A power distribution cabinet monitoring method and system based on cloud edge fusion technology and an intelligent power distribution cabinet

Abstract

The application discloses a power distribution cabinet monitoring method and system based on cloud-edge fusion technology and an intelligent power distribution cabinet, and relates to the technical field of power distribution cabinet control. The method comprises the following steps: dynamically forming an edge cluster, electing a main cooperative node and at least one backup cooperative node in the edge cluster, fragmenting, redundantly encoding and transmitting abnormal data, and the main cooperative node executing or forwarding control instructions according to the feedback state of the cloud platform. Through the dynamic election mechanism of the intelligent gateway, the cooperative responsibilities in the edge cluster can be circulated according to the instant state of each node, thereby improving or even avoiding the problem of long-term overload of individual intelligent gateways caused by uneven time and space distribution of abnormal events. Through the introduction of the fragmentation and redundancy coding technology of abnormal data, the transmission task originally concentrated on a single link is decomposed and dispersed to multiple independent nodes and links, thereby reducing the systemic risk of local surface failure caused by the single-point failure diffusion of the cooperative gateway in the traditional scheme.

Description

Technical Field

[0001] This invention relates to the field of power distribution cabinet control technology, and in particular to a power distribution cabinet monitoring method, system and intelligent power distribution cabinet based on cloud-edge fusion technology. Background Technology

[0002] Traditional power distribution monitoring adopts either local isolation or pure cloud mode. The former has limited intelligence level, while the latter faces network latency and bandwidth pressure, making it difficult to achieve rapid response and in-depth analysis. Cloud-edge fusion technology solves this problem by building real-time edge processing and cloud intelligence. At the power distribution site, data is processed directly by edge devices to perform fault location, isolation and protection, ensuring that core control does not rely on the network. At the same time, it uploads the simplified data to the cloud, where it gathers information from the entire network and is responsible for macro-analysis such as model training, energy efficiency optimization, and predictive maintenance. It also continuously distributes optimization algorithms to the edge, enabling the system as a whole to evolve continuously.

[0003] Chinese patent application number CN202511518791.1 discloses an efficient data interaction method for intelligent power distribution cabinets. This invention adopts a redundant transmission mechanism of multi-intelligent gateway cooperation. When an intelligent gateway detects a local anomaly, it does not rely solely on its own potentially unstable uplink, but forwards the abnormal data packet within the edge cluster, where it is coordinated by other intelligent gateways with healthy uplinks in the edge cluster.

[0004] However, in actual power distribution systems, abnormal events (such as overload and short circuit) often occur more frequently on specific lines, under specific loads, or during specific time periods (such as peak electricity consumption). Using a collaborative gateway approach means that some gateways located in fault-prone areas will repeatedly become abnormal gateways. When an abnormal gateway (let's say gateway A) detects an anomaly, it must find a collaborative gateway that is currently in its reporting cycle. This collaborative gateway is determined by its fixed, preset reporting time slot. If the reporting time slot of gateway B happens to be aligned with the high-incidence period of gateway A's anomalies, then gateway B will be continuously selected as a collaborative gateway, resulting in its overload. Other gateways with misaligned reporting time slots may remain idle for a long time.

[0005] More importantly, the health status of a gateway's uplink is constantly changing. When a gateway is selected as a cooperating gateway, its uplink may be healthy, but during the execution of forwarding tasks, the link may suddenly fail due to external interference (such as instantaneous fading of wireless signals, network congestion, fluctuations in operator networks, etc.) or its own problems (such as module overheating, occasional software bugs, etc.). If it fails or its uplink is interrupted, it will not only lose its own data, but also cause the abnormal data of other gateways it is responsible for forwarding to be lost as well, evolving from a single point of failure into a local area failure.

[0006] To address these issues, this invention proposes a power distribution cabinet monitoring method, system, and intelligent power distribution cabinet based on cloud-edge fusion technology. Summary of the Invention

[0007] The purpose of this invention is to provide a method, system, and intelligent power distribution cabinet for monitoring power distribution cabinets based on cloud-edge fusion technology, so as to solve the technical problems mentioned in the background.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a power distribution cabinet monitoring method based on cloud-edge fusion technology, comprising the following steps:

[0009] Dynamically assemble edge clusters to enable multiple electrically connected smart gateways to interact collaboratively;

[0010] Within the edge cluster, a primary collaborative node and at least one backup collaborative node are dynamically elected based on the real-time load and health of each smart gateway.

[0011] When a smart gateway in the edge cluster detects abnormal local data, it segments and redundantly encodes the abnormal data to generate multiple data segments, and sends the different segments to the main coordinating node and one or more selected backup coordinating nodes respectively.

[0012] The primary and backup collaborative nodes that receive the fragments upload the received fragments to the cloud platform in parallel using their respective independent uplinks.

[0013] The primary collaborative node initiates a control instruction proposal based on the abnormal data. After obtaining verification from other nodes in the edge cluster, if no control instruction is received from the cloud platform within a preset time, the primary collaborative node sends the control instruction to the target power distribution cabinet. If a control instruction is received from the cloud platform, the primary collaborative node executes the control instruction from the cloud platform first.

[0014] Preferably, the dynamic election step specifically includes:

[0015] Periodically or when the node status changes exceed a threshold, each smart gateway broadcasts its own status indicators, which include at least real-time load and health.

