A substation cluster processing method and system based on multi-machine redundancy

Abstract

The present application relates to the substation operation and maintenance technical field, especially relates to a kind of substation cluster processing method and system based on multi-machine redundancy, the present application is marked by main node according to preset label and its small label combination inspection data, form the data group of unique identification, combine preset weight calculation value fraction and divide data group grade, backup node adopts different synchronization strategy to realize data synchronization, by integrating backup node idle cache resource to build cache pool, utilize data grouping and abnormality determination to delete no abnormal duplicate data, when main node fails, enable backup node as new main node, based on switching timestamp to demarcate time sequence window, retrieve cache pool corresponding data sorting integration form time sequence chain, locate breakpoint after according to video frame, sensing data type uses adaptive algorithm to complete.The present application solves the data breakpoint in the data synchronization process of backup node, reduces bandwidth occupancy, guarantees the stable operation and safe operation and maintenance of substation multi-machine redundancy system.

Description

Technical Field

[0001] This invention relates to the field of substation operation and maintenance technology, and in particular to a substation cluster processing method and system based on multi-machine redundancy. Background Technology

[0002] In intelligent inspection scenarios with multi-machine redundancy in substations, when the primary node fails, the system quickly switches to the backup node. Although this switching process can be considered seamless, it introduces millisecond-level implicit latency. During this process, temporary data interruptions can easily occur, leading to the loss of critical inspection data and causing the system to miss equipment defects. Current technologies synchronize incremental inspection data (such as video and sensor data) in real time between the primary and backup nodes, and pre-load AI inference models on the backup node, directly calling cached data and the pre-loaded model during switching. However, when the backup node retrieves the last batch of data before the primary node's failure, the retrieval operation is delayed, core data segments are missing, leading to misjudgments. Furthermore, real-time incremental synchronization of all inspection data by the backup node consumes a significant amount of cluster communication bandwidth, increasing output transmission latency and disrupting the continuity of critical time-series data, resulting in data interruptions.

[0003] Therefore, it is necessary to propose a substation cluster processing method and system based on multi-machine redundancy to solve the above problems. Summary of the Invention

[0004] The purpose of this invention is to provide a substation cluster processing method and system based on multi-machine redundancy, in order to solve the problems of time delay in pull operation, missing core data segments leading to misjudgment, and the real-time incremental synchronization of all inspection data by standby nodes consuming a large amount of cluster communication bandwidth, increasing output transmission delay, and disrupting the continuity of key time-series data, resulting in data breakpoints.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A substation cluster processing method based on multi-machine redundancy, the method comprising: S1: Based on the inspection data cached by the master node, the inspection data is marked, processed and combined by preset tags to obtain multiple data groups. The standby node uses a synchronization strategy to synchronize the data groups in real time. S2: Integrate the idle cache resources of all standby nodes to build a cache pool, and store the synchronization data of standby nodes in the cache pool; S3: When a primary node failure is detected, a backup node is activated as the new primary node. The new primary node retrieves data from the cache pool for synchronization and completion.

[0006] Preferably, the step of marking and combining inspection data using preset tags to obtain multiple data groups, and then using a synchronization strategy to synchronize the data groups in real time by the backup node, includes: The inspection data in the data group is marked according to the sub-labels corresponding to each label in the preset labels to obtain tagged data. The preset labels include equipment labels, defect labels and time series labels. Based on the sub-labels of each preset label and the corresponding tagged data, multiple data groups with unique sub-label combinations are obtained. Based on the value level, the standby nodes use a synchronization strategy to synchronize the data groups in real time.

[0007] Preferably, the step of using a synchronization strategy to perform real-time synchronization of data groups by backup nodes according to value level includes: The value score of each data group is calculated based on the combination of sub-tags, including equipment tags, defect tags, and time sequence tags, and the preset weights. Based on the value score, the data groups are divided into value levels to obtain the data group value levels, wherein the data group value levels include high-level data groups, medium-level data groups and low-level data groups. The backup node performs real-time synchronization of the data groups according to the value level of the data groups and adopts a synchronization strategy.

