Edge computing gateway-based switching power supply abnormality processing method, device and medium

Abstract

The application provides a kind of based on edge computing gateway switching power supply exception handling method, equipment and medium, related to switching power supply exception handling technical field, method includes: according to the communication data between edge computing gateway and all terminal devices connected, establish the network knowledge graph corresponding to edge computing gateway;According to the data statistical characteristics in the first-in first-out queue of each edge, generate the sub data feature vector corresponding to each edge;Using a predetermined model to extract features from network knowledge graph, obtain the network knowledge graph corresponding to network knowledge graph feature;Network knowledge graph features are input into the preset anomaly degree prediction model, to obtain the anomaly degree θ corresponding to the current network knowledge graph;According to θ, send several communication data to cloud server, to carry out the preset depth anomaly detection through cloud server;The application can improve the accuracy of switching power supply system anomaly detection, the convenience of fault tracing and the intelligent level of operation and maintenance management.

Description

TECHNICAL FIELD

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

[0002] In a high-frequency switching power supply system, with the application of edge computing technology, an edge computing gateway needs to interact with a large number of terminal devices (such as power supply modules, sensors, control units, etc.) at a high frequency to collect running parameters such as current, voltage, temperature and communication instructions between devices in real time. However, there are many technical problems at present: on the one hand, the communication data between the terminal device and the gateway is large and has a time sequence. If all the raw data is uploaded to the cloud server in real time without screening, it will occupy a large amount of network bandwidth, not only increasing the data transmission cost, but also affecting the timeliness of abnormal response due to transmission delay, especially for high-frequency switching power supply systems which have very high real-time requirements, which may lead to fault diffusion; on the other hand, the existing abnormality detection methods mainly focus on the isolated parameters of a single device (such as the output voltage fluctuation of a single power supply module), ignoring the correlation formed by data interaction between devices (such as a single sensor abnormality may cause coordinated failure of multiple power supply modules), which is difficult to capture systematic faults caused by changes in network topology or abnormal data flow, resulting in false or missed detection of abnormalities; in addition, when an abnormality is detected, the existing scheme usually only uploads the abnormal data at the current time, lacking the ability to trace the development process of the abnormality, and cannot provide a complete abnormal evolution trajectory for cloud deep analysis, resulting in difficulty in fault location and low processing efficiency. These problems seriously restrict the intelligent operation and maintenance level of the switching power supply system, and an efficient solution is needed that can balance data transmission efficiency, correlation analysis and abnormality tracing. SUMMARY

[0003] To solve the above technical problems, the technical solution adopted by the present application is as follows:

[0004] According to a first aspect of the present application, a switching power supply abnormality processing method based on an edge computing gateway is provided, which comprises the following steps:

[0005] S100, a network knowledge graph corresponding to the edge computing gateway is established according to the communication data between the edge computing gateway and all connected terminal devices; the network knowledge graph comprises a plurality of nodes and directed edges connecting the nodes; each node corresponds to a terminal device, and the direction of the edge is the data flow direction between the corresponding terminal devices; each edge corresponds to a first-in-first-out queue;

[0006] S200, every interval of a preset time length, a sub-data feature vector corresponding to each edge is generated according to the data statistical characteristics in the first-in-first-out queue of each edge;

[0007] In S300, a preset model is used to extract features of the network knowledge graph, and network knowledge graph features corresponding to the network knowledge graph are obtained.

[0008] In S400, the network knowledge graph features are input into a preset abnormality degree prediction model, and an abnormality degree theta corresponding to the current network knowledge graph is obtained.

[0009] In S500, if theta < gamma 1, data statistical features corresponding to data entering each queue in a preset time length are sent to a cloud server, where gamma 1 is a first preset abnormality degree threshold.

[0010] In S600, if theta >= gamma 1, each historical abnormality degree is sequentially traversed until a target historical abnormality degree theta' is determined, where theta' <= gamma 2, gamma 2 < gamma 1, and gamma 2 is a second preset abnormality degree threshold.

[0011] In S700, a plurality of communication data after a time point corresponding to theta' are sent to the cloud server, so that a preset deep abnormality detection is performed by the cloud server.

[0012] According to another aspect of the present application, a non-transitory computer readable storage medium is also provided, which stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by a processor to implement the above-mentioned edge computing gateway-based switch power supply abnormality processing method.

[0013] According to another aspect of the present application, an electronic device is also provided, which includes a processor and the above-mentioned non-transitory computer readable storage medium.