[0016] Each smart gateway calculates the comprehensive score of each candidate node based on all received status indicators and according to a preset weight model.

[0017] Using a distributed consensus algorithm, the node with the highest overall score is elected as the primary coordinating node, and at least one node with the next highest score is elected as the backup coordinating node.

[0018] Preferably, the step of fragmenting and redundancy coding the abnormal data specifically includes:

[0019] The abnormal data packet is divided into K raw data fragments;

[0020] Based on the K original data fragments, M redundancy check fragments are calculated using erasure coding algorithm, where K and M are positive integers and satisfy the following condition: any K fragments selected from these K+M fragments can completely reconstruct the abnormal data packet.

[0021] The number of the one or more selected backup cooperative nodes is not less than M.

[0022] Preferably, if the primary collaborative node fails to update its status within a preset heartbeat cycle, or is detected to have a health level below a fault threshold, a fault switching process is triggered, which includes:

[0023] The backup collaborative node with the highest overall score will temporarily assume the responsibilities of the primary collaborative node.

[0024] The temporary node initiates a new round of dynamic election.

[0025] Preferably, before the main collaborative node initiates a control command proposal based on the abnormal data, an edge collaborative verification step is also included:

[0026] The primary collaborative node broadcasts an anomaly verification request to other gateways within the edge cluster. The request includes at least the anomaly type, the anomaly source gateway identifier, and the anomaly timestamp.

[0027] The gateway that receives the request acts as a verification node, retrieves electrical data associated with the abnormal timestamp from its local database, and performs correlation analysis.

[0028] Each verification node will return a verification response, including the correlation analysis results, to the main collaboration node;

[0029] If the proportion of valid responses to the abnormal data confirmed by the main collaborative node exceeds the first threshold, the control command proposal will continue to be initiated; otherwise, the proposal will be suppressed and an abnormal diagnosis report will be reported to the cloud platform.

[0030] Preferably, the correlation analysis is specifically performed in at least one of the following ways:

[0031] Compare whether there are correlated changes in the same electrical quantity monitored before and after the abnormal timestamp;

[0032] Based on the preset electrical topology and circuit model, verify whether the electrical state reflected by the abnormal data is valid in physical logic.

[0033] Check whether other types of abnormal events that have a preset causal relationship with the aforementioned abnormal type occur simultaneously.

[0034] Preferably, when detecting abnormal local data, the smart gateway further includes a data analysis step:

[0035] When any electrical parameter is detected to exceed its corresponding alarm threshold for the first time, a viewing window with a configurable duration is initiated.

[0036] If the electrical parameters continuously exceed the alarm threshold within the observation window, they are determined to be valid abnormal data.

[0037] If the electrical parameters fall below the alarm threshold and remain stable within the observation window, it is determined to be a transient disturbance, and only local log recording is performed without triggering subsequent fragmentation and upload steps.

[0038] Preferably, an emergency threshold higher than the alarm threshold is configured for each electrical parameter;

[0039] If the electrical parameter exceeds the emergency threshold upon first detection, it is immediately identified as emergency abnormal data, and the observation window is skipped to directly proceed with the steps of fragmenting and uploading the abnormal data.

[0040] A power distribution cabinet monitoring system based on cloud-edge fusion technology includes a cloud platform, multiple power distribution cabinets, and intelligent gateways installed in each cabinet;

[0041] The smart gateway is configured with:

[0042] The cluster management module is used to participate in the dynamic formation, status broadcasting, and consensus election of edge clusters;

[0043] The data processing module is used to perform local anomaly detection, data fragmentation, redundancy coding, and fragment distribution.

[0044] The communication module has an independent uplink interface for parallel uploading of fragmented data and receiving instructions;

[0045] The control agent module is used to execute or forward control commands that are verified at the edge or from the cloud platform.

[0046] The cloud platform is equipped with a data reconstruction engine and a strategy center, which are used to reconstruct fragmented data, perform global analysis, and issue control strategies.

[0047] A smart power distribution cabinet includes a cabinet, electrical equipment installed inside the cabinet, and a smart gateway. The smart gateway is communicatively connected to sensors, measuring devices, and actuators inside the cabinet and is connected to an edge cluster as a node.

[0048] The beneficial effects of this invention are:

[0049] This invention utilizes a dynamic election mechanism based on the real-time load and health status of smart gateways. This mechanism enables collaborative responsibilities within the edge cluster to be transferred according to the immediate status of each node, achieving a balanced distribution of load within the edge cluster. This improves upon or even avoids the problem of long-term overload of individual smart gateways caused by uneven spatiotemporal distribution of abnormal events. Furthermore, by introducing fragmentation and redundant coding techniques for abnormal data, combined with a multi-path parallel upload strategy, the transmission tasks originally concentrated on a single link are decomposed and distributed across multiple independent nodes and links. This constructs dual redundancy for paths and data, improving the success rate and timeliness of abnormal data reporting. It also solves the systemic risk of single-point failures spreading into localized area failures caused by the failure of collaborative gateways in traditional solutions. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the control flow of a power distribution cabinet monitoring method based on cloud-edge fusion technology according to the present invention.

[0051] Figure 2 This is a schematic diagram of the control flow for edge collaborative verification and decision-making in this invention. Detailed Implementation

[0052] 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.

