Switching power supply exception handling method and device based on edge computing gateway and medium

Abstract

The invention provides a switching power supply exception handling method and device based on an edge computing gateway and a medium, and relates to the technical field of switching power supply exception handling, and the method comprises the steps: building a network knowledge graph corresponding to the edge computing gateway according to the communication data between the edge computing gateway and all connected terminal devices; generating a sub-data feature vector corresponding to each edge according to the data statistical features in the first-in first-out queue of each edge; performing feature extraction on the network knowledge graph by using a preset model to obtain network knowledge graph features corresponding to the network knowledge graph; inputting the network knowledge graph features into a preset anomaly degree prediction model to obtain an anomaly degree theta corresponding to the current network knowledge graph; sending the plurality of communication data to a cloud server according to theta so as to perform preset depth anomaly detection through the cloud server; according to the invention, the accuracy of anomaly detection, the convenience of fault tracing and the intelligent level of operation and maintenance management of the switching power supply system can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of switching power supply abnormality processing, and in particular to a switching power supply abnormality processing method, device and medium based on an edge computing gateway. Background Art

[0002] In high-frequency switching power supply systems, with the application of edge computing technology, edge computing gateways need to interact with a large number of terminal devices (such as power modules, sensors, control units, etc.) at high frequencies to collect operating parameters such as current, voltage, temperature, and communication instructions between devices in real time. However, there are currently numerous technical pain points. First, the volume of data communicated between terminal devices and gateways is vast and time-series-dependent. Uploading all raw data to cloud servers in real time without filtering would consume significant network bandwidth, increasing data transmission costs and impacting the timeliness of exception responses due to transmission delays. This is particularly true for systems with extremely high real-time requirements, such as high-frequency switching power supplies, which can lead to fault propagation. Second, existing anomaly detection methods often focus on isolated parameters of a single device (e.g., output voltage fluctuations of a single power module), ignoring the correlations formed through data interaction between devices (e.g., a sensor anomaly can trigger coordinated failures of multiple power modules). This makes it difficult to capture systemic failures caused by changes in network topology or abnormal data flows, leading to misjudgments or missed anomalies. Furthermore, when an anomaly is detected, existing solutions typically only upload the anomaly data at the current moment, lacking the ability to trace the anomaly's development process. This inability to provide a complete anomaly evolution trajectory for in-depth cloud analysis makes fault location difficult and processing inefficiency low. These issues severely hinder the intelligent operation and maintenance of switching power supply systems. A solution that balances data transmission efficiency, correlation analysis, and anomaly tracing is urgently needed. Summary of the Invention

[0003] In view of the above technical problems, the technical solution adopted by the present invention is: According to a first aspect of the present application, a method for handling abnormalities in a switching power supply based on an edge computing gateway is provided, the method comprising the following steps: S100: Establish a network knowledge graph corresponding to the edge computing gateway based on communication data between the edge computing gateway and all connected terminal devices; the network knowledge graph includes a number of nodes and directed edges connecting the nodes; each node corresponds to a terminal device, and the direction of the edge is the direction of data flow between the corresponding terminal devices; each edge corresponds to a first-in, first-out queue; S200, generating a sub-data feature vector corresponding to each side according to the statistical features of the data in the first-in-first-out queue of each side at every preset interval; S300, using a preset model to extract features from the network knowledge graph to obtain network knowledge graph features corresponding to the network knowledge graph; S400, inputting the network knowledge graph features into a preset abnormality prediction model to obtain the abnormality θ corresponding to the current network knowledge graph; S500, if θ < γ1, then the data statistical features corresponding to the data entering each queue within the preset time period are sent to the cloud server; γ1 is the first preset abnormality threshold; S600, if θ≥γ1, then traverse each historical abnormality in sequence until a target historical abnormality θ' is determined; θ'≤γ2; γ2<γ1; γ2 is a second preset abnormality threshold; S700 , sending a number of communication data after the time point corresponding to θ′ to the cloud server, so as to perform a preset deep anomaly detection through the cloud server.

[0004] According to another aspect of the present application, a non-transitory computer-readable storage medium is also provided, in which at least one instruction or at least one program is stored, and the at least one instruction or at least one program is loaded and executed by a processor to implement the above-mentioned switching power supply abnormality handling method based on the edge computing gateway.

[0005] According to another aspect of the present application, an electronic device is provided, including a processor and the above-mentioned non-transitory computer-readable storage medium.

