Non-intrusive State Intelligent Sensing and Fault Warning System for Circuit Breaker Operating Mechanism

Abstract

The present invention discloses a non-intrusive state intelligent perception and fault warning system for a circuit breaker operating mechanism, which relates to the technical field of power equipment monitoring. In order to solve the problems that in the process of data processing, there are still poor stability and low transmission speed, which cannot meet the requirements of real-time response, and the complexity and cost of the system are high due to multiple levels and various technologies, increasing the difficulty of implementation and maintenance; the non-intrusive state intelligent perception and fault warning system for a circuit breaker operating mechanism includes a data acquisition unit, a data processing and analysis unit, a fault warning and diagnosis unit, and a visualization interaction unit; through the collaborative work of the cloud, the edge side, and the device side, the real-time processing, analysis, and AI decision-making of data are realized. The application of edge computing reduces the burden on the cloud and improves the response speed, while the centralized management of the cloud ensures the security and consistency of data, timely discovers faults or abnormalities, and improves the security and reliability of the power system.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment monitoring, and particularly to a non-intrusive state intelligent perception and fault warning system for a circuit breaker operating mechanism. Background Art

[0002] As an important device in the power system, the operating state of the operating mechanism of a circuit breaker directly affects the safety and stability of the entire power system. Regarding the intelligent perception and fault warning of circuit breakers, a Chinese patent with the publication number CN113904443B discloses a multi-dimensional space visualization on-site substation equipment monitoring and warning system, including equipment and the surrounding environment. The equipment includes transformers, circuit breakers (GIS), and lightning arresters. The system is divided into four parts, namely, the perception layer, the network layer, the platform layer, and the application layer. The comprehensive perception, data preprocessing, and simple state evaluation of the Internet of Things of substation equipment are realized by using different types of intelligent sensors and nodes; the high-speed and reliable transmission of the sensor data is realized by using the equipment and network in the network layer; the configuration and distribution of data and equipment management, data storage, and edge computing are realized by using the equipment in the platform layer; finally, the application layer realizes the advanced analysis and application of the collected data, realizes information sharing, and performs state evaluation and warning according to the collected information, and provides auxiliary decision-making.

[0003] Although the above patent uses a non-intrusive method for perception and monitoring, in the process of data processing, there are still problems of poor stability and low transmission speed, which cannot meet the requirements of real-time response. Multiple levels and various technologies lead to high complexity and cost of the system, increasing the difficulty of implementation and maintenance. Summary of the Invention

[0004] The purpose of the present invention is to provide a non-intrusive state intelligent perception and fault warning system for a circuit breaker operating mechanism. Through the collaborative work of the cloud side, the edge side, and the device side, real-time processing, analysis, and AI decision-making of data are realized. The application of edge computing reduces the burden on the cloud side and improves the response speed, while the centralized management of the cloud side ensures the security and consistency of data, so as to solve the problems raised in the above background art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions:

[0006] A non-intrusive state intelligent perception and fault warning system for a circuit breaker operating mechanism, including:

[0007] A data acquisition unit, used to install at least one sensor on the circuit breaker operating mechanism, collect the original operation data of the circuit breaker operating mechanism in real time, and transmit the collected original operation data to the cloud platform master station;

[0008] A data processing and analysis unit, which is used to preprocess the received original operation data, extract features from the preprocessed data, identify key information related to the operating mechanism status and potential faults, and determine the feature vector.

[0009] A fault warning and diagnosis unit, which is used to input the feature vector into a trained fault diagnosis model for discrimination, evaluate the operating status of the circuit breaker operating mechanism, judge whether there are faults or abnormalities. If the model judges that there are faults or abnormalities, the model outputs the corresponding diagnosis results. At the same time, it triggers an early warning mechanism to generate fault warning information.

[0010] A visualization interaction unit, which is used to display the fault warning information, diagnosis results and real-time status data to the user through a visualization interface. The user can view historical data, set parameters, receive warning notifications and interact based on the display interface.

[0011] Further, the data acquisition unit includes a cloud end, an edge end and a device end, where:

[0012] The cloud end runs on the cloud platform main station, which is used to manage the terminal devices of the edge nodes, build a data interaction channel, and send the application to the edge end for calculation and processing based on the data interaction channel.

[0013] At the same time, it is also used to store the data uploaded by the device end, and process, analyze and perform AI training on the uploaded data.

