Power internet of things anomaly monitoring system and method based on topological constraint modal representation
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TIANJIN BAIZE TECH CO LTD
- Filing Date
- 2026-04-27
- Publication Date
- 2026-07-21
AI Technical Summary
Existing power IoT anomaly monitoring solutions ignore electrical topology constraints, making it difficult to distinguish multimodal data quality. This results in insufficient ability to differentiate between deviations in equipment physical state and topology consistency, and a lack of effective differentiation of anomaly types.
By constructing a topology graph, acquiring multimodal data and preprocessing it, using topological constraint modal representation for neighborhood aggregation and quality perception fusion, and combining power mechanism consistency verification, physical deviation scores and consistency deviation indicators are generated to distinguish anomaly types.
It improves the ability to detect linkage anomalies and cascading effects, reduces the false alarm rate, enhances the ability to distinguish between physical faults, sensor/communication anomalies, and topology configuration anomalies, and strengthens the interpretability and reliability of the system.
Smart Images
Figure CN122437243A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power system condition monitoring and intelligent analysis technology, specifically to a power Internet of Things anomaly monitoring system and method based on topological constraint modal characterization. Background Technology
[0002] In the ubiquitous power Internet of Things (IoT) environment, online measurement devices for voltage, current, active / reactive power, and frequency are widely deployed in transmission, substation, distribution, and important user sectors, along with environmental and status sensors for temperature, humidity, smoke, and vibration. These are complemented by visible light and infrared video cameras, microphones, and systems for recording operation logs, alarms, and maintenance data. These terminals connect to the power IoT platform via communication networks, forming high-dimensional, multimodal, and strongly time-correlated data streams.
[0003] From a mechanistic perspective, the operating state of equipment can be viewed as an "implicit state" that evolves over time, while multimodal data represents different observation channels for this state: measurement channels reflect electrical quantities and load disturbances, image and infrared channels reflect appearance and temperature rise, sound channels reflect mechanical and discharge characteristics, text channels carry event semantics, and environmental channels characterize external conditions. These observations are not isolated; their spatial correlation is determined by the electrical topology structure composed of buses, lines, transformers, and switches; the same anomaly often exhibits interpretable interconnected characteristics within its topological neighborhood.
[0004] The existing technology has the following defects or problems: Existing anomaly monitoring solutions mainly have the following problems: First, it ignores electrical topology constraints and often uses a single device or sensor as the modeling unit, making it difficult to determine whether the anomaly is reasonable in terms of topology and whether it is propagable. Secondly, there is a lack of explicit characterization of modal data quality. Multimodal features are often simply spliced together or fixedly weighted. When the camera is blocked, the sensor drifts, or there is packet loss in communication, it is difficult to distinguish between data anomalies and physical anomalies. Third, anomaly scores are mostly based on a single indicator, making it difficult to simultaneously characterize deviations in the physical state of equipment and its deviations from the topology consistency of surrounding equipment. This results in insufficient ability to distinguish between different anomaly types such as sensor failures, topology configuration errors, and actual physical failures.
[0005] Therefore, there is a need for a power Internet of Things (IoT) anomaly monitoring technology that simultaneously considers electrical topology, multimodal data quality, and anomaly type differentiation.
[0006] It should be noted that the above content falls within the inventor's technical knowledge and does not necessarily constitute prior art. Summary of the Invention
[0007] To address the shortcomings of existing technologies, this invention provides a power Internet of Things anomaly monitoring system and method based on topological constraint modal characterization, which solves the current problems.
[0008] To achieve the above objectives, the present invention provides the following technical solution: a power Internet of Things anomaly monitoring method based on topological constraint modal characterization, comprising the following steps: Step 1: Obtain electrical topology information of the target area and construct a topology graph G=(V,E), and at the same time obtain multimodal data of each node of the topology graph within a preset time window; Step 2: Preprocess and time-align the multimodal data to construct a node-1 window and multimodal samples; Step 3: Encode each modal data to obtain modal representations, and perform neighborhood aggregation on the topology graph to obtain topological constraint representations; Step 4: Extract the quality features of each modality and calculate the modality confidence coefficient. Based on the modality confidence coefficient, perform quality perception fusion on the modality representations to obtain the fused representation. Step 5: Based on the topological constraint representation and fusion representation, jointly reconstruct the modal data and features, calculate the reconstruction error, and generate a physical deviation score; Step 6: Based on the consistency of topological neighborhood and the consistency of power mechanism, obtain the consistency deviation index; Step 7: Construct an anomaly profile based on the physical deviation score and consistency deviation index, determine the threshold, output an anomaly alarm, and at least distinguish between physical fault anomalies, sensor / communication anomalies, and topology configuration anomalies.