[0006] The present invention has at least the following beneficial effects: The switching power supply abnormality processing method based on edge computing gateway of the present invention can organically integrate the scattered device communication data and the relationship between devices by constructing a network knowledge graph including terminal device nodes, data flow edges and corresponding first-in-first-out queues. By extracting and analyzing the graph features with the help of a preset model, it breaks through the limitations of traditional single device parameter monitoring and fully explores the correlation and timing laws of data interaction between devices, thereby more accurately identifying systemic faults caused by network topology changes or data flow abnormalities, and greatly reducing the probability of abnormal misjudgment and missed judgment; through a hierarchical data uploading mechanism based on abnormality degree, only traditional data is uploaded when the system is running smoothly (with low abnormality degree). By collecting characteristic data, the original data transmission volume can be significantly reduced, network bandwidth pressure can be alleviated, data transmission costs can be reduced, and the real-time delivery of key information can be guaranteed. When a significant anomaly is detected, by tracing back to the normal state time point and uploading the relevant communication data thereafter, a complete anomaly evolution trajectory can be provided for cloud-based deep detection, avoiding the loss of key information and helping to quickly locate the root cause of the fault. Overall, this solution takes into account real-time performance, relevance, and data transmission efficiency, effectively improving the accuracy of anomaly detection in the switching power supply system, the convenience of fault tracing, and the intelligent level of operation and maintenance management, providing more reliable protection for systems such as high-frequency switching power supplies that have extremely high stability requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0008] Figure 1 A flowchart of a method for handling abnormalities in a switching power supply based on an edge computing gateway provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0009] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0010] It should be noted that, based on this disclosure, those skilled in the art will appreciate that an aspect described herein can be implemented independently of any other aspect, and that two or more of these aspects can be combined in various ways. For example, any number of the aspects described herein can be used to implement an apparatus and / or practice a method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement such an apparatus and / or practice such a method.

[0011] The following will refer to Figure 1 The flowchart of the switching power supply abnormality processing method based on the edge computing gateway shown in the figure introduces a switching power supply abnormality processing method based on the edge computing gateway.

[0012] The method in this embodiment is applicable to high-frequency switching power supply network systems that include edge computing gateways and multiple types of terminal devices. It is particularly well-suited for industrial-grade high-frequency switching power supply cluster operation and maintenance scenarios with stringent real-time and stability requirements, such as high-frequency switching power supply arrays in data centers, power supply systems for communication base stations, and high-frequency power supply module groups in industrial automation production lines. In such scenarios, the edge computing gateway must frequently exchange data with multiple terminal devices, including the high-frequency switching power supply itself, various sensors (current, voltage, temperature, vibration, etc.), the power supply's internal control unit (MCU / PLC), drive circuits, load monitoring equipment, cooling systems, and local monitoring terminals. Complex data flows form an interconnected network between these devices, and overall power system fluctuations are easily caused by device coordination anomalies, data transmission failures, or local component failures during system operation. Accurate anomaly detection and efficient data processing mechanisms are essential to ensure stable operation of the power system.

[0013] The method for handling abnormalities of a switching power supply based on an edge computing gateway may include the following steps: S100, based on the communication data between the edge computing gateway and all connected terminal devices, establish a network knowledge graph corresponding to the edge computing gateway; the network knowledge graph includes several nodes and directed edges connecting the nodes; each node corresponds to a terminal device, and the direction of the edge is the data flow between the corresponding terminal devices; each edge corresponds to a first-in-first-out queue.

[0014] Node definition: First, through the initial communication handshake information between the edge computing gateway and the terminal device, all associated terminal devices (such as high-frequency switching power supply body, current / voltage sensor, power supply built-in MCU, cooling fan sensor, load monitor, etc.) are identified, and each terminal device is mapped as a node in the network knowledge graph. At the same time, the edge computing gateway itself is included in the graph as a special node.

[0015] Directed edge construction: Based on the actual data transmission records between devices (such as the sensor sending collected data to the gateway, the gateway sending control instructions to the power supply MCU, the power supply body feeding back output status to the load monitor, etc.), directed edges are established between corresponding nodes. The direction of the edge strictly follows the data flow (for example, "current sensor node → edge gateway node" indicates that the sensor transmits data to the gateway).

[0016] FIFO queue configuration: Bind an independent first-in, first-out queue to each directed edge, storing the communication data between the two devices corresponding to that edge in chronological order (for example, the queue for the "gateway → power MCU" edge stores the voltage adjustment command at time t1 and the frequency control signal at time t2, etc.).

[0017] For example: In the power supply system of a communication base station, the network knowledge graph may contain four nodes: "edge gateway", "base station power module", "temperature sensor", and "load monitor"; the directed edges are "temperature sensor → edge gateway" (transmitting temperature data), "edge gateway → base station power module" (transmitting control instructions), and "base station power module → load monitor" (transmitting output current data). The FIFO queue of each edge stores the corresponding data flow records.

[0018] By presenting the relationship between devices and the data interaction logic in a structured manner through graphs, we break through the limitations of traditional "isolated device monitoring" and lay the foundation for capturing system-level correlation anomalies (such as a sensor anomaly causing coordinated failures of multiple devices); the FIFO queue retains the timing characteristics of the data, providing the original basis for subsequent timing feature analysis.

[0019] Furthermore, step S100 may include the following steps: S110, establishing a node corresponding to each terminal device connected to the edge computing gateway.

