An intelligent high-voltage switch cabinet monitoring system and method based on the Internet of Things

Abstract

This invention relates to the field of switchgear monitoring technology, specifically an intelligent high-voltage switchgear monitoring system and method based on the Internet of Things (IoT). The monitoring system includes: a distributed model training module: each edge node independently trains a prediction model using locally collected monitoring data and uploads the training results to the cloud in encrypted form; the cloud aggregates the training results uploaded by each edge node to generate a global model; a dynamic resource negotiation module: dynamically adjusts the allocation strategy of monitoring resources based on the real-time load status and fault risk level of each device; and a contribution measurement module: calculates the contribution weight of each edge node based on the amount of resources invested and the fault suppression effect during the collaboration process and dynamically allocates the collaboration benefits. This invention transforms multiple high-voltage switchgear from "individual operation" to "intelligent collaboration," effectively preventing the spread of a single device fault to regional power grid paralysis. It meets power grid safety requirements; resource allocation is adjusted in real time according to the power grid status, avoiding waste.

Description

Technical Field

[0001] This invention relates to the field of switchgear monitoring technology, specifically an intelligent high-voltage switchgear monitoring system and method based on the Internet of Things. Background Technology

[0002] High-voltage switchgear plays a vital protective role in power grids. When equipment or lines in the power grid fail, high-voltage switchgear can quickly take countermeasures to prevent the fault from escalating, thereby protecting the normal operation of other equipment in the power grid.

[0003] In the power grid, high-voltage switchgear is usually deployed in clusters. Due to the data confidentiality requirements in power grid operation, the existing monitoring systems are mostly isolated. Each switchgear corresponds to its own monitoring system, and data is not shared. This can lead to a chain reaction (such as overload transmission and voltage drop) caused by the failure of a single device, and there is a lack of collaborative early warning and dynamic resource allocation capabilities between devices. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent high-voltage switchgear monitoring system and method based on the Internet of Things to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent high-voltage switchgear monitoring system based on the Internet of Things, the monitoring system comprising:

[0006] Distributed model training module: Edge nodes are deployed on multiple high-voltage switchgear. Each edge node independently trains a prediction model using locally collected monitoring data and uploads the training results to the cloud in encrypted form. The cloud aggregates the training results uploaded by each edge node to generate a global model and distributes it to each edge node to update the local model.

[0007] Dynamic resource negotiation module: Communicatively connected to the distributed model training module, used to establish cooperation rules among multiple devices, and dynamically adjust the allocation strategy of monitoring resources according to the real-time load status and fault risk level of each device. The monitoring resources include communication bandwidth, computing resources and task priority.

[0008] Low-latency collaborative network module: It uses time-sensitive communication protocol to connect each high-voltage switchgear, and supports the synchronous transmission of equipment status data and collaborative instructions;

[0009] Contribution Measurement Module: Calculates the contribution weight of each edge node and dynamically allocates collaboration benefits based on the amount of resources invested and the fault suppression effect during the collaboration process.

[0010] In the distributed model training module:

[0011] The training results uploaded by each edge node are model parameter updates processed with homomorphic encryption to ensure that the original monitoring data does not leave the local area;

[0012] The cloud uses an adversarial feature alignment method to eliminate the impact of differences in data distribution between different devices on the global model;

[0013] Model compression technology is used to convert complex models generated in the cloud into lightweight versions that are adapted to the computing capabilities of edge nodes.

[0014] In the dynamic resource negotiation module:

[0015] The generation of resource allocation strategies includes the following steps:

[0016] Construct a multi-dimensional state space that includes device load rate, health status of adjacent devices, and communication quality;

[0017] Define a set of executable actions, including increasing the frequency of local data collection, requesting external computing support, or releasing idle resources;

[0018] A reward mechanism will be established based on historical collaborative experience, prioritizing strategies that can both improve local monitoring accuracy and reduce the risk of global failures.

[0019] Through iterative optimization, each device can autonomously achieve a balanced state of resource allocation.

[0020] The low-latency cooperative network further includes:

[0021] A device trustworthiness assessment mechanism based on historical communication records automatically downgrades the weight of nodes that frequently fail to transmit.

