data prediction module and the multi-dimensional data prediction module is input into the trained second classifier model. The model will classify the input temperature data as normal or abnormal based on the learned temperature anomaly pattern. If the prediction result is abnormal, the anomaly detection module will generate temperature anomaly data containing outlier temperature values ​​and temperature anomaly locations for subsequent analysis and processing.

[0074] The early warning device 103 is used to generate and send an early warning notification and display the fault early warning information.

[0075] In an optional embodiment, the early warning device includes:

[0076] A monitoring module, used to analyze temperature anomaly data and fault type according to the fault warning information, display a temperature distribution map according to the temperature anomaly data, and identify the fault location and fault type in the temperature distribution map;

[0077] The early warning module is used to generate and send an early warning notification according to the fault early warning information.

[0078] In an optional embodiment, the fault warning system further includes:

[0079] The data management device is used to manage the storage and reading of the real-time data.

[0080] In this optional embodiment, the data collected by the data collection device can be stored in the data management device, and the fault warning device can read the real-time data from the data management device.

[0081] See also Figure 3 As shown, Figure 3 1 is a flow chart of a fault warning method for a nuclear power instrumentation and control system disclosed in an embodiment of the present invention. Figure 3As shown, a fault warning method for a nuclear power instrumentation and control system, using the fault warning system for a nuclear power instrumentation and control system as described in any of the above embodiments, the fault warning method may include:

[0082] S301, collecting real-time data of each PCB board in the nuclear power instrumentation and control system, wherein the real-time data includes one-dimensional circuit parameter data and two-dimensional temperature data including position and temperature;

[0083] S302, based on a preset time series deep learning model and a preset space-time series deep learning model, predict whether there is a fault in the nuclear power instrumentation and control system, the fault type and the fault location according to the circuit parameter data and the temperature data;

[0084] In an optional embodiment, based on a preset time series deep learning model and a preset space-time series deep learning model, the step of predicting whether there is a fault, the fault type and the fault location in the nuclear power instrumentation and control system according to the circuit parameter data and the temperature data includes:

[0085] For the real-time data of each PCB board, based on a preset time series deep learning model and a preset spatiotemporal series deep learning model, a temperature anomaly prediction is performed on the PCB board to obtain temperature anomaly data, wherein the temperature anomaly data includes an outlier temperature value and a temperature anomaly position;

[0086] Based on the preset first classifier model, the Mahalanobis distance method is used to predict whether there is a fault in the nuclear power instrumentation and control system, the fault type and the fault location according to the temperature abnormality data of all PCB boards.

[0087] In this optional embodiment, for the Mahalanobis distance method and the introduction of the classifier model, please refer to the description of the relevant parts of the fault warning system of the nuclear power instrumentation and control system of the present invention, which will not be repeated here.

[0088] In this optional embodiment, the first classifier model and the second classifier model may be classifier models of the same type, which is not limited by the embodiment of the present invention.

[0089] In another optional embodiment, for the real-time data of each PCB board, based on a preset time series deep learning model and a preset spatiotemporal series deep learning model, the step of performing temperature anomaly prediction on the PCB board to obtain temperature anomaly data includes:

[0090] Based on a preset time series deep learning model, according to the one-dimensional circuit parameter data of each PCB board, the temperature of the PCB board is predicted to obtain a first predicted temperature;

[0091] Based on a preset spatiotemporal sequence deep learning model, and according to the two-dimensional temperature data of each PCB board, predict a plurality of second predicted temperatures of the PCB board at a plurality of preset positions;

[0092] Based on the first predicted temperature and the plurality of second predicted temperatures, according to a preset second classifier model, it is predicted whether the temperature is abnormal, and when a temperature anomaly is predicted, temperature anomaly data including an outlier temperature value and a temperature anomaly position is generated.

[0093] In this optional embodiment, the time series deep learning model performs time series prediction on one-dimensional circuit parameter data to capture the event dependencies and trend changes of the data.