[0014] The edge end runs on the terminal device, which is used to connect to the cloud platform main station through the cloud end, manage the edge computing applications sent by the cloud end, collect device data in real time and perform edge caching, edge data preprocessing and analysis, and perform AI decision-making.

[0015] The device end is used to dock with the edge end through the device protocol, report the collected device data to the edge end for caching and edge processing, and at the same time respond to the edge-side linkage events of the edge nodes.

[0016] Further, after the data interaction channel is constructed, it includes:

[0017] Extract the data interaction frequency corresponding to the unit time of the data interaction channel and the amount of data that needs to be interacted within the unit time.

[0018] Extract the benchmark channel capacity of the preset data interaction channel.

[0019] Obtain a capacity adjustment coefficient according to the data interaction frequency and the amount of data that needs to be interacted within the unit time; where, the capacity adjustment coefficient is obtained through the following formula:

[0020]

[0021] Among them, K represents the capacity adjustment coefficient; n represents the number of unit times corresponding to the historical data interaction experienced by the data interaction channel; f i represents the data interaction frequency corresponding to the i-th unit time; C i represents the data volume of the data interaction corresponding to the i-th unit time; C z represents the reference channel capacity of the data interaction channel; C min represents the data volume corresponding to the minimum data interaction frequency per unit time; C max represents the data volume corresponding to the maximum data interaction frequency per unit time;

[0022] Obtain the regional space capacity corresponding to the buffer of the data interaction channel according to the capacity adjustment coefficient in combination with the reference channel capacity of the data interaction channel; among them, the regional space capacity is obtained through the following formula:

[0023]

[0024] Among them, C h represents the regional space capacity corresponding to the buffer; K represents the capacity adjustment coefficient; C z represents the reference channel capacity of the data interaction channel;

[0025] Construct the buffer corresponding to the data interaction channel according to the regional space capacity.

[0026] Further, the data acquisition unit further includes:

[0027] A performance optimization module, configured to determine the performance loss weight of the edge node, and adjust the deployment and data processing strategy of the edge computing application according to the performance loss weight;

[0028] A task scheduling module, configured to construct a task scheduling strategy based on the performance loss weight of the edge node and the obtained data processing strategy, and perform task allocation based on the task scheduling strategy.

[0029] Further, the determination of the performance loss weight is specifically:

[0030] Monitor the resource consumption data of the edge node during data acquisition, preprocessing, and analysis;

[0031] Calculate the performance loss weight of the edge node based on the monitoring data, and reflect the performance loss degree of the edge node when processing different tasks in real time, where the performance loss weight expression is:

[0032]

[0033] In the formula: W(U i) represents the performance loss weight of the $i$-th edge node; $M(U i ) represents the memory capacity of the $i$-th edge node; $V m (U i ) represents the memory utilization rate of the $i$-th edge node; $C(U i ) represents the CPU processing rate of the $i$-th edge node; $V c (U i ) represents the CPU utilization rate of the $i$-th edge node; $Cnt(U i ) represents the number of CPUs of the $i$-th edge node; $N(U i ) represents the network bandwidth of the $i$-th edge node; $V n (U i ) represents the network bandwidth utilization rate of the $i$-th edge node; $t$ represents the time coefficient;

[0034] Dynamically adjust the deployment strategy of the edge computing application according to the performance loss weight, and allocate the computing tasks to the edge nodes with less performance loss.

[0035] Furthermore, task allocation is performed based on the task scheduling strategy, which specifically includes:

[0036] Determine multiple task sets, allocate corresponding weight sets to each task, and determine the task weights of all tasks. The task weight formula is:

[0037]

[0038] When there is a new connection request, select a server for task scheduling based on the result of the task scheduling strategy judgment. Among them, the task scheduling strategy judgment condition is:

[0039]

[0040] In the formula, $B ij represents the number of the $j$-th type of task processed on the edge node $U i ; $M j represents the weight of the $j$-th type of task; $U m represents the edge server node to be selected;

[0041] Among them, when the ratio of the sum of the weights of all tasks on the edge node $U i to its performance loss weight is less than the corresponding ratio on the edge node $U m , select $U m as the edge node for processing new tasks.