[0022] A chain-based data verification method is used to ensure the integrity and authenticity of collaborative instructions.

[0023] In the contribution metric module:

[0024] The calculation logic for contribution weights includes:

[0025] Statistics on the computing resources, communication bandwidth, and amount of key early warning data provided by each device during the collaboration process;

[0026] Evaluate the improvement in fault suppression effectiveness before and after the equipment participates in the collaboration;

[0027] The basic contribution value is dynamically adjusted based on resource cost weights and fault severity levels.

[0028] The revised contribution values ​​are mapped proportionally to resource revenue allocation weights.

[0029] A monitoring method for intelligent high-voltage switchgear based on the Internet of Things, the monitoring method comprising the following steps:

[0030] Step S1: An edge node is deployed for each high-voltage switchgear. Each edge node uses local data to train and generate a local monitoring model. The local monitoring model is then encrypted and uploaded to the cloud to generate a global model.

[0031] Step S2: When any edge node detects an abnormal situation, it initiates a multi-device resource negotiation request and selects the optimal resource allocation scheme based on the real-time status data of each edge node.

[0032] Step S3: Quantify the resource input and fault suppression contribution of each edge node in the collaboration process, calculate the contribution weight of each edge node, and dynamically adjust the subsequent collaboration rights allocation.

[0033] Step S4: Send the optimal resource allocation plan to the edge node where the abnormal situation occurs and execute it.

[0034] Step S1 includes: the edge node is a local intelligent computing device deployed on the high-voltage switchgear, used to perform local detection data processing, local monitoring model training and collaborative decision-making tasks, and to interact with the cloud system. The local intelligent computing device includes a data acquisition interface, an embedded processor and a communication unit.

[0035] The input to the local monitoring model is the local detection data of the high-voltage switchgear, and its output is the fault type of the high-voltage switchgear; the local detection data includes electrical parameters, temperature parameters, mechanical parameters, insulation parameters, and environmental parameters; the fault types include overload, insulation aging, and partial discharge.

[0036] The global model is used to integrate the feature patterns of local monitoring data from each edge node to form unified monitoring knowledge, and the global model is updated periodically.

[0037] Step S2 includes:

[0038] Step S201: When the local detection data of a certain edge node triggers the threshold, it is marked as an abnormal situation, and a multi-device resource negotiation request is issued;

[0039] Step S202: After receiving a multi-device resource negotiation request from an edge node, the cloud obtains the fault type of the abnormal situation and generates a negotiation request message, which includes:

[0040] Required resource types: computing resources, communication bandwidth, sensor sampling frequency;

[0041] Urgency level: classified according to the type of fault;

[0042] Expected collaboration duration: The time window for failure development estimated based on the predictive model;

[0043] Step S203: Send a request to adjacent edge nodes and the cloud through the low-latency collaborative network. After receiving the request, each edge node calculates the idle resources of its device. The idle resources include computing resource availability, communication bandwidth margin, and sensor load. Generate a response message based on the idle resources. The response message includes: the type and quantity of available required resources and its own status label. The own status label includes high reliability nodes and low load nodes.

[0044] Step S204: Collect all response messages and calculate the weighted score of each edge node that sends a response message. The weighted score is calculated as: (computing resources × computing power weight + bandwidth × bandwidth weight) × reliability coefficient / response delay. The response delay is the time from when each edge node receives the request until it sends a response message. Arrange the weighted scores in descending order to form a node sequence. Send a joint collaboration instruction to each edge node in the order of the node sequence from top to bottom. The edge node that first confirms the response to the joint collaboration instruction is designated as the collaboration node, and the collaboration node is assigned to run the analysis task.

[0045] The calculation process for calculating the contribution weight of the edge nodes in step S3 is as follows:

[0046] Step S301: Extract the record of the collaborative instructions completed by the edge node as a collaborating node, and obtain the contribution data of the edge node in the collaborative instructions. The contribution data includes the amount of computing resources invested, the bandwidth usage value, and the response timeliness. The response timeliness is the delay time from receiving the request to executing the collaborative instructions.

