Method and system for automatically checking relay protection setting value of power plant based on edge calculation

Abstract

The invention relates to the field of electric power, and particularly discloses a power plant relay protection setting value automatic checking method and system based on edge computing, and the method comprises the steps: carrying out the credibility self-evaluation of an operation state of an edge computing node through the edge computing node, and generating a credibility grade label; the node dynamically selects and executes a strategy of complete checking, simplified checking or calling of a cache conclusion according to the label, and outputs a graded checking result with a mark; the aggregation gateway or the master station performs differentiation processing according to the result mark, and starts collaborative redundancy check for a low-credibility conclusion; performing closed-loop optimization on the node evaluation model and the global algorithm by utilizing arbitration data generated by cooperative verification; the system comprises a node credibility self-evaluation module, a grading check decision and execution module, a constant value processing and collaborative scheduling module and a group collaboration and model evolution module. According to the invention, by introducing an execution main body credibility evaluation and dynamic cooperative verification mechanism, endogenous safety, decision robustness and continuous self-evolution of relay protection constant value verification are realized.

Description

Technical Field

[0001] This invention relates to the field of power technology, and specifically to a method and system for automatic verification of relay protection settings in power plants based on edge computing. Background Technology

[0002] The accuracy of relay protection settings is the first line of defense for ensuring the safe and stable operation of the power grid. With the high proportion of new energy integration and the increasing complexity of the power grid structure, traditional setting verification methods face severe challenges. Current mainstream technologies mainly rely on two modes: one is periodic manual verification and offline calculation. This method is slow to respond, has a large workload, cannot adapt to real-time changes in power grid operation, and is difficult to detect hidden setting defects caused by parameter input errors or model simplification; the other is an online verification system based on a cloud master station. Although it improves timeliness, its core calculations are concentrated in the cloud, resulting in large data transmission delays and high communication bandwidth pressure. It also poses a risk of service interruption when the plant network is abnormal and cannot meet the stringent millisecond-level real-time requirements of relay protection.

[0003] In recent years, edge computing technology has been introduced into power systems, aiming to decentralize computing power to reduce latency. However, existing edge computing-based solutions generally suffer from two fundamental limitations: First, they blindly trust edge nodes, treating them as ideal computing terminals, completely ignoring the fact that in real industrial environments, edge nodes may be in a "sub-healthy" state due to hardware aging, software anomalies, resource contention, or network disturbances, resulting in unreliable verification conclusions. This risk of "unreliable execution subject" has not been systematically identified and addressed by any existing solution. Second, nodes operate in isolation, with each edge node only performing local computation, lacking mechanisms for intelligent collaboration and cross-verification with other nodes or higher-level systems when necessary. When a single node fails or its conclusions are questionable, a "blind spot" appears in the verification tasks within its protection scope, resulting in insufficient overall system robustness.

[0004] Therefore, there is an urgent need for a new generation of automatic verification method and system for relay protection settings that can deeply integrate the real-time advantages of edge computing, proactively ensure the reliability of the computing entity, and achieve intelligent collaboration and self-evolution. Summary of the Invention

[0005] The purpose of this invention is to provide an automatic verification method and system for power plant relay protection settings based on edge computing, so as to solve the problems of implicit verification errors in relay protection settings and single-point vulnerability of the system caused by blindly trusting the reliability of edge computing nodes and lacking inter-node coordination mechanisms in the prior art.

[0006] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution: The automatic verification method for power plant relay protection settings based on edge computing includes the following steps: S1. Edge computing nodes use a credibility assessment model to evaluate their own status and generate a node credibility level; S2. The edge computing nodes perform verification according to the node trust level selection strategy, and generate hierarchical verification results with corresponding tags; S3. Based on the markers in the hierarchical verification results, the system performs differentiated value processing and initiates collaborative verification; S4. Utilize the collaborative arbitration conclusions and process data generated by collaborative verification to provide feedback for optimizing the evaluation model and global verification algorithm.

[0007] As a preferred embodiment of the present invention, S1 specifically includes: S11. Before each setpoint verification calculation, the edge computing nodes within the power plant automatically trigger a self-evaluation process; S12. Edge computing nodes monitor their own hardware computing unit load rate, memory data consistency verification results, and network latency jitter values ​​with the upper-level master station or adjacent nodes in real time. S13. The node takes the hardware computing unit load rate, memory data consistency verification result and network latency jitter value as inputs and inputs them into the preset credibility evaluation model; the credibility evaluation model assigns dynamic weights to each indicator and uses a piecewise function to score them, and outputs a quantitative node credibility score after comprehensive calculation; S14. The node maps its credibility score to a high-credibility, medium-credibility, or low-credibility edge node self-evaluation result, i.e., a credibility level label, based on a preset score threshold range.

[0008] As a preferred embodiment of the present invention, S12 specifically includes: S121. Hardware processing unit load rate calculation: in: for The load rate of the hardware computing unit at any given time, and its range. ; They are respectively The actual computing resource usage of CPU, GPU, and FPGA at any given time; These represent the total computing resource capacity of the CPU, GPU, and FPGA, respectively. For the weight coefficients of the hardware units, satisfying Pre-configured according to the node hardware architecture; S122. Calculation of memory data consistency check results: in: for The memory data consistency check score at any given time, and its range. ; In the evaluation cycle The total number of CRC checks performed on critical memory regions; In the evaluation cycle Number of internal CRC check failures; In the evaluation cycle The number of successful internal CRC checks, and ; S123. Network latency jitter is quantified using the standard deviation of round-trip time (RTT): Among them, average round-trip delay for: ; In the formula: for Network latency jitter value at any given time. For the first Round-trip delay sample value of each heartbeat data packet In the evaluation cycle Heart rate sampling (Recommended value: 20) To evaluate the average round-trip time within the evaluation period.

[0009] As a preferred embodiment of the present invention, S2 specifically includes: S21. The edge computing node reads the edge node self-evaluation result generated in step S14, parses the confidence level label in it, and then selects the corresponding verification execution strategy; S22. If the confidence level label is high confidence, the node calls the complete local real-time setting calculation engine, performs a complete verification calculation including all protection setting items based on the latest real-time power grid operation data, and generates a standard verification conclusion including specific setting deviation conclusions and adjustment suggestions. S23-S24. If the label is medium or low confidence, only a simplified verification of the critical protection setting is performed or the cached verification conclusion of the most recent high confidence state is called, and an uncertainty flag corresponding to the confidence level is attached to the generated conclusion. S25. Finally, output a unified hierarchical verification result, the content of which may be an unmarked standard verification conclusion, or a simplified conclusion or cached conclusion with uncertainty marking.

