Lightning arrester on-line monitoring platform and detection method
By combining multimodal sensing and intelligent diagnostic modules, high-precision sensing and intelligent diagnosis of surge arrester status are achieved, solving the problem of insufficient intelligence in existing technologies and improving the real-time performance and accuracy of surge arrester monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-31
- Publication Date
- 2026-04-07
AI Technical Summary
Existing surge arrester condition monitoring technologies lack intelligence, have limited multi-source data fusion and modeling capabilities, exhibit high system closure, and struggle to dynamically adapt to complex operating conditions. Traditional monitoring methods are also unable to achieve high-precision perception and multi-dimensional intelligent diagnosis.
It employs a multimodal perception module, an edge intelligent analysis module, a cloud-edge collaboration module, and an intelligent diagnosis module, combined with a lightweight state detection intelligent model and a fault diagnosis model, to achieve multi-dimensional feature extraction, collaborative training and enhancement, supporting high-precision perception and intelligent diagnosis.
This improves the real-time performance and accuracy of surge arrester status monitoring, enhances operation and maintenance efficiency, and provides technical support for the safe and stable operation of the power grid.
Smart Images
Figure CN121804579A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial control safety technology, and in particular to an online monitoring platform and testing method for surge arresters. Background Technology
[0002] As a core component of power system overvoltage protection, surge arresters directly impact the reliability of power grid operation and the safety of electrical equipment. Traditional surge arrester condition assessment primarily relies on preventative tests after periodic power outages (such as DC leakage current testing and resistive current measurement) and manual inspections, which have significant limitations. In recent years, online monitoring technology has gradually replaced traditional methods, achieving condition tracking by real-time acquisition of parameters such as surge arrester leakage current, temperature, and ambient humidity. However, existing technologies lack sufficient algorithmic intelligence; most systems rely on threshold alarms or simple linear analysis, lacking the ability to fuse and model multi-source data (such as current, temperature, and weather data); furthermore, the systems are highly closed, with existing platforms often employing proprietary protocols and isolated architectures, resulting in insufficient exploitation of data value.
[0003] Furthermore, with the high proportion of new energy sources being integrated and the grid load fluctuations intensifying, the frequency and intensity of transient overvoltages experienced by surge arresters have increased significantly. Traditional monitoring methods are unable to dynamically adapt to complex operating conditions, and there is an urgent need for an online monitoring platform with high-precision sensing, multi-dimensional intelligent diagnosis, and deep collaborative capabilities to achieve the transformation from "periodic inspection" to "state management" and provide technical support for proactive operation and maintenance and risk prevention and control of the power system. Summary of the Invention
[0004] This invention provides an online monitoring platform for surge arresters, comprising: a multimodal sensing module, an edge intelligent processing module, a cloud-edge collaboration module, an intelligent diagnostic module, and a large visualization screen; The multimodal sensing module is deployed at the surge arrester installation site to monitor the electrical, thermal, mechanical, environmental, and life parameters of the surge arrester simultaneously through multiple channels. The edge intelligent analysis module is deployed on each node of the edge network to receive parameters monitored by the multimodal perception module and use the state detection intelligent model on the node to detect anomalies in the surge arrester. The cloud-edge collaboration module, deployed on a cloud server, is used to collaboratively train and enhance the state detection intelligent model on the edge node using global lightning arrester monitoring data. The intelligent diagnostic module, deployed on a cloud server, is used to generate fault diagnosis results for the surge arrester based on the anomaly detection results returned by the edge intelligent analysis module. A large visualization screen, deployed on a cloud server, is used to visualize the physical location and status of all surge arresters within the monitoring area.
[0005] As described above, the online monitoring platform for surge arresters includes an edge intelligent analysis module that specifically comprises: The monitoring parameter preprocessing submodule is used to extract features from the received monitoring parameters and generate a multi-dimensional feature vector. The state detection model inference submodule is used to run a lightweight state detection intelligent model to infer the health status of the surge arrester in real time based on multi-dimensional feature vectors. The local decision-making submodule is used to dynamically adjust the sampling frequency and storage mode based on the inference results of the surge arrester's health status.
