Method for monitoring panoramic state of power secondary equipment across security partitions

Abstract

The invention is applicable to the technical field of power system monitoring, and provides a cross-security-partition power secondary equipment panoramic state monitoring method, which comprises the following steps of: dividing an equipment network into a plurality of equipment sub-graphs based on electrical topology and functional logic; acquiring real-time operation data of each equipment sub-graph through a data acquisition agent at a production control area side, and dynamically switching and transmitting the real-time operation data to a management information area through a transmission strategy comprising a main path and a standby path; generating a state feature subsequence of each equipment sub-graph on the management information area side based on the real-time data, and further aggregating the state feature subsequence into a panoramic state feature sequence; a feature coding model is used for coding the feature vectors into low-dimensional potential feature vectors, and an equipment state knowledge graph is formed; and sequentially performing health state evaluation, anomaly detection and associated risk analysis on the feature vectors in the map to generate a structured panoramic state monitoring report. The integrity of state evaluation, the timeliness of anomaly detection and the accuracy of risk identification are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system monitoring, in particular to a panoramic state monitoring method for power secondary equipment across security partitions. BACKGROUND

[0002] Power secondary equipment is the core of ensuring the safe and stable operation of the power system, and its state monitoring is crucial for preventing faults and optimizing operation and maintenance. Currently, the state monitoring of secondary equipment is mainly based on multi-source data such as online monitoring devices, preventive test data, live detection results, and manual daily patrol records. These data are processed and analyzed in independent systems or security partitions, such as real-time operation data in the production control area, equipment inventory and historical defects in the management information area, and state evaluation of individual equipment or local parameters based on threshold and expert experience.

[0003] However, due to the diverse sources, heterogeneous formats of power secondary equipment state parameters, and strict security isolation between the production control area and the management information area, real-time operation data and static management data cannot be effectively integrated. Secondly, it is difficult to build a unified model reflecting the internal correlation and overall operation state of the equipment, and the state evaluation is limited to local equipment, lacking of collaborative analysis of the panoramic state of equipment clusters and their associated impacts. This leads to problems such as fragmented state evaluation, insufficient information utilization, delayed abnormality detection, and difficulty in identifying associated risks, which cannot meet the demand for precise health management of power secondary equipment. SUMMARY

[0004] The present application provides a panoramic state monitoring method for power secondary equipment across security partitions, which solves the problem of not meeting the demand for precise health management of power secondary equipment.

[0005] The present application provides a panoramic state monitoring method for power secondary equipment across security partitions, which is applied to a power monitoring system containing a production control area and a management information area, comprising: Based on the electrical topology and functional logical correlation of power secondary equipment, the equipment network is divided into multiple device subgraphs; In the production control area, real-time operation data of each device subgraph is collected by the deployed data collection agent; according to a preset transmission strategy containing primary and backup paths, the real-time operation data is transmitted to the management information area, and the transmission strategy is dynamically switched based on the connectivity state of the data channel; In the management information area, a state feature subsequence corresponding to each device subgraph is generated based on the real-time operation data; aggregate all the state feature subsequences corresponding to the device subgraphs to generate a panoramic state feature sequence; and process the panoramic state feature sequence using a preset feature encoding model to encode the panoramic state feature sequence into a latent feature vector with a dimension lower than that of the panoramic state feature sequence, thereby forming a device state knowledge graph; process the latent feature vector to sequentially perform device health state evaluation, anomaly detection, and associated risk analysis; generate and output a structured panoramic state monitoring report dynamically according to the results of the device health state evaluation, anomaly detection, and associated risk analysis.

[0006] Further, each device subgraph includes a core secondary device and a slave device associated with the core secondary device, and the core secondary devices included in any two different device subgraphs have no intersection.

[0007] Further, the preset transmission strategy including a primary and backup path comprises: The data collection agent is configured with a first transmission path and a second transmission path, the first transmission path leads to the management information area via a first forward isolation device, the second transmission path is a backup path via an encrypted virtual private network, and the transmission priority of the first transmission path is higher than that of the second transmission path. The real-time transmission state of the first transmission path and the second transmission path is monitored, and the real-time transmission state is defined according to the transmission requirements of the power monitoring data stream; When the real-time transmission state of the first transmission path does not meet the preset transmission requirements for the real-time operation data, the transmission of the subsequent real-time operation data is dynamically switched to the second transmission path.