[0010] As a preferred embodiment of the present invention, S23-S24 specifically includes: S23. If the confidence level label is medium confidence, the node starts a simplified verification mode, and only selects key protection settings from the complete set value list for rapid calculation. The key protection settings include at least the main protection instantaneous trip setting value and the backup protection time limit value related to system stability. The generated simplified conclusion will be automatically marked with confidence downgrade and incomplete calculation uncertainty, forming a graded verification result. S24. If the credibility level label is low credibility, the node does not perform real-time calculation, but instead calls the most recently generated and cached historical standard verification conclusion from the local secure storage area under a high credibility state; after the call, a credibility downgrade and uncertainty mark for using cached data are added to the conclusion to form a graded verification result.

[0011] As a preferred embodiment of the present invention, S3 specifically includes: S31. The factory-level edge computing aggregation gateway or cloud-based fixed-value management master station receives the hierarchical verification results from all edge computing nodes in the network and parses the result content and marking status. S32. For the hierarchical verification results confirmed after parsing without uncertainty markers, if the results indicate that the set value needs to be adjusted, the aggregation gateway or the main station shall, according to the preset security policy, directly issue a set value modification instruction to the corresponding protection device, or grant the edge computing node that generated the hierarchical verification results the permission to automatically perform the adjustment. S33. For graded verification results that are confirmed to have uncertainty markers after parsing, a two-level processing mechanism is initiated; S34. The above-mentioned redundancy verification mechanism will continue until the source node completes self-check and recovery and re-evaluates and generates a high-confidence level label through step S1. Only then will the system remove the pending confirmation status and terminate the redundancy verification process initiated for it.

[0012] As a preferred embodiment of the present invention, S33 specifically includes: Send a mandatory deep self-check and system recovery command to the source edge computing node that generated the result, and at the same time mark all the verification conclusions reported by the node this time as pending confirmation in its internal database and suspend its automatic execution permission. Dynamically adjust the verification task allocation strategy for the power grid protection domain: immediately issue redundant verification tasks for the protection range of the faulty node to one or more edge computing nodes that are physically adjacent to the node and whose current trust level label is high trust; if there are no available high trust adjacent nodes, the cloud-based value management master station will directly perform synchronous verification calculations on the protection range.

[0013] As a preferred embodiment of the present invention, S4 specifically includes: S41. The edge computing aggregation gateway triggers collaborative verification of a specific protection device or associated protection link under at least one of the following conditions: reaching a preset periodic verification cycle; receiving a graded verification result with an uncertainty marker; or the system detects a significant change in the operating mode of the protection range; S42. The aggregation gateway selects two or more edge computing nodes from the node list that have data collection capabilities and verification permissions for the target protection scope and whose current trust level label is the highest, and uses them as members of the collaborative verification group. S43-S44. How does the aggregation gateway assign dynamic confidence weights to the conclusions of each node based on the state micro-change data and historical confidence records before and after the execution of this task, and generate collaborative arbitration conclusions based on a weighted arbitration mechanism? S45. The aggregation gateway uses the dynamic confidence weight to perform weighted arbitration on the independent verification conclusions of each node, and generates the final collaborative arbitration conclusion; S46. The collaborative arbitration conclusion and the entire process data of this collaborative verification are used in the following two aspects: feeding back to the local credibility assessment model of the relevant nodes, using the deviation between the node conclusion and the collaborative arbitration conclusion as a correction parameter to update the internal weights of the model; and uploading the verification data package as a training sample with high confidence labels to the main station analysis system in the cloud for iterative optimization of the global fixed value verification algorithm model.

[0014] As a preferred embodiment of the present invention, S43-S44 specifically includes: S43. For each verification group member The aggregation gateway collects its state change data and calculates the normalized deviation: Task response delay deviation: Average response time: Memory usage deviation: in: For nodes The relative deviation of task response latency; For nodes The response time of this collaborative verification task; This represents the average response time of all nodes within the collaborative verification group. For nodes The deviation in memory usage; For nodes Peak memory usage during the verification calculation process; For nodes The normal baseline value for memory usage during the verification calculation process; For nodes Total memory capacity; S44. Calculate each node Dynamic confidence weights : Among them, historical credibility factor : In the formula: For nodes Dynamic confidence weights in collaborative verification; This is the penalty coefficient for small state changes, used to control the strength of the impact of response latency and memory bias on the weights; For nodes The current credibility score; For nodes The historical credibility factor reflects its long-term credibility performance; This is the length of the historical data sampling window; For the first The timestamp of the previous historical verification; This represents the execution time of the current collaborative verification task; This is a historical credibility decay factor used to reduce the weight of outdated historical records.

[0015] An edge computing-based automatic verification system for power plant relay protection settings is used to implement an edge computing-based automatic verification method for power plant relay protection settings, including: The node trustworthiness self-assessment module is deployed in each edge computing node within the power plant to monitor and assess the node's own operating status and generate a quantified node trustworthiness level. The hierarchical verification decision and execution module is deployed in the edge computing node and communicates with the node credibility self-evaluation module. It is used to select and execute the corresponding verification strategy according to the node credibility level and generate hierarchical verification results with or without uncertainty markers. The setpoint processing and collaborative scheduling module is deployed at the plant-level edge computing aggregation gateway or cloud-based setpoint management master station. It is used to receive and parse the hierarchical verification results of each node, execute differentiated setpoint processing instructions according to the marked status, and initiate collaborative verification tasks for low-confidence conclusions. The group collaboration and model evolution module is deployed on the edge computing aggregation gateway and the cloud-based value management master station. It is used to organize multiple nodes to perform collaborative verification, generate collaborative arbitration conclusions, and use this conclusion data to back-optimize the evaluation model and global verification algorithm model of the node credibility self-evaluation module.

[0016] Compared with the prior art, the present invention has the following advantages: 1. Building inherent security: Through an innovative edge node trustworthiness self-assessment mechanism, unreliable computing nodes are proactively identified and isolated, eliminating the hidden verification risks caused by abnormal execution entities from the source.

[0017] 2. Enhance decision robustness: The flexible verification architecture based on credibility-driven dynamic triggering of local hierarchical execution and group collaborative arbitration makes the conclusions independent of a single point, significantly enhancing fault tolerance and reliability in complex scenarios.