[0020] The edge computing gateway initiates a handshake with connected terminal devices through an initialization communication protocol (such as MQTT or Modbus), obtaining each device's unique identifier (such as device ID and MAC address) and basic attributes (such as device type, rated parameters, and installation location). Each terminal device is mapped as an independent node in the network knowledge graph, with the node attributes storing the device's unique identifier and basic attributes (for example, the attributes of the "Current Sensor A" node are [ID=CS001, Type=Current Sensor, Range=0-50A]). The edge computing gateway itself is also included in the graph as a special node to clarify its role as a transit node in data flow.

[0021] By mapping terminal devices to unique nodes, the "entity" basis of the network knowledge graph is clarified. The storage of node attributes provides the inherent information of the device itself for subsequent feature analysis (such as the sensor range affecting the data anomaly judgment standard), avoiding device confusion and laying the foundation for the accurate construction of the graph.

[0022] S120, for any two nodes D1 and D2, if the data sent by the terminal device corresponding to D1 is sent to the terminal device corresponding to D2 through the edge computing gateway, a directed edge is established from D1 to D2.

[0023] The edge computing gateway records the source address (sending device) and destination address (receiving device) of all packets forwarded by it in real time. For any two nodes D1 (corresponding to device X) and D2 (corresponding to device Y), if it detects that a packet sent by device X is forwarded to device Y via the gateway (i.e., the packet's source address is X's ID, its destination address is Y's ID, and it is relayed via the gateway), a directed edge from D1 to D2 is created in the graph, with the edge attribute marked "forwarded via gateway."

[0024] The indirect data flow of "device → gateway → device" is accurately depicted through directed edges, clarifying the one-way interactive relationship of data transferred through the gateway. This provides a topological basis for subsequent analysis of "the impact of upstream device anomalies on downstream devices" (such as abnormal sensor data causing HMI display errors).

[0025] S130, if the data sent by the terminal device corresponding to D2 is sent to the terminal device corresponding to D1 through the edge computing gateway, a directed edge is established from D2 to D1.

[0026] Symmetrical to the logic of S120, when the edge gateway detects that the data packet sent by device Y (corresponding to node D2) is forwarded by itself to device X (corresponding to node D1), a directed edge from D2 to D1 is established in the graph. The attribute of the edge is also marked as "forwarded via gateway" and exists independently of the "D1→D2" edge that may exist in S120 (that is, two-way data interaction corresponds to two directed edges in opposite directions).

[0027] It supports the depiction of two-way data interaction, avoiding omission of reverse data flow from "downstream devices to upstream devices" (such as control instructions and status feedback), and fully presents the closed-loop interaction formed between devices through the gateway (such as HMI issuing instructions → power supply execution → HMI receiving feedback), providing a complete topology for analyzing the "impact of abnormal control instructions on device response."

[0028] S140: If there is no data transmission between the terminal device corresponding to D1 and the terminal device corresponding to D2, no edge connection is established between D1 and D2.

[0029] The edge gateway collects statistics on the communication records between all terminal devices within a preset monitoring period (such as 10 minutes after initialization). If there has never been any data packet forwarded by the gateway between nodes D1 (device X) and D2 (device Y) (including both directions from X to Y and from Y to X), no edge connection is established between the two nodes, and the nodes remain independent.

[0030] By simplifying the graph structure by following the rule of "no connection without data transmission", we can avoid redundant edges interfering with subsequent feature extraction (for example, edges between unrelated devices will dilute effective associations), reduce graph complexity, and improve the efficiency and accuracy of feature extraction from the GNN model.

[0031] S200 , generating a sub-data feature vector corresponding to each side according to the statistical features of the data in the first-in-first-out queue of each side at every preset time interval.

[0032] Furthermore, step S200 may include the following steps: S210, obtain the initial sub-data feature vector E = (E1, E2, ..., E j ,…,E m ), j = 1, 2, …, m; E j is the empty vector corresponding to the j-th preset communication protocol, and m is the number of preset communication protocols.

[0033] First, we define the commonly used preset communication protocols between the terminal devices and the edge gateway in the switching power supply system (such as Modbus, MQTT, Profinet, EtherCAT, etc. commonly used in industrial scenarios). Let the total number of protocols be m. We assign an empty vector to each protocol, and the dimension of the empty vector is consistent with the number of communication parameters that need to be counted for the protocol (for example, if each protocol needs to count 3 parameters, then E j is a 3-dimensional empty vector); the empty vectors of all protocols are combined in a fixed order to form the initial sub-data feature vector E.

[0034] By presetting the protocol and initial vectors with fixed dimensions, the structure of the sub-data feature vectors is standardized, ensuring that the feature vectors of different edges (data interaction links between devices) have a unified format, avoiding feature dimension confusion caused by protocol differences, and providing a consistent input basis for feature aggregation of subsequent graph neural networks (GNNs).

[0035] S220, for any QR on the side, obtain the communication parameters of the data in the first-in-first-out queue of the QR under different communication protocols at every preset time interval; the communication parameters include: total communication data volume, total number of communication data items, and communication frequency.