[0006] As described above, the online monitoring platform for surge arresters has two storage modes: merged storage and individual storage. When the surge arrester health status inference result H is higher than the health threshold for a long period, the merged storage mode is adopted. In the merged storage mode, the monitoring time and monitoring data will be periodically merged into a range interval, and the monitoring data in each merging cycle occupies only one storage unit. When the H value is close to or below the threshold, the individual storage mode is adopted. In the individual storage mode, each monitoring data occupies one storage unit.
[0007] As described above, the online monitoring platform for surge arresters includes a cloud-edge collaboration module that specifically comprises: The edge model evaluation submodule is used to periodically verify the accuracy of the state monitoring intelligent model on each edge node; The data distribution query submodule is used to query the distribution characteristics of the data stored in each edge node; The edge model aggregation submodule is used to aggregate edge models whose data distribution feature similarity is higher than a threshold into a general model. The edge model enhancement training submodule is used to enhance the training of the aggregated general model, and then push it to the source node after training is completed.
[0008] As described above, in an online monitoring platform for surge arresters, the aggregation process of the edge model is as follows: Calculate the aggregate weighted value based on the node data volume and the edge model accuracy; Aggregate model parameters using aggregated weighted values; A new general model is generated based on the aggregated model parameters.
[0009] As described above, the online monitoring platform for surge arresters includes two parts for abnormal detection results: a status code and additional information. If the status code indicates that the equipment is operating normally, it is ignored. If the status code indicates that the equipment is operating abnormally, the content of the additional information is extracted and input into the fault diagnosis model to generate the fault diagnosis result of the surge arrester.
[0010] The present invention also provides a method for online monitoring of surge arresters, comprising: Step 1: Use a sensor network to synchronously collect electrical, thermal, mechanical, environmental, and lifespan parameters of the lightning protection system through multiple channels. Step 2: Each node of the edge network receives parameters collected by the sensor network in real time and uses the state detection intelligent model on its node to detect anomalies in the surge arrester. Step 3: The cloud server uses global lightning arrester monitoring data to collaboratively train and enhance the state detection intelligent model on the edge nodes; Step 4: Generate fault diagnosis results for the surge arrester based on the anomaly detection results returned by the edge nodes; Step 5: Visualize the physical location and status of all surge arresters within the monitoring area.
[0011] The beneficial effects achieved by this invention are as follows: it combines high-precision sensing, intelligent diagnosis and deep collaboration capabilities, which can comprehensively improve the real-time performance, accuracy and operation and maintenance efficiency of surge arrester status monitoring, and provide strong technical support for the safe and stable operation of the power grid. Attached Figure Description
[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings.
[0013] Figure 1 This is a schematic diagram of an online monitoring platform for surge arresters provided in Embodiment 1 of this application; Figure 2 This is a flowchart of an online monitoring method for surge arresters provided in Embodiment 2 of this application. Detailed Implementation
[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0015] Example 1 like Figure 1 As shown, Embodiment 1 of this application provides an online monitoring platform for surge arresters, including: a multimodal perception module 11, an edge intelligent processing module 12, a cloud-edge collaboration module 13, an intelligent diagnostic module 14, and a visualization screen 15; The multimodal sensing module 11 is deployed at the surge arrester installation site to monitor the electrical, thermal, mechanical, environmental, and life parameters of the surge arrester simultaneously through multiple channels. A sensor network is deployed at the surge arrester installation site, and a multi-channel data acquisition card is used to acquire various monitoring parameters of the surge arrester in real time. These parameters include: electrical parameters such as leakage current, resistive / capacitive current components, harmonic distortion rate, and partial discharge signal; thermal parameters such as surface temperature distribution, internal hot spot temperature, and temperature rise rate; mechanical parameters such as vibration frequency, mechanical deformation, and porcelain bushing stress and strain; environmental parameters such as ambient temperature and humidity, pollution level, wind speed, and rainfall; and life parameters such as the number of operations and the integral of the current carrying capacity.