[0018] 3. Achieve system self-evolution: Utilize the closed-loop data flow generated by collaborative verification to continuously optimize the node evaluation model in reverse and train the global algorithm in forward, enabling the system to learn and adapt autonomously during operation. Attached Figure Description

[0019] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the method described in Embodiment 1 of the present invention.

[0021] Figure 2 This is an architecture diagram of the credibility assessment model in the method described in Embodiment 1 of the present invention.

[0022] Figure 3 This is a framework diagram of the system described in Embodiment 2 of the present invention.

[0023] Figure 4 This is a comparison diagram of the effects of the method of the present invention and the prior art method in Embodiment 3 of the present invention. Detailed Implementation

[0024] 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.

[0025] The concepts involved in this application will first be described with reference to the accompanying drawings. It should be noted that the following descriptions of various concepts are only for the purpose of making the content of this application easier to understand and do not constitute a limitation on the scope of protection of this application; furthermore, the embodiments and features in the embodiments of this application can be combined with each other unless otherwise specified. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] Example 1 like Figure 1-2 As shown, this invention provides a method and system for automatic verification of power plant relay protection settings based on edge computing, including the following steps: S1. Edge computing nodes use a credibility assessment model to evaluate their own status and generate a node credibility level; specifically including: S11. Before each setpoint verification calculation, the edge computing nodes in the power plant automatically trigger a self-evaluation process to ensure that the reliability of the node status is confirmed in real time before the verification calculation.

[0027] S12. Edge computing nodes obtain the hardware computing unit load rate by calling the local operating system interface, obtain the memory data consistency check result by periodically performing cyclic redundancy checks on key memory areas, and calculate the network latency jitter value by sending synchronization heartbeat data packets to the superior master station or adjacent nodes and statistically analyzing response time fluctuations. Specifically: S121. Hardware processing unit load rate calculation: in: for The load rate of the hardware computing unit at any given time, and its range. ; They are respectively The actual computing resource usage of CPU, GPU, and FPGA at any given time; These represent the total computing resource capacity of the CPU, GPU, and FPGA, respectively. For the weight coefficients of the hardware units, satisfying The configuration is pre-configured based on the node's hardware architecture.

[0028] S122. Calculation of memory data consistency check results: in: for The memory data consistency check score at any given time, and its range. ; In the evaluation cycle The total number of CRC checks performed on critical memory regions; In the evaluation cycle Number of internal CRC check failures; In the evaluation cycle The number of successful internal CRC checks, and .

[0029] S123. Network latency jitter is quantified using the standard deviation of round-trip time (RTT): Among them, average round-trip delay for: ; In the formula: for Network latency jitter value at any given time. For the first Round-trip delay sample value of each heartbeat data packet In the evaluation cycle Heart rate sampling (Recommended value: 20) To evaluate the average round-trip time within the evaluation period.

[0030] S13. The node takes the hardware computing unit load rate, memory data consistency verification result, and network latency jitter value as inputs to a pre-set credibility assessment model. The credibility assessment model assigns dynamic weights to each indicator and uses a piecewise function for scoring. After comprehensive calculation, it outputs a quantified node credibility score. Specifically: S131. Node Trustworthiness Score Calculated using a weighted comprehensive scoring model: in: for The node credibility score at time point, with a range of values. ; Represent Input indicator values; These are dynamic weighting coefficients for hardware load, memory consistency, and network jitter, respectively, satisfying... ; These are the piecewise scoring functions for the three indicators, and they output normalized score values.

[0031] S132. Weighting Adaptive adjustment based on node historical status: in: For the first The historical variance of an indicator within a sliding time window reflects the stability of the indicator; Weight sensitivity adjustment coefficient (recommended value) ), used to control the strength of the influence of variance on weights. S133. The piecewise scoring function is defined as follows: Hardware load rate scoring function : Memory consistency scoring function : Network jitter scoring function : S14. Based on a preset score threshold range, the node's credibility score is mapped to the edge node's self-evaluation result, i.e., a credibility level label, which is a high-credibility, medium-credibility, or low-credibility score. Specifically: The credibility level label mapping function is: in: The final output credibility level label (enumerated values: high credibility / medium credibility / low credibility); Assign a node credibility score; This is a rank mapping function that converts continuous scores into discrete rank labels; The high confidence threshold is set at 0.85 (recommended value). The low confidence threshold is set (recommended value: 0.60).

[0032] S2. Edge computing nodes select a verification strategy based on node trust level to generate hierarchical verification results with corresponding labels; specifically including: S21. The edge computing node reads the edge node self-evaluation result generated in step S14, parses the confidence level label, and then... Select the execution strategy path The function for verifying the execution strategy selection is: in: Credibility level labels; Select a function for the strategy and map the trust level label to the specific execution path; To ensure a complete verification of the execution path; To simplify the verification process; To cache the execution path of the call.

[0033] S22. If the credibility level label is high credibility, then execute the full verification execution path. Specifically, the node calls the complete localized real-time setting calculation engine, performs a complete verification calculation including all protection setting items based on the latest real-time power grid operation data, and generates a standard verification conclusion that includes specific setting deviation conclusions and adjustment suggestions.

[0034] S23. If the confidence level label is Medium Confidence, then execute the simplified verification execution path. Specifically: S231. In the simplified verification mode, from the complete protection setting set Selecting key subsets : Among them, the keyness scoring function Defined as: In the formula: This is the complete set of setting values ​​for all protection devices within the power plant. This is a subset of the key protection settings selected through screening. For the first Protection settings (such as generator differential protection instantaneous overcurrent setting, line backup protection time setting, etc.); For constant terms Keyness score, range of values ; The keyness screening threshold (recommended value 0.75) meets the following requirements. The fixed values ​​were included in the simplified verification; The main protection identifier function, if If it belongs to the main protection, the value is 1; otherwise, it is 0. For instantaneous overcurrent protection flag function, if If the value is a quick-break setting, it is 1; otherwise, it is 0. For the system stability-related identification function, if If the system's stability is affected, the value is 1; otherwise, it is 0. Let be the weighting coefficient, satisfying This is used to quantify the importance of various types of protection; For constant terms The historical operation frequency weight is calculated based on the normalized number of actual operations of the protection in the past year, reflecting its actual significance in the operation of the power grid.