[0094] The time series deep learning model can be a recurrent neural network (RNN), a long short-term memory network (LSTM), a gated recurrent unit (GRU), or a Transformer model, which is not limited in the embodiments of the present invention.

[0095] RNN is a neural network model that specializes in processing sequence data. Unlike traditional feedforward neural networks, RNN introduces recurrent connections, which enables the network to maintain a hidden state and capture long-term dependencies in sequence data. In RNN, the output of each time step depends not only on the input of the current time step, but also on the hidden state of the previous time step.

[0096] LSTM is an improved variant of RNN, which aims to solve the gradient vanishing and gradient exploding problems of ordinary RNN. LSTM introduces a special memory unit, which contains three gating mechanisms: input gate, forget gate and output gate. These gating mechanisms allow LSTM to selectively memorize and forget information, so that it can better capture long-term dependencies. The input gate controls the entry of new information into the memory unit, the forget gate controls the forgetting of historical information, and the output gate controls the flow of information in the memory unit to the next time step. Through this gating mechanism, LSTM is able to learn and maintain key information in long sequences while avoiding the gradient vanishing and gradient exploding problems, enabling it to achieve excellent performance in many sequence modeling tasks.

[0097] GRU is another popular RNN variant, similar to LSTM, but with a simpler structure. GRU combines the forget gate and input gate into an update gate, and merges the cell state and hidden state at the same time. The update gate controls the retention of historical information and the addition of new information, and the reset gate controls the impact of historical information on the current time step. Compared with LSTM, GRU has fewer parameters and lower computational complexity, but can achieve comparable performance to LSTM in many tasks.

[0098] Transformer completely abandons the loop structure and uses the self-attention mechanism to capture long-term dependencies in the sequence. Transformer consists of two parts: the encoder and the decoder. The encoder encodes the input sequence through multi-head self-attention and feedforward neural networks, and the decoder generates the output sequence through a similar structure. The self-attention mechanism allows the model to establish direct dependencies between elements at different positions without the need to pass information step by step like RNN. This parallel computing method enables Transformer to efficiently process long sequences while learning global dependencies in the sequence.

[0099] In this optional embodiment, the spatiotemporal series deep learning model performs spatiotemporal series prediction on two-dimensional temperature data, capturing correlations and pattern changes in space and time.

[0100] The spatiotemporal sequence deep learning model can be a convolutional neural network combined with a recurrent neural network (CNN+RNN), a temporal convolutional network (TCN), a convolutional long short-term memory network (ConvLSTM), a graph attention network (GAT), a spatiotemporal Transformer model (ST-Transformer), a convolutional autoencoder combined with LSTM (CAE+LSTM), and a variational autoencoder combined with LSTM (VAE+LSTM).

[0101] In this optional embodiment, a convolutional neural network is combined with a recurrent neural network (CNN+RNN): the convolutional neural network (CNN) is used to extract local features, and the recurrent neural network (RNN) is used to capture temporal dependencies. CNN is responsible for learning spatial features, and RNN is responsible for learning temporal relationships. The combination of the two can simultaneously model spatiotemporal information.

[0102] Temporal Convolutional Network (TCN) uses causal convolution and dilated convolution to capture long-term dependencies in time series. TCN can be calculated in parallel, has fast training speed, and can process variable-length sequences, making it suitable for a variety of time series modeling tasks.

[0103] Convolutional Long Short-Term Memory (ConvLSTM) network introduces convolution operation into LSTM network, making it able to process spatiotemporal sequence data. Convolutional LSTM network adds convolution layer on the basis of LSTM network, which can extract spatiotemporal features and temporal dependencies at the same time.

[0104] Graph Attention Network (GAT): It learns and predicts on graph structured data through the attention mechanism. The graph attention network can adaptively assign different weights to different nodes, effectively capture the complex relationship between nodes, and is suitable for spatiotemporal graph modeling tasks such as traffic flow prediction.

[0105] The spatiotemporal Transformer model (ST-Transformer) applies the Transformer architecture to spatiotemporal sequence modeling. ST-Transformer uses a self-attention mechanism to learn spatiotemporal dependencies, has high parallel computing efficiency, and can handle long sequences and multivariate spatiotemporal data.