[0036] For any directed edge QR in the network knowledge graph (corresponding to the data flow between terminal devices Q and R), its bound FIFO queue stores all communication data transmitted between Q and R through the edge gateway in chronological order. At intervals of a preset duration (such as 5 seconds, which can be adjusted dynamically), the edge gateway parses the data in the queue: The data is classified according to the preset communication protocol through the protocol identifier in the data packet header (such as the function code of Modbus and the fixed header field of MQTT).

[0037] Parameter calculation: For each type of protocol data, three communication parameters are calculated separately: Total communication data volume: The total number of bytes of data packets transmitted within the preset time length under the protocol (for example, if 10 data packets are transmitted under the Modbus protocol, the total number of bytes is 200B).

[0038] Total number of communication data items: The total number of data packets within the preset time length under the protocol (for example, a total of 8 data items are transmitted under the MQTT protocol).

[0039] Communication frequency: the ratio of the total number of communication data to the preset duration (for example, the Profinet protocol transmits 20 data items within 5 seconds, with a frequency of 4 items per second).

[0040] For the communication data stored in the queue, key statistical features are calculated, including but not limited to: data transmission frequency (the number of data packets per unit time), mean / variance of data values ​​(such as the degree of fluctuation of current sensor data), proportion of outliers (the proportion of data packets outside the normal range), standard deviation of data transmission intervals (reflecting transmission stability), etc.

[0041] For example, if edge QR is "power module Q → load monitor R," and data packets within its FIFO queue are parsed within 5 seconds, the Modbus protocol transmits 12 voltage monitoring data items (144 bytes total, at a rate of 12 / 5 = 2.4 items / second); the MQTT protocol transmits 5 status feedback data items (60 bytes total, at a rate of 1 item / second); and the Profinet protocol transmits no data. Therefore, the Modbus parameters are (144 bytes, 12 items, 2.4 items / second), and the MQTT parameters are (60 bytes, 5 items, 1 item / second).

[0042] By distinguishing communication protocols and extracting core parameters, the data interaction intensity (total data volume), density (total number of entries), and stability (frequency) between devices under different protocols are accurately characterized. This avoids feature ambiguity caused by mixing data from different protocols (such as the real-time requirements of Modbus and the asynchronous characteristics of MQTT), and provides a more fine-grained basis for anomaly detection (for example, a sudden drop in the frequency of a certain protocol may indicate an anomaly in the link).

[0043] S230, filling the communication parameters under different communication protocols into the corresponding empty vectors in E to obtain the sub-data feature vector corresponding to QR.

[0044] The communication parameters of each protocol calculated in S220 are filled into the corresponding empty vectors in the initial vector E of S210 in the order of the protocols: if a protocol has no data within the preset time length, the corresponding empty vector is filled with a zero vector; finally, a complete sub-data feature vector corresponding to the edge QR is formed.

[0045] Integrating the scattered protocol parameters into structured feature vectors not only retains the independent features of different protocols (facilitating the analysis of anomalies in a certain protocol), but also achieves comparability of cross-edge features through a unified format (such as comparing the MQTT protocol frequency differences between the "sensor→gateway" and "gateway→power" edges). This provides standardized, high-information-density input for subsequent GNN extraction of global network features, improving the efficiency and accuracy of feature aggregation.

[0046] S300, using a preset model to extract features of the network knowledge graph to obtain network knowledge graph features corresponding to the network knowledge graph.

[0047] Furthermore, the preset model is GNN.

[0048] Graph neural network (GNN, such as GCN or GAT) is used as the feature extraction model because it can effectively process graph structure data and capture the relationship between nodes and edges.

[0049] Feature input: The network knowledge graph and the sub-data feature vectors of the edges generated by S200 are used as model input.

[0050] Graph feature aggregation: GNN uses a "message passing mechanism" to allow each node to aggregate the features of its neighboring nodes and connected edges (for example, the "edge gateway" node aggregates the features of neighbors such as temperature sensors and power modules), and ultimately outputs a global vector, namely the network knowledge graph feature, which contains the topological relationship and data interaction status information of the entire network.

[0051] In a data center power array, GNN aggregates the node attributes of "power module A" (rated power 500W), the feature vector of the "power module A→load B" edge (transmission current average 2A, small fluctuation), and the feature vector of the "edge gateway→power module A" edge (stable control instruction frequency) to generate a graph feature that reflects the operating status of the entire power array.

[0052] Through the association modeling capabilities of GNN, scattered device attributes and edge features are integrated into global graph features, effectively capturing the collaborative relationships between devices (such as the chain reaction of a sensor anomaly on multiple downstream power modules), overcoming the one-sidedness of traditional single device feature analysis and improving the global representativeness of the features.

[0053] S400: Input the network knowledge graph features into a preset abnormality prediction model to obtain the abnormality θ corresponding to the current network knowledge graph.

[0054] Abnormality prediction model training: Use historical operation and maintenance data (including the knowledge graph of normal and various abnormal states, the sub-data feature vectors corresponding to each edge, and the corresponding manually labeled abnormality levels) to train a regression model (such as a neural network or random forest). The model input is the network knowledge graph features, and the output is the abnormality degree θ between 0 and 1 (the closer θ is to 1, the higher the degree of abnormality).

