Nuclear power enclosed bus intelligent monitoring model based on big data

Abstract

The invention relates to a nuclear power enclosed bus intelligent monitoring model based on big data. The model specifically comprises a plurality of technical links such as data acquisition, cleaning preprocessing, feature extraction and selection, modeling and algorithm design, health condition evaluation and monitoring and the like. The intelligent monitoring module designed by the invention realizes comprehensive, real-time and intelligent monitoring of the operation state of the enclosed bus through integrating a plurality of links such as data acquisition, cleaning, feature extraction, modeling prediction and the like; real-time monitoring and early warning functions are realized by combining an Internet of Things technology and a cloud computing platform; a user is helped to quickly know the equipment state and make a scientific and reasonable operation and maintenance decision through display and decision support functions; potential faults are found in time through comprehensive monitoring and intelligent evaluation of the operation state of the enclosed bus, and effective measures are taken for processing, so that the safety and reliability of equipment are enhanced, and long-term stable operation of a nuclear power station is guaranteed; the intelligent monitoring model adopts a modular design and is easy to adapt to different monitoring requirements and scenes.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment monitoring and big data analysis, and in particular to a nuclear power closed bus intelligent monitoring model based on big data. The model integrates multiple links such as data collection, data cleaning, feature extraction, model prediction, health status assessment and visual display, aiming to achieve comprehensive, real-time and intelligent monitoring of the operating status of the nuclear power closed bus, and improve the operating safety and reliability of the nuclear power plant. Background Art

[0002] With the continuous growth of global energy demand and the increasing awareness of environmental protection, nuclear energy, as a clean and efficient energy form, occupies an increasingly important position in the energy structure. As the main carrier of nuclear energy utilization, the safe and stable operation of nuclear power plants is of great significance to ensuring energy supply and environmental protection. As an important part of the power transmission system of nuclear power plants, the health of closed busbars directly affects the overall operating efficiency and safety of nuclear power plants.

[0003] The traditional closed bus monitoring method mainly relies on manual inspection and regular maintenance. This method has many shortcomings, such as low monitoring efficiency, slow response speed, and low prediction accuracy. With the rapid development of technologies such as big data, cloud computing, and artificial intelligence, new ideas and methods have been provided for power equipment monitoring. Using big data technology, the comprehensive collection and efficient processing of closed bus operation data can be achieved; through machine learning algorithms, the potential laws in the data can be mined and accurate prediction models can be established; combined with visualization technology, the equipment status can be intuitively displayed to provide users with convenient decision support.

[0004] Therefore, developing an intelligent monitoring model for nuclear power closed busbar based on big data is of great significance to improving the operational safety and reliability of nuclear power plants. Summary of the invention

[0005] The purpose of the present invention is to provide a nuclear power closed bus intelligent monitoring model based on big data, which realizes comprehensive, real-time and intelligent monitoring of the operating status of the nuclear power closed bus by integrating multiple links such as data collection, data cleaning, feature extraction, model prediction, health status assessment and visualization display.

[0006] To achieve the above object, the present invention provides the following technical solutions: A nuclear power closed bus intelligent monitoring model based on big data includes a data acquisition and storage module, a data cleaning and preprocessing module, a feature extraction and selection module, a modeling and algorithm design module, and a health status assessment and monitoring module.

[0007] The data acquisition and storage module is used to collect the operating data of the nuclear power closed bus in real time, including current, voltage, temperature, humidity, vibration and other indicators, and transmit the collected data to the data center for storage.

[0008] The data cleaning and preprocessing module is used to remove noise, process outliers, fill in missing values, unify data formats, standardize or normalize the collected raw data to improve data quality.

[0009] The feature extraction and selection module is used to extract features useful for predicting the health status of the closed busbar from the cleaned and preprocessed data, and perform feature selection to determine the most representative feature set.

[0010] The modeling and algorithm design module is used to establish a prediction model using a machine learning algorithm to predict the health status of the closed busbar. The machine learning algorithm includes but is not limited to a support vector machine, a decision tree, a random forest, a neural network or a deep learning algorithm, and an integrated learning method can be used to improve the accuracy and stability of the prediction model.