[0014] The present application has at least the following beneficial effects:

[0015] The edge computing gateway-based switch power supply abnormality processing method of the present application can organically integrate the scattered device communication data and the association relationship between devices by constructing a network knowledge graph containing terminal device nodes, data flow edges and corresponding first-in-first-out queues, and by means of the extraction and analysis of the graph features by the preset model, the limitations of traditional single device parameter monitoring are broken through, the correlation and timing regularity of data interaction between devices are fully tapped, so that the systematic failure caused by network topology changes or data flow abnormality can be more accurately identified, and the probability of abnormal misjudgment and omission can be greatly reduced; through the hierarchical data uploading mechanism based on abnormality degree, only statistical feature data is uploaded when the system runs smoothly (the abnormality degree is low), which can significantly reduce the amount of raw data transmission, relieve the network bandwidth pressure, reduce the data transmission cost, and at the same time guarantee the real-time performance of key information; when significant abnormality is detected, the relevant communication data after the normal state time point is uploaded, which can provide a complete abnormal evolution track for cloud deep detection, avoid missing key information, and help to quickly locate the fault root cause; overall, the scheme takes into account the real-time performance, correlation and data transmission efficiency, effectively improves the accuracy of switch power supply system abnormality detection, the convenience of fault tracing and the intelligent level of operation and maintenance management, and provides more reliable protection for high-frequency switch power supplies and other systems with extremely high stability requirements. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor.

[0017] Figure 1 The flowchart of the edge computing gateway-based switch power supply abnormality processing method provided by the embodiments of the present application. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0019] It is noted that, based on the present disclosure, one skilled in the art will appreciate that one aspect described herein can be implemented independently of any other aspect and that com bination of two or more of these aspects can be made in accordance with the present disclosure. For example, an apparatus can be implemented or a method can be practiced using any number of the aspects set forth herein. In addition, such an apparatus can be implemented or such a method can be practiced using other structure and / or functionality in addition to or other than one or more of the aspects set forth herein.

[0020] Reference will now be made to the drawings Figure 1 A flowchart of the edge computing gateway-based switching power supply abnormality processing method is shown, and an edge computing gateway-based switching power supply abnormality processing method is introduced.

[0021] The method in this embodiment is applicable to a high-frequency switching power supply network system scenario including an edge computing gateway and multiple types of terminal devices, and is particularly applicable to an industrial-grade high-frequency switching power supply cluster operation and maintenance scenario with strict requirements for real-time performance and stability, such as a high-frequency switching power supply array of a data center, a power supply system of a communication base station, a high-frequency power module group of an industrial automation production line, and the like. In such a scenario, the edge computing gateway needs to perform high-frequency data interaction with a high-frequency switching power supply body, various sensors (current, voltage, temperature, vibration, and the like), a power supply internal control unit (MCU / PLC), a driving circuit, a load monitoring device, a heat dissipation system, and a local monitoring terminal and other types of terminal devices. The devices are associated through a complex data flow, and the system is prone to fluctuations in the overall power supply system due to device coordination abnormalities, data transmission faults, or local device failures during operation, and needs to ensure stable operation of the power supply system through precise abnormality detection and efficient data processing mechanisms.

[0022] The edge computing gateway-based switching power supply abnormality processing method can include the following steps:

[0023] S100, a network knowledge graph corresponding to the edge computing gateway is established according to communication data between the edge computing gateway and all connected terminal devices; the network knowledge graph includes a plurality of nodes and directed edges connecting the nodes; each node corresponds to a terminal device, and the direction of the edge is the data flow direction between the corresponding terminal devices; and each edge corresponds to a first-in-first-out queue.

[0024] Node definition: first, through the initial communication handshake information between the edge computing gateway and the terminal devices, all associated terminal devices (such as high-frequency switching power supply bodies, current / voltage sensors, power supply built-in MCUs, heat dissipation fan sensors, load monitors, and the like) are identified, each terminal device is mapped to a node of the network knowledge graph, and the edge computing gateway itself is included in the graph as a special node.

[0025] Directed edge construction: Based on the actual data transmission records between devices (such as sensors sending collected data to gateways, gateways sending control instructions to power MCU, power body feeding back output status to load monitor, etc.), directed edges are established between corresponding nodes, and the direction of the edges strictly follows the data flow direction (such as "current sensor node → edge gateway node" indicating that the sensor transmits data to the gateway).

[0026] FIFO queue configuration: Bind an independent first-in-first-out queue for each directed edge, and store the communication data between the two devices corresponding to the edge in chronological order (such as the queue of the "gateway → power MCU" edge, which stores the voltage adjustment instruction at t1 and the frequency control signal at t2, etc.).

[0027] For example: In the communication base station power supply system, the network knowledge graph may include "edge gateway", "base station power module", "temperature sensor", "load monitor" four nodes; the directed edges are "temperature sensor → edge gateway" (transmit temperature data), "edge gateway → base station power module" (transmit control instruction), "base station power module → load monitor" (transmit output current data), and the FIFO queue of each edge stores the corresponding data flow record.

[0028] Through the graph structure, the association relationship and data interaction logic between devices are presented, breaking through the limitations of traditional "isolated device monitoring", and laying the foundation for capturing system-level associated anomalies (such as a sensor anomaly causing multiple device coordinated failure); the FIFO queue retains the time sequence characteristics of the data, providing original basis for subsequent time sequence feature analysis.