[0053] Example 1

[0054] Because the abnormal data reporting of edge gateways is highly random and easily conflicts with periodic reporting tasks, some existing technologies build edge clusters to achieve the reporting of abnormal data to the cloud platform through the cooperation of multiple edge gateways. However, in actual power distribution systems, abnormal events often occur more frequently on specific lines, under specific loads, or during specific time periods. Using a collaborative gateway approach means that some gateways located in fault-prone areas will repeatedly become abnormal gateways. When an abnormal gateway (let's say gateway A) detects an anomaly, it must find a collaborative gateway that is currently in its reporting cycle. This collaborative gateway is determined by its fixed, preset reporting time slot. If the reporting time slot of gateway B happens to be aligned with the high-incidence period of gateway A's anomalies, then gateway B will be continuously selected as a collaborative gateway, resulting in excessive load on it. Other gateways with misaligned reporting time slots may remain idle for a long time.

[0055] Furthermore, the health status of a gateway's uplink is constantly changing. When a gateway is selected as a cooperating gateway, its uplink may be healthy, but during the execution of forwarding tasks, the link may suddenly fail due to external interference or its own problems. If it fails or its uplink is interrupted, it will not only lose its own data, but also cause the abnormal data of other gateways that it is responsible for forwarding to be lost as well, evolving from a single point of failure into a local area failure.

[0056] To address the aforementioned issues, this embodiment is specifically invented. This embodiment is based on a typical industrial park power distribution system scenario and is described in detail. In this scenario, a 10kV feeder is connected to five intelligent distribution cabinets, which are responsible for powering different factories within the park. They are numbered sequentially as cabinet 1 to cabinet 5. Each distribution cabinet is equipped with an intelligent gateway (corresponding to the numbers G1 to G5). These gateways have the capabilities of data acquisition, edge computing, multi-mode communication, and command execution. In the initial state, these gateways work independently and periodically package and upload the electrical parameters such as current, voltage, and power of their respective cabinets to a remote cloud monitoring platform.

[0057] Please see Figure 1 and Figure 2 As shown, an embodiment of the present invention provides a power distribution cabinet monitoring method based on cloud-edge fusion technology, which specifically includes the following steps:

[0058] S100 dynamically assembles edge clusters, enabling multiple electrically connected smart gateways to interact collaboratively.

[0059] In practice, each smart gateway first achieves microsecond-level time synchronization through a precision clock protocol. Then, each smart gateway continuously monitors the effective value of the bus voltage of its distribution cabinet and sets a dynamic threshold based on historical operating data statistics (e.g., ±10% of the nominal voltage). When the monitored voltage value continuously exceeds the dynamic threshold for more than a preset duration (e.g., 2 power frequency cycles), it is recorded as a voltage event. The system captures the precise timestamp of the event's start, the depth of the voltage drop or rise, the duration of the event, and a snapshot containing waveform data from several cycles before and after the event. When multiple smart gateways report voltage events within a similar time window (e.g., 1-2 milliseconds), the system further analyzes whether these events originate from the same source.

[0060] Specific criteria for determining the source of an event include: the difference between the start timestamps of each event is within a preset small time window; the similarity of distortion features (such as specific harmonic content and sag waveform profile) in the waveform snapshots recorded by each gateway exceeds a preset threshold (e.g., 85%); and the change in event amplitude conforms to the expected attenuation model propagating along the electrical line. Smart gateways corresponding to voltage events that satisfy all the criteria for the source of an event will be dynamically assigned to the same edge cluster by the system. All smart gateways in this edge cluster will be interconnected through a low-latency local communication network to form a logical unit that can work collaboratively. At the same time, by comparing the timestamps of when the source of an event is detected on each gateway, the approximate electrical connection order between the distribution cabinets in the cluster can be inferred. For example, the gateway that detects the event earliest may be closest to the disturbance source, thus forming a topology recognition from upstream to downstream. This topology information will serve as a reference for subsequent processes.

[0061] S200. Within the edge cluster, a primary collaborative node and at least one backup collaborative node are dynamically elected based on the real-time load and health of each smart gateway.

[0062] Once the edge cluster is established, the system initiates periodic (e.g., every 30 seconds) or triggered node role elections. Each smart gateway periodically broadcasts its status metrics, which constitute a multi-dimensional vector, mainly including two aspects: real-time load and health.

[0063] In this embodiment, real-time load includes the gateway's CPU utilization, memory usage, network I / O throughput, and the length of the current pending data task queue; health includes the hardware status (such as temperature and power stability) obtained through self-testing programs, the survival status of key software processes (such as data acquisition and communication daemons), and the uplink (to the cloud) quality assessed through active probing (such as Ping tests), with specific indicators being latency and packet loss rate.

[0064] After collecting the status broadcasts of all other nodes in the edge cluster, each smart gateway calculates a comprehensive score for each candidate node based on a preset weight model. In this embodiment, the comprehensive score Score can be calculated using the following formula:

[0065]

[0066] in, , , The preset weighting coefficients, and It is usually set to the maximum, such as 0.5, to emphasize the reliability of the uplink;

[0067] F() is a conversion function that transforms utilization rate into a score; the lower the utilization rate, the higher the score.

[0068] G() is a link quality score function calculated based on latency L (milliseconds) and packet loss rate P (percentage). In this embodiment, the link score function is as follows:

[0069]

[0070] Where A and B are adjustment coefficients.