[0042] Furthermore, the data processing and analysis unit includes:

[0043] A data preprocessing module, which is used to clean the acquired raw data, remove outliers and missing values, and apply wavelet denoising method to denoise the vibration signals in the raw data;

[0044] A feature extraction module, which is used to perform time-domain analysis on the preprocessed data, extract global features, calculate the frequency-domain features of the vibration signals, and combine the time-domain and frequency-domain features to form a feature matrix;

[0045] A feature dimensionality reduction module, which is used to normalize the feature matrix, calculate the covariance matrix of the feature matrix, obtain the auto-covariance coefficient and cross-covariance coefficient, perform dimensionality reduction processing on the feature matrix, and extract the main feature vectors.

[0046] Further, a fault warning and diagnosis unit includes:

[0047] A model training module, which is used to construct a corresponding OCELM model based on the data categories corresponding to the feature vectors, extract historical normal state data as input data to train the OCELM model, and determine the decision boundary and decision threshold of the OCELM model;

[0048] An anomaly detection module, which is used to input the main feature vectors extracted from the newly acquired real-time data into the trained OCELM model, and judge whether the device is in a normal state according to the decision boundary and decision threshold of the OCELM model;

[0049] A warning module, which is used to trigger a warning mechanism based on the device anomaly judgment result, divide different warning levels according to the degree and type of the anomaly, including primary warning, intermediate warning and high-level warning, and different levels of warnings correspond to different response measures and emergency levels.

[0050] Further, the model training module also includes classifying newly acquired input samples, specifically:

[0051] Input the newly acquired input samples into not less than one trained OCELM model, where each OCELM model outputs a prediction value;

[0052] Compare the output values of all OCELM models, select the smallest prediction value, and determine the final classification result of the newly acquired input samples based on the category corresponding to the smallest prediction value.

[0053] Further, a visualization interaction unit includes:

[0054] A warning information display module, which is used to display the fault warning information in real time in the visualization interface;

[0055] A historical data analysis module, which is used to provide a visualization analysis function of historical data, allowing users to view historical warning records and information on fault trends;

[0056] An interactive feedback module for providing a parameter setting interface and an interactive interface, adjusting and optimizing the parameters of the OCELM model based on the parameter setting interface, and providing feedback and suggestions based on the interactive interface.

[0057] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0058] By installing a variety of sensors to collect the operation data of the circuit breaker operating mechanism in real time, through the collaborative work of the cloud, edge, and device sides, real-time data processing, analysis, and AI decision-making are achieved. The application of edge computing reduces the burden on the cloud and improves the response speed, while the centralized management of the cloud ensures the security and consistency of the data. The feature extraction operation identifies the key information related to the state and potential faults of the operating mechanism, and uses the trained fault diagnosis model to evaluate the operating state of the operating mechanism, enabling timely detection of faults or anomalies and notifying users through the early warning mechanism, which helps prevent potential safety risks, reduce losses caused by faults, and provides users with a comprehensive, efficient, and intelligent fault early warning solution, contributing to improving the security and reliability of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a module diagram of the non-intrusive state intelligent perception and fault early warning system for the circuit breaker operating mechanism of the present invention;

[0060] Figure 2 It is a schematic diagram of the cloud-edge architecture of the data acquisition unit of the present invention;

[0061] Figure 3 It is a flowchart of the edge-side operation of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0062] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0063] To solve the technical problems that in the data processing process, there are still situations of poor stability and low transmission speed, which cannot meet the requirements of real-time response, and the complexity and cost of the system are high due to multiple levels and various technologies, increasing the difficulty of implementation and maintenance, please refer to Figures 1-3 This embodiment provides the following technical solutions:

[0064] A non-intrusive state intelligent perception and fault early warning system for a circuit breaker operating mechanism, comprising:

[0065] The data acquisition unit is used to install at least one sensor on the breaker operating mechanism, collect the original operation data of the breaker operating mechanism in real time, and transmit the collected original operation data to the cloud platform master station. Among them, the sensors include temperature sensors, humidity sensors, vibration sensors, current and voltage sensors, etc.;

[0066] In this embodiment, the data acquisition unit includes a cloud end, an edge end, and a device end, where:

[0067] The cloud end runs on the cloud platform master station, is used to manage the terminal devices of the edge nodes, and build a data interaction channel. Based on the data interaction channel, the application is sent to the edge end for computing and processing;

[0068] In this embodiment, the application is various functions or programs developed based on data analysis and business requirements, and is sent to the edge end through the data interaction channel for computing and processing; it includes data analysis algorithms, business logics, configurations and parameters, model updates, software updates, as well as instructions and control signals, etc., to achieve distributed computing, real-time response, and optimized data processing;

[0069] At the same time, it is also used to store the data uploaded by the device end, and perform processing analysis and AI training on the uploaded data;