[0008] Preferably, the backup node performs real-time synchronization of the data groups according to their value level, using a synchronization strategy. The synchronization strategy includes: The backup node synchronizes the inspection data of high-level data groups using lossless compression and millisecond-level response. The backup node uses a lightweight compression and asynchronous synchronization strategy to synchronize the inspection data of the medium-level data group. The standby node is responsible for the inspection data of the low-level data group. The inspection data of the low-level data group is cached on the primary node. The primary node generates a corresponding data index for the inspection data of the low-level data group, and the standby node caches the data index.

[0009] Preferably, the step of integrating the idle cache resources of all standby nodes to construct a cache pool, and storing the synchronization data of the standby nodes in the cache pool, includes: By using cluster resource monitoring tools, the idle cache capacity, read / write speed and load status information of standby nodes are collected in real time to obtain available idle cache resources; Based on a distributed caching framework, idle cache resources are integrated to obtain a cache pool; Based on the cache pool, the synchronization data of the standby nodes is stored in the cache pool.

[0010] Preferably, the step of storing the synchronization data of the standby node in the cache pool according to the cache pool includes: After receiving the synchronization data, the cache pool groups the synchronization data according to a preset time period to obtain multiple synchronization data groups. Based on the new synchronization data group received from the synchronization data group and the cache pool, anomaly detection is performed by comparing and analyzing the data values ​​within the data group. If the analysis results are abnormal, the newly synchronized data group is retained, and the cache pool caches the new synchronized data group; If the analysis results are normal, delete the new data group.

[0011] Preferably, the step of activating a backup node as the new master node when a master node failure is detected, and the new master node retrieving data from the cache pool for synchronization and completion, includes: Obtain the timestamp for the switchover between the primary and backup nodes. Based on the timestamp, obtain the time sequence window for the time before and after the timestamp, wherein the time sequence window includes a previous time sequence window and a subsequent time sequence window. The new master node retrieves data from the cache pool in the previous time window and receives data from the cache pool in the subsequent time window in real time. The new master node sorts and integrates the data from the previous and subsequent time-series windows according to a unified timestamp to obtain a time-series chain and corresponding continuous data segments. Based on the time-series chain, the continuous data segments are then completed.

[0012] Preferably, the step of completing the continuous data segments according to the time sequence includes: Based on the time sequence chain and continuous data segments, the breakpoints in the continuous data segments and the start and end time ranges of the breakpoints are located to obtain the breakpoint data segments. The breakpoint data segments are classified according to the data type of the data group corresponding to the breakpoint data segments, wherein the data types include video frames and sensor data; Based on the data type corresponding to the breakpoint data segment, an algorithm that matches the data type is used to complete the breakpoint data segment.

[0013] A substation cluster processing system based on multi-machine redundancy, used to implement the substation cluster processing method based on multi-machine redundancy described above, the system comprising: The data processing module is used to mark and combine the inspection data cached by the master node according to the preset tags, classify the data groups obtained by the marking and combination into value levels, and use different strategies to cache the data groups of different levels. The management module is used to build a cache pool based on the idle resources of the standby nodes, receive synchronization data from the standby nodes through the cache pool, process the synchronization data, and delete groups of duplicate data without abnormalities. The switching module is used to monitor the status of the master node, trigger the standby node to switch to the new master node, and obtain the switching timestamp. The data completion module is used to divide the time sequence window according to the switching timestamp, control the new master node to retrieve the time sequence window data from the cache pool, generate continuous data segments, locate breakpoints and start and end ranges, and use an appropriate algorithm to complete the data segments.

[0014] The technical effects and advantages of the present invention in the above technical solution are as follows: This invention uses preset tags to mark and combine inspection data, and classifies the value levels according to small tag combinations and preset weights. The standby node synchronizes the data through a differentiated synchronization strategy, avoiding the problem of excessive cluster bandwidth consumption caused by full data synchronization, and ensuring the real-time and integrity of high-value key data. By integrating the idle cache resources of the standby node to build a cache pool and deleting abnormal duplicate data, efficient utilization of storage resources is achieved. When the master node fails, the new master node can directly retrieve the historical data of the previous time window and receive the real-time data of the next time window from the cache pool without pulling data from the original master node. By using breakpoint location and adaptation algorithms to complete the data, the problem of inspection data breakpoints caused by implicit latency during switching is solved, reducing the risk of missed or misjudged equipment defects, and providing reliable data support for the safe operation and maintenance of substations. Attached Figure Description

[0015] Figure 1 This is a flowchart of a substation cluster processing method based on multi-machine redundancy according to the present invention.