[0055] Real-time prediction: Input the current network knowledge graph features extracted by S300 into the trained model to obtain the abnormality degree θ of the current system.

[0056] For example, if the current graph features show a sudden drop in the transmission frequency of the "temperature sensor → gateway" edge (from 5 times / second to 1 time / second) and an increase in the control command variance of the "gateway → power module" edge, the model will output θ = 0.6 (indicating an upper-moderate abnormality). If the features of all edges are stable within the normal range, θ may be 0.1 (close to normal).

[0057] The quantitative abnormality degree θ is used to objectively evaluate the system operating status, avoiding the limitations of traditional "black or white" abnormality judgment (such as ignoring the cumulative effect of minor abnormalities) and providing an accurate decision-making basis for subsequent graded processing.

[0058] S500, if θ<γ1, then the data statistical features corresponding to the data entering each current queue within the preset time length are sent to the cloud server; γ1 is the first preset abnormality threshold.

[0059] When θ<γ1 (e.g., γ1=0.9, indicating that the system has no abnormalities), the edge gateway only packages the sub-data feature vectors of each edge generated by S200 (instead of the original communication data) and uploads them to the cloud server for basic status monitoring in the cloud.

[0060] For example, if θ = 0.7 (< γ1 = 0.9), the gateway only uploads statistical features such as the mean / variance of the "temperature sensor → gateway" route and the transmission frequency of the "power module → load" route, but does not upload every piece of raw temperature data or current data.

[0061] When the system is stable, the amount of original data transmission is reduced (the amount of statistical feature data is only 1 / 10-1 / 100 of the original data), significantly reducing network bandwidth usage and cloud storage pressure. At the same time, key monitoring information is retained through statistical features, balancing real-time performance and resource efficiency.

[0062] S600: If θ≥γ1, each historical abnormality is traversed forward in sequence until a target historical abnormality θ' is determined; θ'≤γ2; γ2<γ1; γ2 is a second preset abnormality threshold.

[0063] When θ≥γ1 (for example, θ=0.96≥0.9), the edge gateway retrieves the historical anomaly records stored locally (saved in chronological order) and traverses forward (tracing back from the current moment) until it finds the first historical time point that satisfies θ'≤γ2 (γ2=0.6<γ1, representing the normal state of the system). This time point is the starting reference point for the anomaly evolution.

[0064] For example, if the current value θ = 0.95 ≥ γ1 = 0.9, and the forward traversal reveals that 10 minutes ago θ = 0.4 (≤ γ2 = 0.1), and 8 minutes ago θ = 0.7 (> γ2), then the time point corresponding to θ' is determined to be 10 minutes ago.

[0065] Accurately locating the starting point of the system's transition from "normal" to "abnormal" avoids the problem of blindly uploading massive historical data, defines an efficient time range for subsequent abnormality tracing, and improves the targeted nature of data screening.

[0066] S700 , sending a number of communication data after the time point corresponding to θ′ to the cloud server, so as to perform a preset deep anomaly detection through the cloud server.

[0067] Furthermore, step S700 may include the following steps: S710, obtaining each historical abnormality after θ' to obtain a historical abnormality list τ = (τ1, τ2, ..., τ i ,…,τ n ), i=1, 2,…, n; τ i is the i-th historical anomaly degree after θ', and n is the number of historical anomalies after θ'.

[0068] θ' is the time point corresponding to the "normal system state" determined in step S600 (θ' ≤ γ2). The edge gateway extracts all historical anomaly levels from the locally stored anomaly records from the time point corresponding to θ' up to the current moment, arranges them in chronological order, and forms a historical anomaly list τ.

[0069] The historical anomaly list in the form of a time series fully records the system's evolution from "normal" to "current anomaly," providing a quantitative basis for subsequent analysis of anomaly development trends (such as slow accumulation or sudden outbreaks), avoiding a fragmented understanding of the anomaly process.

[0070] The cloud server has powerful computing capabilities and is pre-installed with high-precision analysis algorithms. It can perform in-depth calculations and processing on the entire amount of data quickly to determine the cause of the anomaly.

[0071] S720: For any side QR, obtain the historical sub-data feature vectors of QR within the preset time period corresponding to each historical abnormality in τ, so as to obtain the historical sub-data feature vector list A=(A1, A2, ..., A i ,…,A n );A i is QR in τ i The corresponding historical sub-data feature vector within the preset time length.

[0072] For any directed edge QR in the network knowledge graph (such as "temperature sensor → edge gateway"), combined with the historical anomaly list in S710, extract the sub-data feature vector of the QR edge within the preset time period corresponding to each anomaly: each historical anomaly corresponds to a preset time period; extract the communication data within the preset time period from the FIFO queue of the QR edge, and generate a sub-data feature vector (including communication parameters of different protocols) according to the logic of step S200; arrange all historical sub-data feature vectors in chronological order to form list A.