[0070] The edge end runs on the terminal device, is used to connect to the cloud platform master station through the cloud end, manage the edge computing applications sent by the cloud end, collect device data in real time and perform edge caching, edge data preprocessing and analysis, and perform AI decision-making;

[0071] The device end is used to dock with the edge end through the device protocol, report the collected device data to the edge end for caching and edge processing, and at the same time respond to the edge-side linkage events of the edge nodes to meet the requirements of real-time response;

[0072] The data processing and analysis unit is used to preprocess the received original operation data, including cleaning, removing noise data such as outliers and missing values, applying methods such as wavelet denoising to denoise the vibration signal, improving the data quality, extracting features from the preprocessed data, identifying key information related to the operating mechanism state and potential faults of the breaker, and determining the feature vector;

[0073] The fault warning and diagnosis unit is used to input the feature vector into the trained fault diagnosis model for discrimination, evaluate the operating state of the breaker operating mechanism, judge whether there are faults or abnormalities. If the model judges that there are faults or abnormalities, the model outputs the corresponding diagnosis result. At the same time, it triggers the warning mechanism to generate fault warning information;

[0074] A visual interaction unit is used to display fault warning information, diagnostic results, and real-time status data to users through a visual interface. Users can view historical data, set parameters, receive warning notifications, and interact based on the display interface.

[0075] Among them, the real-time status data is obtained based on the data acquisition unit, reflecting the operating status of the circuit breaker operating mechanism at the current moment, including parameters such as temperature, pressure, current, voltage, position, and speed.

[0076] Specifically, after the data interaction channel is constructed, it includes:

[0077] Extract the data interaction frequency corresponding to the unit time of the data interaction channel and the amount of data that needs to be interacted within the unit time.

[0078] Extract the benchmark channel capacity of the data interaction channel set in advance.

[0079] Obtain a capacity adjustment coefficient according to the data interaction frequency and the amount of data that needs to be interacted within the unit time. Among them, the capacity adjustment coefficient is obtained through the following formula:

[0080]

[0081] Among them, K represents the capacity adjustment coefficient; n represents the number of unit times corresponding to the historical data interaction experienced by the data interaction channel; f i represents the data interaction frequency corresponding to the i-th unit time; C i represents the amount of data interacted corresponding to the i-th unit time; C z represents the benchmark channel capacity of the data interaction channel; C min represents the amount of data interacted corresponding to the minimum data interaction frequency within the unit time; C max represents the amount of data interacted corresponding to the maximum data interaction frequency within the unit time;

[0082] Obtain the regional space capacity corresponding to the buffer of the data interaction channel according to the capacity adjustment coefficient combined with the benchmark channel capacity of the data interaction channel. Among them, the regional space capacity is obtained through the following formula:

[0083]

[0084] Among them, C h represents the regional space capacity corresponding to the buffer; K represents the capacity adjustment coefficient; C z represents the benchmark channel capacity of the data interaction channel;

[0085] Construct the buffer corresponding to the data interaction channel according to the regional space capacity.

[0086] The technical effects of the above technical solution are as follows: By extracting the data interaction frequency and data volume per unit time of the data interaction channel and using these data to calculate the capacity adjustment coefficient, this technical solution can dynamically adjust the size of the buffer according to the real-time data interaction requirements. This dynamic adjustment enables the data interaction channel to better cope with the data interaction pressure in different time periods or different scenarios, improving the adaptability and stability of the system.

[0087] By calculating the capacity adjustment coefficient and setting the regional space capacity of the buffer accordingly, this technical solution can allocate system resources more reasonably. When the data interaction demand is low, the buffer capacity can be reduced to save resources; when the data interaction demand is high, the buffer capacity is increased to ensure the smooth progress of data interaction. This optimization of resource allocation can improve the overall performance and efficiency of the system.

[0088] Dynamically adjusting the buffer capacity according to the data interaction frequency and data volume can ensure that the buffer has enough space to accommodate and process data during the peak period of data interaction, thus avoiding data loss or delay problems caused by buffer overflow or untimely processing. At the same time, reducing the buffer capacity during the low period of data interaction can reduce the system overhead and delay, further improving the efficiency of data interaction.

[0089] By constructing a reasonable buffer and dynamically adjusting its capacity, this technical solution can enhance the stability of the system. When facing sudden data interaction pressure, the system can maintain the continuity and stability of data interaction through the adjustment function of the buffer, thus avoiding system crashes or performance degradation problems caused by poor data interaction.