[0029] Further, step S100 can include the following steps:

[0030] S110, establishing a node corresponding to each terminal device connected by the edge computing gateway.

[0031] The edge computing gateway initializes the communication protocol (such as MQTT, Modbus) to handshake with the connected terminal devices, obtains the unique identifier (such as device ID, MAC address) and basic attributes (such as device type, rated parameter, installation location, etc.) of each device; maps each terminal device to an independent node in the network knowledge graph, and stores the unique identifier and basic attributes of the device in the node attributes (for example, the attributes of the "current sensor A" node are [ID = CS001, type = current sensor, range = 0-50A]). At the same time, the edge computing gateway itself is also included in the graph as a special node, which is used to clarify its role in data flow.

[0032] By mapping terminal devices through unique nodes, the "entity" basis of the network knowledge graph is clarified, and the storage of node attributes provides inherent information of the device itself for subsequent feature analysis (such as sensor range affecting data anomaly judgment criteria), avoiding device confusion and laying a foundation for accurate construction of the graph.

[0033] 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 from D1 to D2 is established.

[0034] The edge computing gateway records in real time the source address (sending device) and destination address (receiving device) of all data packets forwarded through itself. For any two nodes D1 (corresponding to device X) and D2 (corresponding to device Y), if it is detected that the data packet sent by device X is forwarded to device Y through the gateway (i.e. the source address of the data packet is the ID of X, the destination address is the ID of Y, and it is transferred through the gateway), a directed edge from D1 to D2 is established in the graph, and the attribute of the edge is marked as "forwarded through the gateway".

[0035] Through the directed edge, the indirect data flow of "device → gateway → device" is accurately described, and the one-way interaction relationship of data transfer through the gateway is clarified, providing topological basis for subsequent analysis of the influence of "upstream device anomaly on downstream device" (such as sensor data anomaly causing HMI display error).

[0036] 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 from D2 to D1 is established.

[0037] Symmetrically with S120, when the edge gateway detects that the data packet sent by device Y (corresponding to node D2) is forwarded to device X (corresponding to node D1) through itself, a directed edge from D2 to D1 is established in the graph, and the attribute of the edge is also marked as "forwarded through the gateway", and exists independently from the "D1→D2" edge that may exist in S120 (i.e. two directed edges in opposite directions correspond to bidirectional data interaction).

[0038] Supporting the description of bidirectional data interaction avoids missing the reverse data flow from "downstream device to upstream device" (such as control instructions, state feedback), and completely presents the closed-loop interaction between devices through the gateway (such as HMI issuing instructions → power executing → HMI receiving feedback), providing complete topology for analyzing the influence of "control instruction anomaly on device response".

[0039] S140, if there is no data transmission between the terminal device corresponding to D1 and the terminal device corresponding to D2, no edge connection between D1 and D2 is established.

[0040] The edge gateway counts the communication records between all terminal devices within a preset monitoring period (such as 10 minutes after initialization). If there has never been a data packet (including both X to Y and Y to X directions) between node D1 (device X) and D2 (device Y) that is forwarded through the gateway, no edge connection is established between the two nodes, and the nodes remain independent.

[0041] The graph structure is simplified by the rule of "no connection without data transmission", which avoids the interference of redundant edges on subsequent feature extraction (such as edges between irrelevant devices diluting effective associations), reduces the complexity of the graph, and improves the efficiency and accuracy of GNN model feature extraction.

[0042] S200, every interval, according to the data statistics features in the first-in-first-out queue of each edge, generate a corresponding sub-data feature vector for each edge.

[0043] Further, step S200 can include the following steps:

[0044] S210, obtaining an 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 jth preset communication protocol, and m is the number of preset communication protocols.

[0045] First, the common preset communication protocols between terminal devices and edge gateways in switching power supply systems are determined (such as Modbus, MQTT, Profinet, EtherCAT, etc. in industrial scenarios), and the total number of protocols is m; an empty vector is allocated to each protocol, and the dimension of the empty vector is consistent with the number of communication parameters that need to be counted for that protocol (for example, if 3 parameters need to be counted for each protocol, then E j is a 3-dimensional empty vector); all protocol empty vectors are combined in a fixed order to form an initial sub-data feature vector E.

[0046] Through the preset protocol and the initial vector of fixed dimension, the structure of the sub-data feature vector 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 subsequent graph neural network (GNN) feature aggregation.

[0047] S220, for any edge QR, every interval, obtain the communication parameters of the data in the first-in-first-out queue of QR under different communication protocols; the communication parameters include: total communication data volume, total communication data number, communication frequency.

[0048] For any directed edge QR in the network knowledge graph (corresponding to the data flow between terminal devices Q and R), the FIFO queue associated with it stores all the communication data transmitted between Q and R through the edge gateway in chronological order. Every interval of a preset duration (e.g., 5 seconds, which can be dynamically adjusted), the edge gateway parses the data in the queue:

[0049] By the protocol identifier in the data packet header (e.g., the function code of Modbus, the fixed header field of MQTT), the data is classified according to the preset communication protocol.