[0071] Subsequently, the edge cluster makes decisions through a fast election mechanism based on priority and heartbeat confirmation. The node with the highest score calculated by each smart gateway is elected as the primary coordinating node, and this result is broadcast via heartbeat messages. If a node receives more than half of the total number of recommended votes in the cluster, the election is successful, and that node is officially confirmed as the new primary coordinating node. Within the same election cycle, each gateway will default to the node with the second highest comprehensive score calculated by itself as the backup coordinating node. If no node receives more than half of the recommended votes within a complete election cycle, the election is considered a failure, and the system will take the following recovery strategy based on the current state:

[0072] If the system has a valid previous master coordinating node that is still online and healthy, it continues to maintain its responsibilities and immediately shortens the interval of the next election cycle in order to quickly re-elect a new leader.

[0073] If the system is in its initial state, or if the previous master coordinating node has been marked as faulty, a new election will be initiated immediately.

[0074] The aforementioned dynamic election mechanism ensures that the nodes responsible for key forwarding and control coordination are the most capable in the current cluster, avoiding the problem of load being concentrated on a few nodes for a long time due to fixed assignment or simple round-robin.

[0075] S300. When a smart gateway in the edge cluster detects local abnormal data, it segments and redundantly encodes the abnormal data to generate multiple data segments, and sends the different segments to the main collaborative node and one or more selected backup collaborative nodes respectively.

[0076] Each smart gateway continuously runs an anomaly detection algorithm locally. For example, by comparing electrical parameters such as current and voltage with preset alarm thresholds, when a parameter is detected to be continuously exceeding the standard, such as the overcurrent duration exceeding the preset continuous judgment duration (e.g., 100 milliseconds), the smart gateway (hereinafter referred to as the source gateway) marks this event as a valid anomaly and encapsulates it into a structured anomaly data packet. To improve transmission reliability and avoid single points of failure, the source gateway then performs fragmentation and redundancy coding on the data packet.

[0077] Specifically, it uses erasure coding techniques, such as Reed-Solomon coding, and fragmentation strategy parameters. The selection criteria are as follows: the K value is usually determined based on the size of the data packets to be distributed and the reasonable transmission overhead of a single fragment; the M value (redundancy) can be dynamically adjusted based on the current scale of the edge cluster and historical link reliability statistics. For example, when the number of gateways N is large, it can be set to... In order to achieve a balance between reliability and cost.

[0078] In this embodiment, for clarity, a definition is provided. The source gateway first divides the original abnormal data packet into three equal-sized original data fragments. Then, using an encoding algorithm, it calculates two redundant check fragments based on these three original fragments. The encoding characteristics ensure that, among these five fragments, as long as any three fragments (which can be original fragments or check fragments) are successfully received subsequently, the original abnormal data packet can be completely reconstructed through the decoding algorithm. After the fragments are generated, the source gateway selects the target node for distribution based on the current cluster status. It will at least select the primary coordinating node and, based on the redundancy parameter M (which is 2 here), select one or more nodes with relatively light load and high health from the list of backup coordinating nodes. For example, it may send original fragment 1 and redundant fragment 1 to the primary coordinating node, send original fragment 2 and redundant fragment 2 to the first backup coordinating node, and retain original fragment 3 itself. Each fragment comes with necessary metadata, including a globally unique abnormal event ID and the total number of fragments. The shard's index, and a hash value used to verify integrity for subsequent reassembly.

[0079] S400: The primary and backup collaborative nodes that receive the fragments upload the received fragments to the cloud platform in parallel using their respective independent uplinks.

[0080] The primary coordinating node and the backup coordinating node that received the fragment immediately initiate independent upload tasks after confirming receipt of the fragment. They do not wait for other nodes or synchronize; instead, each uses its own dedicated, physically independent uplink communication link (e.g., node A uses a 5G network, and node B uses wired fiber optic Ethernet) to package the received fragment (along with its metadata) and upload it directly to the remote cloud platform. This process is parallel and asynchronous. The cloud platform's data receiving service continuously listens for upload streams from different gateways and IP addresses. When it receives a fragment belonging to the same exception event ID, it caches it. Once it has collected a sufficient number (i.e., at least K) of fragments, regardless of where these fragments come from... From whichever gateway it is, the cloud platform's data reconstruction engine will trigger the decoding algorithm to fully recover the original abnormal data packet for subsequent global analysis, storage, and visualization. The core advantage of this multi-path parallel upload mechanism is that it completely breaks the absolute dependence of data reporting on a single gateway or a single link. Even if the uplink of the primary coordinating node is suddenly interrupted during transmission, as long as a backup coordinating node successfully uploads its fragment, plus other possible routes, the data still has a high probability of being successfully reconstructed. The failure of a single node or link will only cause the loss of the individual fragments it is responsible for, and will not cause the entire abnormal data to be lost, thus eliminating the risk of single-point failure propagation at the architecture level.

[0081] S500: The main collaborative node initiates a control instruction proposal based on the abnormal data. After obtaining verification from other nodes in the edge cluster, if no control instruction is received from the cloud platform within a preset time, the control instruction is sent to the target power distribution cabinet. If a control instruction is received from the cloud platform, the control instruction from the cloud platform is executed first.

[0082] While abnormal data is reported in a sharded manner, the control coordination process on the edge side is started simultaneously. After the main coordination node, as one of the shard receivers, or after sensing a valid abnormal event through the internal state of the cluster, does not immediately issue control commands, it will first generate a preliminary control command proposal based on the pre-set local security rule base (for example, if the downstream cabinet has continuous overcurrent, it will instruct its upstream cabinet to trip and isolate).