[0090] In summary, the above technical solution optimizes system resource allocation and improves data interaction efficiency by dynamically adjusting the buffer capacity of the data interaction channel, thereby enhancing the adaptability and stability of the system.

[0091] In this embodiment, the visual interaction unit includes:

[0092] An early warning information display module for real-time displaying fault early warning information in the visual interface, including the early warning time, early warning level, possible fault reasons, etc.;

[0093] A historical data analysis module for providing a visual analysis function of historical data, allowing users to view historical early warning records and information on fault trends;

[0094] An interaction feedback module for providing a parameter setting interface and an interaction interface, adjusting and optimizing the parameters of the OCELM model based on the parameter setting interface to adapt to different application scenarios and requirements, and providing feedback and suggestions based on the interaction interface to help improve and optimize the performance and accuracy of the fault early warning system.

[0095] In this embodiment, by installing a variety of sensors to collect the operation data of the circuit breaker operating mechanism in real time, the collected raw data is preprocessed to remove noise and outliers, improving the data quality. Further, the feature extraction operation identifies key information related to the state of the operating mechanism and potential faults. Through the collaborative work of the cloud, edge, and device sides, real-time data processing, analysis, and AI decision-making are achieved. The application of edge computing reduces the burden on the cloud and improves the response speed, while the centralized management of the cloud ensures the security and consistency of the data. Using the trained fault diagnosis model to evaluate the operating state of the operating mechanism can timely detect faults or anomalies and notify the user through the early warning mechanism, which helps to prevent potential safety risks and reduce losses caused by faults. The visualization interface enables the user to intuitively view historical data, set parameters, receive early warning notifications, and interact, providing the user with a comprehensive, efficient, and intelligent fault early warning solution, which helps to improve the security and reliability of the power system.

[0096] In this embodiment, the data acquisition unit further includes:

[0097] A performance optimization module for determining the performance loss weight of the edge node and adjusting the deployment and data processing strategies of the edge computing application according to the performance loss weight, including adjusting the data acquisition frequency, selecting appropriate denoising algorithms and feature extraction methods, to reduce the system power consumption while ensuring the data quality, so as to optimize the system performance and reduce the power consumption;

[0098] In this embodiment, the determination of the performance loss weight is specifically as follows:

[0099] Monitor the resource consumption data of the edge node during data acquisition, preprocessing, and analysis, including CPU usage, memory occupancy, network bandwidth, etc.;

[0100] Calculate the performance loss weight of the edge node based on the monitoring data and reflect the performance loss degree of the edge node when processing different tasks in real time;

[0101] According to the performance loss weight, dynamically adjust the deployment strategy of the edge computing application, and allocate the computing tasks to the edge nodes with less performance loss to optimize the system performance;

[0102] Among them, the performance loss weight expression is:

[0103] W(U i ) = k1 * M(U i ) * V m (U i ) * t + k2 * C(U i ) * V c (Ui )

[0104] *Cnt(U i )*t + k3*N(U i )*V n (U i )*t

[0105] Where: k1 is the memory utilization rate V m (U i ) For W(U i ) The weight coefficient; k2 is the CPU utilization rate V c (U i ) For W(U i ) The weight coefficient; k3 is the network bandwidth utilization rate V n (U i ) For W(U i ) The weight coefficient, and k1 + k2 + k3 = 1, V m (U i ) ∈ (0, 1), V c (U i ) ∈ (0,

[0106] Let k1 = 1 / 3, k2 = 1 / 3, k3 = 1 / 3, then the complete performance loss weight expression is:

[0107]

[0108] Where: W(U i ) Represents the performance loss weight of the i-th edge node; M(U i ) Represents the memory capacity of the i-th edge node; V m (U i ) Represents the memory utilization rate of the i-th edge node; C(U i ) Represents the CPU processing rate of the i-th edge node; V c (U i ) Represents the CPU utilization rate of the i-th edge node; Cnt(U i ) Represents the number of CPUs of the i-th edge node; N(U i ) Represents the network bandwidth of the i-th edge node; V n (U i ) Represents the network bandwidth utilization rate of the i-th edge node; t represents the time coefficient.