[0008] Further, the real-time operation data is transmitted to the management information area according to the preset transmission strategy including a primary and backup path, comprising: The data collection agent encapsulates the collected real-time operation data based on the currently selected transmission path; In the encapsulation process, a message authentication code is generated and attached to the encapsulated data based on the shared key with the management information area side; The encapsulated data with the attached message authentication code is transmitted to the management information area through the currently selected transmission path.

[0009] Further, the state feature subsequence corresponding to each device subgraph is generated based on the real-time operation data, comprising: Based on the device unique identifier of the core secondary device and the slave device included in each device subgraph, the corresponding device inherent reference data is obtained; The real-time operating data is associated and matched with the inherent baseline data of the device according to the unique identifier of the device to form a multi-dimensional state data set of the device sub-graph; A state feature subsequence is generated based on the multidimensional state data set. The dynamic feature sequence includes logical coordinate encoding, inter-device correlation matrix, and dynamic state vector; wherein: The logical coordinate encoding is used to identify the center position of the core secondary device in the device sub-diagram and the logical hierarchy relationship between the core secondary device and each subordinate device. The device association matrix represents the electrical connections, communication topology, and logical dependencies between the core secondary device and each subordinate device. The dynamic state vector is formed by quantizing the multi-dimensional state data set of the core secondary device and each subordinate device.

[0010] Furthermore, the aggregation of the state feature subsequences corresponding to all device subgraphs to generate a panoramic state feature sequence includes: All the state feature subsequences corresponding to the device subgraphs are sorted and spliced ​​according to a preset global logical order; wherein the preset global logical order is determined based on at least one or more of the physical location, voltage level or functional importance of each device subgraph in the power network.

[0011] Furthermore, the preset feature encoding model is a variational quantization autoencoder.

[0012] Furthermore, the step of processing the panoramic state feature sequence using a preset feature encoding model to encode it into a latent feature vector with a dimension lower than that of the panoramic state feature sequence includes: The panoramic state feature sequence is input into the encoder of the variational quantization autoencoder; The panoramic state feature sequence is mapped to a continuous latent space through multi-layer neural network processing of the encoder. The vectors in the latent space are quantized, and the closest quantized vector is selected as the output latent feature vector.

[0013] Furthermore, the formation of the device state knowledge graph includes: Using the core secondary device of each device subgraph as the root node and the subordinate devices of the core secondary device as child nodes, a graph node structure reflecting the electrical connection and communication topology is constructed. The potential feature vectors are associated with the nodes in the graph, which serve as the state attributes of the devices corresponding to the nodes; In the knowledge graph, the connection edges between nodes are assigned connection types and weights defined by the device association matrix.

[0014] Further, the processing of the potential feature vector sequentially performs device health state evaluation, anomaly detection and associated risk analysis, comprising: inputting the potential feature vector into a pre-trained evaluation analysis model, outputting a quantitative device health state score based on the comparison results of real-time data in the dynamic state vector and device inherent benchmark data; matching the potential feature vector with a pre-set normal working condition feature library and calculating the deviation degree of the matching, identifying the abnormal device subgraph and specific abnormal parameters deviating from the normal mode based on the device health state score and the calculated deviation degree; combining the electrical connection relationship between nodes in the knowledge graph and the communication topology link and its weight, analyzing the influence of the abnormal state of the abnormal device subgraph on the associated device subgraph, and generating an associated risk analysis report.

[0015] From the above technical solutions, the present application has the following advantages: The present application divides the device network into multiple device subgraphs based on electrical topology and functional logic; real-time running data of each device subgraph is obtained through data acquisition agent on the production control area side, and is dynamically switched to the management information area through transmission strategy containing main and backup paths; the state feature subsequence of each device subgraph is generated based on real-time data on the management information area side, and then aggregated into a panoramic state feature sequence; the feature coding model is used to encode it into a low-dimensional potential feature vector, forming a device state knowledge graph; the feature vector in the graph is sequentially subjected to health state evaluation, anomaly detection and associated risk analysis, and a structured panoramic state monitoring report is generated. The present application unifies the representation form of multi-source heterogeneous data through device subgraph division and feature sequence construction, providing a structured basis for overall analysis; the constructed knowledge graph can represent the internal association between devices, supporting local to global state collaborative analysis; effectively improving the overallity of state evaluation, the timeliness of anomaly detection and the accuracy of risk identification, thereby providing effective technical support for safe and stable operation and intelligent operation and maintenance of power systems. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 is a flowchart of a panoramic state monitoring method of power secondary equipment across security partitions in the present application; Figure 2 is a flowchart of transmitting real-time running data to the management information area in the present application; Figure 3 is a flowchart of generating a state feature subsequence in the present application; Figure 4 is a flowchart of encoding by a pre-set feature coding model in the present application; Figure 5 A flowchart for forming a device state knowledge graph in the present application is shown in Figure 1. Figure 6 A flowchart for performing device health state evaluation, anomaly detection and associated risk analysis in the present application is shown in Figure 2. DETAILED DESCRIPTION