[0016] Figure 2 This is a structural diagram of a substation cluster processing system based on multi-machine redundancy according to the present invention. Detailed Implementation

[0017] 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.

[0018] Example 1, such as Figure 1 As shown in the figure, this embodiment provides a substation cluster processing method based on multi-machine redundancy, the method including: S1: Based on the inspection data cached by the master node, the inspection data is marked, processed and combined by preset tags to obtain multiple data groups. The standby node uses a synchronization strategy to synchronize the data groups in real time. S2: Integrate the idle cache resources of all standby nodes to build a cache pool, and store the synchronization data of standby nodes in the cache pool; S3: When a primary node failure is detected, a backup node is activated as the new primary node. The new primary node retrieves data from the cache pool for synchronization and completion. In one embodiment of the present invention, the step of marking and combining inspection data with preset tags to obtain multiple data groups, and the standby node using a synchronization strategy to synchronize the data groups in real time includes: S11: Mark the inspection data in the data group according to the sub-labels corresponding to each label in the preset labels to obtain labeled data. The preset labels include equipment labels, defect labels and time sequence labels. S12: Based on the sub-labels of each preset label and the corresponding tagged data, multiple data groups with unique sub-label combinations are obtained; S13: Based on the value level, the standby node uses a synchronization strategy to synchronize the data groups in real time. In this embodiment of the invention, the system first uses preset equipment tags, defect tags, and time-series tags. Equipment tags primarily reflect equipment priority, defect tags primarily reflect whether a defect is confirmed based on feature data, and time-series tags primarily reflect the time sequence of defect development. Each tag has its own corresponding sub-tags. For example, equipment tags include p1, p2, and p3, where p1 indicates core equipment such as main transformers and circuit breakers, p2 indicates secondary equipment such as instrument transformers, and p3 indicates auxiliary facilities such as fence lighting equipment. Defect tags include D0, D1, D2, and D3, where D0 indicates worthless data. According to the data, D1 is used to indicate the normal state, D2 is used to indicate the critical state of the equipment, and D3 is used to indicate the suspected defect. The time sequence labels include T1 and T2. T1 is used to indicate the critical time sequence of defect development, and T2 is used to indicate the ordinary time sequence. Through these small labels, each inspection data cached by the master node is marked. For example, for the main transformer in the inspection data, because it is a key equipment in the substation, it can be marked by the label p1D3T1. Each inspection data has a different combination of labels, thus obtaining a data group with labels. The standby node uses a synchronization strategy to perform real-time synchronization caching of the data group based on the value level.

[0019] In one embodiment of the present invention, the step of the backup node performing real-time synchronization of data groups according to the value level using a synchronization strategy includes: S131: Calculate the value score of each data group based on the combination of sub-tags of equipment tags, defect tags, and time sequence tags, and the preset weights; S132: Based on the value score, the data group is divided into value levels to obtain the data group value level, which includes high-level data group, medium-level data group and low-level data group. S133: The standby node uses a synchronization strategy to synchronize the data groups in real time according to the value level of the data groups; In this embodiment of the invention, the value score of each data group is calculated based on the combination of small tags and preset weights. By combining equipment tags, defect tags and time series tags into calculable data values, the actual value of each group of inspection data for substation defect identification and maintenance is reflected. This allows the standby node to adopt an appropriate synchronization strategy based on the actual value. The data groups are divided into high-level, medium-level and low-level data groups based on the value score. For different levels of data groups, the standby node adopts different synchronization strategies to synchronize, thus solving the problem of excessive resource consumption during full synchronization of the standby node. The preset weights are assigned basic weight priorities to equipment tags, defect tags, and time-series tags based on the core objectives of substation inspection. According to the sub-tags under each type of tag, the contribution of different sub-tags to defect identification is statistically analyzed based on the historical inspection data. For example, the defect identification contribution rate of the sub-tags corresponding to core equipment in the equipment tags reaches 85%, while that of auxiliary equipment is only 12%. Sub-weights are assigned to each sub-tag, such as a weight of 3 for core equipment, a weight of 2 for important equipment, and a weight of 1 for auxiliary equipment. Finally, through multiple simulation tests and on-site trial runs, the weight values ​​of each sub-tag are iteratively adjusted based on the comprehensive results of data synchronization under different weight combinations, and the optimal weight is finally obtained as the preset weight.