[0083] To ensure security and consistency, the proposal will be validated within the edge cluster. The primary coordinating node will broadcast the proposal to other nodes (especially the backup coordinating node and the source gateway). The validation logic is as follows:

[0084] Upon receiving a proposal, the node first checks whether the target distribution cabinet involved in the proposal has a direct electrical measurement connection with itself (such as being on the same line). If so, it extracts relevant data from the local node within a specific time window before and after the abnormal timestamp (such as the current flowing to the faulty branch extracted by the upstream node), and calculates whether its changes are consistent with the abnormal mode described in the proposal in terms of amplitude, direction, and timing. If there is no direct connection, it performs logical reasoning verification based on the known topology relationship.

[0085] Each verification node responds with its verification result (agree / disagree, confidence level) to the main coordinating node within a very short time window (e.g., 200 milliseconds). If the number of nodes that agree to the proposal exceeds a preset threshold (e.g., more than half), the proposal is considered to have achieved consensus in the edge cluster.

[0086] However, it's important to note that even if a proposal gains consensus across all collaborating nodes, the primary collaborating node won't execute it immediately. It will initiate a delay timer (e.g., 500 milliseconds) to await the cloud platform's decision. After reorganizing the data, the cloud may make a judgment based on more comprehensive information (such as multiple cluster states, historical patterns, and operational strategies) and issue control commands via the downlink channel. If the primary collaborating node receives a control command from the cloud platform during the delay period, regardless of whether the command content is consistent with the edge proposal, the system will prioritize executing the cloud platform command to ensure the uniformity of the global control strategy. If no cloud command is received after the timeout, the primary collaborating node will act as a backup, executing the control proposal that has already gained edge consensus and issuing the command to the actuator (such as a smart circuit breaker) of the target distribution cabinet. This dual decision-making mechanism of cloud priority and edge backup respects the cloud's global optimization capabilities while ensuring critical rapid protection capabilities in extreme situations such as network outages or cloud response delays through edge autonomy, achieving an optimal balance between control reliability, response timeliness, and global decision-making.

[0087] In summary, this invention, through a dynamic election mechanism for the real-time load and health of smart gateways, enables collaborative responsibilities within the edge cluster to be transferred according to the immediate status of each node, achieving balanced load distribution within the edge cluster. This improves upon or even avoids the problem of long-term overload of individual smart gateways caused by uneven spatiotemporal distribution of abnormal events. Furthermore, by introducing fragmentation and redundant coding techniques for abnormal data, combined with a multi-path parallel upload strategy, the transmission tasks originally concentrated on a single link are decomposed and distributed across multiple independent nodes and links, constructing dual redundancy for paths and data. This improves the success rate and timeliness of abnormal data reporting and solves the systemic risk of single-point failures spreading to localized area failures caused by the failure of collaborative gateways in traditional solutions.

[0088] In addition, the established cloud-first, edge-backup collaborative control mechanism, while respecting the authority of global decision-making in the cloud, empowers the edge cluster with autonomous protection capabilities in extreme situations, achieving the optimal balance of control reliability, response timeliness, and global strategy, and significantly improving the resilience and availability of the entire power distribution monitoring system.

[0089] Example 2

[0090] Given the complexity of abnormal signal sources in power distribution networks—including real faults that must be quickly isolated (such as persistent short circuits) as well as numerous harmless transient disturbances (such as brief voltage drops) or sensor false alarms—if the system lacks the ability to differentiate and triggers a complete reporting and control chain for all exceeding signals indiscriminately, it will not only cause a large number of meaningless transient events to occupy valuable communication and computing resources and interfere with operation and maintenance judgments, but may also lead to the system overreacting to non-fault events. For example, if the edge collaboration mechanism does not issue instructions in the cloud because it determines that the event is a transient anomaly, it may still perform local circuit breaker operations, causing unnecessary power outages. Therefore, how to enable the cloud-edge converged system to have the ability to intelligently distinguish the authenticity, persistence, and severity of abnormal events, and greatly suppress false triggering while ensuring safe response, has become the key to improving system availability and intelligence.

[0091] This embodiment is a further improvement on the above embodiment.

[0092] Please see Figure 1 and Figure 2 As shown, when the smart gateway detects abnormal local data, it also includes a data analysis step, the details of which are as follows:

[0093] S600: When any electrical parameter is detected to exceed its corresponding alarm threshold for the first time, a viewing window with a configurable duration is started.

[0094] In practice, each smart gateway maintains a parameter-threshold mapping table and associated judgment strategy. This table not only defines the alarm thresholds for various electrical parameters, but also presets an independent, dynamically configurable observation window duration for each parameter.

[0095] Furthermore, the setting of the above observation window duration follows these principles: For fast electrical quantities such as current and voltage that reflect electromagnetic transient processes, the window should be able to cover the typical minimum duration of this type of fault (e.g., the minimum operating time of protection equipment or the steady-state establishment time of a short-circuit fault), typically set to 2-10 power frequency cycles (i.e., 40-200 milliseconds); For thermally slow-changing parameters such as temperature, the window should be able to filter out normal temperature fluctuations caused by equipment start-up and shutdown and cooling fan operation, typically set to tens of seconds to several minutes. The specific duration can be uniformly distributed by the cloud strategy center based on the analysis of historical abnormal events of this cluster and optimized through machine learning models.