[0011] The health status assessment and monitoring module is used to assess and monitor the health status of the closed busbar based on the output results of the prediction model, including real-time monitoring and early warning functions, and timely issue alarm information when abnormalities or potential faults are detected.

[0012] Preferably, the data acquisition and storage module is specifically: This module is responsible for real-time collection of operating data of nuclear power closed busbars, including current, voltage, temperature, humidity, vibration and other indicators. Data collection is achieved through a sensor network installed on the closed busbar, which includes various types of sensors, such as current sensors, voltage sensors, temperature sensors, humidity sensors and vibration sensors. These sensors can monitor the operating status of the closed busbar in real time and transmit the collected data to the data center for storage. The data center uses high-performance storage devices and cloud computing platforms to ensure fast data processing and efficient storage. The stored data includes raw data and preliminarily processed data, which provides a basis for subsequent data analysis and model training.

[0013] Step S11: Real-time collection of operating data of nuclear power closed busbar, including but not limited to current, voltage, temperature, humidity, vibration and other indicators, marked as {I t , V t , T t , RH t , A t}, where I t Expressed as the current value at time point t, V t Expressed as the voltage value at time point t, Tt Expressed as the temperature value at time point t, RH t Expressed as the relative humidity value at time point t, A t Expressed as the vibration level at time point t; Step S12: preprocessing the collected data, including but not limited to filtering, denoising, etc., to improve data quality; Step S13: Packing the preprocessed data into a specific data format; Step S14: Transmitting the packaged data to the data center via wired or wireless communication; Step S15: In the data center, the received data is stored in a network database for subsequent data analysis and processing.

[0014] Preferably, the data cleaning and preprocessing module is specifically: The data cleaning and preprocessing module uses a filtering algorithm to remove high-frequency noise and random interference in the data, uses statistical methods to identify and remove outliers, uses interpolation or machine learning-based methods to fill in missing values, and performs data format unification and data standardization or normalization. The module includes the following steps: Step S21, high frequency noise and random interference removal: The data cleaning and preprocessing module first uses a filtering algorithm to process the collected raw data to remove high-frequency noise and random interference in the data; Step S22, outlier identification and elimination: Statistical methods are used to automatically identify and remove outliers in the data. By calculating the mean, standard deviation and other statistical quantities of the data, reasonable thresholds are set, and data that exceeds the threshold range is regarded as outliers and removed. Furthermore, algorithms based on distance, density or clustering are used to further identify and process abnormal data points. Step S23, missing value filling: For missing values ​​in the data, this module uses interpolation to fill them; Step S23, missing value filling: For missing values ​​in the data, this module uses a machine learning-based method to fill them; Step S24, data format is unified: After completing the above processing, the module will further unify the data format to ensure that all input data meets the requirements of model processing, including but not limited to data type conversion, timestamp alignment, unit unification and other operations, so as to facilitate subsequent data analysis and processing; Step S25, data standardization and normalization processing: The data is standardized and normalized to eliminate the impact of different dimensions and numerical ranges on model training.