[0109] The task scheduling module is used to construct a task scheduling strategy based on the performance loss weights of edge nodes and the obtained data processing strategy, and perform task allocation based on the task scheduling strategy, including real-time monitoring of the operating status of the circuit breaker operating mechanism, collecting and analyzing relevant data, predicting possible faults, etc.;

[0110] In this embodiment, the performance optimization module calculates the performance loss weights by monitoring the resource consumption data of edge nodes, and adjusts the deployment and data processing strategy of edge computing applications based on these weights. While ensuring data quality, it reduces system power consumption and optimizes the system. By adjusting the data acquisition frequency, selecting appropriate denoising algorithms and feature extraction methods, effective utilization of resources can be achieved, the service life of the device can be extended, and energy waste can be reduced at the same time. By real-time monitoring the operating status of the circuit breaker operating mechanism, collecting and analyzing relevant data, predicting possible faults, and performing task allocation according to the task scheduling strategy, it can ensure that computing tasks are assigned to edge nodes with less performance loss, achieve load balancing, and improve the overall performance and response speed of the system.

[0111] In this embodiment, performing task allocation based on the task scheduling strategy specifically includes:

[0112] Determine multiple task sets \(R = \{R_1, R_2, \ldots, R_t\}\), where \(t\) is the total number of task types, and assign a corresponding weight set \(M = \{M_1, M_2, \ldots, M_t\}\) to each task, and determine the task weights of all tasks. The task weight formula is:

[0113]

[0114] When there is a new connection request, select a server for task scheduling based on the judgment result of the task scheduling strategy. Among them, the task scheduling strategy judgment condition is:

[0115]

[0116] In the formula, \(B\) ij represents the number of the \(j\)-th type of task processed on the edge node \(U\) i ; \(M\) j represents the weight of the \(j\)-th type of task; \(U\) m represents the edge server node to be selected;

[0117] Among them, when the ratio of the sum of the weights of all tasks on the edge node \(U\) i to its performance loss weight is less than the corresponding ratio on the edge node \(U\) m , select \(U\) m as the edge node for processing new tasks.

[0118] In this embodiment, through the combination of performance optimization and task scheduling, various complex scenarios and emergencies can be better handled. When the task volume is large or the system load is high, the task scheduling module can timely adjust the task allocation strategy to ensure that critical tasks are given priority. At the same time, the performance optimization module can reduce the system power consumption and reduce equipment failures caused by overheating and other reasons, thereby improving the reliability and stability of the system. Significant improvements have been made in aspects such as data acquisition, processing, analysis, and application, improving the overall performance and reliability of the system, reducing power consumption and costs, and providing a strong guarantee for the safe and stable operation of the power system.

[0119] In this embodiment, the data processing and analysis unit includes:

[0120] The data preprocessing module is used to clean the acquired raw data, remove outliers and missing values, and apply the wavelet denoising method to denoise the vibration signals in the raw data to ensure data quality;

[0121] The feature extraction module is used to perform time-domain analysis on the preprocessed data, extract global features such as skewness coefficient, kurtosis coefficient, and peak coefficient, calculate the frequency-domain features of the vibration signals, and combine the time-domain and frequency-domain features to form a feature matrix;

[0122] The feature dimensionality reduction module is used to perform normalization processing on the feature matrix to eliminate the dimensional differences between different features, calculate the covariance matrix of the feature matrix, obtain the auto-covariance coefficient and cross-covariance coefficient, and use the PCA method to perform dimensionality reduction processing on the feature matrix to extract the main eigenvectors and reduce the computational complexity.

[0123] In this embodiment, the fault warning and diagnosis unit includes:

[0124] The model training module is used to construct a corresponding OCELM model based on the data categories corresponding to the eigenvectors, extract historical normal state data as input data to train the OCELM model, and determine the decision boundary and decision threshold of the OCELM model;

[0125] The model training module also includes classifying newly acquired input samples. Specifically:

[0126] Input the newly acquired input samples into no less than one trained OCELM model. Among them, each OCELM model outputs a prediction value, indicating the distance of the sample from the decision boundary of this category;

[0127] Compare the output values of all OCELM models, select the smallest prediction value, and determine the final classification result of the newly acquired input sample based on the category corresponding to the smallest prediction value;

[0128] Anomaly detection module, which is used to input the main feature vectors extracted from newly acquired real-time data into the trained OCELM model, and determine whether the device is in a normal state according to the decision boundary and decision threshold of the OCELM model;

[0129] Early warning module, which is used to trigger an early warning mechanism based on the device anomaly judgment result, divide different early warning levels according to the degree and type of anomaly, including primary early warning, intermediate early warning and advanced early warning. Different levels of early warning correspond to different response measures and emergency levels. The hierarchical early warning mechanism can flexibly adjust the response measures and emergency levels according to the degree and type of anomaly, which helps to detect and handle potential faults in time and avoid the expansion of equipment failures.