[0035] S232. The generated simplified conclusions will be automatically marked with uncertainty tags indicating reduced credibility and incomplete calculations, forming a graded verification result. The uncertainty tags in the graded verification result... Credibility level labels and verification integrity factor Joint decision: in: This is an uncertainty-labeled vector containing three components; Generate functions for uncertainty labeling; The verification completeness factor represents the proportion of the actual verification set of values ​​to the complete set of values. This is a confidence level label; a higher value indicates lower confidence. ; The integrity marker component indicates the degree of coverage of the verification settings. ; For a complete set of protection settings The total number of constants in the data; For critical protection set subset The number of constant values ​​in the data; To mark the timeliness of the data, record the timestamp of the conclusion's generation or caching. ; The timestamp of the current verification execution time; The original generation timestamp of the cached conclusion being invoked.

[0036] S24. If the trust level label is low trust, then execute the cached call execution path. Specifically: In a low-trust state, nodes do not perform real-time computation, but instead retrieve data from local secure storage. Recall the most recent historical standard verification conclusion generated and cached under a high-trust state. : in: This is the set of historical verification conclusions stored in the local secure storage area. For a single cached historical verification conclusion item; The selected optimal cache conclusion; For caching conclusions Credibility score of the node corresponding to the generation time (from step S13); For caching conclusions The original generation timestamp; The timestamp of the current moment; Let be the weighting coefficient, satisfying This is important for balancing historical reliability with data timeliness; The time decay factor (recommended value 0.1) controls the rate at which the influence of outdated data decays. After the call, a confidence downgrade and uncertainty flag indicating the use of cached data are added to the conclusion to form a graded verification result.

[0037] S25. Final output of graded verification results Use structured data format: The result data body The content varies depending on the execution path: In the formula: It is a structured data tuple representation; The data structure for the final output hierarchical verification results; The verification result data includes core information such as the set value deviation and adjustment suggestions; This is the deviation vector for all fixed-value items under the complete verification mode; To simplify the deviation vector of key fixed-value terms in the verification mode; This is a set of setpoint adjustment suggestions under the complete verification mode; A set of key setpoint adjustment suggestions to simplify the verification mode.

[0038] S3. Based on the markers in the hierarchical verification results, the system performs differentiated value processing and initiates collaborative verification; specifically including: S31. A factory-level edge computing aggregation gateway or cloud-based fixed-value management master station receives hierarchical verification results from all edge computing nodes in the network. And analyze and extract the uncertainty marker vector. And determine the marked state: in: The result is marked as the state determination result (0 indicates no uncertainty mark, 1 indicates uncertainty mark). For analytical decision functions; The graded verification results from step S25; Uncertainty label vector The confidence level label component (from step S23, 0 indicates high confidence, 1 indicates medium confidence, 2 indicates low confidence).

[0039] S32. For the graded verification results confirmed after analysis to be without uncertainty markers, i.e. If the result indicates that a setpoint adjustment is needed, the aggregation gateway or main station will generate a processing decision. : Among them, the maximum constant deviation The calculation is as follows: In the formula: The processing decisions generated by the system; To process decision functions; This represents the maximum relative percentage deviation of all specified values ​​in the verification results. For the first The actual operating value of the protection setting; For the first The setting value of the protection setting; This represents the total number of setpoints (when performing a full check) or the number of critical setpoints (when performing a simplified check). Safety deviation threshold (recommended value) (This can be dynamically adjusted according to the protection type). When the maximum deviation does not exceed this threshold, the node can be authorized to handle it autonomously. The current credibility level label for edge nodes used to generate hierarchical verification results.

[0040] S33. For the hierarchical verification results that are confirmed to have uncertainty markers after analysis, i.e. The result then triggers a two-level processing mechanism: S331. Generate node control instructions : Send a mandatory deep self-check and system recovery command to the source edge computing node that generates the hierarchical verification results, and at the same time mark all verification conclusions reported by the node this time as pending confirmation in its internal database and suspend its automatic execution permission; S332. Dynamically adjust the verification task allocation strategy for the power grid protection domain: Immediately issue redundant verification tasks for the protection range of the faulty node to one or more edge computing nodes that are physically adjacent to the node and whose current trust level label is high trust; if there are no available high trust adjacent nodes, the cloud-based value management master station will directly perform synchronous verification calculations on the protection range.

[0041] S34. The source node receives Afterwards, a self-check recovery is performed, and upon completion, step S1 is re-executed to generate a new credibility level label. System monitoring recovery status : Self-inspection integrity score The calculation is as follows: Redundant verification process termination conditions: in: The trust level label is regenerated after the source node completes its self-check and recovery. This is a node recovery status flag (1 indicates recovery, 0 indicates non-recovery); The integrity score for node depth self-check, with a value range of... ; Set a threshold for self-test integrity (recommended value 0.95); This represents the total number of test items included in the deep self-test. For the first The weighting coefficients of the self-inspection items meet the requirements. ; For the first The execution result of each self-check item (1 indicates pass, 0 indicates failure); Terminate is the marker for terminating the redundancy verification process. (represents logical OR). The redundant verification process has been ongoing; Set a timeout threshold for the redundancy verification process (recommended value: 300 seconds, which can be dynamically adjusted according to the protection level).

[0042] S4. Utilize the collaborative arbitration conclusions and process data generated by collaborative verification to provide feedback for optimizing the evaluation model and global verification algorithm; specifically including: S41. The edge computing aggregation gateway triggers collaborative verification of a specific protection device or associated protection link under at least one of the following conditions: Triggering condition function Output a logical truth value if any of the following conditions are met: in: This is the collaborative verification trigger flag (1 indicates triggering, 0 indicates not triggering); This is the current system timestamp; This is the timestamp of the last collaborative verification completion. The preset periodic verification cycle (recommended value: 3600 seconds); The uncertainty labeling result (from step S31, 1 indicates labeled); To quantify changes in the operating mode within the protection scope, topology analysis is used. Set the threshold for triggering changes in operating mode (recommended value: 0.3). For logical OR operator.

[0043] S42. The aggregation gateway collects data from the set of available nodes. Optimal verification group selection : Constraints: in: The final set of members of the collaborative verification group. ; This represents the complete set of currently available edge computing nodes. Number the candidate nodes; For nodes The current credibility score (from step S13); For the function to verify permission identifiers, if the node If the user has verification authority over the target protection scope, the value is 1; otherwise, it is 0. For nodes Network topology distance to the target protection device; This is a function to identify data acquisition capabilities; if the node... Capable of collecting target protection range The real-time data is set to 1 if it is 1, and 0 otherwise. To select weighting coefficients, satisfying This is used to balance the importance of trustworthiness, permissions, and distance; A subset of candidate nodes For set The number of edge computing nodes included in it; Minimum number of nodes in the collaborative verification group (recommended value: 2); This is the maximum number of nodes in the collaborative verification group (recommended value: 5, to avoid excessive communication overhead due to too many nodes). The power grid range corresponding to the target protection device or protection link in this collaborative verification.