[0016] The edge intelligent analysis module 12 is deployed on each node of the edge network to receive parameters monitored by the multimodal perception module and use the state detection intelligent model on the local node to perform anomaly detection of the surge arrester; specifically, it includes: a monitoring parameter preprocessing submodule, a state detection model inference submodule, a state detection model update submodule, and a local decision submodule. 1. Monitoring parameter preprocessing submodule, used to extract features from the received monitoring parameters and generate a multi-dimensional feature vector; First, the monitored parameters are normalized to unify their dimensions. Then, features are extracted and fused. Finally, the extracted and fused features are concatenated to obtain a multi-dimensional feature vector. For parameters that are linearly related to the health of the surge arrester, including resistive current, leakage current, pollution level, and operational frequency statistics, the formula is used: To extract and fuse, among which This represents the linear parameter fusion feature at time stamp t. This indicates the start timestamp of the current detection period. This represents the fusion weight of the i-th linear parameter. Indicates the i-th linear parameter in The value at point i ranges from 1 to n, where n is the total number of linear parameters; for other parameters that have a non-linear relationship with the arrester's health, the formula is used: To extract and fuse, among which This represents the nonlinear parameter fusion feature at time stamp t. The fusion weight for the j-th nonlinear parameter is... Indicates the j-th nonlinear parameter in The value at the point, Let represent a single transformation function for the j-th nonlinear parameter, where j takes values from 1 to m, and m is the total number of nonlinear parameters. Then, the multidimensional feature vector generated at time stamp t would be represented as... .
[0017] 2. The state detection model inference submodule is used to run a lightweight state detection intelligent model and infer the health status of the surge arrester in real time based on multi-dimensional feature vectors. The lightweight state detection intelligent model can take into account the differences between linear and nonlinear features simultaneously, and is expressed as: Where H represents the health status inference result output by the model. This is the weight matrix of the output layer. This is the bias vector for the output layer. and It is an adjustable parameter. This represents the activation function of a linear branching neural network. These are the weight matrix and bias vector of a linear branch neural network, respectively. For the attention weights of a non-linear branching neural network, Represents element-wise product. A temporal convolutional network used for extraction Temporal dependency features, These are the two components of a multidimensional feature vector.
[0018] 3. Local decision-making submodule, used to dynamically adjust the sampling frequency and storage mode based on the reasoning results of the surge arrester's health status; The closer the surge arrester's health status inference result H is to 1, the healthier the equipment is; conversely, the more likely an abnormality is to occur. Therefore, the closer the H value is to 1, the lower the sampling frequency; the closer it is to 0, the higher the sampling frequency. The storage modes are divided into two types: merged storage and individual storage. When the H value is higher than the health threshold for a long period of time, the merged storage mode is adopted. In the merged storage mode, the monitoring time and monitoring data will be merged into a range periodically. The monitoring data in each merging cycle only occupies one storage unit. When the H value is close to or lower than the threshold, the individual storage mode is adopted. In the individual storage mode, each monitoring data occupies one storage unit. When the H value is higher than the health threshold, the edge node returns a "normal operation" status code to the cloud server. When the H value is lower than or equal to the health threshold, it returns an "abnormal" status code, along with the inference result set of the recent surge arrester.
[0019] The cloud-edge collaboration module 13 is deployed on a cloud server and is used to collaboratively train and enhance the state detection intelligent model on the edge node using global lightning arrester monitoring data. Specifically, it includes: an edge model evaluation submodule, a data distribution query submodule, an edge model aggregation submodule, and an edge model enhancement training submodule. 1. Edge Model Evaluation Submodule: This module is used to periodically verify the accuracy of the intelligent status monitoring model on each edge node. First, a verification period is defined, and the accuracy of the state monitoring intelligent model is verified by comparing the output results of the model with the actual results within the verification period.
[0020] 2. Data distribution query submodule, used to query the distribution characteristics of the data stored in each edge node; The cloud server can obtain the stored data in each edge node through the data distribution query submodule, which is the global monitoring data. The distribution characteristics of the stored data in each edge node is the classification identifier of the global monitoring data. The distribution characteristic identifier is P, P=(c,e,s,g), where c,e,s,g are the mean, variance, skewness and kurtosis of the stored data (after normalization), respectively.