[0050] Parameter calculation: For each type of protocol data, three communication parameters are calculated:

[0051] Total communication data volume: The total number of bytes of data packets transmitted within the preset duration under this protocol (e.g., 10 data packets are transmitted under the Modbus protocol, with a total byte size of 200B).

[0052] Total communication data count: The total number of data packets within the preset duration under this protocol (e.g., 8 data packets are transmitted under the MQTT protocol).

[0053] Communication frequency: The ratio of the total communication data count to the preset duration (e.g., 20 data packets are transmitted within 5 seconds under the Profinet protocol, with a frequency of 4 packets per second).

[0054] For the communication data stored in the queue, key statistical features are calculated, including but not limited to: data transmission frequency (number of data packets per unit time), mean / variance of data values (e.g., fluctuation degree of current sensor data), proportion of outliers (proportion of data packets outside the normal range), standard deviation of data transmission interval (reflecting transmission stability), etc.

[0055] For example: Edge QR is "power module Q -> load monitor R", and the data packets in its FIFO queue within 5 seconds are parsed: Modbus protocol transmits 12 voltage monitoring data (total byte size 144B, frequency 12 / 5 = 2.4 packets per second); MQTT protocol transmits 5 status feedback data (total byte size 60B, frequency 1 packet per second); Profinet protocol has no data. The parameters of Modbus are (144B, 12, 2.4 packets per second), and the parameters of MQTT are (60B, 5, 1 packet per second).

[0056] By distinguishing communication protocols and extracting core parameters, the data interaction intensity (total data volume), density (total count), and stability (frequency) between devices under different protocols are accurately described, avoiding feature ambiguity caused by mixing different protocol data (e.g., differences in real-time requirements of Modbus and asynchronous characteristics of MQTT), providing a more fine-grained basis for anomaly detection (e.g., a sudden drop in the frequency of a protocol may indicate an abnormal link).

[0057] S230, fill the communication parameters under different communication protocols into the corresponding empty vector in E to obtain the sub-data feature vector corresponding to QR.

[0058] Fill the communication parameters of each protocol calculated in S220 into the corresponding empty vector in the initial vector E of S210 according to the protocol order: if there is no data for a protocol within a preset time length, the corresponding empty vector is filled with a zero vector; finally, the complete sub-data feature vector corresponding to the edge QR is formed.

[0059] Integrating the scattered protocol parameters into a structured feature vector not only retains the independent features of different protocols (convenient for analyzing the abnormality of a protocol), but also realizes the comparability of cross-edge features through a unified format (such as comparing the frequency differences of the MQTT protocol of the "sensor-gateway" and "gateway-power supply" edges), providing standardized and high information density inputs for subsequent GNN to extract global features of the network, and improving the efficiency and accuracy of feature aggregation.

[0060] 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.

[0061] Further, the preset model is GNN.

[0062] 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 association between nodes and edges.

[0063] Feature input: the network knowledge graph and the sub-data feature vector of the edge generated by S200 as the model input.

[0064] Graph feature aggregation: GNN aggregates the features of its neighbor nodes and connected edges through the "message passing mechanism" (such as the "edge gateway" node aggregating the features of its neighbors such as temperature sensors, power modules, etc.), and finally outputs a global vector, i.e. the network knowledge graph feature, which contains the topological relationship and data interaction state information of the entire network.

[0065] In the data center power array, GNN will aggregate the node attributes of "power module A" (rated power 500W), the feature vector of "power module A-load B" edge (average transmission current 2A, small fluctuation), and the feature vector of "edge gateway-power module A" edge (stable control instruction frequency) to generate a graph feature reflecting the running state of the entire power array.

[0066] Through the association modeling capability of GNN, the scattered device attributes and edge features are fused into global graph features, effectively capturing the collaborative relationship between devices (such as the chain effect of a sensor anomaly on multiple power modules downstream), overcoming the one-sidedness of traditional single device feature analysis, and improving the global representativeness of the features.

[0067] S400, input the network knowledge graph features into the preset anomaly degree prediction model to obtain the anomaly degree θ corresponding to the current network knowledge graph.

[0068] Anomaly degree prediction model training: using historical operation and maintenance data (including knowledge graphs of normal state and various abnormal states, sub-data feature vectors corresponding to each edge, and corresponding artificial labeling of abnormal levels), a regression model (such as neural network or random forest) is trained, the model input is the network knowledge graph feature, and the output is the anomaly degree θ between 0 and 1 (the closer θ is to 1, the higher the abnormality degree is).

[0069] Real-time prediction: input the current network knowledge graph features extracted by S300 into the trained model to obtain the anomaly degree θ of the current system.

[0070] For example: if the current graph feature shows that the transmission frequency of the "temperature sensor→gateway" edge has dropped sharply (from 5 times / sec to 1 time / sec), and the control instruction variance of the "gateway→power module" edge has increased, the model will output θ=0.6 (indicating moderate to high abnormality); if all edge features are stable within the normal range, θ may be 0.1 (close to normal).