[0047] The contribution data is normalized by normalizing the data on computing resources, bandwidth usage, and response timeliness into scores in the range of 0 to 1. The computing resources and bandwidth ratios are directly mapped to the actual values, and the response latency score increases as the latency decreases.

[0048] The scores for computing resources, bandwidth usage, and response timeliness are weighted and summed to generate a resource input score, with computing resources having the highest weight, followed by communication bandwidth, and response delay having the lowest weight.

[0049] Step S302: Select evaluation indicators based on the fault type:

[0050] Overload-related faults: Measured by the percentage of load reduction and the number of affected devices;

[0051] Insulation-related faults: Indicators include the percentage decrease in discharge intensity and the lead time of the early warning.

[0052] After normalizing the performance of each edge node under different indicators, the weighted sum is calculated to generate the fault suppression contribution value. In overload faults, the weight of load reduction is higher than the impact range, and in insulation faults, the weight of discharge suppression is higher than the warning timeliness.

[0053] Step S303: Based on the current stability level of the power grid, set the integration ratio of resource input and fault suppression: the lower the stability, the higher the contribution of fault suppression.

[0054] The resource input score and the fault suppression contribution score are combined proportionally to generate a comprehensive contribution score.

[0055] The overall contribution of each node is converted into a percentage of the total contribution, which is then used as the weight for the allocation of rights and interests in subsequent collaboration cycles.

[0056] In step S4, the process of executing the optimal resource allocation scheme is as follows: select the corresponding level channel to send control commands according to the urgency of the fault type; after receiving the command, the cooperating node executes the analysis task and monitors the execution status of the command in real time.

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

[0058] 1. This invention transforms multiple high-voltage switchgear from "individual operation" to "intelligent collaboration," effectively preventing the spread of single equipment failures to regional power grid paralysis. It also has privacy protection functions, ensuring that data does not leave the local area and meeting power grid security requirements. Resource allocation is adjusted in real time according to the power grid status, avoiding waste. It has self-learning capabilities, and the equipment becomes more intelligent with use, eliminating the need for frequent manual adjustments to strategies. Attached Figure Description

[0059] Figure 1 This is a flowchart illustrating an intelligent high-voltage switchgear monitoring method based on the Internet of Things according to the present invention. Detailed Implementation

[0060] 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 embodiments of the present invention, and not all embodiments. 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.

[0061] Example: Figure 1 As shown, this invention provides an intelligent high-voltage switchgear monitoring system based on the Internet of Things, the monitoring system comprising:

[0062] Distributed model training module: Edge nodes are deployed on multiple high-voltage switchgear. Each edge node independently trains a prediction model using locally collected monitoring data and uploads the training results to the cloud in encrypted form. The cloud aggregates the training results uploaded by each edge node to generate a global model and distributes it to each edge node to update the local model.

[0063] Dynamic resource negotiation module: Communicatively connected to the distributed model training module, used to establish cooperation rules among multiple devices, and dynamically adjust the allocation strategy of monitoring resources according to the real-time load status and fault risk level of each device. The monitoring resources include communication bandwidth, computing resources and task priority.

[0064] Low-latency collaborative network module: It uses time-sensitive communication protocol to connect each high-voltage switchgear, and supports the synchronous transmission of equipment status data and collaborative instructions;

[0065] Contribution Measurement Module: Calculates the contribution weight of each edge node and dynamically allocates collaboration benefits based on the amount of resources invested and the fault suppression effect during the collaboration process.

[0066] In the distributed model training module:

[0067] The training results uploaded by each edge node are model parameter updates processed with homomorphic encryption to ensure that the original monitoring data does not leave the local area;

[0068] The cloud uses an adversarial feature alignment method to eliminate the impact of differences in data distribution between different devices on the global model;

[0069] Model compression technology is used to convert complex models generated in the cloud into lightweight versions that are adapted to the computing capabilities of edge nodes.