[0015] Preferably, the feature extraction and selection module is specifically: The feature extraction and selection module extracts features by using methods such as time domain analysis, frequency domain analysis, wavelet transform, principal component analysis, etc., and performs feature selection by using statistical methods, machine learning methods, or deep learning methods, specifically: Step S31, feature extraction: The module first extracts features from the data using a variety of methods, including but not limited to: Method S311, time domain analysis method: using the statistical characteristics of time series data in the time domain, such as mean, variance, skewness, kurtosis, etc., as well as the difference and trend term of the time series as features; Method S312, frequency domain analysis method: converting the time domain signal to the frequency domain through Fourier transform (FT) or fast Fourier transform (FFT) to extract the spectrum features; Method S313, Wavelet Transform: Apply wavelet transform techniques, such as continuous wavelet transform (CWT) or discrete wavelet transform (DWT), to analyze signals at multiple scales and extract wavelet coefficients as features to capture local characteristics and detail information of the signal; Method S314, principal component analysis (PCA): dimensionality reduction is performed on the original feature set, and multiple variables are converted into a few principal components through linear transformation. These principal components retain most of the information of the original data while reducing the redundancy and noise of the data; Step S32, feature selection: The extracted features are further screened to remove redundant features. The module uses one or more of the following methods for feature selection: Method S321, statistical-based method: using statistical tests to evaluate the correlation or importance between features and target variables, and selecting features with higher significance; Method S322, a machine learning-based method: using the built-in feature importance evaluation mechanism of the machine learning algorithm to select the features that contribute most to the model prediction performance; Method S323, a method based on deep learning: In a deep learning model, the most representative features are automatically learned and extracted through the training process, or feature selection is indirectly achieved through dropout and L1 / L2 regularization techniques.

[0016] Example of feature selection formula (taking the chi-square test based on statistics as an example): Assume that the chi-square statistic between feature X and target variable Y is χ 2 (X,Y), then the feature selection criteria can be expressed as: If x 2 (X,Y)>threshold, then select feature X The threshold is a value set based on the specific situation of the data and the requirements of feature selection. The choice of threshold is subjective, but it is based on the understanding of the data set and the setting of model performance requirements.

[0017] χ 2 (X,Y) > Threshold: This means that the correlation between feature X and target variable Y is strong enough that we can consider feature X to be helpful for the model’s prediction. Hence, we should select this feature and include it in the model’s training process.

[0018] χ 2 (X,Y) is less than or equal to the threshold: This indicates that the correlation between feature X and target variable Y is weak and may not have a significant impact on the predictive performance of the model. Therefore, we can choose to ignore this feature to reduce the complexity and computational cost of the model.

[0019] Preferably, the modeling and algorithm design module is specifically: The prediction model in the modeling and algorithm design module adopts a support vector machine algorithm or a deep learning algorithm. The support vector machine algorithm also needs to combine multiple base learners through an integrated learning method to improve the overall performance, specifically: Method S41: When the support vector machine algorithm is used, the algorithm separates data points of different categories by finding an optimal hyperplane, and its objective function is expressed as:

[0020] in, is the i-th feature, m is the number of features, express For the target value, w is the normal vector of the hyperplane, b is the intercept, C is the regularization parameter, ξ i is the slack variable, Indicates the limiting condition; and, in order to improve the overall performance, multiple different support vector machine base learners can be trained, each base learner uses a different kernel function, regularization parameter or data subset; the prediction results of all base learners are aggregated, and the aggregation method can be simple average, weighted average, voting method or more complex strategies such as Stacking; the aggregated results are used as the final prediction output to monitor the operating status of the nuclear power closed bus system; Method S42: When a deep learning algorithm is used, the algorithm uses a multi-layer structure of a neural network to automatically learn an abstract representation of data, including but not limited to convolutional neural networks (CNN), recurrent neural networks (RNN), Transformer and their variants; the deep learning model optimizes network parameters through a back-propagation algorithm to minimize prediction errors.

[0021] Preferably, the health status assessment and monitoring module is specifically: The health status assessment and monitoring module assesses the health status of the closed busbar by setting a threshold or a probability-based method, monitors the equipment status in real time, and issues an alarm message in a timely manner when an abnormality or potential failure is detected. The alarm message is sent to relevant personnel through a preset communication channel, as follows: Method S51: Setting a threshold evaluation method, setting a preset threshold range for each key operating parameter, when the real-time data exceeds the range, it is determined to be abnormal or a potential fault, and an alarm mechanism is triggered; Method S52: Probability-based evaluation method, using probability models, such as Bayesian networks, hidden Markov models, etc., combined with historical data and real-time data, to calculate the probability of the closed bus system being in different health states, and when the probability of an unhealthy state exceeds a preset threshold, an alarm is issued. The specific form of the probability model can be expressed as a part of the Bayesian network, and its conditional probability calculation formula is as follows:

[0022] Where H represents a certain health state of the closed bus system, D represents the observed data (i.e., the real-time operating parameters after preprocessing), P(H|D) is the posterior probability that the system is in state H given the observed data D, P(D|H) is the likelihood probability of observing data D given the system is in state H, P(H) is the prior probability that the system is in state H, and P(D) is the marginal probability of observing data D, which is usually used as a normalization factor; The alarm information is sent to relevant personnel through a preset communication channel, including but not limited to a wireless network, a wired network, a text message service, an email or a dedicated monitoring system interface, and the alarm information contains a specific description of the abnormality or potential fault, the time of occurrence, location information and recommended response measures.

[0023] Technical effects and advantages The big data-based intelligent monitoring model for nuclear power closed busbars proposed in the present invention has the following technical effects and advantages: 1. Improve monitoring efficiency and accuracy: By integrating multiple links such as data collection, cleaning, feature extraction, modeling and prediction, comprehensive, real-time and intelligent monitoring of the operating status of the closed busbar is achieved. Compared with traditional manual inspection and regular maintenance methods, the monitoring efficiency and accuracy are greatly improved.

[0024] 2. Realize fault warning and rapid response: Combine IoT technology and cloud computing platform to realize real-time monitoring and warning functions. When the system detects an abnormality or potential fault, it can quickly issue an alarm message and notify relevant personnel so that they can take timely measures to deal with it, thereby effectively avoiding the impact of equipment failure on the operation of the nuclear power plant.

[0025] 3. Provide decision support to optimize operation and maintenance strategies: Through display and decision support functions, users can quickly understand the status of equipment and make scientific and reasonable operation and maintenance decisions. The optimized operation and maintenance solutions and decision suggestions provided by the system are based on a large amount of historical data and real-time monitoring results. They are highly accurate and practical, and help reduce operation and maintenance costs and improve operation and maintenance efficiency.

[0026] 4. Enhance equipment safety and reliability: Through comprehensive monitoring and intelligent evaluation of the closed busbar operating status, potential faults can be discovered in a timely manner and effective measures can be taken to deal with them, thereby enhancing the safety and reliability of equipment and ensuring the long-term stable operation of the nuclear power plant.

[0027] 5. Strong adaptability and easy expansion: The intelligent monitoring model of the present invention adopts a modular design that is easy to integrate different sensors and algorithm modules to adapt to different monitoring needs and scenarios. At the same time, with the continuous development of technology and the continuous expansion of application scenarios, the system can also easily introduce new algorithms and technical means for upgrading and optimization to continuously improve monitoring performance and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a schematic diagram of the system module process connection of the present invention; Figure 2 This is a schematic diagram of the connection between the data acquisition and storage modules of the present invention; Figure 3 It is a schematic diagram of the connection of the data cleaning and preprocessing modules of the present invention. DETAILED DESCRIPTION

[0029] Implementation Phase 1: Implementation of Data Collection and Storage Module In this embodiment, the data acquisition and storage module is implemented by combining a sensor network and a cloud computing platform. The sensor network includes various types of sensors such as current sensors, voltage sensors, temperature sensors, humidity sensors, and vibration sensors, which can monitor various operating parameters of the closed bus in real time. The sensor network transmits the collected data to the data center for storage through wireless communication technology.

[0030] The data center uses high-performance storage devices and cloud computing platforms to ensure fast data processing and efficient storage. The stored data includes raw data and data that has been preliminarily processed, such as data after noise removal and standardized data. These data provide a solid foundation for subsequent data analysis and model training.

[0031] Implementation Phase 2: Implementation of Data Cleaning and Preprocessing Modules In this embodiment, the data cleaning and preprocessing module uses a variety of technical means to process the original data. First, the high-frequency noise and random interference in the data are removed through the filtering algorithm; then the data points that are obviously deviated from the normal range are identified and removed as outliers using statistical methods; finally, the missing values ​​are filled by interpolation or machine learning-based methods.