[0017] The terms "first", "second", "third", "fourth" and the like in the description of the application and in the claims hereof, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the use of the terms so termed herein is solely for descriptive purposes and not for pronouncing the limitations of the application described herein. For example, a first element billed as a "second" element, billed as a "third" element, and so on, can perform substantially the same function in functionally similar or different embodiments or use cases. Also, the terms "comprises", "comprising", "corresponds" and "corresponding", along with their derivatives, can be used herein to specify the presence of stated features, integers, steps or components but not preclude the presence or addition of one or more other features, integers, steps, components, or groups thereof. It is to be understood that the terms so used are to be interpreted in their broadest possible manner, consistent with the context of the application.

[0018] Embodiment One The method implemented in this embodiment can be implemented in a system, and can be implemented in a server or in a terminal, and the specific implementation is not limited. From the perspective of system implementation, the method in the present application will be introduced below. Please refer to Figure 1 The method provided by the embodiment of the present application comprises the following steps: S1. Based on the electrical topology and functional logic association relationship of the power secondary equipment, the device network is divided into a plurality of device subgraphs; The electrical topology refers to the physical connection relationship between the secondary equipment formed by cables, optical fibers and other media, such as the sampling loop between the protection device and the merging unit, and the trip loop between the intelligent terminal; the functional logic association relationship refers to the coordination and dependence relationship between devices for completing a specific system function, such as a complete line protection function which can be realized by the line protection device, the corresponding merging unit, the intelligent terminal and the operation box. The device network refers to the whole formed by all secondary equipment and their mutual connection relationship in a substation or power plant. The whole complex network is divided into a plurality of relatively independent and functionally cohesive analysis subsets, so as to facilitate subsequent fine data acquisition, feature extraction and state evaluation for each functional unit.

[0019] In this embodiment, each device subgraph contains a core secondary equipment and a slave device associated with the core secondary equipment, and the core secondary equipment contained in any two different device subgraphs has no intersection.

[0020] Specifically, the core secondary device refers to a device that plays a leading role in a specific protection, control or monitoring function and undertakes core logic processing, such as a line protection device, a transformer protection device, a bus protection device or a safety automatic device. The slave device refers to a device that directly cooperates with the core secondary device, provides data for the core secondary device or executes instructions of the core secondary device, such as a merging unit that provides current and voltage signals for the core secondary device, an intelligent terminal that receives and executes a trip and close command of the core secondary device, and a fault recorder. The association relationship is determined based on the electrical circuit connection and the established function configuration. It is stipulated that the core secondary devices included in any two different device subgraphs have no intersection, so as to ensure that the analysis subject of each functional unit is unique and avoid the state of the same core device being repeatedly or contradictorily evaluated in different subgraphs.

[0021] S2. On the production control area side, real-time operation data of each device subgraph is collected by the deployed data collection agent; and the real-time operation data is transmitted to the management information area according to a preset transmission strategy including a main and backup path, and the transmission strategy is dynamically switched based on the connectivity state of the data channel. The purpose of this step is to achieve safe and reliable transmission of real-time state data of secondary devices in the production control area to the management information area under the premise of meeting the safety partitioning regulations of the power monitoring system. The production control area side refers to a network area divided in the power monitoring system for realizing real-time monitoring and control of power production, and secondary devices such as protection devices and measurement and control devices are deployed in the production control area. The data collection agent is a software or soft and hardware integrated module deployed in the production control area, which is used to communicate with the above-mentioned secondary devices according to the established protocol, collect and aggregate real-time operation data thereof. The real-time operation data at least includes the action signal of the protection device, the current setting value, the soft pressure plate on-off state, the self-check alarm information and the metadata of the fault recording file. The preset transmission strategy including the main and backup paths is a data transmission rule set to cope with the uncertainty of cross-area networks, which configures two transmission paths of different principles, and selects the optimal path according to the real-time performance of the path, so as to ensure that the data flow can still be continuously and stably sent to the management information area when facing single path failure or performance degradation.