[0044] S43. The aggregation gateway broadcasts the verification task to all verification group members and synchronously shares the latest real-time operational data snapshot within the protection scope; each member node independently performs the setpoint verification calculation based on this shared data and returns its independent verification conclusion to the aggregation gateway; specifically: For each member of the verification group The aggregation gateway collects its state change data and calculates the normalized deviation: Task response delay deviation: Average response time: Memory usage deviation: in: For nodes The relative deviation of task response latency; For nodes The response time of this collaborative verification task; This represents the average response time of all nodes within the collaborative verification group. For nodes The deviation in memory usage; For nodes Peak memory usage during the verification calculation process; For nodes Normal baseline value of memory usage during the verification calculation process (from historical performance profile). For nodes Total memory capacity.

[0045] S44. After receiving the conclusions from each node, the aggregation gateway combines the state change data of each node during the execution of this verification task (including task response delay and memory usage deviation during the calculation process) and its historical reliability records to calculate the reliability of each node. Dynamic confidence weights : Among them, historical credibility factor : In the formula: For nodes Dynamic confidence weights in collaborative verification ); Penalty coefficient for data with small state changes (recommended value) ), used to control the intensity of the impact of response latency and memory bias on weights; For nodes The current credibility score (from step S13); For nodes The historical credibility factor reflects its long-term credibility performance; The length of the historical data sampling window (recommended value 50, representing the 50 most recent verification records); For the first The timestamp of the previous historical verification; This represents the execution time of the current collaborative verification task; This is a historical credibility decay factor (recommended value 0.05), used to reduce the weight of old historical records.

[0046] S45. The aggregation gateway uses dynamic confidence weights. Independent verification conclusions for each node A weighted arbitration process is conducted to generate a final, coordinated arbitration conclusion. : Arbitration of setpoint deviation (taking the setpoint deviation vector as an example): in: This is the overall deviation vector after arbitration; For nodes The vector of constant deviation obtained through independent verification; Arbitration is recommended (using a weighted voting mechanism): For each adjustment suggestion Calculate its weighted support: in: Recommendations for adjustment The weighted support score; For nodes Recommendations The voting function, if the node The suggestions include The value is 1 if it is 1, otherwise it is 0. The final arbitration conclusion includes satisfying The proposed adjustments include, among which To adjust the recommendation adoption threshold (recommendation value 0.6), recommendations with a weighted support exceeding this threshold were included in the final conclusion.

[0047] S46. The collaborative arbitration conclusion and the data from the entire collaborative verification process were used for the following two purposes: S461. Feedback to the local credibility assessment model of relevant nodes uses the deviation between the node's conclusion and the collaborative arbitration conclusion as a correction parameter to update its internal model weights: Node Local credibility assessment model parameters (From step S13) Update based on the co-validation deviation: Among the verification deviation The calculation is as follows: In the formula: For nodes The The updated values ​​of the evaluation indicator weights; For nodes The The weight of each evaluation indicator at the current moment; Update the learning rate for the model (recommended value 0.1) and control the update step size; For nodes The degree of deviation between the verification conclusion and the collaborative arbitration conclusion; For nodes Independently verified constant deviation vector; The comprehensive setpoint deviation vector generated for collaborative arbitration; The L2 norm (Euclidean distance) of the vector; Small constant (recommended value) ), to prevent the denominator from being zero; For reference weight values, the average weight of other high-confidence nodes in the collaborative verification group is usually taken.

[0048] S462. Send this verification data packet. As training samples with high confidence labels, they are uploaded to the main analysis system in the cloud for iterative optimization of the global constant value verification algorithm model. Where: input features Labeling Sample quality score ; In the formula: For the training sample data structure; Input feature vectors into the model; To share real-time data corresponding to hardware load, memory consistency, and network jitter metrics (collected during collaborative verification). The power grid operating status vector within the protection range (including real-time measured values ​​such as voltage, current, and power); Output labels for the model, namely the setpoint deviation vector in the collaborative arbitration conclusion. ; Score the sample quality (range of values) ), used for sample weighting during cloud training; The standard deviation of the dispersion of independent conclusions for all nodes reflects the degree of consistency within the group (the smaller the standard deviation, the higher the sample quality).

[0049] Example 2 like Figure 3 As shown, the edge computing-based automatic verification system for power plant relay protection settings is used to implement an edge computing-based automatic verification method for power plant relay protection settings, including: A node trustworthiness self-assessment module, deployed in each edge computing node within the power plant, is used to monitor and assess the node's own operational status and generate a quantified node trustworthiness level; specifically including: The status monitoring unit is used to collect data in real time, including the load rate of the node's hardware computing unit, the memory data consistency verification results, and the network latency jitter value. The credibility calculation unit, which has an embedded credibility evaluation model, is used to receive data from the status monitoring unit and calculate the node credibility score. The rating mapping unit is used to map the node credibility score to a credibility level label of high credibility, medium credibility, or low credibility according to a preset threshold, and output the edge node self-evaluation result.

[0050] The hierarchical verification decision and execution module, deployed in edge computing nodes, communicates with the node credibility self-assessment module. It selects and executes the corresponding verification strategy based on the node's credibility level, generating hierarchical verification results with or without uncertainty markers. Specifically, it includes: The strategy selection unit is used to read the credibility level label and determine the verification mode; The complete verification calculation unit is invoked when the label is high confidence to perform localized real-time fixed value calculation and generate standard verification conclusions; The simplified / cached verification unit is invoked when the label is medium or low confidence. It is used to perform simplified calculations of key values ​​or call historical cached conclusions, and automatically attaches the corresponding uncertainty flags. The result encapsulation unit is used to encapsulate the output of the above units into the hierarchical verification result.

[0051] The setpoint processing and collaborative scheduling module, deployed at the plant-level edge computing aggregation gateway or cloud-based setpoint management master station, is used to receive and parse the hierarchical verification results of each node, execute differentiated setpoint processing instructions based on the marked status, and initiate collaborative verification tasks for low-confidence conclusions; specifically including: The result parsing and routing unit is used to receive and parse the hierarchical verification results, and route them to different processing channels according to whether they contain uncertainty markers. The instruction issuing unit, connected to the result parsing and routing unit, is used to generate and issue setpoint adjustment instructions for unlabeled reliable results; The exception handling and redundancy scheduling unit, connected to the result parsing and routing unit, is used to perform operations such as sending self-check instructions to the source node, setting its conclusion to a pending confirmation state, and initiating redundancy verification tasks to adjacent high-trust nodes or the master station for tagged results.