[0070] In the dynamic resource negotiation module:

[0071] The generation of resource allocation strategies includes the following steps:

[0072] Construct a multi-dimensional state space that includes device load rate, health status of adjacent devices, and communication quality;

[0073] Define a set of executable actions, including increasing the frequency of local data collection, requesting external computing support, or releasing idle resources;

[0074] A reward mechanism will be established based on historical collaborative experience, prioritizing strategies that can both improve local monitoring accuracy and reduce the risk of global failures.

[0075] Through iterative optimization, each device can autonomously achieve a balanced state of resource allocation.

[0076] The low-latency cooperative network further includes:

[0077] A device trustworthiness assessment mechanism based on historical communication records automatically downgrades the weight of nodes that frequently fail to transmit.

[0078] A chain-based data verification method is used to ensure the integrity and authenticity of collaborative instructions.

[0079] In the contribution metric module:

[0080] The calculation logic for contribution weights includes:

[0081] Statistics on the computing resources, communication bandwidth, and amount of key early warning data provided by each device during the collaboration process;

[0082] Evaluate the improvement in fault suppression effectiveness before and after the equipment participates in the collaboration;

[0083] The basic contribution value is dynamically adjusted based on resource cost weights and fault severity levels.

[0084] The revised contribution values ​​are mapped proportionally to resource revenue allocation weights.

[0085] A monitoring method for intelligent high-voltage switchgear based on the Internet of Things, the monitoring method comprising the following steps:

[0086] Step S1: An edge node is deployed for each high-voltage switchgear. Each edge node uses local data to train and generate a local monitoring model. The local monitoring model is then encrypted and uploaded to the cloud to generate a global model.

[0087] Step S2: When any edge node detects an abnormal situation, it initiates a multi-device resource negotiation request and selects the optimal resource allocation scheme based on the real-time status data of each edge node.

[0088] Step S3: Quantify the resource input and fault suppression contribution of each edge node in the collaboration process, calculate the contribution weight of each edge node, and dynamically adjust the subsequent collaboration rights allocation.

[0089] Step S4: Send the optimal resource allocation plan to the edge node where the abnormal situation occurs and execute it.

[0090] Step S1 includes: the edge node is a local intelligent computing device deployed on the high-voltage switchgear, used to perform local detection data processing, local monitoring model training and collaborative decision-making tasks, and to interact with the cloud system. The local intelligent computing device includes a data acquisition interface, an embedded processor and a communication unit.

[0091] The input to the local monitoring model is the local detection data of the high-voltage switchgear, and its output is the fault type of the high-voltage switchgear; the local detection data includes electrical parameters, temperature parameters, mechanical parameters, insulation parameters, and environmental parameters; the fault types include overload, insulation aging, and partial discharge.

[0092] The global model is used to integrate the feature patterns of local monitoring data from each edge node to form unified monitoring knowledge, and the global model is updated periodically.

[0093] Step S2 includes:

[0094] Step S201: When the local detection data of a certain edge node triggers the threshold, it is marked as an abnormal situation, and a multi-device resource negotiation request is issued;

[0095] Step S202: After receiving a multi-device resource negotiation request from an edge node, the cloud obtains the fault type of the abnormal situation and generates a negotiation request message, which includes:

[0096] Required resource types: computing resources, communication bandwidth, sensor sampling frequency;

[0097] Urgency level: classified according to the type of fault (e.g., overload = emergency level, partial discharge = high-risk level).

[0098] Expected collaboration duration: The time window for failure development estimated based on the predictive model;

[0099] Step S203: Send a request to adjacent edge nodes and the cloud through the low-latency collaborative network. After receiving the request, each edge node calculates the idle resources of its device. The idle resources include computing resource availability, communication bandwidth margin, and sensor load. Generate a response message based on the idle resources. The response message includes: the type and quantity of available required resources and its own status label. The own status label includes high reliability nodes and low load nodes.

[0100] Step S204: Collect all response messages and calculate the weighted score of each edge node that sends a response message. The weighted score is calculated as: (computing resources × computing power weight + bandwidth × bandwidth weight) × reliability coefficient / response delay. The response delay is the time from when each edge node receives the request until it sends a response message. Arrange the weighted scores in descending order to form a node sequence. Send a joint collaboration instruction to each edge node in the order of the node sequence from top to bottom. The edge node that first confirms the response to the joint collaboration instruction is designated as the collaboration node, and the collaboration node is assigned to run the analysis task.