[0071] By quantifying the anomaly degree θ, the system running state is objectively evaluated, avoiding the limitations of traditional "black and white" abnormality judgment (such as ignoring the cumulative effect of slight abnormalities), and providing accurate decision basis for subsequent hierarchical processing.

[0072] S500, if θ<γ1, the data statistical features corresponding to the data entering each queue within a preset time length are sent to the cloud server; γ1 is the first preset anomaly degree threshold.

[0073] When θ<γ1 (such as γ1=0.9, indicating that the system is normal), the edge gateway only uploads the sub-data feature vectors (not the original communication data) of each edge generated by S200 to the cloud server for basic state monitoring by the cloud.

[0074] For example: if θ=0.7 (<γ1=0.9), the gateway only uploads the mean / variance of "temperature sensor→gateway" and the transmission frequency of "power module→load", etc. statistical features, without uploading each raw temperature data or current data.

[0075] The original data transmission amount (the data amount of the statistical characteristics is only 1 / 10-1 / 100 of the original data) is reduced when the system is stable, the network bandwidth occupation and the cloud storage pressure are significantly reduced, the key monitoring information is retained through the statistical characteristics, and the real-time performance and the resource efficiency are balanced.

[0076] S600, if the theta is greater than or equal to gamma 1, each historical abnormality degree is sequentially traversed forward until a target historical abnormality degree theta' is determined; the theta' is less than or equal to gamma 2; the gamma 2 is less than the gamma 1; and the gamma 2 is a second preset abnormality threshold.

[0077] When the theta is greater than or equal to the gamma 1 (for example, the theta=0.96 is greater than or equal to 0.9), the edge gateway calls the historical abnormality record (saved in time sequence) stored locally, and traverses forward (traces back from the current time) until the first historical time point that satisfies the theta' is less than or equal to the gamma 2 (the gamma 2=0.6 is less than the gamma 1, representing a normal state of the system) is found, and the time point is the starting reference point of the abnormal evolution.

[0078] For example, the current theta=0.95 is greater than or equal to the gamma 1=0.9, the forward traversal finds that the theta=0.4 (less than or equal to the gamma 2=0.1) 10 minutes ago, and the theta=0.7 (greater than the gamma 2) 8 minutes ago, and it is determined that the time point corresponding to the theta' is 10 minutes ago.

[0079] The accurate positioning system avoids the problem of blindly uploading a large amount of historical data, and defines an efficient time range for subsequent abnormality tracing, thereby improving the pertinence of data screening.

[0080] S700, a plurality of communication data after the time point corresponding to the theta' are sent to a cloud server, so that the cloud server performs a preset deep abnormality detection.

[0081] Further, the step S700 can include the following steps:

[0082] S710, each historical abnormality degree after the theta' is obtained to obtain a historical abnormality degree list tau=(tau 1, tau 2,..., tau n), i=1, 2,..., n; tau i is the i th historical abnormality degree after the theta', and n is the number of historical abnormality degrees after the theta'. i n i

[0083] The theta' is the time point corresponding to the “normal state of the system” determined in the step S600 (the theta' is less than or equal to the gamma 2). The edge gateway extracts all historical abnormality degrees after the time point corresponding to the theta' until the current time from the abnormality record stored locally, arranges them in time sequence, and forms a historical abnormality degree list tau.

[0084] ​​​Through the historical abnormality degree list in the form of time series, the evolution process of the system from "normal" to "current abnormality" is recorded completely, which provides a quantitative basis for subsequent analysis of the development trend of abnormality (such as slow accumulation or sudden outbreak), and avoids fragmented understanding of the abnormal process.

[0085] The cloud server has powerful computing capability, and is provided with high-precision analysis algorithm, so that full-amount data can be calculated and processed quickly in depth, and the abnormality reason is determined.

[0086] S720, for any edge QR, the historical sub-data feature vector of QR in each preset time length corresponding to each historical abnormality degree in τ is obtained, so as to obtain the historical sub-data feature vector list A = (A1, A2,..., A i ,..., A n ) corresponding to QR; A i is the historical sub-data feature vector of QR in the preset time length corresponding to τ i .

[0087] For any directed edge QR (such as "temperature sensor → edge gateway") in the network knowledge graph, the sub-data feature vector of the edge QR in the preset time length corresponding to each abnormality degree is extracted in combination with the historical abnormality degree list in S710: each historical abnormality degree corresponds to a preset time length; the communication data in the preset time length is extracted from the FIFO queue of the edge QR, and the sub-data feature vector (containing communication parameters of different protocols) is generated according to the logic of step S200; all historical sub-data feature vectors are arranged in time sequence to form a list A.

[0088] The sub-data feature vector of the edge is associated with the time sequence of the historical abnormality degree one by one, so that the corresponding relationship between the "communication feature change of the edge" and the "overall abnormality degree change of the system" (such as whether the abnormality of the edge feature is earlier than the increase of the system abnormality degree) is clearly presented, which provides fine-grained time sequence feature support for locating the "abnormal source edge".