[0096] When the real-time sampling and calculation module of the smart gateway detects that a parameter value exceeds its alarm threshold for the first time, the local judgment state machine is immediately activated. It does not immediately assert a fault, but marks the parameter as a suspected abnormal state. At the same time, it starts a dedicated observation window timer for the parameter event. During this period, the smart gateway will automatically increase the high-speed sampling frequency of the parameter in order to record its dynamic trajectory more delicately and accumulate high-quality data for subsequent judgment.

[0097] S700. If the electrical parameters continuously exceed the alarm threshold within the observation window, they are determined to be valid abnormal data.

[0098] If the electrical parameters fall below the alarm threshold and remain stable within the observation window, it is determined to be a transient disturbance, and only local log recording is performed without triggering subsequent fragmentation and upload steps.

[0099] Specifically, after the observation window is launched, the system enters a crucial judgment phase. The continuous monitoring data within the window serves as the basis for judgment, addressing different scenarios, including:

[0100] Scenario 1: If a valid anomaly is confirmed, and the value of the target electrical parameter remains above the alarm threshold throughout the entire observation window, or if it fluctuates slightly but fails to fall below the threshold and remain stable for more than a very short period (e.g., five consecutive sampling points are below the threshold), then when the observation window timer expires, the analysis engine will make a final judgment: this is a valid anomaly or a persistent anomaly. This means that the anomaly is not accidental and is very likely to indicate a real, developing fault or abnormal state. At this time, the smart gateway ends the local analysis phase and immediately triggers the complete set of collaborative processing procedures described in Implementation Example 1: constructing a structured data packet containing detailed timestamps, parameter trajectories, anomaly types, etc., then performing fragmentation and redundant coding, and preparing to distribute it to the dynamically elected primary and backup collaborative nodes.

[0101] Scenario 2: Identified as a transient disturbance. More commonly, after a parameter exceeds a threshold, it falls back and stabilizes below the alarm threshold before the observation window ends. For example, a voltage drop caused by switching on or off a nearby device may automatically recover within 2-3 cycles. If the parameter value falls back below the threshold within the window period and remains stable for the remaining window time (or an additionally defined short stabilization period, such as 20 milliseconds), the analysis engine determines that the event is a transient disturbance. For such events, the gateway generates a concise event log, recording the parameter type. The maximum out-of-limit value, occurrence time, and duration are stored in local circular storage or non-volatile memory for possible local traceability analysis. Most importantly, the processing ends here, without activating any subsequent steps involving edge cluster collaboration or cloud platform reporting. This achieves noise filtering at the source, digesting massive amounts of harmless transient interference locally, preventing them from flooding into subsequent valuable communication links and computing resources. This greatly improves the signal-to-noise ratio of the entire system's data flow and fundamentally eliminates the possibility of triggering unnecessary control links due to transient disturbances.

[0102] However, the security defense of power distribution systems needs to balance accuracy and timeliness. For extreme situations that may deteriorate rapidly and endanger equipment or even personal safety, such as a sudden increase in current caused by a metallic short circuit, waiting for a complete observation window before making a judgment and reporting will delay valuable time for accident isolation. Therefore, this embodiment adds a step for processing emergency abnormal data, specifically including:

[0103] S800: Configure an emergency threshold higher than the alarm threshold for each electrical parameter;

[0104] If the electrical parameter exceeds the emergency threshold upon first detection, it is immediately identified as emergency abnormal data, and the observation window is skipped, directly triggering the subsequent fragmentation and uploading steps.

[0105] For each electrical parameter, a higher emergency threshold is set above the alarm threshold. The setting of this threshold is often directly related to the physical limits of the equipment and the instantaneous trip setting value of the protection device. For example, the emergency threshold of the current may be set to 200% or higher of the rated value, directly pointing to the characteristics of a short circuit fault. When the monitoring module finds that a parameter value exceeds the emergency threshold on the first sampling, the local judgment engine will immediately start the highest priority emergency path. The design principle of this path is time priority. The local judgment engine will bypass the above-mentioned suspected anomaly marking, observation window start and waiting process, and make the final judgment of emergency anomaly directly within a very short system cycle (e.g., milliseconds).

[0106] Once this determination is made, it is immediately inserted at the front of the gateway's processing queue as a highest priority task event. It wakes up and drives all subsequent collaborative processing modules in Implementation Example 1 with almost no delay. Operations such as encapsulation, fragmentation, encoding, and distribution of abnormal data packets will be executed with the highest priority, seizing system resources to ensure that information about this critical situation can enter the edge collaborative network as quickly as possible and attempt to report it to the cloud. This ensures the fastest perception and response speed when a real crisis occurs, providing a crucial time prerequisite for achieving rapid fault isolation and protection.

[0107] In summary, this invention deploys an analysis engine with multi-level thresholds and configurable observation windows on each smart gateway to identify and filter massive amounts of self-recovering transient disturbances or sporadic sensor noise locally. These disturbances are only logged without triggering subsequent collaborative processes, significantly reducing the impact of invalid data on communication, computing resources, and uplink network bandwidth within the edge cluster. This greatly improves the signal-to-noise ratio of the entire system's data flow, allowing maintenance personnel to focus on real threats.

[0108] Furthermore, an emergency threshold is set for serious faults that may deteriorate rapidly, allowing the system to skip the regular observation process and immediately initiate reporting and response with the highest priority, ensuring rapid detection of critical events such as short circuits and buying critical time for protection actions.