[0021] 3. Edge model aggregation submodule, used to aggregate edge models with data distribution feature similarity higher than a threshold into a general model; First, the cosine similarity algorithm is used to calculate the similarity of the data distribution features. Then, edge models with a similarity higher than 0.8 (the threshold can be set as needed) are aggregated into a general model. The aggregation process of edge models is as follows: ① Calculate the aggregate weighted value based on the node data volume and the edge model accuracy; It should be noted that the node data volume here also needs to be normalized before participating in the calculation. The formula for calculating the aggregate weighted value is expressed as follows: ,in This represents the aggregate weighted value of the k-th edge model. These represent the amount of monitoring data stored in the k-th node and the accuracy of the k-th edge model, respectively. , representing the reward factors for node data volume and model accuracy respectively, and G is the set of edge nodes to be aggregated.
[0022] ② Use aggregated weighted values to aggregate model parameters; Using formula Complete the aggregation of model parameters, where These are the aggregated model parameters. These are the parameters before the aggregation of the k-th edge model. Let G represent the aggregate weighted value of the k-th edge model, where G is the set of edge nodes to be aggregated.
[0023] ③ Generate a new general model based on the aggregated model parameters; Once all model parameters are aggregated, a new general model can be generated.
[0024] For each set of data distribution feature similarity nodes above 0.8, the edge models deployed within those nodes can be aggregated into a new general model, with the aggregation process being consistent.
[0025] 4. Edge model enhancement training submodule, used to enhance the training of the aggregated general model, and then push it to the source node after training is completed; Nodes with a distribution feature similarity of more than 80% are stored and processed into a training dataset. The aggregated general model is then enhanced and trained. The trained general model is applicable to all edge nodes, i.e., source nodes, where the edge model was located before aggregation.
[0026] The intelligent diagnostic module 14, deployed on the cloud server, is used to generate fault diagnosis results for the surge arrester based on the anomaly detection results returned by the edge intelligent analysis module. The anomaly detection result is divided into two parts: status code and additional information. If the status code indicates that the equipment is operating normally, it is ignored. If the status code indicates that the equipment is operating abnormally, the content of the additional information is extracted and input into the fault diagnosis model to generate the fault diagnosis result of the surge arrester. The fault diagnosis model can diagnose the fault type of the surge arrester based on the recent health status inference data. The model is established as follows: a training dataset is created based on the historical fault detection data of the surge arrester. The output set in the training dataset is the historical fault type of the surge arrester, and the input set is the health status inference data of the surge arrester in the detection cycle before each historical fault occurred. A lightweight time-series classification model is used as the base model, and a supervised training method is used to train and optimize the base model to obtain the final fault diagnosis model.
[0027] The diagnostic results generated by the fault diagnosis model will be uploaded to the visualization dashboard in real time.
[0028] The visualization screen 15, deployed on a cloud server, is used to visualize the physical location and status of all surge arresters in the monitoring area. The large visual screen can display the physical location of all surge arresters within the monitoring area, supports multi-level zoom and view switching, and displays the operating status of each surge arrester. The status bar of a surge arrester operating normally is green, while the status bar of a surge arrester operating abnormally is red. Clicking the corresponding icon can also display detailed monitoring data and fault type.
[0029] Example 2 like Figure 2 As shown, Embodiment 2 of this application provides an online monitoring method for surge arresters, including: Step S10: Use a sensor network to synchronously collect electrical, thermal, mechanical, environmental, and life parameters of the lightning protection system through multiple channels; A sensor network is deployed at the surge arrester installation site, and a multi-channel data acquisition card is used to acquire various monitoring parameters of the surge arrester in real time. These parameters include: electrical parameters such as leakage current, resistive / capacitive current components, harmonic distortion rate, and partial discharge signal; thermal parameters such as surface temperature distribution, internal hot spot temperature, and temperature rise rate; mechanical parameters such as vibration frequency, mechanical deformation, and porcelain bushing stress and strain; environmental parameters such as ambient temperature and humidity, pollution level, wind speed, and rainfall; and life parameters such as the number of operations and the integral of the current carrying capacity.