[0089] S730, the similarity between each historical sub-data feature vector in A and the preset standard sub-data feature vector XL QR is obtained, so as to obtain the similarity list B = (B1, B2,..., B i ,..., B n ) corresponding to QR; B i is the similarity between A i and XL QR .

[0090] The preset standard sub-data feature vector XL QR: For edge QR, based on historical normal running data (sub-data eigenvectors when θ≤γ2), the standard eigenvector of the edge is generated by statistical average; for example, the average parameters of the Modbus protocol of the "temperature sensor gateway" edge in the normal state are [100B, 8, 1.6 per second], XL QR =([100, 8, 1.6], [0, 0, 0],...); the similarity calculation uses cosine similarity or similarity converted by Euclidean distance (the value is closer to 1, the higher the similarity).

[0091] Through similarity calculation with the standard eigenvector, the deviation degree of the edge QR from the "normal state" in the abnormal evolution process is quantified, avoiding the deviation of subjective judgment "whether the feature is abnormal", and providing an objective quantitative index for subsequent screening of "truly abnormal edges".

[0092] 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.

[0093] Calculate the average similarity WB: take the average of all similarities in list B to get the overall deviation WB of edge QR in the abnormal evolution process (the smaller WB, the more significant the overall deviation of the edge from the normal state).

[0094] Threshold comparison and data sending: 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 means that the edge significantly deviates from the normal in the abnormal evolution process, and all its communication data after the time point corresponding to θ' (including the original data and sub-data eigenvectors in the FIFO queue) are sent to the cloud; otherwise, it means that the edge does not deviate significantly, and there is no need to send its data.

[0095] Through average similarity screening, only the communication data of the edge strongly related to abnormal evolution is uploaded to the cloud, avoiding the bandwidth waste and cloud processing pressure caused by "full data upload", while ensuring that the cloud can focus on truly abnormal device interaction data, greatly improving the efficiency and accuracy of deep anomaly detection (such as quickly locating the root cause of temperature sensor data anomaly).

[0096] Further, step S700 can further include the following steps:

[0097] S750, send all communication data of edges after the time point corresponding to θ' to the cloud server.

[0098] The edge gateway packs all the relevant communication data (including the original data in the FIFO queue) from the time point corresponding to θ' (e.g., 10 minutes ago) to the current time, and uploads them to the cloud server in priority, so that the cloud can conduct deep anomaly detection (e.g., locate the specific faulty equipment and analyze the abnormal propagation path) through a deep learning model, etc.

[0099] For example, the "temperature sensor original data", "gateway control instruction record", "power module feedback data", etc. from 10 minutes ago to the present are uploaded, and the cloud finds through analysis that the temperature sensor data started to jump 10 minutes ago, which caused the gateway control instruction to be disordered, and finally led to the abnormal output of the power module.

[0100] Providing the cloud with complete abnormal evolution trajectory data can avoid missing key information (e.g., the traditional scheme of uploading only the current abnormal data may miss early abnormal signals), help the cloud quickly locate the root cause of the fault (e.g., a specific sensor failure or communication link abnormality), and greatly improve the efficiency and accuracy of abnormal handling.

[0101] Further, after step S700, the method further comprises the following steps:

[0102] S800, if θ≥γ1, adjust the preset time interval T to T'; wherein 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.

[0103] When θ is between γ1 and 1, T' is nonlinearly shortened as θ increases (the larger β is, the faster the shortening speed is); the determination of the sensitivity coefficient β needs to be combined with the abnormal response requirements of the high-frequency switching power supply system, resource constraints and historical fault characteristics, and is comprehensively set through the way of "scene adaptation + data verification", and the core goal is to match the adjustment rate of the preset time interval with the sensitivity of the system to the abnormality.

[0104] High real-time scenario (e.g., high-frequency switching power supply of medical equipment, precision instrument power supply system): In this kind of scenario, abnormal propagation may lead to serious consequences (e.g., equipment downtime, data loss), and the sampling interval needs to be shortened quickly to capture abnormal details, so β should take a larger value (usually 3-5).

[0105] General real-time scenario (e.g., ordinary industrial production line power supply, communication base station standby power supply): The abnormal propagation speed is slower, and the sensitivity can be moderately reduced to balance resource consumption, so β takes a smaller value (usually 2-3).

[0106] This step has at least the following beneficial effects:

[0107] Precise matching of abnormal degree: through a nonlinear adjustment formula, the preset time is dynamically shortened as the abnormality increases, the more serious the abnormality, the shorter the sampling interval, ensuring that the abnormal details (such as data mutation in the fault propagation process) can be captured at high frequency, meeting the real-time requirements of high-frequency switching power supply for abnormal response.

[0108] Avoid wasting resources: through T min Limit the shortest interval to prevent excessive sampling frequency (such as multiple samples per second) when the abnormality is extremely high, causing edge gateway computing / storage resources to be overloaded, balancing abnormal monitoring accuracy and system stability.