[0101] The calculation process for calculating the contribution weight of the edge nodes in step S3 is as follows:

[0102] Step S301: Extract the record of the collaborative instructions completed by the edge node as a collaborating node, and obtain the contribution data of the edge node in the collaborative instructions. The contribution data includes the amount of computing resources invested, the bandwidth usage value, and the response timeliness. The response timeliness is the delay time from receiving the request to executing the collaborative instructions.

[0103] The contribution data is normalized by normalizing the data on computing resources, bandwidth usage, and response timeliness into scores in the range of 0 to 1. The computing resources and bandwidth ratios are directly mapped to the actual values, and the response latency score increases as the latency decreases.

[0104] The scores for computing resources, bandwidth usage, and response timeliness are weighted and summed to generate a resource input score, with computing resources having the highest weight, followed by communication bandwidth, and response delay having the lowest weight.

[0105] Step S302: Select evaluation indicators based on the fault type:

[0106] Overload-related faults: Measured by the percentage of load reduction and the number of affected devices;

[0107] Insulation-related faults: Indicators include the percentage decrease in discharge intensity and the lead time of the early warning.

[0108] After normalizing the performance of each edge node under different indicators, the weighted sum is calculated to generate the fault suppression contribution value. In overload faults, the weight of load reduction is higher than the impact range, and in insulation faults, the weight of discharge suppression is higher than the warning timeliness.

[0109] Step S303: Based on the current stability level of the power grid, set the integration ratio of resource input and fault suppression: the lower the stability, the higher the contribution of fault suppression.

[0110] The resource input score and the fault suppression contribution score are combined proportionally to generate a comprehensive contribution score.

[0111] The overall contribution of each node is converted into a percentage of the total contribution, which is then used as the weight for the allocation of rights and interests in subsequent collaboration cycles.

[0112] For example: Switch cabinet A's temperature soared to 90℃ (threshold 80℃) due to excessive load.

[0113] A asked his neighbors B and C for help: "I need more bandwidth to upload data, who is willing to help me?"

[0114] B responded: "I can provide the bandwidth, but I need you to prioritize analyzing the vibration data for me next time."

[0115] C is not responding (its own load has reached 90%).

[0116] Collaborative execution:

[0117] B allocates its free bandwidth to A, and A quickly uploads data to the cloud for analysis.

[0118] The cloud-based model determines that the load on line A needs to be reduced and automatically switches some of the current to other lines.

[0119] Contribution Allocation:

[0120] Of the benefits of avoiding power outages, A (identifying the problem) accounts for 40%, and B (providing resources) accounts for 60%.

[0121] In step S4, the process of executing the optimal resource allocation scheme is as follows: select the corresponding level channel to send control commands according to the urgency of the fault type; after receiving the command, the cooperating node executes the analysis task and monitors the execution status of the command in real time.

[0122] Example: Each switch cabinet trains a small AI model (such as a fault prediction model) using its own data (such as temperature and current).

[0123] Encrypted upload: Only the model's "learning results" (gradient parameters) are encrypted and uploaded to the cloud; the original data is not uploaded.

[0124] Global integration: The learning results of all models are aggregated in the cloud to generate a more powerful global model, which is then redistributed to all switch cabinets;

[0125] When cabinet A detects overheating, it needs more bandwidth to upload data.

[0126] If the load on cabinet B is low at this time, it can sell its idle bandwidth to cabinet A in exchange for priority assistance from cabinet A when cabinet B needs it in the future.

[0127] Record contributions: In each collaboration, record the resources invested by each switch cabinet (such as how much computing power is provided).

[0128] Calculate weights: Use the Shapley value algorithm (a classic method in cooperative game theory) to allocate payouts based on historical contributions.

[0129] Example:

[0130] The cooperation of switchgear A, B, and C prevented a power outage:

[0131] A provides key data (contributing 50%), B provides computing power (30%), and C provides bandwidth (20%).