[0032] In addition, data format unification and data standardization were performed during data preprocessing to ensure that data from different sources and types can be analyzed and compared under the same framework. Data standardization used the Z-score standardization method to convert the original data into a standard normal distribution with a mean of 0 and a standard deviation of 1 for subsequent processing and analysis.

[0033] Implementation Phase 3: Implementation of Feature Extraction and Selection Module In this embodiment, the feature extraction and selection module uses multiple methods such as time domain analysis, frequency domain analysis and wavelet transform to extract effective features from the cleaned data. Time domain analysis mainly focuses on the trend and law of the change of parameters such as current and voltage over time; frequency domain analysis converts time domain signals to frequency domain for analysis through methods such as Fourier transform to extract frequency domain features; wavelet transform can simultaneously analyze the time-frequency characteristics of signals at different scales to extract time-frequency features.

[0034] After feature extraction, feature selection is performed using statistical methods (such as chi-square test) and machine learning methods (such as recursive feature elimination) to determine the most representative feature set for subsequent modeling analysis. Feature selection can effectively reduce model complexity and improve prediction accuracy.

[0035] Implementation Phase 4: Implementation of Modeling and Algorithm Design Module In this embodiment, the modeling and algorithm design module uses the support vector machine (SVM) algorithm to establish a prediction model to predict the health status of the closed busbar. The SVM algorithm is a machine learning algorithm based on the principle of structural risk minimization, which has good generalization and nonlinear processing capabilities and is suitable for classification and regression problems of small sample high-dimensional data.

[0036] In order to improve the accuracy and stability of the prediction model, an ensemble learning method is used to combine multiple SVM base learners to form an integrated SVM model. The integrated SVM model improves the overall performance and reduces the risk of overfitting by combining the prediction results of multiple base learners.

[0037] In addition, the possibility of introducing deep learning algorithms (such as convolutional neural networks (CNNs)) to model and analyze complex data is also considered. However, since deep learning algorithms require a large amount of labeled data for training and have high computational complexity, deep learning algorithms are not used for modeling and analysis in this embodiment, but the possibility of introducing deep learning algorithms later is retained.

[0038] Implementation Phase 5: Implementation of the Health Assessment and Monitoring Module In this embodiment, the health status assessment and monitoring module assesses and monitors the health status of the closed busbar by setting thresholds. Reasonable threshold ranges are set for different health indicators based on historical data and expert experience. When the output of the prediction model exceeds these thresholds, it is considered that the equipment may be in an abnormal state, and the system will trigger an early warning mechanism.

[0039] The realization of the early warning mechanism relies on the integration of IoT technology and cloud computing platform. IoT technology ensures that the sensor network can transmit data to the cloud in real time, while the cloud computing platform provides powerful data processing and analysis capabilities to support the rapid response of real-time monitoring and early warning functions. When the system detects an abnormality or potential failure, it will promptly send the alarm information to relevant personnel through preset communication channels (such as SMS, email, APP push, etc.) so that they can take prompt measures to deal with it.

[0040] In addition, the health status assessment and monitoring module also has the function of recording and analyzing historical data. The system will save each monitoring result and warning record to form a complete equipment health file. By analyzing these historical data, we can further explore the laws of equipment failures, optimize the prediction model and warning strategy, and improve the accuracy and reliability of the system.

[0041] Finally: The big data-based intelligent monitoring model for nuclear power closed busbars proposed in the present invention realizes comprehensive, real-time and intelligent monitoring of the operating status of nuclear power closed busbars by integrating multiple links such as data collection, cleaning, feature extraction, modeling prediction and health status assessment. The model has the advantages of high monitoring efficiency, good accuracy, timely fault warning and strong decision support, which can effectively improve the operating safety and reliability of nuclear power plants and has important application value and promotion prospects.