[0030] Step S20: Each node in the edge network receives parameters collected by the sensor network in real time and uses the state detection intelligent model on its node to perform anomaly detection of the surge arrester; specifically, it is divided into the following sub-steps: Step S21: Extract features from the received monitoring parameters to generate a multi-dimensional feature vector; First, the monitored parameters are normalized to unify their dimensions. Then, features are extracted and fused. Finally, the extracted and fused features are concatenated to obtain a multi-dimensional feature vector. For parameters that are linearly related to the health of the surge arrester, including resistive current, leakage current, pollution level, and operational frequency statistics, the formula is used: To extract and fuse, among which This represents the linear parameter fusion feature at time stamp t. This indicates the start timestamp of the current detection period. This represents the fusion weight of the i-th linear parameter. Indicates the i-th linear parameter in The value at point i ranges from 1 to n, where n is the total number of linear parameters; for other parameters that have a non-linear relationship with the arrester's health, the formula is used: To extract and fuse, among which This represents the nonlinear parameter fusion feature at time stamp t. The fusion weight for the j-th nonlinear parameter is... Indicates the j-th nonlinear parameter in The value at the point, Let represent a single transformation function for the j-th nonlinear parameter, where j takes values from 1 to m, and m is the total number of nonlinear parameters. Then, the multidimensional feature vector generated at time stamp t would be represented as... .
[0031] Step S22: Used to run a lightweight condition detection intelligent model to infer the health status of the surge arrester in real time based on multi-dimensional feature vectors; The lightweight state detection intelligent model can take into account the differences between linear and nonlinear features simultaneously, and is expressed as: Where H represents the health status inference result output by the model. This is the weight matrix of the output layer. This is the bias vector for the output layer. and It is an adjustable parameter. This represents the activation function of a linear branching neural network. These are the weight matrix and bias vector of a linear branch neural network, respectively. For the attention weights of a non-linear branching neural network, Represents element-wise product. A temporal convolutional network used for extraction Temporal dependency features, These are the two components of a multidimensional feature vector.
[0032] Step S23: Dynamically adjust the sampling frequency and storage mode based on the reasoning results of the surge arrester's health status; For the inference result H of the surge arrester's health status, the closer the value is to 1, the healthier the equipment is, and the more likely it is to be abnormal. Therefore, the closer the H value is to 1, the lower the sampling frequency should be, and the closer it is to 0, the higher the sampling frequency should be. The storage modes are divided into two types: merged storage and individual storage. When the H value is higher than the health threshold for a long period of time, the merged storage mode is adopted. In the merged storage mode, the monitoring time and monitoring data will be merged into a range periodically. The monitoring data in each merging cycle only occupies one storage unit. When the H value is close to or lower than the threshold, the individual storage mode is adopted. In the individual storage mode, each monitoring data occupies one storage unit. When the H value is higher than the health threshold, the edge node returns a "normal operation" status code to the cloud server. When the H value is lower than or equal to the health threshold, it returns an "abnormal" status code, along with the inference result set of the recent surge arrester.
[0033] Step S30: The cloud server uses global lightning arrester monitoring data to collaboratively train and enhance the state detection intelligent model on the edge nodes; specifically, it consists of the following sub-steps: Step S31: Periodically verify the accuracy of the state monitoring intelligent model on each edge node; First, a verification period is defined, and the accuracy of the state monitoring intelligent model is verified by comparing the output results of the model with the actual results within the verification period.
[0034] Step S32: Query the distribution characteristics of the stored data in each edge node; The cloud server can obtain the stored data in each edge node through the data distribution query submodule, which is the global monitoring data. The distribution characteristics of the stored data in each edge node is the classification identifier of the global monitoring data. The distribution characteristic identifier is P, P=(c,e,s,g), where c,e,s,g are the mean, variance, skewness and kurtosis of the stored data (after normalization), respectively.