[0109] Example 3

[0110] Based on the two embodiments described above, this implementation provides a power distribution cabinet monitoring system based on cloud-edge fusion technology, including a cloud platform, multiple power distribution cabinets, and intelligent gateways installed in each cabinet. These components are interconnected through a network, wherein the intelligent gateways are connected through a local area network to form the internal communication backbone of the edge cluster; each intelligent gateway has its own independent uplink communication link for connecting to the Internet, and finally establishes a secure data tunnel with the cloud platform. The cloud platform acts as a centralized service center, providing powerful computing, storage, and analysis capabilities.

[0111] The smart gateway is configured with a cluster management module, a data processing module, a communication module, and a control agent module.

[0112] The aforementioned cluster management module is responsible for running a precision time synchronization protocol and maintaining a local high-precision clock; continuously executing voltage event detection and homology analysis algorithms to dynamically discover electrically related neighbors and maintain a local cluster member list and topology table; periodically collecting local status (CPU, memory, uplink quality, etc.), encapsulating it into status indicator messages, and broadcasting them through the local network; simultaneously, it also receives status broadcasts from other nodes, maintains a real-time updated neighbor status table, and when an election is triggered, this module calculates a score based on a preset weight model and runs a distributed consensus algorithm to participate in the voting and confirmation process of the primary and backup cooperating nodes, ensuring that the cluster always has a healthy and reliable organizational structure.

[0113] The data processing module comprises several sub-functional units: the data acquisition and standardization unit is responsible for reading raw data from devices such as smart meters, temperature sensors, and humidity sensors connected within the cabinet via protocols such as RS-485 and Modbus TCP, and performing cleaning, calibration, and format unification; the anomaly detection and analysis unit runs a local detection algorithm, compares electrical parameters with multi-level thresholds (alarm thresholds, emergency thresholds) in real time, and executes intelligent analysis logic with a configurable observation window to distinguish between transient disturbances and valid anomalies. Upon confirmation of a valid or emergency anomaly, the fragmentation and encoding unit is activated. This unit, based on the current cluster size and configured strategy (such as (K, M) parameters), uses a specific erasure coding library to fragment and redundantly encode structured abnormal data packets, generating data fragments and verification fragments. Subsequently, the fragmentation and distribution unit, based on the current primary and backup node information and their health scores provided by the cluster management module, packages different fragments according to a predetermined strategy and sends them to the target gateway via the local network.

[0114] The communication module manages one or more physically independent uplink interfaces. For example, a gateway can be configured with a 5G wireless module and a wired fiber optic Ethernet port at the same time. The communication module continuously monitors the health status (latency, packet loss rate) of each uplink. When it receives fragmented data from the data processing module, it immediately encapsulates the fragmented data into TCP / UDP packets through the currently optimal one or more active uplinks and uploads them to the designated receiving endpoint of the cloud platform in parallel and asynchronously. At the same time, the module also listens to the downlink port of the cloud platform, receives control commands or configuration updates from the cloud, and reliably forwards them to the control agent module.

[0115] The control agent module receives two types of instruction inputs: first, local control instruction proposals from the gateway cluster management module that have been verified by edge collaboration and not rejected by the cloud; second, control instructions directly from the cloud platform via the communication module. This module has an embedded instruction arbitration logic, the core principle of which is that cloud instructions have absolute priority. When both types of instructions are received simultaneously, the cloud instruction is executed first; if only the edge instruction is received and no cloud instruction is received within the preset waiting time, the edge instruction is executed. The control agent module converts the instructions into specific device control protocols and sends them to the actuators in the distribution cabinet, such as smart circuit breakers, load switches, and contactors, through the corresponding output interfaces to realize operations such as opening, closing, and interlocking.

[0116] The cloud platform is deployed on cloud servers or private data centers. Its data reconstruction engine is responsible for receiving data streams asynchronously uploaded from tens of thousands of smart gateways across the network. For regular periodic data, it is directly parsed and stored in the database. For abnormal data reported using fragmentation, the engine can group fragments belonging to the same event from different IP addresses and at different times in the cache based on information such as the abnormal event ID, total number of fragments, and index in the fragment metadata. Once a sufficient number (K) of fragments are collected, the corresponding erasure coding decoding algorithm is called to completely reconstruct the original abnormal data packet. Subsequently, the complete data packet is sent for further analysis. The pipeline and strategy center are the intelligent core of the cloud platform, containing multiple analytical microservices: the global situational analysis service can correlate events across clusters and regions to identify systemic risks; historical data mining and machine learning models run continuously to optimize the threshold parameters, observation window duration, and sharding strategy parameters (K, M) issued to each edge cluster; the rule engine maintains high-level protection and control policies, and when analysis confirms that intervention is needed, it generates specific control instructions. The strategy center issues control policies or instructions to the main coordinating node or specific intelligent gateway of the target edge cluster in real time and in a targeted manner through a reliable message queue or API interface.