[0052] The group collaboration and model evolution module, deployed on the edge computing aggregation gateway and the cloud-based value management master station, is used to organize multiple nodes to perform collaborative verification, generate collaborative arbitration conclusions, and use this conclusion data to back-optimize the evaluation model of the node credibility self-assessment module and the global verification algorithm model; specifically including: The collaborative task triggering and organizing unit, deployed on the edge computing aggregation gateway, is used to organize a collaborative verification team when conditions are met. The dynamic weighted arbitration unit is used to calculate dynamic confidence weights based on the real-time status micro-change data and historical records of the verification group members, and to perform weighted arbitration on the conclusions of each node to generate a collaborative arbitration conclusion. The model feedback optimization unit is used to feed back the deviation data between the collaborative arbitration conclusion and the node conclusion to the credibility evaluation model of the relevant node for parameter updates, and at the same time upload high-quality verification data samples to the cloud for optimization of the global algorithm model.

[0053] Example 3 1. Test Scenario and Configuration A gas-fired wind power hybrid power plant connected to a large-scale offshore wind power system was selected as the verification environment in a coastal area. The power grid structure of the plant is complex, and its fault characteristics are greatly affected by the fluctuations in the output of new energy sources.

[0054] The deployment plan is as follows: 12 edge computing nodes are deployed at the protection screens of 4 main transformers and the protection devices of 8 key collection lines in the plant; 1 edge computing aggregation gateway is deployed at the plant monitoring center; and a setpoint management master station is deployed on the group cloud platform.

[0055] The method of this invention is used to verify the improvement effect of the accuracy and reliability of the setpoint verification in scenarios involving switching of power grid operation modes and artificial injection of node faults.

[0056] 2. Specific Implementation Process Scenario Simulation: Simulates the working condition of "main transformer under maintenance, one collector line fault". At the same time, the memory utilization of node numbered Edge-07 (responsible for the protection of one collector line) is artificially increased to 95%, and the network latency jitter is increased to ±50ms to simulate its sub-health state.

[0057] Step S1 execution: All nodes initiate self-assessment before performing the verification. Edge-07 node detected a 98% load rate on its hardware computing unit, three error blocks were found in memory data consistency verification, and network latency jitter reached ±52ms. After inputting the above data into its local trustworthiness assessment model, the node trustworthiness score was calculated to be 42 points (out of 100). According to the mapping rules (≥80 for high trustworthiness, 60-79 for medium trustworthiness, <60 for low trustworthiness), Edge-07 generated a low-trust self-assessment result for the edge node. All other healthy nodes generated high-trust results.

[0058] Step S2 execution: The Edge-07 node automatically triggers a simplified / cached verification strategy based on the low-trust label. Instead of performing real-time computation, it retrieves a cached verification conclusion from its local secure storage (when it was in a high-trust state) from 5 minutes prior, appends a trust downgrade and uncertainty flag indicating the use of cached data to the conclusion, and generates its tiered verification result, which it then reports. The remaining healthy nodes perform full real-time computation, generate an unlabeled standard verification conclusion, and report it.

[0059] Step S3 execution: The aggregation gateway receives the results from each node. For the tagged results reported by Edge-07, the gateway immediately initiates two-level processing: it sends a forced self-check and restart command to Edge-07 and sets its conclusion status to pending confirmation. The strategy is dynamically adjusted, and redundancy verification tasks are issued to the physically adjacent and highly trusted Edge-05 and Edge-08 nodes, requiring a re-verification of the protection range of Edge-07. Within 2 milliseconds, the two highly trusted nodes return new verification conclusions based on real-time data. The conclusions are consistent and show significant differences from the cached conclusions of Edge-07 in key values, indicating that the cached data has become outdated due to the power grid mode switch.

[0060] Step S4 Execution (Collaborative Verification and Evolution): Upon receiving inconsistent conclusions (two real-time conclusions vs. one outdated cached conclusion), the aggregation gateway automatically triggers collaborative verification for the protection scope. The gateway organizes Edge-05, Edge-08, and another highly reliable Edge-02 node into a verification group. After independent verification based on shared real-time data, the three nodes return highly consistent conclusions. Based on the response speed of the three nodes in this task (all <1ms) and their consistently high reliability history, the aggregation gateway assigns extremely high dynamic confidence weights to their conclusions, generates a final collaborative arbitration conclusion after weighted arbitration, and immediately authorizes the execution of value fine-tuning.

[0061] Feedback and Evolution: The conclusion of the Edge-07 node deviated significantly from the collaborative arbitration conclusion. This deviation was fed back to its local credibility assessment model. Based on this, the model strengthened the rule weight of "significantly reducing the score when high memory usage and high network jitter occur simultaneously".

[0062] The high-quality data package documenting the entire process from "anomaly detection" to "collaborative arbitration" was uploaded to the cloud main station. The main station's analysis system used this data package to optimize the model parameters in the global verification algorithm related to "short-circuit current calculation under sudden drops in renewable energy," improving the estimated calculation accuracy for similar scenarios by approximately 1.5%.

[0063] 3. Effect Comparison The effects of this invention compared with existing technologies are shown in Table 1 below. Figure 4 As shown: Table 1: Comparison of Results 4. Summary This invention, through a closed loop of self-assessment → hierarchical execution → collaborative handling → evolutionary feedback, successfully transforms a potential missetting risk caused by node unreliability into a successful system collaborative correction and self-learning event. This not only avoids protection malfunctions that may be caused by using outdated cached data, but also optimizes the assessment models of relevant nodes and the global algorithm through this event. Compared with existing technologies, this invention elevates setting verification from a simple computational task to an intelligent system process with self-awareness, dynamic defense, collective intelligence, and continuous evolution capabilities, providing an innovative and effective technical path for solving the problem of accurate relay protection verification faced by power plants in new power systems.