[0042] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0043] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A nuclear power closed bus intelligent monitoring model based on big data, characterized in that: include: Data acquisition and storage module, used to collect real-time operation data of nuclear power closed busbar, including current, voltage, temperature, humidity, vibration and other indicators, and transmit the collected data to the data center for storage; Data cleaning and preprocessing module, used to remove noise, process outliers, fill missing values, unify data formats, standardize or normalize data, etc., to improve data quality; Feature extraction and selection module, used to extract features useful for predicting the health status of the closed busbar from the cleaned and preprocessed data, and perform feature selection to determine the most representative feature set; A modeling and algorithm design module, which is used to establish a prediction model using a machine learning algorithm to predict the health status of the closed busbar. The machine learning algorithm includes but is not limited to a support vector machine, a decision tree, a random forest, a neural network or a deep learning algorithm, and an integrated learning method can be used to improve the accuracy and stability of the prediction model; The health status assessment and monitoring module is used to assess and monitor the health status of the closed busbar based on the output results of the prediction model, including real-time monitoring and early warning functions, and timely issue alarm information when abnormalities or potential faults are detected.

2. The nuclear power closed bus intelligent monitoring model according to claim 1 is characterized in that: The data acquisition and storage module is specifically: Data collection is achieved through a sensor network installed on the closed bus. The sensor network includes various types of sensors such as current sensors, voltage sensors, temperature sensors, humidity sensors, and vibration sensors. The sensor network transmits the collected data to the data center for storage through wireless communication technology; Step S11: Real-time collection of operating data of nuclear power closed busbar, including but not limited to current, voltage, temperature, humidity, vibration and other indicators, marked as {I t , V t , T t , RH t , A t }, where I t Expressed as the current value at time point t, V t Expressed as the voltage value at time point t, T t Expressed as the temperature value at time point t, RH t Expressed as the relative humidity value at time point t, A t Expressed as the vibration level at time point t; Step S12: preprocessing the collected data, including but not limited to filtering, denoising, etc., to improve data quality; Step S13: Packing the preprocessed data into a specific data format; Step S14: Transmitting the packaged data to the data center via wired or wireless communication; Step S15: In the data center, the received data is stored in a network database for subsequent data analysis and processing.

3. The nuclear power closed bus intelligent monitoring model according to claim 1 is characterized in that: The data cleaning and preprocessing module uses a filtering algorithm to remove high-frequency noise and random interference in the data, uses statistical methods to identify and remove outliers, uses interpolation or machine learning-based methods to fill in missing values, and unifies the data format and performs data standardization and normalization. The module includes the following steps: Step S21, high frequency noise and random interference removal: The data cleaning and preprocessing module first uses a filtering algorithm to process the collected raw data to remove high-frequency noise and random interference in the data; Step S22, outlier identification and elimination: Statistical methods are used to automatically identify and remove outliers in the data. By calculating the mean, standard deviation and other statistical quantities of the data, reasonable thresholds are set, and data that exceeds the threshold range is regarded as outliers and removed. Furthermore, algorithms based on distance, density or clustering are used to further identify and process abnormal data points. Step S23, missing value filling: For missing values ​​in the data, this module uses interpolation to fill them; Step S23, missing value filling: For missing values ​​in the data, this module uses a machine learning-based method to fill them; Step S24, data format is unified: After completing the above processing, the module will further unify the data format to ensure that all input data meets the requirements of model processing, including but not limited to data type conversion, timestamp alignment, unit unification and other operations, so as to facilitate subsequent data analysis and processing; Step S25, data standardization and normalization processing: The data is standardized and normalized to eliminate the impact of different dimensions and numerical ranges on model training.