[0106] Convolutional autoencoder combined with LSTM (CAE+LSTM): Use convolutional autoencoder to learn the compressed representation of spatiotemporal data, and then input the compressed representation into the long short-term memory network for time series modeling. Convolutional autoencoder (CAE) is responsible for learning the high-level features of the data, and the long short-term memory network is responsible for capturing time dependencies. The combination of the two can effectively model spatiotemporal sequences.

[0107] Variational Autoencoder combined with LSTM (VAE+LSTM): Combining variational autoencoder and LSTM for the generation and prediction of spatiotemporal sequences. Variational autoencoder (VAE) learns the probability distribution of data, and LSTM captures the temporal dependency. Combining the two can generate spatiotemporal sequence data.

[0108] In an optional embodiment, the pre-training method of the first classifier model includes:

[0109] Test the PCB board under predefined conditions to obtain experimental data;

[0110] Collecting operation data of nuclear power instrumentation and control systems;

[0111] The selected first classifier model is trained in combination with the experimental data and the operating data to obtain a trained first classifier model.

[0112] In this optional embodiment, the test under predefined conditions refers to performing abnormal condition tests such as heating and humidification on the PCB board or power module in a laboratory environment, simulating a fault state, and collecting temperature and circuit parameter data under abnormal condition tests.

[0113] S303: When the prediction result is a fault, display fault warning information, generate and send a warning notification.

[0114] In this optional embodiment, the fault warning information refers to a temperature distribution map, in which the fault location and the fault type are marked. The warning notification includes the fault location and the fault type, and is pushed to a preset person.

[0115] In summary, the present invention discloses a fault warning method and system for a nuclear power instrumentation and control system, which fully utilizes the value of multi-source heterogeneous data in the nuclear power instrumentation and control system, combines the powerful characterization ability of deep learning technology, and realizes high-precision and real-time fault warning, providing strong support for ensuring the safe and stable operation of industrial production. Therefore, the present invention effectively overcomes the various shortcomings of the prior art and has a high industrial utilization value.

[0116] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A fault warning system for a nuclear power instrumentation and control system, characterized in that: The fault warning system comprises: A plurality of data acquisition devices, used to collect real-time data of the PCB board of the nuclear power instrumentation and control system connected thereto, the real-time data including one-dimensional circuit parameter data and two-dimensional temperature data including position and temperature; A fault warning device is used to predict whether the nuclear power instrumentation and control system has a fault, the fault type and the fault location according to the circuit parameter data and the temperature data through a preset time series deep learning model and a preset space-time series deep learning model, and generate fault warning information when there is a fault; The early warning device is used to generate and send early warning notifications and display the fault early warning information.

2. A fault warning system for a nuclear power instrumentation and control system according to claim 1, characterized in that: The data acquisition device comprises: A plurality of temperature sensors are arranged at a plurality of preset positions of the PCB board, and are used to measure the temperature at the positions and generate two-dimensional temperature data; The thermometer is arranged above the PCB board and is used to measure the temperature of the surface of the PCB board and generate two-dimensional temperature data; The data acquisition and transmission module is used to collect the one-dimensional circuit parameter data of the PCB board and integrate the two-dimensional temperature data and the one-dimensional circuit parameter data into real-time data.

3. A fault warning system for a nuclear power instrumentation and control system according to claim 1, characterized in that: The fault warning device includes: A plurality of timing analysis modules, each corresponding to a PCB board; used to predict temperature anomaly according to real-time data of the corresponding PCB board through a preset time series deep learning model and a preset spatiotemporal series deep learning model, to obtain temperature anomaly data, wherein the temperature anomaly data includes an outlier temperature value and a temperature anomaly position; The system health analysis module uses the preset first classifier model and the Mahalanobis distance method to predict whether a fault occurs in the nuclear power instrumentation and control system, as well as the fault type and location when a fault occurs, based on the temperature abnormality data of all PCB boards. When a fault occurs, the temperature abnormality data, fault type and fault location are used to generate fault warning information.