[0073] By associating the sub-data feature vectors of the edge with the historical anomaly time series one by one, we can clearly present the correspondence between the "changes in the communication characteristics of the edge" and the "changes in the overall anomaly of the system" (for example, whether the anomaly of a certain edge feature occurs earlier than the increase in the system anomaly), providing fine-grained temporal feature support for locating the "abnormal source edge".

[0074] S730, obtaining each historical sub-data feature vector in A and a preset standard sub-data feature vector XL QR The similarity between them is used to obtain the similarity list B=(B1,B2,…,B i ,…,B n ); B i A i With XL QR The similarity between them.

[0075] Preset standard sub-data feature vector XL QR : For edge QR, based on historical normal operation data (sub-data feature vector when θ≤γ2), the standard feature vector of the edge is generated by statistical averaging; for example, if the "temperature sensor → gateway" edge is in normal state, the average parameters of the Modbus protocol are [100B, 8, 1.6 / s], then XL QR =([100,8,1.6],[0,0,0],...); Similarity is calculated using cosine similarity or similarity converted from Euclidean distance (the closer the value is to 1, the higher the similarity).

[0076] By calculating the similarity with the standard eigenvector, the degree of deviation of the edge QR from the "normal state" during the abnormal evolution process is quantified, avoiding the deviation of subjective judgment of "whether the feature is abnormal" and providing an objective quantitative indicator for the subsequent screening of "truly abnormal edges".

[0077] S740, if the average similarity WB corresponding to B is less than WB', all communication data of QR after the time point corresponding to θ' is sent to the cloud server; otherwise, the communication data of QR after the time point corresponding to θ' is not sent.

[0078] Calculate the average similarity WB: average all similarities in list B to obtain the overall deviation WB of edge QR during the abnormal evolution process (the smaller the WB, the more significant the overall deviation of the edge from the normal state).

[0079] Threshold comparison and data transmission: Compare WB with WB’ (such as 0.6, calibrated according to historical fault data): If WB < WB’ (such as WB = 0.4 < 0.6), it indicates that this edge significantly deviates from normal during the abnormal evolution process. Send all communication data after the time point corresponding to θ’ (including the original data in the FIFO queue and the sub-data feature vectors) to the cloud; otherwise, it indicates that the deviation of this edge is not significant and there is no need to send its data.

[0080] Through average similarity screening, only the communication data of the edges strongly related to the abnormal evolution are uploaded to the cloud, avoiding the bandwidth waste and cloud processing pressure caused by "uploading all data", while ensuring that the cloud can focus on the truly abnormal device interaction data, greatly improving the efficiency and accuracy of deep anomaly detection (such as quickly locating that the abnormal temperature sensor data is the root cause of the fault).

[0081] Furthermore, step S700 may further include the following steps: S750, send the communication data of all edges after the time point corresponding to θ’ to the cloud server.

[0082] The edge gateway packages all relevant communication data from the time point corresponding to θ’ (such as 10 minutes ago) to the current moment (including the original data in the FIFO queue) and preferentially uploads it to the cloud server for the cloud to perform deep anomaly detection through deep learning models, etc. (such as locating specific faulty devices and analyzing the abnormal propagation path).

[0083] Example: Upload the "original temperature sensor data", "gateway control instruction records", "power module feedback data", etc. from 10 minutes ago to the current moment. Through analysis, the cloud discovers that the temperature sensor data started to jump 10 minutes ago, which in turn led to the disorder of the gateway control instructions and ultimately caused the abnormal output of the power module.

[0084] Provide the cloud with complete abnormal evolution trajectory data, avoiding the lack of key information (such as the traditional scheme of only uploading current abnormal data may miss early abnormal signals), helping the cloud quickly locate the root cause of the fault (such as specific sensor failure or communication link abnormality), and greatly improving the efficiency and accuracy of anomaly handling.

[0085] Furthermore, after step S700, the method further includes the following steps: S800, if θ ≥ γ₁, then adjust the preset duration T to T’; where T’ = T min + (T - T min )×(1 - (θ - γ₁) / (1 - γ₁)) β ; T min is the preset minimum sampling interval; β is the preset sensitivity coefficient; β > 1.

[0086] When θ is between γ1 and 1, as θ increases, T' shortens nonlinearly (the larger β is, the faster the shortening rate is). The sensitivity coefficient β must be determined in combination with the abnormal response requirements, resource constraints, and historical fault characteristics of the high-frequency switching power supply system, and is comprehensively set through "scenario adaptation + data verification". The core goal is to match the adjustment rate of the preset duration with the system's sensitivity to abnormalities.

[0087] Highly real-time scenarios (such as high-frequency switching power supplies for medical equipment and power supply systems for precision instruments): In these scenarios, the spread of anomalies may lead to serious consequences (such as equipment downtime and data loss). The sampling interval needs to be quickly shortened to capture abnormal details. β should take a larger value (usually 3-5).

[0088] In general real-time scenarios (such as ordinary industrial production line power supplies and communication base station backup power supplies), the anomaly diffusion rate is slow, and the sensitivity can be moderately reduced to balance resource consumption. β takes a small value (usually 2-3).