[0130] In this embodiment, through time-domain and frequency-domain analysis of the preprocessed data, rich global features are extracted. The feature dimension reduction module uses the PCA method to perform dimension reduction processing on the feature matrix, which not only retains the main feature information but also reduces the computational complexity and improves the efficiency of subsequent model training and classification. The data-driven modeling method can make full use of the information in historical data to improve the accuracy of fault early warning and diagnosis. At the same time, by comparing the output values of multiple OCELM models, the category of the input sample can be judged more accurately, further enhancing the reliability of early warning and diagnosis. Through the collaborative work of the data processing and analysis unit and the fault early warning and diagnosis unit, comprehensive optimization is achieved in aspects such as data acquisition, processing, analysis and early warning, which can not only improve the safety and reliability of the power system but also reduce the operation and maintenance costs and improve the overall performance of the system.

[0131] The above is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and all should be covered by the protection scope of the present invention.

Claims

1. The non-intrusive intelligent state perception and fault warning system for circuit breaker operating mechanism is characterized by: include: A data acquisition unit, which is used to install at least one sensor on the circuit breaker operating mechanism, collect the operation data of the circuit breaker operating mechanism in real time, and transmit the collected raw data to the cloud platform master station; A data processing and analysis unit is used to pre-process the received raw data, extract features from the pre-processed data, identify key information related to the state of the operating mechanism and potential faults, and determine feature vectors; The fault warning and diagnosis unit is used to input the feature vector into the trained fault diagnosis model for judgment, evaluate the operating status of the circuit breaker operating mechanism, and judge whether there is a fault or abnormality. If the model judges that there is a fault or abnormality, the warning mechanism is triggered to generate fault warning information; Visualization interaction unit, used to display fault warning information, diagnosis results and real-time status data to users through a visualization interface. Users can view historical data, set parameters, receive warning notifications and interact based on the display interface. The data collection unit includes the cloud, edge, and device ends, where: The cloud runs on the cloud platform master station and is used to manage the terminal devices of the edge nodes and build data interaction channels. Based on the data interaction channels, applications are sent to the edge for computing and processing. At the same time, it is also used to store data uploaded by the device, and to process, analyze and train AI on the uploaded data; The edge runs on the terminal device and is used to connect to the cloud platform master station through the cloud connection, manage the edge computing applications issued by the cloud, collect device data in real time and perform edge caching, edge data preprocessing and analysis, and make AI decisions; The device side is used to connect to the edge side through the device protocol, report the collected device data to the edge side for caching and edge processing, and respond to the edge node's edge linkage events; After completing the construction of the data interaction channel, it includes: Extract the data interaction frequency corresponding to the unit time of the data interaction channel and the amount of data that needs to be interacted with per unit time; Extracting a preset reference channel capacity of the data interaction channel; The capacity adjustment coefficient is obtained according to the data interaction frequency and the amount of data that needs to be interacted with per unit time; wherein the capacity adjustment coefficient is obtained by the following formula: ； Where K represents the capacity adjustment coefficient; n represents the number of unit times corresponding to the historical data interactions experienced by the data interaction channel; fi represents the data interaction frequency corresponding to the i-th unit time; C i represents the amount of data interaction corresponding to the i-th unit time; C z Indicates the benchmark channel capacity of the data exchange channel; C min Indicates the data interaction volume corresponding to the minimum value of data interaction frequency per unit time; C max Indicates the data interaction volume corresponding to the maximum data interaction frequency per unit time; The regional space capacity corresponding to the buffer of the data interaction channel is obtained according to the capacity adjustment coefficient and the reference channel capacity of the data interaction channel; wherein the regional space capacity is obtained by the following formula: ； Among them, C h represents the regional space capacity corresponding to the buffer zone; K represents the capacity adjustment coefficient; C z Indicates the benchmark channel capacity of the data exchange channel; A buffer corresponding to the data interaction channel is constructed according to the regional space capacity.

2. The non-intrusive intelligent state perception and fault warning system for circuit breaker operating mechanism according to claim 1, characterized in that: The data acquisition unit also includes: The performance optimization module is used to determine the performance loss weight of the edge node and adjust the deployment and data processing strategy of the edge computing application according to the performance loss weight; The task scheduling module is used to build a task scheduling strategy based on the performance loss weight of the edge node and the acquired data processing strategy, and to allocate tasks based on the task scheduling strategy.