[0035] Step S33: Aggregate edge models with data distribution feature similarity higher than a threshold into a general model; First, the cosine similarity algorithm is used to calculate the similarity of the data distribution features. Then, edge models with a similarity higher than 0.8 (the threshold can be set as needed) are aggregated into a general model. The aggregation process of edge models is as follows: ① Calculate the aggregate weighted value based on the node data volume and the edge model accuracy; It should be noted that the node data volume here also needs to be normalized before participating in the calculation. The formula for calculating the aggregate weighted value is expressed as follows: ,in This represents the aggregate weighted value of the k-th edge model. , These represent the amount of monitoring data stored in the k-th node and the accuracy of the k-th edge model, respectively. , representing the reward factors for node data volume and model accuracy respectively, and G is the set of edge nodes to be aggregated.
[0036] ② Use aggregated weighted values to aggregate model parameters; Using formula Complete the aggregation of model parameters, where These are the aggregated model parameters. These are the parameters before the aggregation of the k-th edge model. Let G represent the aggregate weighted value of the k-th edge model, where G is the set of edge nodes to be aggregated.
[0037] ③ Generate a new general model based on the aggregated model parameters; Once all model parameters are aggregated, a new general model can be generated.
[0038] For each set of data distribution feature similarity nodes above 0.8, the edge models deployed within those nodes can be aggregated into a new general model, with the aggregation process being consistent.
[0039] Step S34: Perform augmentation training on the aggregated general model, and then push it to the source node after training is complete; Nodes with a distribution feature similarity of more than 80% are stored and processed into a training dataset. The aggregated general model is then enhanced and trained. The trained general model is applicable to all edge nodes, i.e., source nodes, where the edge model was located before aggregation.
[0040] Step S40: Generate fault diagnosis results for the surge arrester based on the anomaly detection results returned by the edge nodes; The anomaly detection result is divided into two parts: status code and additional information. If the status code indicates that the equipment is operating normally, it is ignored. If the status code indicates that the equipment is operating abnormally, the content of the additional information is extracted and input into the fault diagnosis model to generate the fault diagnosis result of the surge arrester. The fault diagnosis model can diagnose the fault type of the surge arrester based on the recent health status inference data. The model is established as follows: a training dataset is created based on the historical fault detection data of the surge arrester. The output set in the training dataset is the historical fault type of the surge arrester, and the input set is the health status inference data of the surge arrester in the detection cycle before each historical fault occurred. A lightweight time-series classification model is used as the base model, and a supervised training method is used to train and optimize the base model to obtain the final fault diagnosis model.
[0041] Step S50: Visualize the physical location and status of all surge arresters within the monitoring area; The system visualizes the physical location of all surge arresters within the monitoring area, supports multi-level zoom and view switching, and displays the operating status of each surge arrester. Surge arresters operating normally have a green status bar, while those operating abnormally have a red status bar. Clicking the corresponding icon will also display detailed monitoring data and fault type.
[0042] Corresponding to the above embodiments, the present invention provides a computer storage medium, including: at least one memory and at least one processor; The memory is used to store one or more program instructions; A processor is used to run one or more program instructions to execute an online monitoring method for surge arresters.
[0043] Corresponding to the above embodiments, the present invention provides a computer-readable storage medium containing one or more program instructions, which are executed by a processor to provide an online monitoring method for surge arresters.
[0044] The embodiments disclosed in this invention provide a computer-readable storage medium storing computer program instructions, which, when executed on a computer, cause the computer to perform the above-described online monitoring method for surge arresters.
[0045] In this embodiment of the invention, the processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.
[0046] The various methods, steps, and logic diagrams disclosed in the embodiments of this invention can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The processor reads information from the storage medium and, in conjunction with its hardware, completes the steps of the above methods.
[0047] The storage medium can be memory, such as volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.
[0048] Among them, non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory.
[0049] Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (Synchlink DRAM, SLDRAM), and direct memory bus RAM (DRRAM).
[0050] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.
[0051] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in this invention can be implemented using a combination of hardware and software. When applied as software, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of computer programs from one place to another. Storage media can be any available medium that can be accessed by a general-purpose or special-purpose computer.