[0020] In one embodiment of the present invention, the standby node performs real-time synchronization of data groups according to the value level of the data groups, using a synchronization strategy. The synchronization strategy includes: S1331: The standby node uses lossless compression and millisecond-level response to synchronize the inspection data of high-level data groups; S1332: The standby node uses a lightweight compression and asynchronous synchronization strategy to synchronize the inspection data of the medium-level data group; S1333: The standby node is for the inspection data of the low-level data group. The inspection data of the low-level data group is cached on the master node. The master node generates the corresponding data index for the inspection data of the low-level data group, and the standby node caches the data index. In this embodiment of the invention, the standby node uses lossless compression and millisecond-level response to synchronize the inspection data of the high-level data group. This is because the high-level data group corresponds to the core equipment in the substation, and it is necessary to keep monitoring the core equipment at all times to ensure that there is no loss of inspection data. Therefore, lossless compression and millisecond-level response are used for synchronization. Specifically, the data is processed by lossless compression algorithms such as FLAC. Without losing any data value of the original data, the resources occupied by data transmission are reduced, and the integrity of the core data is guaranteed so that it will not be lost due to compression. The entire process of data processing by the master node is controlled within 100ms or less. For example, the processing and compression time of the master node is within 1-10ms, and the decompression and verification time of the standby node is within 1-20ms, so as to realize the real-time synchronization of the core data. For the inspection data of the medium-level data group, the backup node uses a lightweight compression and asynchronous synchronization strategy for synchronization. The lightweight compression uses lightweight algorithms such as GZIP to quickly compress the data, reducing the data volume with a compression time of 1-5ms while allowing for the loss of some non-core details. The synchronization method uses asynchronous communication, so the master node does not need to wait for the backup node to receive feedback before it can continue to process subsequent data. The backup node caches the data to be synchronized through a queue and processes it in batches according to its own load status, avoiding the occupation of communication link resources by synchronization operations. Furthermore, the periodic verification between the master node and the backup node ensures the integrity of data synchronization and effectively controls the network bandwidth utilization. The standby node is for inspection data of low-level data groups. The inspection data of low-level data groups is cached on the master node. The master node generates corresponding data indexes for the inspection data of low-level data groups, and the standby node caches the data indexes. For some inspection data that has little impact on the operation of the substation, the standby node only synchronizes the data indexes of these inspection data, which further reduces the problem of too much data cached by the standby node due to full data synchronization. The process by which the master node generates a corresponding data index for the inspection data of low-level data groups includes: The master node receives and caches low-level data groups, extracts the location information of the data group, including the small tag corresponding to the device tag, the timestamp of data collection, and the address of the data cached locally on the master node. The master node integrates the extracted location information data according to the preset structured index format to obtain a data index with distinctive characteristics; The master node verifies the timestamp format, storage location, and other information of the data index. If the verification is correct, the index is pushed to the standby node, and a copy of the index is kept in the local cache space of the master node.

[0021] In one embodiment of the present invention, the step of integrating the idle cache resources of all standby nodes to construct a cache pool, and storing the synchronization data of the standby nodes in the cache pool, includes: S21: Use cluster resource monitoring tools to collect real-time information on the idle cache capacity, read / write speed, and load status of standby nodes to obtain available idle cache resources; S22: Based on a distributed caching framework, idle cache resources are integrated to obtain a cache pool; S23: Based on the cache pool, store the synchronization data of the standby node in the cache pool; In this embodiment of the invention, a cluster resource monitoring tool is used to collect real-time information on the idle cache capacity, read / write speed, and load status of standby nodes to obtain available idle cache resources. The cluster resource monitoring tool uses monitoring components built into the distributed caching framework, such as the INFO command of the caching framework Redis and the stats tool of Memcached. The cache pool stores the synchronization data of high- and medium-level data groups and the associated indexes of low-level data groups according to a differentiated synchronization strategy, reducing the redundant occupation of synchronization data. When the master node fails and a switchover is triggered, the new master node (i.e., the standby node) does not need to pull data from the original master node, but directly retrieves data from the cache pool quickly. This solves the latency problem of traditional pull operations, provides data source support for data synchronization and completion, ensures the continuity of inspection data, solves the problem of defect omissions and misjudgments caused by data breakpoints during the switchover, and improves the overall stability and operation and maintenance efficiency of the multi-machine redundant system to a certain extent.