[0132] Benefits (such as savings in maintenance costs) are distributed in a 5:3:2 ratio.

[0133] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An intelligent high-voltage switchgear monitoring system based on the Internet of Things, characterized in that: The monitoring system includes: Distributed model training module: Edge nodes are deployed on multiple high-voltage switchgear. Each edge node independently trains a prediction model using locally collected monitoring data and uploads the training results to the cloud in encrypted form. The cloud aggregates the training results uploaded by each edge node to generate a global model and distributes it to each edge node to update the local model. Dynamic resource negotiation module: Communicatively connected to the distributed model training module, used to establish cooperation rules among multiple devices, and dynamically adjust the allocation strategy of monitoring resources according to the real-time load status and fault risk level of each device. The monitoring resources include communication bandwidth, computing resources and task priority. In the dynamic resource negotiation module: The generation of resource allocation strategies includes the following steps: Construct a multi-dimensional state space that includes device load rate, health status of adjacent devices, and communication quality; Define a set of executable actions, including increasing the frequency of local data collection, requesting external computing support, or releasing idle resources; A reward mechanism will be established based on historical collaborative experience, prioritizing strategies that can both improve local monitoring accuracy and reduce the risk of global failures. Through iterative optimization, each device can autonomously achieve a balanced state of resource allocation. Low-latency collaborative network module: It uses time-sensitive communication protocol to connect each high-voltage switchgear, and supports the synchronous transmission of equipment status data and collaborative instructions; Contribution Measurement Module: Calculates the contribution weight of each edge node and dynamically allocates collaboration benefits based on the amount of resources invested and the fault suppression effect during the collaboration process. In the contribution metric module: The calculation logic for contribution weights includes: Statistics on the computing resources, communication bandwidth, and amount of key early warning data provided by each device during the collaboration process; Evaluate the improvement in fault suppression effectiveness before and after the equipment participates in the collaboration; The basic contribution value is dynamically adjusted based on resource cost weights and fault severity levels. The revised contribution values ​​are mapped proportionally to resource revenue allocation weights.

2. The intelligent high-voltage switchgear monitoring system based on the Internet of Things according to claim 1, characterized in that: In the distributed model training module: The training results uploaded by each edge node are model parameter updates processed with homomorphic encryption to ensure that the original monitoring data does not leave the local area; The cloud uses an adversarial feature alignment method to eliminate the impact of differences in data distribution between different devices on the global model; Model compression technology is used to convert complex models generated in the cloud into lightweight versions that are adapted to the computing capabilities of edge nodes.

3. The intelligent high-voltage switchgear monitoring system based on the Internet of Things according to claim 1, characterized in that: The low-latency cooperative network further includes: A device trustworthiness assessment mechanism based on historical communication records automatically downgrades the weight of nodes that frequently fail to transmit. A chain-based data verification method is used to ensure the integrity and authenticity of collaborative instructions.

4. A method for monitoring intelligent high-voltage switchgear based on the Internet of Things, applied to the intelligent high-voltage switchgear monitoring system based on the Internet of Things as described in any one of claims 1-3, characterized in that: The monitoring method includes the following steps: Step S1: An edge node is deployed for each high-voltage switchgear. Each edge node uses local data to train and generate a local monitoring model. The local monitoring model is then encrypted and uploaded to the cloud to generate a global model. Step S2: When any edge node detects an abnormal situation, it initiates a multi-device resource negotiation request and selects the optimal resource allocation scheme based on the real-time status data of each edge node. Step S3: Quantify the resource input and fault suppression contribution of each edge node in the collaboration process, calculate the contribution weight of each edge node, and dynamically adjust the subsequent collaboration rights allocation. Step S4: Send the optimal resource allocation plan to the edge node where the abnormal situation occurs and execute it.