[0089] This step has at least the following beneficial effects: Accurately match the degree of abnormality: Through a nonlinear adjustment formula, the preset duration is dynamically shortened as the degree of abnormality increases. The more severe the abnormality, the shorter the sampling interval. This ensures that abnormal details (such as data mutations during the fault propagation process) can be captured at high frequency, meeting the real-time response requirements of high-frequency switching power supplies to abnormalities.

[0090] Avoid waste of resources: Through T min Limiting the minimum interval prevents over-frequent sampling (e.g., multiple samplings per second) when the anomaly level is extremely high, which can lead to overload of edge gateway computing / storage resources. This balances anomaly monitoring accuracy and system stability.

[0091] Smooth transition adjustment: In the formula, “(TT min )×nonlinear term”, so that the duration smoothly transitions from the current value T to T min , avoiding the damage to the continuity of data time series caused by sudden adjustments (such as a sudden jump from 10 seconds to 2 seconds may cause feature discontinuity), and ensuring the accuracy of subsequent feature extraction and anomaly calculation.

[0092] This adjustment mechanism enables the system to "sample frequently on demand" under abnormal conditions, while taking into account resource efficiency and providing more intensive and timely data support for abnormality tracing and fault location.

[0093] The method in this embodiment, by constructing a network knowledge graph including terminal device nodes, data flow edges and corresponding first-in-first-out queues, can organically integrate the scattered device communication data and the relationship between devices. By extracting and analyzing the graph features with the help of a preset model, it breaks through the limitations of traditional single device parameter monitoring and fully explores the correlation and timing laws of data interaction between devices, thereby more accurately identifying systemic failures caused by changes in network topology or abnormal data flow, and greatly reducing the probability of misjudgment and missed judgment of abnormalities; through a hierarchical data upload mechanism based on abnormality degree, only statistical feature data is uploaded when the system is running smoothly (with low abnormality degree), which can be used for fault diagnosis and omission detection. It significantly reduces the amount of original data transmission, alleviates network bandwidth pressure, reduces data transmission costs, and ensures the real-time delivery of critical information. When a significant anomaly is detected, by tracing back to the normal state time point and uploading relevant communication data thereafter, it can provide a complete anomaly evolution trajectory for cloud-based deep detection, avoid the loss of key information, and help quickly locate the root cause of the fault. Overall, this solution takes into account real-time, correlation, and data transmission efficiency, effectively improving the accuracy of anomaly detection in switching power supply systems, the convenience of fault tracing, and the intelligent level of operation and maintenance management, providing more reliable protection for systems such as high-frequency switching power supplies that have extremely high stability requirements.

[0094] Furthermore, although the steps of the method of the present disclosure are described in a particular order in the accompanying drawings, this does not require or imply that the steps must be performed in this particular order, or that all steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0095] An embodiment of the present invention also provides a non-transitory computer-readable storage medium, which can be set in an electronic device to store at least one instruction or at least one program related to implementing a method in a method embodiment. The at least one instruction or the at least one program is loaded and executed by the processor to implement the method provided in the above embodiment.

[0096] The program product may utilize any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or component, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0097] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium that can transmit, propagate, or transfer a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0098] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0099] The program code for performing the operations of the present application can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0100] An embodiment of the present invention further provides an electronic device including a processor and the aforementioned non-transitory computer-readable storage medium.

[0101] The electronic device is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0102] The electronic device is implemented as a general-purpose computing device. Components of the electronic device may include, but are not limited to, the aforementioned at least one processor, the aforementioned at least one memory, and a bus connecting different system components (including the memory and the processor).

[0103] The memory stores program codes, which can be executed by the processor, so that the processor performs the steps of various embodiments described in this specification.

[0104] The memory may include readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory, and may further include read-only memory (ROM).

[0105] The memory may also include a program / utility having a set (at least one) of program modules, such program modules including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0106] The bus may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures.

[0107] The electronic device may also communicate with one or more external devices (e.g., keyboards, pointing devices, Bluetooth devices, etc.), one or more devices that enable a user to interact with the electronic device, and / or any device that enables the electronic device to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface. Furthermore, the electronic device may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter. The network adapter communicates with other modules of the electronic device via a bus. It should be understood that other hardware and / or software modules may be used in conjunction with the electronic device, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0108] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

[0109] An embodiment of the present invention further provides a computer program product comprising program code. When the program product is run on an electronic device, the program code is used to enable the electronic device to execute the steps of the method according to various exemplary embodiments of the present invention described above in this specification.

[0110] Although some specific embodiments of the present invention have been described in detail by way of examples, it should be understood by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present invention. It should also be understood by those skilled in the art that various modifications may be made to the embodiments without departing from the scope and spirit of the present invention.