[0022] In one embodiment of the present invention, the step of storing synchronization data of standby nodes in a cache pool includes: S231: After receiving the synchronization data, the cache pool groups the synchronization data according to a preset time period to obtain multiple synchronization data groups; S232: Based on the new synchronization data group received from the synchronization data group and the buffer pool, perform anomaly detection by comparing and analyzing the data values ​​within the data group. If the analysis results are abnormal, the newly synchronized data group is retained, and the cache pool caches the new synchronized data group; If the analysis results are normal, delete the new data group. In this embodiment of the invention, after receiving synchronization data from the backup node, the buffer pool groups all synchronization data in batches according to a preset fixed time period (e.g., 1 min-5 min), ensuring that each group of data corresponds to a monitoring segment of the same time dimension, resulting in multiple synchronization data groups. When the buffer pool receives new synchronization data and completes the grouping for the same time period, a new synchronization data group is obtained. The core values ​​of the corresponding monitoring indicators (e.g., temperature sensor values ​​of the same device, key feature parameters of the same video frame) are extracted from the new and old synchronization data groups. Then, indicators such as numerical deviation rate and trend similarity are calculated for quantitative comparison, combined with the normal parameter range of substation equipment operation (e.g., core parameters). The system uses the allowable temperature fluctuation range of the core equipment as an auxiliary criterion for judgment. Based on the comparison results, anomaly judgment is made: if the deviation rate between the value of the new synchronized data group and the historical synchronized data group exceeds any of the preset thresholds, changes in trend, etc., and does not conform to the normal pattern, or if the value is within the range of abnormal operating parameters of the equipment, then the analysis result is judged to be abnormal, the cache pool retains the new synchronized data group, and the caching is completed according to the classification storage rules; if the deviation between the value of the new synchronized data group and the historical data is within a reasonable range, the change trend conforms to the normal operating pattern of the equipment, and there are no abnormal characteristics, then the analysis result is judged to be normal. In order to avoid storage redundancy, the new synchronized data group is directly deleted, and the historically valid data group is retained.

[0023] In one embodiment of the present invention, when a primary node failure is detected, a backup node is activated as the new primary node, and the new primary node retrieves data from the cache pool for synchronization and completion. This step includes: S31: Obtain the timestamp of the switchover between the primary and standby nodes. Based on the timestamp, obtain the time series window before and after the timestamp. The time series window includes the previous time series window and the next time series window. S32: The new master node retrieves data from the cache pool in the previous time window and receives data from the cache pool in the subsequent time window in real time. S33: The new master node sorts and integrates the data from the previous and subsequent time windows according to a unified timestamp to obtain the time chain and the corresponding continuous data segments. Based on the time chain, the continuous data segments are completed. The steps for completing consecutive data segments based on the time sequence chain include: S331: Based on the timing chain and continuous data segments, locate the breakpoints in the continuous data segments and the start and end time ranges of the breakpoints to obtain the breakpoint data segments. S332: Classify the breakpoint data segments according to the data type of the data group corresponding to the breakpoint data segments. The data types include video frames and sensor data. S333: Based on the data type corresponding to the breakpoint data segment, use an algorithm that matches the data type to complete the breakpoint data segment; In this embodiment of the invention, when a primary node failure is detected, the system first captures the millisecond-level timestamp T0 at the moment of switching between the primary and backup nodes. Using this timestamp T0 as a reference point, a time-series window is defined. The preceding time-series window covers one second before T0, and the following time-series window covers one second after T0 (including real-time data added after the switchover). Next, the new primary node directly pulls data from the cache pool. Specifically, it quickly retrieves all synchronized data from the preceding time-series window using intelligent indexing. Simultaneously, it receives newly generated data from the following time-series window pushed by the cache pool in real time. The new primary node sorts and integrates the data from the preceding and following time-series windows according to a unified timestamp, obtaining a preliminary time-series chain and corresponding continuous data segments. Based on the timestamp sequence, the system then verifies the integrity and continuity of the data. The system continuously verifies and locates breakpoints and their start and end time ranges within continuous data segments (e.g., 10ms data gaps before and after T0), resulting in clearly defined breakpoint data segments. Based on the original data group type to which the breakpoint data segments belong, they are divided into video frame breakpoint data segments and sensor data breakpoint data segments. Adaptive algorithms are used to complete the breakpoint data segments according to their different types: for video frame breakpoints, the physical structure of the substation equipment (e.g., the heat conduction law of metal components, the continuity of equipment appearance) is considered, and inter-frame interpolation algorithms are used to generate transition frames to fill the breakpoints; for sensor data breakpoints, historical sensor data from the same period and real-time environmental parameters (humidity, temperature) are integrated, and a weighted correction algorithm is used to calculate the missing values, achieving seamless integration of continuous data segments.