[0022] In this embodiment, the preset transmission strategy including the main and backup paths includes the following: 1. The first transmission path and the second transmission path are configured for the data collection agent, the first transmission path passes through the first forward isolation device to the management information area, the second transmission path is a backup path passing through the encrypted virtual private network, and the transmission priority of the first transmission path is higher than that of the second transmission path; 2. The real-time transmission state of the first transmission path and the second transmission path is monitored, and the real-time transmission state is defined according to the transmission requirements of the power monitoring data flow; 3. When the real-time transmission state of the first transmission path does not meet the preset transmission requirement for real-time operation data, the transmission of subsequent real-time operation data is dynamically switched to the second transmission path.

[0023] Specifically, the first forward isolation device is a special hardware device in line with the safety protection regulations of the power system, used to realize one-way physical isolation data ferry from the production control area to the management information area, and ensure one-way safe flow of data. The backup path of the encrypted virtual private network refers to another logically isolated communication channel established outside the first transmission path using authenticated and encrypted virtual private network technology. Encryption refers to the use of algorithms that meet the requirements of national cryptography management to encrypt and protect the transmitted data. The transmission priority of the first transmission path is set to be higher than that of the second transmission path, considering that the forward isolation device, as a standard safety device for the power system, usually has better reliability, stability and security, and is therefore the preferred path under normal circumstances. The real-time transmission state is defined according to the core requirement of reliable real-time performance of power monitoring data flow, and the monitoring indicators include at least the transmission success rate (the proportion of data packets successfully delivered) and the transmission delay (the time difference between data transmission from the production control area and reception in the management information area). The preset transmission requirement for real-time operation data is a quantitative threshold set in advance according to the importance of the business, the type of data and the network conditions when the system is deployed, for example, the transmission success rate is required to be no less than 99.9%, and the average transmission delay is required to be no more than 500 milliseconds. These preset requirements are determined based on the analysis of the demand for power secondary equipment state monitoring business and historical operation statistical data.

[0024] Please refer to Figure 2 , according to the preset transmission strategy including the primary and backup paths, transmit the real-time operation data to the management information area, including the following steps: S21. The data collection agent encapsulates the collected real-time operation data based on the currently selected transmission path; S22. In the encapsulation process, based on the shared key with the management information area side, generate and attach a message authentication code to the encapsulated data; S23. Transmit the encapsulated data with the attached message authentication code to the management information area through the currently selected transmission path.

[0025] Specifically, encapsulating the collected real-time operation data means adding necessary protocol headers, trailers and other information to the data according to the communication protocol supported by the selected transmission path, to form a standard data packet that can be transmitted on the network. Generating and attaching a message authentication code to the encapsulated data based on the shared key with the management information area side means using the security key previously agreed with the information management layer side, calculating a short data string that uniquely represents the content of the data packet and cannot be forged through cryptographic techniques such as hash operation, and attaching it to the data packet. This message authentication code is used to verify whether the data has been tampered with during transmission at the management information area side, and to confirm the legitimacy of the data source. Finally, the data packet is sent out through the currently selected transmission path (first path or second path) based on steps S21-S22 according to the format after encapsulation and signing.

[0026] S3. At the management information area side, generate a state feature sub-sequence corresponding to each device sub-graph based on the real-time operation data; This step is a deep processing and structuring of the real-time operation data transmitted safely from the production control area, which integrates static management information to generate unified, standardized and semantically rich feature data units for subsequent panoramic state analysis. The management information area side refers to the network area in the power monitoring system that is divided for non-real-time data management, advanced application analysis and decision support, where production management systems, device asset management systems and other systems are deployed. Please refer to Figure 3 This step is implemented through the following sub-steps: S31. Based on the unique device identifier of each core secondary device and slave device included in the device sub-graph, obtain the corresponding device inherent reference data; S32. Associate and match the real-time operation data with the device inherent reference data according to the device unique identifier, to form a multi-dimensional state data set of the device sub-graph; S33. Generate a state feature sub-sequence based on the multi-dimensional state data set, including logical coordinate coding, inter-device association matrix and dynamic state vector; wherein: 1. The logical coordinate coding is used to identify the central position of the core secondary device in the device sub-graph and the logical hierarchical relationship between the core secondary device and each slave device; 2. The inter-device association matrix represents the electrical connection, communication topology and logical dependency relationship between the core secondary device and each slave device; 3. The dynamic state vector is composed of the multi-dimensional state data set of the core secondary device and each slave device after quantization processing.