[0109] Smooth transition adjustment: the design of "(T-T min ) x nonlinear term" in the formula makes the time length smoothly transition from the current value T to T min , avoiding sudden changes that disrupt the continuity of the data time sequence (such as jumping from 10 seconds to 2 seconds, which may cause feature breakage), ensuring the accuracy of subsequent feature extraction and abnormality calculation.

[0110] This adjustment mechanism enables the system to "sample at high frequency on demand" in abnormal states, while balancing resource efficiency, providing more intensive and timely data support for abnormal tracing and fault location.

[0111] The method in this embodiment can integrate scattered device communication data and the association between devices by constructing a network knowledge graph containing terminal device nodes, data flow edges, and corresponding first-in-first-out queues. With the extraction and analysis of graph features by a preset model, it breaks through the limitations of traditional single-device parameter monitoring, fully explores the relevance and timing rules of data interaction between devices, and thus more accurately identifies systemic failures caused by changes in network topology or abnormal data flow, significantly reducing the probability of false positives and false negatives. Through a hierarchical data upload mechanism based on abnormality, only statistical feature data is uploaded when the system is running smoothly (with low abnormality), which can significantly reduce the amount of raw data transmission, alleviate network bandwidth pressure, reduce data transmission costs, while ensuring the real-time nature of critical information. When significant abnormalities are detected, the relevant communication data since the normal state time point is uploaded, providing a complete abnormal evolution trajectory for cloud-based deep detection, avoiding missing critical information, and assisting in quickly locating the root cause of the fault. Overall, this scheme balances real-time, relevance, and data transmission efficiency, effectively improving the accuracy of switching power supply system anomaly detection, the convenience of fault tracing, and the intelligent level of operation and maintenance management, providing more reliable protection for high-frequency switching power supplies that require extremely high stability.

[0112] Moreover, although individual steps of the methods in the present disclosure are described in a particular order in the figures, this is not required or implied as to the order of execution of the steps, nor is it required that all of the steps be executed for the desired results. Additionally or alternatively, certain steps can be omitted, combined into a single step, broken into multiple steps, and / or executed in a different order than shown.

[0113] Embodiments of the present application also provide a non-transitory computer readable storage medium, which can be arranged in an electronic device to save at least one instruction or at least one program related to a method in the method embodiments, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the method provided by the above-mentioned embodiments.

[0114] The program product can employ any combination of one or more readable media. The readable media can be a readable signal medium or a readable storage medium. The readable storage medium, for example, can be, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or apparatus, or any suitable combination of the above. More specific examples (a non-exhaustive list) of the readable storage medium include an electrical connection having one or more wires, a portable disc, 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 disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0115] The computer readable signal medium can include a data signal propagated in baseband or propagated as a carrier wave, in which the readable program code is embodied. Such propagated data signal can take multiple forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The readable signal medium can also be any readable medium other than the readable storage medium, which can send, propagate, or transmit the program for use by or in connection with an instruction execution system, apparatus, or device.

[0116] The program code contained on the readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, and the like, or any suitable combination of the above.

[0117] The program code may be implemented in any of various ways, including procedure-based, object-based, and / or class-based environments. Software implemented in such ways includes program instructions that implement the methods of the present application. Program code can be implemented in a high level procedural or object-oriented programming and / or scripting languages, database programming languages, forth-based languages, and / or machine languages. It will be appreciated that the software can be implemented in any combination of firmware, hardware, and / or software. Program code can execute entirely on the user's computing device, partly on the user's computing device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device or entirely on the remote computing device or server. In the latter scenario, the remote computing device can be connected to the user's computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computing device (for example, through the Internet using an Internet Service Provider).

[0118] Embodiments of the present application also provide an electronic device including a processor and the aforementioned non-transitory computer readable storage medium.

[0119] Electronic device is merely an example, and should not bring any limitation to the function and use range of embodiments of the present application.

[0120] The electronic device is in the form of a general computing device. Components of the electronic device can include, but are not limited to, the aforementioned at least one processor, the aforementioned at least one memory, a bus connecting different system components, including the memory and the processor.

[0121] The memory stores program code that can be executed by the processor, so that the processor performs the steps in the various embodiments described in the specification.

[0122] The memory can include a readable medium in the form of volatile memory, such as random access memory (RAM) and / or cache memory, and can further include read-only memory (ROM).

[0123] The memory can also include program / utility programs with a set of (at least one) 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 of which can include implementation of a network environment.

[0124] The bus can be one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor or a local bus using any of a variety of bus structures, using any of a variety of bus structures.

[0125] The electronic device can also communicate with one or more external devices such as a keyboard or a pointing device, through an I / O interface. The electronic device can communicate with one or more devices that enable a user to interact with it through a communication interface. The electronic device can also communicate with one or more devices or networks (e.g., LANs, WANs, and / or the Internet) through a network adapter. The network adapter can communicate with the other modules of the electronic device through the bus. It should be appreciated that other hardware and / or software modules can be used in conjunction with the electronic device. For example, a microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data archival storage systems, etc. can be used with the electronic device.