[0024] Example 2, as Figure 2 As shown, this embodiment provides a substation cluster processing system based on multi-machine redundancy. The system includes: The data processing module is used to mark and combine the inspection data cached by the master node according to the preset tags, classify the data groups obtained by the marking and combination into value levels, and use different strategies to cache the data groups of different levels. The management module is used to build a cache pool based on the idle resources of the standby nodes, receive synchronization data from the standby nodes through the cache pool, process the synchronization data, and delete groups of duplicate data without abnormalities. The switching module is used to monitor the status of the master node, trigger the standby node to switch to the new master node, and obtain the switching timestamp. The data completion module is used to divide the time sequence window according to the switching timestamp, control the new master node to retrieve the time sequence window data from the cache pool, generate continuous data segments, locate breakpoints and start and end ranges, and use an appropriate algorithm to complete the data segments. In this embodiment of the invention, the data processing module marks and combines the inspection data cached by the master node according to the device tag, defect tag, time sequence tag and their corresponding sub-tags to form data groups with unique sub-tag combinations. Then, the value score of each data group is calculated by combining preset weights to obtain the level classification of the data groups. The standby node synchronizes according to the differentiated strategy based on the data group level. The management module collects the idle resources of the standby node in real time through the cluster resource monitoring tool, integrates these idle resources to build a cache pool, uses the cache pool to receive the synchronization data of the standby node, and compares the new synchronized data group with the historical data group for numerical comparison and anomaly detection. The system deletes duplicate data groups without abnormalities to optimize storage resources. At the same time, the switching module continuously monitors the operating status of the master node. When a master node failure is detected, the standby node switches to the new master node and obtains the switching timestamp. The data completion module divides the data into before and after time windows based on the switching timestamp. It controls the new master node to retrieve the data in the before and after time windows from the cache pool, integrates and generates a time chain and corresponding continuous data segments. By comparing the time chain with the continuous data segments, the breakpoint and the start and end time range are located. Then, based on the data type of the breakpoint data segment, an adaptation algorithm is used to complete the breakpoint, so as to achieve the continuity and integrity of the inspection data.

[0025] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A substation cluster processing method based on multi-machine redundancy, characterized in that, The method includes: S1: Based on the inspection data cached by the master node, the inspection data is marked, processed and combined by preset tags to obtain multiple data groups. The standby node uses a synchronization strategy to synchronize the data groups in real time. S2: Integrate the idle cache resources of all standby nodes to build a cache pool, and store the synchronization data of standby nodes in the cache pool; S3: When a primary node failure is detected, a backup node is activated as the new primary node. The new primary node retrieves data from the cache pool for synchronization and completion.

2. The substation cluster processing method based on multi-machine redundancy according to claim 1, characterized in that, The step of marking and combining inspection data using preset tags to obtain multiple data groups, and then using a synchronization strategy to synchronize the data groups in real time by the standby node, includes: The inspection data in the data group is marked according to the sub-labels corresponding to each label in the preset labels to obtain tagged data. The preset labels include equipment labels, defect labels and time series labels. Based on the sub-labels of each preset label and the corresponding tagged data, multiple data groups with unique sub-label combinations are obtained. Based on the value level, the standby nodes use a synchronization strategy to synchronize the data groups in real time.