[0064] As can be seen from the above description, the embodiments of the present invention achieve the following technical effects: Existing technologies assume the reliability of computing terminals, concentrating all risks at the algorithm level. This invention introduces the operational health of edge computing nodes as a core variable into the verification process. By monitoring hardware load, memory consistency, and network latency in real time and generating quantified trust level labels, the system can proactively identify nodes in a sub-healthy or untrustworthy state. This enables the verification system to have self-checking capabilities, allowing it to preemptively downgrade or isolate suspicious nodes before erroneous conclusions are reached. This eliminates hidden miscalibration risks caused by hardware aging, resource overload, or silent data errors—risks that traditional methods cannot detect—at the source, thereby increasing the overall fault tolerance of the system from near zero to over 99%, achieving a fundamental leap from algorithmic security to intrinsic system security.

[0065] Unlike existing technologies with rigid cloud-based centralized or edge-based full-scale computing models, this invention constructs a dynamic and elastic architecture based on trust labels, featuring hierarchical execution and collaborative arbitration. For high-trust nodes, they are authorized to perform millisecond-level (<10ms) local real-time computation, ensuring core speed. For medium and low-trust nodes, simplified verification or cache invocation is triggered, and uncertainty markers are added, intelligently saving 60%-85% of ineffective computing power. More importantly, the system organizes multiple high-trust nodes through an aggregation gateway to independently and in parallel verify and weightedly arbitrate questionable conclusions, generating highly confident collaborative arbitration conclusions. This mechanism is equivalent to configuring a miniature expert review panel for each protection scope, making the verification conclusion no longer dependent on the output of a single node. This effectively resists single-point failures and localized interference, and in complex scenarios such as rapid fluctuations in new energy output, it can stabilize the verification accuracy at over 98%, representing a relative improvement of more than 15%.

[0066] Existing technologies suffer from long model optimization cycles and struggle to adapt quickly to power grid changes. This invention, through a collaborative verification process, naturally generates a large number of high-quality verification data packets with high-confidence annotations. This data is systematically used in two ways: first, to fine-tune the local confidence assessment models of each edge node, making their self-diagnostic capabilities increasingly accurate; and second, to forward train the global verification algorithm model of the cloud master station, ensuring its calculation rules continuously approximate the complex physical characteristics of the real power grid. This creates a continuous learning closed loop of edge awareness – cloud training – edge enhancement, transforming the entire system from a statically deployed tool into an intelligent agent capable of continuously accumulating experience and iteratively evolving from actual operation. This significantly enhances its adaptability to long-term dynamic changes such as power grid topology evolution and the integration of new equipment, achieving a paradigm shift from passive configuration to proactive adaptation.

[0067] The embodiments and / or implementation methods described above are merely preferred embodiments and / or implementation methods for implementing the technology of the present invention, and are not intended to limit the implementation methods of the technology of the present invention in any way. Any person skilled in the art can make some modifications or alterations to other equivalent embodiments without departing from the scope of the technical means disclosed in the present invention, but these should still be regarded as the technology or embodiments that are substantially the same as the present invention.

[0068] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. The above descriptions are only preferred embodiments of this application. It should be noted that due to the limitations of written expression, while there are objectively infinite specific structures, those skilled in the art can make several improvements, modifications, or changes without departing from the principles of this application, and can also combine the above technical features in an appropriate manner. These improvements, modifications, changes, or combinations, or the direct application of the inventive concept and technical solution to other situations without modification, should all be considered within the scope of protection of this application.

Claims

1. An automatic verification method for power plant relay protection settings based on edge computing, characterized in that, Includes the following steps: Edge computing nodes use a credibility assessment model to evaluate their own status and generate a node credibility level. Edge computing nodes select a verification strategy based on the node trust level to generate hierarchical verification results with corresponding tags; Based on the markers in the hierarchical verification results, the system performs differentiated value processing and initiates collaborative verification; The collaborative arbitration conclusions and process data generated by collaborative verification are used to optimize the evaluation model and the global verification algorithm.

2. The automatic verification method for power plant relay protection settings based on edge computing according to claim 1, characterized in that, The edge computing node uses a credibility assessment model to evaluate its own status and generate a node credibility level, specifically including: Before each setpoint verification calculation, the edge computing nodes within the power plant automatically trigger a self-assessment process. Edge computing nodes monitor their own hardware computing unit load rate, memory data consistency verification results, and network latency jitter values ​​with the upper-level master station or adjacent nodes in real time. The node takes the hardware computing unit load rate, memory data consistency verification result and network latency jitter value as inputs to a preset credibility assessment model; the credibility assessment model assigns dynamic weights to each indicator and uses a piecewise function to score them, and outputs a quantitative node credibility score after comprehensive calculation; Based on a preset score threshold range, the node's credibility score is mapped to the edge node's self-evaluation result, which is either high credibility, medium credibility, or low credibility, and thus becomes a credibility level label.

3. The automatic verification method for power plant relay protection settings based on edge computing according to claim 2, characterized in that, The edge computing node monitors its own hardware computing unit load rate, memory data consistency verification results, and network latency jitter values ​​with the upstream master station or adjacent nodes in real time, specifically including: Hardware processing unit load rate calculation: in: for The load rate of the hardware computing unit at any given time; They are respectively The actual computing resource usage of CPU, GPU, and FPGA at any given time; These represent the total computing resource capacity of the CPU, GPU, and FPGA, respectively. These are the weighting coefficients for the hardware units; Calculation of memory data consistency check results: in: for The memory data consistency check score at any given time; In the evaluation cycle The total number of CRC checks performed on critical memory regions; In the evaluation cycle Number of internal CRC check failures; In the evaluation cycle Number of successful internal CRC checks; Network latency jitter is quantified using the standard deviation of round-trip time: Among them, average round-trip delay for: ; In the formula: for Network latency jitter value at any given time. For the first Round-trip delay sample value of each heartbeat data packet In the evaluation cycle Heart rate sampling count within the body, To evaluate the average round-trip time within the evaluation period.

4. The method for automatic verification of power plant relay protection settings based on edge computing according to claim 1, characterized in that, Edge computing nodes perform verification based on the node trust level selection strategy, generating hierarchical verification results with corresponding labels, specifically including: The edge computing node reads the edge node self-evaluation results, parses the credibility level labels, and then selects the corresponding verification execution strategy. If the credibility level label is high credibility, the node calls the complete local real-time setting calculation engine, performs a complete verification calculation including all protection setting items based on the latest real-time power grid operation data, and generates a standard verification conclusion including specific setting deviation conclusions and adjustment suggestions. If the label is medium or low confidence, only a simplified verification of the critical protection setting is performed or the cached verification result of the most recent high confidence state is called, and an uncertainty flag corresponding to the confidence level is attached to the generated result. Finally, a unified hierarchical verification result is output, which may be an unlabeled standard verification conclusion or a simplified or cached conclusion with uncertainty labeling.