3. The non-intrusive intelligent state perception and fault warning system for circuit breaker operating mechanism according to claim 2, characterized in that: The performance loss weight is determined as follows: Monitor resource consumption data of edge nodes during data collection, preprocessing and analysis; The performance loss weight of the edge node is calculated based on the monitoring data, and the performance loss degree of the edge node when processing different tasks is reflected in real time, wherein the performance loss weight expression is: ； Where: It is expressed as the performance loss weight of the i-th edge node; It is represented as the memory capacity of the i-th edge node; Expressed as the memory utilization of the i-th edge node; It is represented as the CPU processing rate of the i-th edge node; It is expressed as the CPU utilization of the i-th edge node; It is represented by the number of CPUs of the ith edge node; It is represented as the network bandwidth of the i-th edge node; It is expressed as the network bandwidth utilization of the i-th edge node; Expressed as time coefficient; According to the performance loss weight, the deployment strategy of the edge computing application is dynamically adjusted to allocate computing tasks to edge nodes with lower performance loss.

4. The non-intrusive intelligent state perception and fault warning system for circuit breaker operating mechanism according to claim 3, characterized in that: Tasks are allocated based on the task scheduling strategy, including: Determine multiple task sets and assign corresponding weight sets to each task, and determine the task weights of all tasks. The task weight formula is: ； When there is a new connection request, a server is selected for task scheduling based on the task scheduling strategy judgment result, where the task scheduling strategy judgment condition is: ； In the formula, Represented as an edge node The number of tasks of the jth type processed; Represented as the weight of the j-th task; Represented as an edge server node to be selected; Among them, when the edge node The ratio of the sum of the weights of all tasks on the edge node to their performance loss weight is less than When the corresponding ratio is Acts as an edge node to process new tasks.

5. The non-intrusive intelligent state perception and fault warning system for circuit breaker operating mechanism according to claim 4, characterized in that: Data processing and analysis unit, including: The data preprocessing module is used to clean the acquired raw data, remove abnormal values ​​and missing values, and apply the wavelet denoising method to denoise the vibration signal in the raw data; The feature extraction module is used to perform time domain analysis on the preprocessed data, extract global features, calculate the frequency domain features of the vibration signal, and combine the time domain and frequency domain features to form a feature matrix; The feature dimension reduction module is used to normalize the feature matrix, calculate the covariance matrix of the feature matrix, obtain the autocovariance coefficient and the cross-covariance coefficient, reduce the dimension of the feature matrix, and extract the main eigenvectors.

6. The non-intrusive intelligent state sensing and fault warning system for circuit breaker operating mechanism according to claim 5, characterized in that: Fault warning and diagnosis unit, including: A model training module is used to build a corresponding OCELM model based on the data category corresponding to the feature vector, extract historical normal state data as input data to train the OCELM model, and determine the decision boundary and decision threshold of the OCELM model; The anomaly detection module is used to input the main feature vectors extracted from the newly acquired real-time data into the trained OCELM model, and determine whether the device is in a normal state according to the decision boundary and decision threshold of the OCELM model; The early warning module is used to trigger the early warning mechanism based on the results of equipment abnormality judgment. Different warning levels are divided according to the degree and type of abnormality, including primary warning, intermediate warning and advanced warning. Different levels of warnings correspond to different response measures and urgency.

7. The non-intrusive intelligent state sensing and fault warning system for circuit breaker operating mechanism according to claim 6, characterized in that: The model training module also includes classifying the newly acquired input samples, specifically: Input the newly acquired input sample into at least one trained OCELM model, where each OCELM model outputs a prediction value; Compare the output values ​​of all OCELM models, select the smallest predicted value, and determine the final classification result of the newly acquired input sample based on the category corresponding to the smallest predicted value.

8. The non-intrusive intelligent state perception and fault warning system for circuit breaker operating mechanism according to claim 7, characterized in that: Visual interaction unit, including: The warning information display module is used to display fault warning information in real time in the visual interface; The historical data analysis module is used to provide visual analysis of historical data, allowing users to view historical warning records and fault trend information; The interactive feedback module is used to provide a parameter setting interface and an interactive interface, adjust and optimize the parameters of the OCELM model based on the parameter setting interface, and provide feedback and suggestions based on the interactive interface.

Images