[0027] Specifically, the device unique identifier is a globally unique identity code pre-allocated to each secondary device in the station, such as a code generated based on the asset ID or the station-voltage level-interval-device type-sequence rule. The device inherent reference data is static data representing the inherent properties and operation and maintenance reference of the device, which is obtained from the production management system or the device asset management system in the management information area, and at least includes: technical specifications of the device, factory test data, major defect or failure records recorded in the historical account, and current anti-accident measure requirement texts or quantitative indicators applicable to the device model and operating environment. The association matching according to the device unique identifier means that the real-time operation data packet with the source device identifier in step S2 is matched and merged with the device inherent reference data organized according to the same identifier in step S31. The multi-dimensional state data set of the device subgraph formed thereby is a structured data set indexed by the device, which contains both the real-time monitoring value and the static reference value of each device entry, providing a data basis for state comparison and analysis. The state feature sub-sequence is a further abstracted and formatted multi-dimensional data set: 1. The logical coordinate code assigns a code representing the center node to the core device according to the functional role of the device in the subgraph, and assigns a code reflecting the hierarchical relationship and connection order of the core device to each slave device, which together constitute a coordinate sequence describing the topological relationship.

[0028] 2. The inter-device association matrix is a two-dimensional matrix, whose rows and columns correspond to the core device and each slave device in the subgraph, respectively. The element values in the matrix are used to quantitatively represent whether there is a connection relationship between the devices, such as sampling, tripping, and interlocking, and the strength thereof.

[0029] 3. The dynamic state vector normalizes and standardizes the key state parameters of each device in the multi-dimensional state data set, and arranges them into a one-dimensional numerical vector in a predetermined order. The vector comprehensively reflects the current operating state of all devices in the device subgraph.

[0030] S4. Aggregate all state feature sub-sequences corresponding to the device subgraphs to generate a panoramic state feature sequence; process the panoramic state feature sequence using a pre-set feature coding model to code it into a latent feature vector with lower dimension than the panoramic state feature sequence, forming a device state knowledge graph; The state features representing the local functional units (device subgraphs) are integrated into a unified data structure representing the operation state of the entire secondary equipment cluster of the substation or station, and a deep learning model is used to compress and abstract the knowledge. The feature encoding model is used to reduce the dimension and encode the panoramic state feature sequence, so as to extract a low-dimensional dense representation containing the essential information of the device health state, filter out the noise and redundancy in the original data, and highlight the key state patterns. The device state knowledge graph is a data model that structurally represents device entities, their attributes (states), and the relationships between entities.

[0031] In this embodiment, all state feature subsequences corresponding to the device subgraphs are aggregated to generate a panoramic state feature sequence, including: sorting and splicing all state feature subsequences corresponding to the device subgraphs according to a preset global logical order; wherein the preset global logical order is determined according to one or more of the physical location, voltage level or functional importance of each device subgraph in the power network.

[0032] Specifically, the preset global logical order is to reflect the actual organization logic of the secondary equipment in the power system in the panoramic sequence, rather than random arrangement. The determination rule is, for example: preferentially sorting by voltage level from high to low, such as arranging the 500kV interval subgraph before the 220kV interval subgraph; under the same voltage level, sorting by physical location or screen cabinet arrangement order, such as the order from the #1 main transformer interval, #1 bus interval to the outgoing line interval of the substation; for device subgraphs that undertake system stability control, bus protection and other key functions, a higher sorting priority can be given according to their functional importance. This sorting method based on the operation logic of the power system makes the generated panoramic state feature sequence contain the topological and functional hierarchy information of the system in structure, which is conducive to the subsequent coding model to learn patterns with physical meaning, and supports segmented analysis by region, voltage level, etc.

[0033] In this embodiment, the preset feature encoding model is a variational quantization autoencoder. The output of this encoder is a distribution parameter in a continuous latent space, and the latent vector is obtained by sampling, which introduces randomness and helps the model to learn more robust feature representation. Secondly, a learnable discrete codebook is also introduced. The latent vector is not directly input into the decoder, but needs to find the discrete coding vector closest to it in the codebook to replace it, and this quantized discrete vector is used as the input of the decoder. The codebook will spontaneously organize itself during the learning process, and different discrete codes may correspond to different typical device state combination patterns.

[0034] The panoramic state feature sequence is processed by the preset feature encoding model to encode it into a latent feature vector with a lower dimension than the panoramic state feature sequence, as shown in Figure 4 , including the following steps: S411. Input the panoramic state feature sequence into the encoder of the variational quantization autoencoder; S412. Map the panoramic state feature sequence to a continuous latent space through the multi-layer neural network processing of the encoder; S413. Quantize the vector in the latent space, and select a closest quantized vector as the latent feature vector output.