4. The nuclear power closed bus intelligent monitoring model according to claim 1 is characterized in that: The feature extraction and selection module extracts features by using methods such as time domain analysis, frequency domain analysis, wavelet transform, principal component analysis, etc., and performs feature selection by using statistical methods, machine learning methods, or deep learning methods, specifically: Step S31, feature extraction: The module first extracts features from the data using a variety of methods, including but not limited to: Method S311, time domain analysis method: using the statistical characteristics of time series data in the time domain, such as mean, variance, skewness, kurtosis, etc., as well as the difference and trend term of the time series as features; Method S312, frequency domain analysis method: converting the time domain signal to the frequency domain through Fourier transform (FT) or fast Fourier transform (FFT) to extract the spectrum features; Method S313, Wavelet Transform: Apply wavelet transform techniques, such as continuous wavelet transform (CWT) or discrete wavelet transform (DWT), to analyze signals at multiple scales and extract wavelet coefficients as features to capture local characteristics and detail information of the signal; Method S314, principal component analysis (PCA): dimensionality reduction is performed on the original feature set, and multiple variables are converted into a few principal components through linear transformation. These principal components retain most of the information of the original data while reducing the redundancy and noise of the data; Step S32, feature selection: The extracted features are further screened to remove redundant features. The module uses one or more of the following methods for feature selection: Method S321, statistical-based method: using statistical tests to evaluate the correlation or importance between features and target variables, and selecting features with higher significance; Method S322, a machine learning-based method: using the built-in feature importance evaluation mechanism of the machine learning algorithm to select the features that contribute most to the model prediction performance; Method S323, a method based on deep learning: In a deep learning model, the most representative features are automatically learned and extracted through a training process, or feature selection is indirectly achieved through dropout and L1 / L2 regularization techniques.

5. The nuclear power closed bus intelligent monitoring model according to claim 1 is characterized in that: The prediction model in the modeling and algorithm design module adopts a support vector machine algorithm or a deep learning algorithm. The support vector machine algorithm also needs to combine multiple base learners through an integrated learning method to improve the overall performance, specifically: Method S41: When the support vector machine algorithm is used, the algorithm separates data points of different categories by finding an optimal hyperplane, and its objective function is expressed as: in, is the i-th feature, m is the number of features, express For the target value, w is the normal vector of the hyperplane, b is the intercept, C is the regularization parameter, ξ i is the slack variable, Indicates the limiting condition; and, in order to improve the overall performance, multiple different support vector machine base learners can be trained, each base learner uses a different kernel function, regularization parameter or data subset; the prediction results of all base learners are aggregated, and the aggregation method can be simple average, weighted average, voting method or more complex strategies such as Stacking; the aggregated results are used as the final prediction output to monitor the operating status of the nuclear power closed bus system; Method S42: When a deep learning algorithm is used, the algorithm uses a multi-layer structure of a neural network to automatically learn an abstract representation of data, including but not limited to convolutional neural networks (CNN), recurrent neural networks (RNN), Transformer and their variants; the deep learning model optimizes network parameters through a back-propagation algorithm to minimize prediction errors.

6. The nuclear power closed bus intelligent monitoring model according to claim 1 is characterized in that: The health status assessment and monitoring module assesses the health status of the closed busbar by setting a threshold or a probability-based method, monitors the equipment status in real time, and issues an alarm message in time when an abnormality or potential failure is detected. The alarm message is sent to relevant personnel through a preset communication channel, as follows: Method S51: Setting a threshold evaluation method, setting a preset threshold range for each key operating parameter, when the real-time data exceeds the range, it is determined to be abnormal or a potential fault, and an alarm mechanism is triggered; Method S52: Probability-based evaluation method, using probability models, such as Bayesian network, hidden Markov model, etc., combined with historical data and real-time data, to calculate the probability of the closed bus system being in different health states, when the probability of an unhealthy state exceeds a preset threshold, an alarm is issued; the specific form of the probability model can be expressed as a part of the Bayesian network, and its conditional probability calculation formula is as follows: Where H represents a certain health state of the closed bus system, D represents the observed data (i.e., the real-time operating parameters after preprocessing), P(H|D) is the posterior probability that the system is in state H given the observed data D, P(D|H) is the likelihood probability of observing data D given the system is in state H, P(H) is the prior probability that the system is in state H, and P(D) is the marginal probability of observing data D, which is usually used as a normalization factor; The alarm information is sent to relevant personnel through a preset communication channel, including but not limited to a wireless network, a wired network, a text message service, an email or a dedicated monitoring system interface, and the alarm information contains a specific description of the abnormality or potential fault, the time of occurrence, location information and recommended response measures.

Images