[0052] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.
Claims
1. An online monitoring platform for surge arresters, characterized in that, include: Multimodal perception module, edge intelligent processing module, cloud-edge collaboration module, intelligent diagnostic module, and visualization screen; The multimodal sensing module is deployed at the surge arrester installation site to monitor the electrical, thermal, mechanical, environmental, and life parameters of the surge arrester simultaneously through multiple channels. The edge intelligent analysis module is deployed on each node of the edge network to receive parameters monitored by the multimodal perception module and use the state detection intelligent model on the node to detect anomalies in the surge arrester. The cloud-edge collaboration module, deployed on a cloud server, is used to collaboratively train and enhance the state detection intelligent model on the edge node using global lightning arrester monitoring data. The intelligent diagnostic module, deployed on a cloud server, is used to generate fault diagnosis results for the surge arrester based on the anomaly detection results returned by the edge intelligent analysis module. A large visualization screen, deployed on a cloud server, is used to visualize the physical location and status of all surge arresters within the monitoring area.
2. The online monitoring platform for surge arresters according to claim 1, characterized in that, The edge intelligent analysis module specifically includes: The monitoring parameter preprocessing submodule is used to extract features from the received monitoring parameters and generate a multi-dimensional feature vector. The state detection model inference submodule is used to run a lightweight state detection intelligent model to infer the health status of the surge arrester in real time based on multi-dimensional feature vectors. The local decision-making submodule is used to dynamically adjust the sampling frequency and storage mode based on the inference results of the surge arrester's health status.
3. The online monitoring platform for surge arresters according to claim 2, characterized in that, The storage modes are divided into two types: merged storage and individual storage. When the surge arrester health status inference result H is higher than the health threshold for a long period of time, the merged storage mode is adopted. In the merged storage mode, the monitoring time and monitoring data will be merged into a range periodically. The monitoring data in each merging cycle only occupies one storage unit. When the H value is close to or lower than the threshold, the individual storage mode is adopted. In the individual storage mode, each monitoring data occupies one storage unit.
4. The online monitoring platform for surge arresters according to claim 1, characterized in that, The cloud-edge collaboration module specifically includes: The edge model evaluation submodule is used to periodically verify the accuracy of the state monitoring intelligent model on each edge node; The data distribution query submodule is used to query the distribution characteristics of the data stored in each edge node; The edge model aggregation submodule is used to aggregate edge models whose data distribution feature similarity is higher than a threshold into a general model. The edge model enhancement training submodule is used to enhance the training of the aggregated general model, and then push it to the source node after training is completed.
5. The online monitoring platform for surge arresters according to claim 4, characterized in that, The aggregation process for edge models is as follows: Calculate the aggregate weighted value based on the node data volume and the edge model accuracy; Aggregate model parameters using aggregated weighted values; A new general model is generated based on the aggregated model parameters.
6. The online monitoring platform for surge arresters according to claim 1, characterized in that, The anomaly detection results are divided into two parts: status code and additional information. If the status code indicates that the equipment is operating normally, it is ignored. If the status code indicates that the equipment is operating abnormally, the content of the additional information part is extracted and input into the fault diagnosis model to generate the fault diagnosis results of the surge arrester.
7. A method for online monitoring of surge arresters, characterized in that, include: Step 1: Use a sensor network to synchronously collect electrical, thermal, mechanical, environmental, and lifespan parameters of the lightning protection system through multiple channels. Step 2: Each node of the edge network receives parameters collected by the sensor network in real time and uses the state detection intelligent model on its node to detect anomalies in the surge arrester. Step 3: The cloud server uses global lightning arrester monitoring data to collaboratively train and enhance the state detection intelligent model on the edge nodes; Step 4: Generate fault diagnosis results for the surge arrester based on the anomaly detection results returned by the edge nodes; Step 5: Visualize the physical location and status of all surge arresters within the monitoring area.
8. A computer storage medium, characterized in that, include: At least one memory and at least one processor; Memory, used to store one or more program instructions; A processor for running one or more program instructions to perform an online monitoring method for a surge arrester as described in claim 7.