[0035] Specifically, the encoder is a deep neural network, the input layer dimension of which matches the length of the panoramic state feature sequence. The multi-layer neural network of the encoder is composed of combinations of fully connected layers, convolutional layers (for capturing sequence local correlation) or long short-term memory network layers (for capturing sequence time sequence dependence), etc., and the specific number of layers and structure are set according to data complexity and experience. The network nonlinearly transforms the high-dimensional input sequence into a vector in a low-dimensional, continuous latent space. The latent space is an abstract feature space learned by the model, and each point (vector) in the space corresponds to a compressed representation of the input panoramic sequence. In theory, similar panoramic states are closer in the space. The model maintains a trainable codebook with K entries, each of which is a vector with the same dimension as the latent vector. The quantization process calculates the Euclidean distance between the continuous latent vector output by the encoder and all K encoding vectors in the codebook, and then selects the closest encoding vector as the quantized vector. The selected quantized vector is the final latent feature vector output. This process realizes the conversion from continuous representation to discrete representation, forcing the model to learn a discrete, information-intensive latent representation.

[0036] Please refer to Figure 5 to form the device state knowledge graph, including the following steps: S421. Take the core secondary device of each device subgraph as the root node, and take the slave devices of the core secondary device as the child nodes to construct a graph node structure reflecting the electrical connection and communication topology; S422. Associate the latent feature vector with the nodes in the graph as the state attributes of the devices corresponding to the nodes; S423. In the knowledge graph, assign connection types and weights defined by the inter-device association matrix to the connection edges between nodes.

[0037] Specifically, the nodes in the graph represent physical equipment entities, and each equipment subgraph is converted into a tree-like or star-like substructure with the core equipment node as the root and its subordinate equipment nodes as the children. These substructures are interconnected through the actual connection relationship between the equipment, and together constitute a graph structure that covers the entire station and reflects the real secondary circuit connection and communication network topology. The potential feature vector obtained in step S4 is associated with the core equipment node and its subordinate equipment nodes corresponding to the equipment subgraph from which the feature sequence is derived. The potential feature vector of the entire subgraph can be taken as a comprehensive state attribute of the root node (core equipment) of the subgraph, or distributed to each subordinate equipment node through a more detailed mapping relationship; for example, different dimensions or different parts obtained after decoding of the potential feature vector as its state attribute. This makes each node in the knowledge graph not only have its identity, but also have a quantitative attribute reflecting its current health and operating condition. The connection edges between the nodes in the graph are defined by the inter-equipment association matrix generated in S3. Each edge can have a type attribute and can have a weight attribute. The equipment state knowledge graph thus becomes a structured knowledge base that integrates equipment entities, real-time state attributes, and rich association relationships, and is queryable, inferable, and computable.

[0038] S5. Processing the potential feature vector, and sequentially performing equipment health state evaluation, anomaly detection, and associated risk analysis; This step is implemented through three logical analysis stages: quantitatively evaluating the overall health condition of the equipment to provide an intuitive comprehensive state indicator; accurately locating the functional unit and specific parameter that appears abnormal based on the evaluation results and feature matching; and analyzing the systemic risks that may be triggered by the local anomaly using the association relationships of the knowledge graph. Please refer to Figure 6 , the implementation process includes the following steps: S51. Input the potential feature vector into a pre-trained evaluation analysis model, and output a quantitative equipment health state score based on the comparison results of the real-time data in the dynamic state vector and the equipment inherent baseline data; The pre-trained evaluation analysis model is a regression model based on a machine learning algorithm. The model is supervised trained using a large amount of historical data, thereby learning the complex mapping relationship from the potential feature to the health state. The model performs comparison in that the potential feature vector input to the model is an advanced abstraction of the original dynamic state vector, and the model has implicitly learned the deviation pattern between the real-time value and the baseline value from it. The output of the model is a quantitative equipment health state score, for example, a value between 0 and 100, and the higher the score, the better the health state. The score can comprehensively reflect the electrical performance, mechanical state, logical function, and compliance with countermeasures requirements of the equipment in multiple dimensions.