Claims

1. A method for handling abnormalities in a switching power supply based on an edge computing gateway, characterized in that: The method comprises the following steps: S100: Establish a network knowledge graph corresponding to the edge computing gateway based on communication data between the edge computing gateway and all connected terminal devices; the network knowledge graph includes a number of nodes and directed edges connecting the nodes; each node corresponds to a terminal device, and the direction of the edge is the direction of data flow between the corresponding terminal devices; each edge corresponds to a first-in, first-out queue; S200, generating a sub-data feature vector corresponding to each side according to the statistical features of the data in the first-in-first-out queue of each side at every preset interval; S300, using a preset model to extract features from the network knowledge graph to obtain network knowledge graph features corresponding to the network knowledge graph; S400, inputting the network knowledge graph features into a preset abnormality prediction model to obtain the abnormality θ corresponding to the current network knowledge graph; S500, if θ < γ1, then the data statistical features corresponding to the data entering each queue within the preset time period are sent to the cloud server; γ1 is the first preset abnormality threshold; S600, if θ≥γ1, then traverse each historical abnormality in sequence until a target historical abnormality θ' is determined; θ'≤γ2; γ2<γ1; γ2 is a second preset abnormality threshold; S700 , sending a number of communication data after the time point corresponding to θ′ to the cloud server, so as to perform a preset deep anomaly detection through the cloud server.

2. The method for handling abnormalities of a switching power supply based on an edge computing gateway according to claim 1, characterized in that: Step S700 includes the following steps: S710, obtaining each historical abnormality after θ' to obtain a historical abnormality list τ = (τ1, τ2, ..., τ i ,…,τ n ), i=1, 2,…, n; τ i is the i-th historical anomaly degree after θ', and n is the number of historical anomalies after θ'; S720: For any side QR, obtain the historical sub-data feature vectors of QR within the preset time period corresponding to each historical abnormality in τ, so as to obtain the historical sub-data feature vector list A=(A1, A2, ..., A i ,…,A n );A i is QR in τ i The corresponding historical sub-data feature vector within the preset time period; S730, obtaining each historical sub-data feature vector in A and a preset standard sub-data feature vector XL QR The similarity between them is used to obtain the similarity list B=(B1,B2,…,B i ,…,B n ); B i A i With XL QR similarity between S740, if the average similarity WB corresponding to B is less than WB', all communication data of QR after the time point corresponding to θ' is sent to the cloud server; otherwise, the communication data of QR after the time point corresponding to θ' is not sent.

3. The method for handling abnormalities of a switching power supply based on an edge computing gateway according to claim 1, characterized in that: Step S700 further includes the following steps: S750: Send the communication data of all edges after the time point corresponding to θ' to the cloud server.

4. The method for handling abnormalities of a switching power supply based on an edge computing gateway according to claim 1, characterized in that: After step S700, the method further includes the following steps: S800, if θ≥γ1, then adjust the preset time length T to T'; where T'=T min + (TT min )×(1-(θ-γ1) / (1-γ1)) β ;T min is the preset minimum sampling interval; β is the preset sensitivity coefficient; β>1.

5. The method for handling abnormalities of a switching power supply based on an edge computing gateway according to claim 1, characterized in that: Step S100 includes the following steps: S110, establishing a node corresponding to each terminal device connected to the edge computing gateway; S120, for any two nodes D1 and D2, if the data sent by the terminal device corresponding to D1 is sent to the terminal device corresponding to D2 through the edge computing gateway, a directed edge is established from D1 to D2; S130, if the data sent by the terminal device corresponding to D2 is sent to the terminal device corresponding to D1 through the edge computing gateway, a directed edge is established from D2 to D1; S140: If there is no data transmission between the terminal device corresponding to D1 and the terminal device corresponding to D2, no edge connection is established between D1 and D2.

6. The method for handling abnormalities of a switching power supply based on an edge computing gateway according to claim 1, characterized in that: Step S200 includes the following steps: S210, obtain the initial sub-data feature vector E = (E1, E2, ..., E j ,…,E m ), j = 1, 2, …, m; E j is the empty vector corresponding to the j-th preset communication protocol, and m is the number of preset communication protocols; S220, for any QR on either side, obtain the communication parameters of the data in the QR's first-in-first-out queue under different communication protocols at every preset interval; the communication parameters include: total communication data volume, total number of communication data items, and communication frequency; S230, filling the communication parameters under different communication protocols into the corresponding empty vectors in E to obtain the sub-data feature vector corresponding to QR.

7. The method for handling abnormalities of a switching power supply based on an edge computing gateway according to claim 1, characterized in that: The preset model is GNN.

8. The method for handling abnormalities of a switching power supply based on an edge computing gateway according to claim 1, characterized in that: The terminal device includes: a high-frequency switching power supply, a current sensor, a voltage sensor and a collaborative edge device.

9. A non-transitory computer-readable storage medium, wherein at least one instruction or at least one program is stored in the storage medium, characterized in that: The at least one instruction or the at least one program is loaded and executed by the processor to implement the switching power supply abnormality handling method based on the edge computing gateway as described in any one of claims 1-8.

10. An electronic device, characterized in that: The device comprises a processor and the non-transitory computer-readable storage medium of claim 9.

Images