5. The automatic verification method for power plant relay protection settings based on edge computing according to claim 4, characterized in that, If the label is medium or low confidence, only a simplified verification of the critical protection settings or the cached verification conclusion from the most recent high confidence state will be performed, and an uncertainty flag corresponding to the confidence level will be appended to the generated conclusion, specifically including: If the credibility level label is medium credibility, the node starts a simplified verification mode, which selects key protection settings from the complete setpoint list for rapid calculation. The key protection settings include at least the main protection instantaneous trip setting and the backup protection time limit setting related to system stability. The generated simplified conclusion will be automatically marked with credibility downgrade and incomplete calculation uncertainty, forming a graded verification result. If the credibility level label is low credibility, the node does not perform real-time calculation, but instead retrieves the most recently generated and cached historical standard verification conclusion from the local secure storage area under a high credibility state. After retrieval, a credibility downgrade and uncertainty mark indicating the use of cached data are added to the conclusion to form a graded verification result.

6. The method for automatic verification of power plant relay protection settings based on edge computing according to claim 1, characterized in that, Based on the markers in the hierarchical verification results, the system performs differentiated value processing and initiates collaborative verification, specifically including: The factory-level edge computing aggregation gateway or cloud-based fixed-value management master station receives the hierarchical verification results from all edge computing nodes in the network and parses the result content and marking status. For hierarchical verification results confirmed after parsing without uncertainty markers, if the result indicates that a setting adjustment is required, the aggregation gateway or main station will directly issue a setting modification instruction to the corresponding protection device according to the preset security policy, or grant the edge computing node that generated the hierarchical verification result the permission to automatically perform the adjustment. For graded verification results that are confirmed to have uncertainty markers after parsing, a two-level processing mechanism is initiated; The aforementioned redundancy verification mechanism will continue until the source node completes its self-check and recovery, and re-evaluates and generates a high-confidence level label. Only then will the system remove the pending confirmation status and terminate the redundancy verification process initiated for it.

7. The automatic verification method for power plant relay protection settings based on edge computing according to claim 6, characterized in that, The two-level processing mechanism specifically includes: Send a mandatory deep self-check and system recovery command to the source edge computing node that generated the result, and at the same time mark all the verification conclusions reported by the node this time as pending confirmation in its internal database and suspend its automatic execution permission. Dynamically adjust the verification task allocation strategy for the power grid protection domain: immediately issue redundant verification tasks for the protection range of the faulty node to one or more edge computing nodes that are physically adjacent to the node and whose current trust level label is high trust; if there are no available high trust adjacent nodes, the cloud-based value management master station will directly perform synchronous verification calculations on the protection range.

8. The method for automatic verification of power plant relay protection settings based on edge computing according to claim 1, characterized in that, The method of using collaborative arbitration conclusions and process data generated by collaborative verification to feed back optimization evaluation models and global verification algorithms specifically includes: The edge computing aggregation gateway triggers collaborative verification of a specific protection device or associated protection link under at least one of the following conditions: reaching a preset periodic verification cycle; receiving a graded verification result with an uncertainty marker; or the system detects a significant change in the operating mode of the protection range. The aggregation gateway selects two or more edge computing nodes from the node list that have data collection capabilities and verification permissions for the target protection scope and whose current trust level label is the highest, as members of the collaborative verification group; How can the aggregation gateway assign dynamic confidence weights to the conclusions of each node based on the state micro-change data and historical confidence records before and after the execution of this task, and generate collaborative arbitration conclusions based on a weighted arbitration mechanism? The aggregation gateway uses the dynamic confidence weight to perform weighted arbitration on the independent verification conclusions of each node, and generates the final collaborative arbitration conclusion. The collaborative arbitration conclusion and the entire process data of this collaborative verification are used in the following two aspects: feeding back to the local credibility assessment model of the relevant nodes, using the deviation between the node conclusion and the collaborative arbitration conclusion as a correction parameter to update the internal weights of the model; and uploading the verification data package as a training sample with high confidence labels to the main station analysis system in the cloud for iterative optimization of the global fixed value verification algorithm model.

9. The automatic verification method for power plant relay protection settings based on edge computing according to claim 8, characterized in that, The aggregation gateway assigns dynamic confidence weights to the conclusions of each node based on the minor state changes and historical reliability records before and after the execution of the current task, and generates collaborative arbitration conclusions based on a weighted arbitration mechanism, specifically including: For each member of the verification group The aggregation gateway collects its state change data and calculates the normalized deviation: Task response delay deviation: Average response time: Memory usage deviation: in: For nodes The relative deviation of task response latency; For nodes The response time of this collaborative verification task; This represents the average response time of all nodes within the collaborative verification group. For nodes The deviation in memory usage; For nodes Peak memory usage during the verification calculation process; For nodes The normal baseline value for memory usage during the verification calculation process; For nodes Total memory capacity; Calculate each node Dynamic confidence weights : Among them, historical credibility factor : In the formula: For nodes Dynamic confidence weights in collaborative verification; The penalty coefficient for data with minor state changes; For nodes The current credibility score; For nodes Historical credibility factor; This is the length of the historical data sampling window; For the first The timestamp of the previous historical verification; This represents the execution time of the current collaborative verification task; This is a historical credibility decay factor.

10. An automatic verification system for power plant relay protection settings based on edge computing, characterized in that, The method for automatically verifying the settings of power plant relay protection based on edge computing as described in any one of claims 1-9 includes: The node trustworthiness self-assessment module is deployed in each edge computing node within the power plant to monitor and assess the node's own operating status and generate a quantified node trustworthiness level. The hierarchical verification decision and execution module is deployed in the edge computing node and communicates with the node credibility self-evaluation module. It is used to select and execute the corresponding verification strategy according to the node credibility level and generate hierarchical verification results with or without uncertainty markers. The setpoint processing and collaborative scheduling module is deployed at the plant-level edge computing aggregation gateway or cloud-based setpoint management master station. It is used to receive and parse the hierarchical verification results of each node, execute differentiated setpoint processing instructions according to the marked status, and initiate collaborative verification tasks for low-confidence conclusions. The group collaboration and model evolution module is deployed on the edge computing aggregation gateway and the cloud-based value management master station. It is used to organize multiple nodes to perform collaborative verification, generate collaborative arbitration conclusions, and use this conclusion data to back-optimize the evaluation model and global verification algorithm model of the node credibility self-evaluation module.

Images