[0039] S52. The potential feature vector is matched with the preset normal operating condition feature library, and the deviation degree of matching is calculated. Based on the device health state score and the calculated deviation degree, the abnormal device subgraph and specific abnormal parameters deviating from the normal mode are identified. The preset normal operating condition feature library is constructed by clustering analysis of all potential feature vectors of the device in a known healthy running state over a long period of history. The feature library defines a set of typical state patterns of the system under normal conditions. The deviation degree of matching is the shortest distance from the current potential feature vector to be analyzed to all cluster centers in the normal operating condition feature library. This shortest distance is defined as the deviation degree. The greater the deviation degree, the greater the difference between the current state and any known normal mode. Abnormality identification is a multi-condition decision process: 1. When the device health state score is lower than the preset attention threshold; and / or 2. When the deviation degree exceeds the preset statistical boundary, it is determined that the device subgraph corresponding to the potential feature vector is an abnormal device subgraph. Further, by analyzing which specific parameters in the dynamic state vector corresponding to the subgraph contribute most to the low score or high deviation degree, the specific abnormal parameters can be located. The purpose of identifying the abnormal device subgraph and the specific abnormal parameters is to provide accurate analysis targets for subsequent steps, avoid indiscriminate scanning of the entire network data, greatly improve the analysis efficiency, and provide direct troubleshooting clues for the operation and maintenance personnel.

[0040] S53. In combination with the electrical connection relationship between nodes in the knowledge graph and the communication topology link and its weight, the influence of the abnormal state of the abnormal device subgraph on the associated device subgraph is analyzed, and an associated risk analysis report is generated.

[0041] The analysis of the influence is a graph-based reasoning: taking the core device node corresponding to the identified abnormal device subgraph as the starting point, traversing along the electrical connection relationship edges and communication topology link edges defined in the knowledge graph, and according to the weights of the edges. Simulate the possible path and influence intensity of the abnormal state propagating to the associated nodes through these edges. For example, analyze whether the abnormal blocking signal of the A interval protection device will cause the protection function of the B interval to be misblocked through the interlocking GOOSE network.

[0042] S6. According to the results of device health state evaluation, abnormality detection and associated risk analysis, a structured panoramic state monitoring report is dynamically generated and output.

[0043] The step dynamically generates and outputs a structured panoramic state monitoring report according to all results obtained by the device health state scores, the abnormal device subgraph list, the specific abnormal parameters and the associated risk analysis report of the preceding steps S1 to S5. The report includes the following parts: 1. a global operation state overview, which displays the health score distribution and the number of abnormal devices in the form of a dashboard; 2. a detailed abnormality and alarm list, which lists all abnormal device subgraphs, their core device identifiers, specific abnormal parameters, deviation degrees and health scores in order of priority; 3. an associated risk prompt section, which abstractly displays the risk propagation path and the affected associated devices; and a knowledge graph-based visual topology diagram, which highlights abnormal nodes and risk paths. The report is automatically pushed to the relevant operation and maintenance management system through a standard data interface or a visual interface, providing operation and maintenance personnel with a clear and actionable decision view, so that the complex technical analysis results described above can be efficiently and accurately converted into direct evidence for guiding on-site inspection, periodic inspection or defect elimination.

[0044] It can be understood that those skilled in the art can combine various embodiments in the above embodiments under the guidance of the above embodiments to obtain various embodiments of the technical solutions.

[0045] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for panoramic status monitoring of secondary power equipment across safety zones, characterized in that, Power monitoring systems applied to areas that include production control zones and management information zones include: Based on the electrical topology and functional logic relationships of secondary power equipment, the equipment network is divided into multiple equipment sub-graphs; On the production control area side, real-time operating data of each of the equipment sub-graphs is collected by deployed data acquisition agents; according to a preset transmission strategy including primary and backup paths, the real-time operating data is transmitted to the management information area, and the transmission strategy is dynamically switched based on the connectivity status of the data channel; On the management information area side, a state feature subsequence corresponding to each device sub-graph is generated based on the real-time operation data; Aggregate the state feature subsequences corresponding to all device subgraphs to generate a panoramic state feature sequence; process the panoramic state feature sequence using a preset feature encoding model to encode it into a latent feature vector with a dimension lower than the panoramic state feature sequence, forming a device state knowledge graph; The potential feature vectors are processed, and equipment health status assessment, anomaly detection, and associated risk analysis are performed sequentially. Based on the results of the equipment health status assessment, anomaly detection, and associated risk analysis, a structured panoramic status monitoring report is dynamically generated and output.

2. The method for panoramic status monitoring of secondary power equipment across safety zones according to claim 1, characterized in that, Each device subgraph contains a core secondary device and subordinate devices associated with the core secondary device. The core secondary devices contained in any two different device subgraphs do not overlap.