[0126] From the above description of 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 software in combination with necessary hardware. Therefore, the technical solutions according to the embodiments of the present disclosure can be embodied in the form of a software product. The software product can be stored in a non-volatile storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.) or a network, and includes a number of instructions to make a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) execute the methods according to the embodiments of the present disclosure.

[0127] The embodiments of the present disclosure also provide a computer program product, which includes program codes for causing an electronic device to perform the steps in the methods according to various example embodiments of the present disclosure described above in the specification when the program product is run on the electronic device.

[0128] Although some specific embodiments of the present disclosure have been described in detail by way of examples, it should be appreciated that the above examples are only for illustration, but not to limit the scope of the present disclosure. It should also be appreciated by those skilled in the art that various modifications can be made to the embodiments without departing from the scope and spirit of the present disclosure.

Claims

1. A method for handling abnormalities in a switching power supply based on an edge computing gateway, characterized in that, The method includes 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 direction between the corresponding terminal devices; each edge corresponds to a first-in-first-out queue. S200, at each preset time interval, generate a sub-data feature vector corresponding to each side based on the data statistical characteristics in the first-in-first-out queue of each side; the data statistical characteristics include: data transmission frequency, mean / variance of data values, proportion of outliers, and standard deviation of data transmission interval; S300 uses a preset model to extract features from the network knowledge graph to obtain the network knowledge graph features corresponding to the network knowledge graph; S400: Input the features of the network knowledge graph into the preset anomaly prediction model to obtain the anomaly θ corresponding to the current network knowledge graph; S500, if θ < γ1, then only the data statistical features corresponding to the data that entered within the preset time in each queue are sent to the cloud server; γ1 is the first preset anomaly threshold. S600, if θ≥γ1, then iterate forward through each historical anomaly until the target historical anomaly θ' is determined; θ'≤γ2; γ2<γ1; γ2 is the second preset anomaly threshold. S700 sends a number of communication data after the time point corresponding to θ' to the cloud server for preset deep anomaly detection through the cloud server; Step S700 includes the following steps: S710, obtain each historical anomaly degree after θ' to obtain a list of historical anomalies τ = (τ1, τ2, ..., τ... i , ..., τ n ), i=1,2,…,n; τ i Let θ' be the i-th historical outlier after θ', and n be the number of historical outliers after θ'. S720, for any side QR, obtain the historical sub-data feature vectors of QR within a preset time period corresponding to each historical anomaly degree in τ, so as to obtain the list of historical sub-data feature vectors A = (A1, A2, ..., A...). i A n A i For QR at τ i The corresponding historical sub-data feature vector within the preset time period; S730, obtain the feature vector of each historical sub-data in A and the 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 For A i With XL QR The similarity between them; S740, if the average similarity WB corresponding to B is less than WB', then all communication data of QR after the time point corresponding to θ' will be sent to the cloud server; otherwise, no communication data of QR after the time point corresponding to θ' will be sent.

2. The method for handling abnormal power supply based on edge computing gateway according to claim 1, characterized in that, Following 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 +(TT) min )×(1-(θ-γ1) / (1-γ1)) β ;T min β is the preset minimum sampling interval; β is the preset sensitivity coefficient; β > 1.

3. The method for handling abnormal power supply based on edge computing gateway according to claim 1, characterized in that, Step S100 includes the following steps: S110, establish the node corresponding to each terminal device connected by 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, then a directed edge from D1 to D2 is established. 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, then a directed edge from D2 to D1 is established. S140, if there is no data transmission between the terminal device corresponding to D1 and the terminal device corresponding to D2, then no edge connection is established between D1 and D2.

4. The method for handling abnormal power supply based on 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 Let m be the empty vector corresponding to the j-th preset communication protocol, and m be the number of preset communication protocols. S220, for any QR, at preset intervals, acquires the communication parameters of the data in the QR's first-in-first-out queue under different communication protocols; the communication parameters include: total communication data volume, total number of communication data entries, and communication frequency; S230, fill the communication parameters under different communication protocols into the corresponding empty vector in E to obtain the sub-data feature vector corresponding to QR.

5. The method for handling abnormal power supply based on edge computing gateway according to claim 1, characterized in that, The preset model is GNN.

6. The method for handling abnormal power supply based on edge computing gateway according to claim 1, characterized in that, The terminal equipment includes: a high-frequency switching power supply, a current sensor, a voltage sensor, and collaborative edge devices.

7. A non-transitory computer-readable storage medium, wherein the storage medium stores at least one instruction or at least one program segment, characterized in that, The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the power supply anomaly handling method based on an edge computing gateway as described in any one of claims 1-6.

8. An electronic device, characterized in that, Includes a processor and the non-transitory computer-readable storage medium as described in claim 7.

Images