5. The method for monitoring intelligent high-voltage switchgear based on the Internet of Things according to claim 4, characterized in that: Step S1 includes: the edge node is a local intelligent computing device deployed on the high-voltage switchgear, used to perform local detection data processing, local monitoring model training and collaborative decision-making tasks, and to interact with the cloud system. The local intelligent computing device includes a data acquisition interface, an embedded processor and a communication unit. The input to the local monitoring model is the local detection data of the high-voltage switchgear, and its output is the fault type of the high-voltage switchgear; the local detection data includes electrical parameters, temperature parameters, mechanical parameters, insulation parameters, and environmental parameters; the fault types include overload, insulation aging, and partial discharge. The global model is used to integrate the feature patterns of local monitoring data from each edge node to form unified monitoring knowledge, and the global model is updated periodically.

6. The method for monitoring intelligent high-voltage switchgear based on the Internet of Things according to claim 4, characterized in that: Step S2 includes: Step S201: When the local detection data of a certain edge node triggers the threshold, it is marked as an abnormal situation, and a multi-device resource negotiation request is issued; Step S202: After receiving a multi-device resource negotiation request from an edge node, the cloud obtains the fault type of the abnormal situation and generates a negotiation request message, which includes: Required resource types: computing resources, communication bandwidth, sensor sampling frequency; Urgency level: classified according to the type of fault; Expected collaboration duration: The time window for failure development estimated based on the predictive model; Step S203: Send a request to adjacent edge nodes and the cloud through the low-latency collaborative network. After receiving the request, each edge node calculates the idle resources of its device. The idle resources include computing resource availability, communication bandwidth margin, and sensor load. Generate a response message based on the idle resources. The response message includes: the type and quantity of available required resources and its own status label. The own status label includes high reliability nodes and low load nodes. Step S204: Collect all response messages and calculate the weighted score of each edge node that sends a response message. The weighted score is calculated as: (computing resources × computing power weight + bandwidth × bandwidth weight) × reliability coefficient / response delay. The response delay is the time from when each edge node receives the request until it sends a response message. Arrange the weighted scores in descending order to form a node sequence. Send a joint collaboration instruction to each edge node in the order of the node sequence from top to bottom. The edge node that first confirms the response to the joint collaboration instruction is designated as the collaboration node, and the collaboration node is assigned to run the analysis task.

7. The method for monitoring intelligent high-voltage switchgear based on the Internet of Things according to claim 4, characterized in that: The calculation process for calculating the contribution weight of the edge nodes in step S3 is as follows: Step S301: Extract the record of the collaborative instructions completed by the edge node as a collaborating node, and obtain the contribution data of the edge node in the collaborative instructions. The contribution data includes the amount of computing resources invested, the bandwidth usage value, and the response timeliness. The response timeliness is the delay time from receiving the request to executing the collaborative instructions. The contribution data is normalized by normalizing the data on computing resources, bandwidth usage, and response timeliness into scores in the range of 0 to 1. The computing resources and bandwidth ratios are directly mapped to the actual values, and the response latency score increases as the latency decreases. The scores for computing resources, bandwidth usage, and response timeliness are weighted and summed to generate a resource input score, with computing resources having the highest weight, followed by communication bandwidth, and response delay having the lowest weight. Step S302: Select evaluation indicators based on the fault type: Overload-related faults: Measured by the percentage of load reduction and the number of affected devices; Insulation-related faults: Indicators include the percentage decrease in discharge intensity and the lead time of the early warning. After normalizing the performance of each edge node under different indicators, the weighted sum is calculated to generate the fault suppression contribution value. In overload faults, the weight of load reduction is higher than the impact range, and in insulation faults, the weight of discharge suppression is higher than the warning timeliness. Step S303: Based on the current stability level of the power grid, set the integration ratio of resource input and fault suppression: the lower the stability, the higher the contribution of fault suppression. The resource input score and the fault suppression contribution score are combined proportionally to generate a comprehensive contribution score. The overall contribution of each node is converted into a percentage of the total contribution, which is then used as the weight for the allocation of rights and interests in subsequent collaboration cycles.

8. The method for monitoring intelligent high-voltage switchgear based on the Internet of Things according to claim 4, characterized in that: In step S4, the process of executing the optimal resource allocation scheme is as follows: select the corresponding level channel to send control commands according to the urgency of the fault type; after receiving the command, the cooperating node executes the analysis task and monitors the execution status of the command in real time.

Images