3. The substation cluster processing method based on multi-machine redundancy according to claim 2, characterized in that, The step of using a synchronization strategy to perform real-time synchronization of data groups by backup nodes according to value level includes: The value score of each data group is calculated based on the combination of sub-tags, including equipment tags, defect tags, and time sequence tags, and the preset weights. Based on the value score, the data groups are divided into value levels to obtain the data group value levels, wherein the data group value levels include high-level data groups, medium-level data groups and low-level data groups. The backup node performs real-time synchronization of the data groups according to the value level of the data groups and adopts a synchronization strategy.

4. The substation cluster processing method based on multi-machine redundancy according to claim 3, characterized in that, The backup node performs real-time synchronization of the data groups according to their value level, using a synchronization strategy. The synchronization strategy includes: The backup node synchronizes the inspection data of high-level data groups using lossless compression and millisecond-level response. The backup node uses a lightweight compression and asynchronous synchronization strategy to synchronize the inspection data of the medium-level data group. The standby node is responsible for the inspection data of the low-level data group. The inspection data of the low-level data group is cached on the primary node. The primary node generates a corresponding data index for the inspection data of the low-level data group, and the standby node caches the data index.

5. The substation cluster processing method based on multi-machine redundancy according to claim 1, characterized in that, The step of integrating the idle cache resources of all standby nodes to construct a cache pool, and storing the synchronization data of the standby nodes in the cache pool, includes: By using cluster resource monitoring tools, the idle cache capacity, read / write speed and load status information of standby nodes are collected in real time to obtain available idle cache resources; Based on a distributed caching framework, idle cache resources are integrated to obtain a cache pool; Based on the cache pool, the synchronization data of the standby nodes is stored in the cache pool.

6. The substation cluster processing method based on multi-machine redundancy according to claim 5, characterized in that, The step of storing the synchronization data of the standby node in the cache pool according to the cache pool includes: After receiving the synchronization data, the cache pool groups the synchronization data according to a preset time period to obtain multiple synchronization data groups. Based on the new synchronization data group received from the synchronization data group and the cache pool, anomaly detection is performed by comparing and analyzing the data values ​​within the data group. If the analysis results are abnormal, the newly synchronized data group is retained, and the cache pool caches the new synchronized data group; If the analysis results are normal, delete the new data group.

7. The substation cluster processing method based on multi-machine redundancy according to claim 1, characterized in that, The step of activating a standby node as the new master node when a master node failure is detected, and the new master node retrieving data from the cache pool for synchronization and completion, includes: Obtain the timestamp for the switchover between the primary and backup nodes. Based on the timestamp, obtain the time sequence window for the time before and after the timestamp, wherein the time sequence window includes a previous time sequence window and a subsequent time sequence window. The new master node retrieves data from the cache pool in the previous time window and receives data from the cache pool in the subsequent time window in real time. The new master node sorts and integrates the data from the previous and subsequent time-series windows according to a unified timestamp to obtain a time-series chain and corresponding continuous data segments. Based on the time-series chain, the continuous data segments are then completed.

8. The substation cluster processing method based on multi-machine redundancy according to claim 7, characterized in that, The step of completing consecutive data segments according to the time sequence chain includes: Based on the time sequence chain and continuous data segments, the breakpoints in the continuous data segments and the start and end time ranges of the breakpoints are located to obtain the breakpoint data segments. The breakpoint data segments are classified according to the data type of the data group corresponding to the breakpoint data segments, wherein the data types include video frames and sensor data; Based on the data type corresponding to the breakpoint data segment, an algorithm that matches the data type is used to complete the breakpoint data segment.

9. A substation cluster processing system based on multi-machine redundancy, used to implement the substation cluster processing method based on multi-machine redundancy as described in any one of claims 1-8, characterized in that, The system includes: The data processing module is used to mark and combine the inspection data cached by the master node according to the preset tags, classify the data groups obtained by the marking and combination into value levels, and use different strategies to cache the data groups of different levels. The management module is used to build a cache pool based on the idle resources of the standby nodes, receive synchronization data from the standby nodes through the cache pool, process the synchronization data, and delete groups of duplicate data without abnormalities. The switching module is used to monitor the status of the master node, trigger the standby node to switch to the new master node, and obtain the switching timestamp. The data completion module is used to divide the time sequence window according to the switching timestamp, control the new master node to retrieve the time sequence window data from the cache pool, generate continuous data segments, locate breakpoints and start and end ranges, and use an appropriate algorithm to complete the data segments.

Images