[0117] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for monitoring power distribution cabinets based on cloud-edge fusion technology, characterized in that, Includes the following steps: Dynamically assemble edge clusters to enable multiple electrically connected smart gateways to interact collaboratively; Within the edge cluster, a primary collaborative node and at least one backup collaborative node are dynamically elected based on the real-time load and health of each smart gateway. When a smart gateway in the edge cluster detects abnormal local data, it segments and redundantly encodes the abnormal data to generate multiple data segments, and sends the different segments to the main coordinating node and one or more selected backup coordinating nodes respectively. The primary and backup collaborative nodes that receive the fragments upload the received fragments to the cloud platform in parallel using their respective independent uplinks. The primary collaborative node initiates a control instruction proposal based on the abnormal data. After obtaining verification from other nodes in the edge cluster, if no control instruction is received from the cloud platform within a preset time, the primary collaborative node sends the control instruction to the target power distribution cabinet. If a control instruction is received from the cloud platform, the primary collaborative node executes the control instruction from the cloud platform first.

2. The power distribution cabinet monitoring method based on cloud-edge fusion technology according to claim 1, characterized in that, The steps of the dynamic election specifically include: Periodically or when the node status changes exceed a threshold, each smart gateway broadcasts its own status indicators, which include at least real-time load and health. Each smart gateway calculates the comprehensive score of each candidate node based on all received status indicators and according to a preset weight model. Using a distributed consensus algorithm, the node with the highest overall score is elected as the primary coordinating node, and at least one node with the next highest score is elected as the backup coordinating node.

3. The method for monitoring power distribution cabinets based on cloud-edge fusion technology according to claim 1, characterized in that, The process of fragmenting and redundancy encoding the abnormal data specifically includes: The abnormal data packet is divided into K raw data fragments; Based on the K original data fragments, M redundancy check fragments are calculated using erasure coding algorithm, where K and M are positive integers and satisfy the following condition: any K fragments selected from these K+M fragments can completely reconstruct the abnormal data packet. The number of the one or more selected backup cooperative nodes is not less than M.

4. The power distribution cabinet monitoring method based on cloud-edge fusion technology according to claim 1, characterized in that, If the primary collaborative node fails to update its status within a preset heartbeat cycle, or is detected to have a health level below a fault threshold, a fault switching process is triggered, which includes: The backup collaborative node with the highest overall score will temporarily assume the responsibilities of the primary collaborative node. The temporary node initiates a new round of dynamic election.

5. A method for monitoring power distribution cabinets based on cloud-edge fusion technology according to claim 1, characterized in that, Before the primary collaborative node initiates a control command proposal based on the abnormal data, an edge collaborative verification step is also included: The primary collaborative node broadcasts an anomaly verification request to other gateways within the edge cluster. The request includes at least the anomaly type, the anomaly source gateway identifier, and the anomaly timestamp. The gateway that receives the request acts as a verification node, retrieves electrical data associated with the abnormal timestamp from its local database, and performs correlation analysis. Each verification node will return a verification response, including the correlation analysis results, to the main collaboration node; If the proportion of valid responses to the abnormal data confirmed by the main collaborative node exceeds the first threshold, the control command proposal will continue to be initiated; otherwise, the proposal will be suppressed and an abnormal diagnosis report will be reported to the cloud platform.

6. A method for monitoring power distribution cabinets based on cloud-edge fusion technology according to claim 5, characterized in that, The correlation analysis specifically includes at least one of the following methods: Compare whether there are correlated changes in the same electrical quantity monitored before and after the abnormal timestamp; Based on the preset electrical topology and circuit model, verify whether the electrical state reflected by the abnormal data is valid in physical logic. Check whether other types of abnormal events that have a preset causal relationship with the aforementioned abnormal type occur simultaneously.

7. A method for monitoring power distribution cabinets based on cloud-edge fusion technology according to claim 1, characterized in that, When detecting abnormal local data, the smart gateway also includes a data analysis step: When any electrical parameter is detected to exceed its corresponding alarm threshold for the first time, a viewing window with a configurable duration is initiated. If the electrical parameters continuously exceed the alarm threshold within the observation window, they are determined to be valid abnormal data. If the electrical parameters fall below the alarm threshold and remain stable within the observation window, it is determined to be a transient disturbance, and only local log recording is performed without triggering subsequent fragmentation and upload steps.

8. A method for monitoring power distribution cabinets based on cloud-edge fusion technology according to claim 7, characterized in that, Configure an emergency threshold higher than the alarm threshold for each electrical parameter; If the electrical parameter exceeds the emergency threshold upon first detection, it is immediately identified as emergency abnormal data, and the observation window is skipped to directly proceed with the steps of fragmenting and uploading the abnormal data.

9. A power distribution cabinet monitoring system based on cloud-edge fusion technology, used to implement the monitoring method according to any one of claims 1-8, characterized in that, This includes a cloud platform, multiple power distribution cabinets, and smart gateways installed in each cabinet; The smart gateway is configured with: The cluster management module is used to participate in the dynamic formation, status broadcasting, and consensus election of edge clusters; The data processing module is used to perform local anomaly detection, data fragmentation, redundancy coding, and fragment distribution. The communication module has an independent uplink interface for parallel uploading of fragmented data and receiving instructions; The control agent module is used to execute or forward control commands that are verified at the edge or from the cloud platform. The cloud platform is equipped with a data reconstruction engine and a strategy center, which are used to reconstruct fragmented data, perform global analysis, and issue control strategies.

10. An intelligent power distribution cabinet for performing the monitoring method according to any one of claims 1-8, characterized in that, It includes a cabinet, electrical equipment installed inside the cabinet, and a smart gateway. The smart gateway is communicatively connected to sensors, measuring devices, and actuators inside the cabinet and is connected to the edge cluster as a node.

Images