4. A fault warning system for a nuclear power instrumentation and control system according to claim 3, characterized in that: The timing analysis module comprises: A one-dimensional data prediction module, used to calculate a first predicted temperature according to the time series deep learning model based on the one-dimensional circuit parameter data of the corresponding PCB board; A multidimensional data prediction module, used to calculate the second predicted temperature of different preset positions of the corresponding PCB board according to the spatiotemporal sequence deep learning model based on the two-dimensional temperature data of the corresponding PCB board; The anomaly detection module is used to predict whether the temperature is abnormal based on the predicted temperature and according to a preset second classifier model. When a temperature anomaly is predicted, temperature anomaly data including an outlier temperature value and a temperature anomaly position is generated.

5. A fault warning system for a nuclear power instrumentation and control system according to claim 1, characterized in that: Early warning devices include: A monitoring module, used to analyze temperature anomaly data and fault type according to the fault warning information, display a temperature distribution map according to the temperature anomaly data, and identify the fault location and fault type in the temperature distribution map; The early warning module is used to generate and send an early warning notification according to the fault early warning information.

6. A fault warning system for a nuclear power instrumentation and control system according to claim 1, characterized in that: The fault warning system also includes: The data management device is used to manage the storage and reading of the real-time data.

7. A fault warning method for a nuclear power instrumentation and control system, characterized in that: Using the fault early warning system of the nuclear power instrumentation and control system as described in any one of claims 1 to 6, the fault early warning method comprises: Collecting real-time data of each PCB board in the nuclear power instrumentation and control system, the real-time data includes one-dimensional circuit parameter data and two-dimensional temperature data including position and temperature; Based on a preset time series deep learning model and a preset space-time series deep learning model, according to the circuit parameter data and temperature data, predict whether there is a fault in the nuclear power instrumentation and control system, the fault type and the fault location; When the prediction result is a fault, the fault warning information is displayed, and a warning notification is generated and sent.

8. A fault warning method for a nuclear power instrumentation and control system according to claim 7, characterized in that: Based on a preset time series deep learning model and a preset space-time series deep learning model, the steps of predicting whether there is a fault, the fault type and the fault location in the nuclear power instrumentation and control system according to the circuit parameter data and the temperature data include: For the real-time data of each PCB board, based on a preset time series deep learning model and a preset spatiotemporal series deep learning model, a temperature anomaly prediction is performed on the PCB board to obtain temperature anomaly data, wherein the temperature anomaly data includes an outlier temperature value and a temperature anomaly position; Based on the preset first classifier model, the Mahalanobis distance method is used to predict whether there is a fault in the nuclear power instrumentation and control system, the fault type and the fault location according to the temperature abnormality data of all PCB boards.

9. A fault warning method for a nuclear power instrumentation and control system according to claim 8, characterized in that: For the real-time data of each PCB board, based on a preset time series deep learning model and a preset spatiotemporal series deep learning model, the temperature anomaly prediction of the PCB board is performed, and the step of obtaining the temperature anomaly data includes: Based on a preset time series deep learning model, according to the one-dimensional circuit parameter data of each PCB board, the temperature of the PCB board is predicted to obtain a first predicted temperature; Based on a preset spatiotemporal sequence deep learning model, and according to the two-dimensional temperature data of each PCB board, predict a plurality of second predicted temperatures of the PCB board at a plurality of preset positions; Based on the first predicted temperature and the plurality of second predicted temperatures, according to a preset second classifier model, it is predicted whether the temperature is abnormal, and when a temperature anomaly is predicted, temperature anomaly data including an outlier temperature value and a temperature anomaly position is generated.

10. A fault warning method for a nuclear power instrumentation and control system according to claim 8, characterized in that: The pre-training method of the first classifier model includes: Test the PCB board under predefined conditions to obtain experimental data; Collecting operation data of nuclear power instrumentation and control systems; The selected first classifier model is trained in combination with the experimental data and the operating data to obtain a trained first classifier model.

Images