3. The method for panoramic status monitoring of secondary power equipment across safety zones according to claim 1, characterized in that, The preset transmission strategy, which includes primary and backup paths, includes: Configure a first transmission path and a second transmission path for the data acquisition agent. The first transmission path leads to the management information area via a first forward isolation device, and the second transmission path is a backup path via an encrypted virtual private network. The transmission priority of the first transmission path is higher than that of the second transmission path. The real-time transmission status of the first transmission path and the second transmission path is monitored, and the real-time transmission status is defined according to the transmission requirements of the power monitoring data stream. When the real-time transmission status of the first transmission path does not meet the preset transmission requirements for the real-time running data, the subsequent transmission of the real-time running data will be dynamically switched to the second transmission path.

4. The method for panoramic status monitoring of secondary power equipment across safety zones according to claim 3, characterized in that, The step of transmitting the real-time operating data to the management information area according to a preset transmission strategy including primary and backup paths includes: The data acquisition agent encapsulates the collected real-time running data based on the currently selected transmission path; During the encapsulation process, a message authentication code is generated and attached to the encapsulated data based on the key shared with the management information area side; The encapsulated data with the attached message authentication code is transmitted to the management information area via the currently selected transmission path.

5. The method for panoramic status monitoring of secondary power equipment across safety zones according to claim 1, characterized in that, The step of generating a state feature subsequence corresponding to each device subgraph based on the real-time operating data includes: Based on the unique device identifiers of the core secondary devices and subordinate devices contained in each device sub-graph, obtain the corresponding inherent baseline data of the device. The real-time operating data is associated and matched with the inherent baseline data of the device according to the unique identifier of the device to form a multi-dimensional state data set of the device sub-graph; A state feature subsequence is generated based on the multidimensional state data set. The dynamic feature sequence includes logical coordinate encoding, inter-device correlation matrix, and dynamic state vector; wherein: The logical coordinate encoding is used to identify the center position of the core secondary device in the device sub-diagram and the logical hierarchy relationship between the core secondary device and each subordinate device. The device association matrix represents the electrical connections, communication topology, and logical dependencies between the core secondary device and each subordinate device. The dynamic state vector is formed by quantizing the multi-dimensional state data set of the core secondary device and each subordinate device.

6. The method for panoramic status monitoring of secondary power equipment across safety zones according to claim 1, characterized in that, The aggregation of the state feature subsequences corresponding to all device subgraphs to generate a panoramic state feature sequence includes: All the state feature subsequences corresponding to the device subgraphs are sorted and spliced ​​according to a preset global logical order; wherein the preset global logical order is determined based on at least one or more of the physical location, voltage level or functional importance of each device subgraph in the power network.

7. The method for panoramic status monitoring of secondary power equipment across safety zones according to claim 1, characterized in that, The preset feature encoding model is a variational quantization autoencoder.

8. The method for panoramic status monitoring of secondary power equipment across safety zones according to claim 7, characterized in that, The step of processing the panoramic state feature sequence using a preset feature encoding model to encode it into a latent feature vector with a dimension lower than that of the panoramic state feature sequence includes: The panoramic state feature sequence is input into the encoder of the variational quantization autoencoder; The panoramic state feature sequence is mapped to a continuous latent space through multi-layer neural network processing of the encoder. The vectors in the latent space are quantized, and the closest quantized vector is selected as the output latent feature vector.

9. The method for panoramic status monitoring of secondary power equipment across safety zones according to claim 5 or 8, characterized in that, The formation of the device status knowledge graph includes: Using the core secondary device of each device subgraph as the root node and the subordinate devices of the core secondary device as child nodes, a graph node structure reflecting the electrical connection and communication topology is constructed. The potential feature vectors are associated with the nodes in the graph, which serve as the state attributes of the devices corresponding to the nodes; In the knowledge graph, the connection edges between nodes are assigned connection types and weights defined by the device association matrix.

10. The method for panoramic status monitoring of secondary power equipment across safety zones according to claim 9, characterized in that, The process of processing the potential feature vectors, including sequentially performing equipment health status assessment, anomaly detection, and associated risk analysis, includes: The latent feature vector is input into a pre-trained evaluation and analysis model. Based on the comparison between real-time data in the dynamic state vector and the inherent benchmark data of the device, a quantitative device health status score is output. The potential feature vectors are matched with a preset normal operating condition feature library and the deviation of the match is calculated. Based on the equipment health status score and the calculated deviation, abnormal equipment sub-graphs and specific abnormal parameters that deviate from the normal mode are identified. By combining the electrical connections and communication topology links between nodes in the knowledge graph and their weights, the potential impact of the abnormal state of the abnormal device subgraph on the associated device subgraph is analyzed, and an associated risk analysis report is generated.

Images