[0009] In some embodiments, the multimodal data includes time-series measurement data, device operation data, text data, and environmental and state sensor data, wherein: The time-series measurement data includes voltage, current, active power, reactive power, and frequency; The device operation data includes visible light images and device operation sounds; The text data includes operation logs, alarm records, and maintenance records; The environmental and condition sensor data includes temperature, humidity, smoke concentration, and vibration level.
[0010] In some embodiments, the topological constraint representation is obtained by performing neighborhood aggregation on the node representation on the topological graph G. The neighborhood aggregation considers at least the information of adjacent nodes and the node itself, and sets the adjacency weight and on / off attributes according to the electrical connection relationship.
[0011] In some embodiments, the modal quality features include at least one or more of the following: missing rate, saturation / truncation rate, signal-to-noise ratio, outlier ratio, short-time stability, and time alignment bias. The modal confidence coefficient is output by the confidence evaluation network based on the modal quality characteristics, and its value range is [0,1].
[0012] In some embodiments, the quality-perceived fusion uses modal scoring and modal credibility to jointly determine modal weights, and performs a weighted summation of each modal representation to obtain a fused representation, wherein the modality with lower modal credibility has a smaller weight.
[0013] In some embodiments, the power mechanism consistency check includes: The residual of state estimation, power flow trend consistency deviation, and power balance deviation are calculated based on measurement data and topology constraints.
[0014] In some embodiments, the exception type distinction includes at least: When both the physical deviation score and the consistency deviation index exceed the threshold, it is judged as a physical fault anomaly; when the physical deviation score is mainly contributed by low confidence modes and the consistency deviation index does not exceed the threshold, it is judged as a sensor / communication anomaly; when the consistency deviation index exceeds the threshold but the physical deviation score does not increase significantly, it is judged as a topology configuration anomaly.
[0015] Another technical problem to be solved by this invention is to propose a power Internet of Things anomaly monitoring system based on topological constraint modal characterization, comprising: The topology and data acquisition module is used to acquire electrical topology relationships and multimodal data in real time, thereby solving the data silo problem and providing a complete topology view and raw data foundation for all subsequent analyses. The topology and data acquisition module is configured to acquire the connection relationships of power grid buses, branches and equipment in real time as electrical topology information, which includes telemetry data, status data, image data, environmental parameter data, communication traffic datasets and text alarm information. The preprocessing and window building module is used to preprocess multimodal data and build a sliding time window; The preprocessing and window building module includes one or more of the following: a data cleaning unit, a noise reduction unit, a missing value completion unit, and an outlier suppression unit, for performing preprocessing; and includes a time alignment unit and a window segmentation unit, for aligning multimodal data and building sliding windows according to a uniform time scale; A multimodal encoding module is used to encode the multimodal data within the sliding time window to obtain the feature representation of each modality; A topology constraint representation module is used to apply topology constraints to the feature representations of each mode based on the electrical topology information, thereby generating a topology-aware multimodal representation. The topology constraint characterization module is configured to introduce one or more of the following during the feature representation process: adjacency constraint, electrical distance constraint, and power flow direction constraint, so that the generated topology-aware multimodal characterization has the ability to perceive the physical topology of the power grid. The modal credibility assessment module is used to dynamically assess the credibility of each modality in the current scenario and obtain the credibility weight of each modality. The modal credibility assessment module is configured to dynamically quantify the relative credibility of each modality at the current moment and generate the credibility weights based on one or more of the following: the current level of interference with the modality, the probability of sensor failure, communication quality, signs of potential network attacks, and the aging state of the equipment. A quality-aware fusion module is used to perform quality-aware fusion of the multimodal representations of topology awareness according to the confidence weights of each modality, and generate a fused representation. The quality perception fusion module adopts an adaptive weighted fusion mechanism, which makes the modal with higher confidence weight occupy a larger proportion in the fusion, while the weight of the modal with lower confidence weight is reduced or even suppressed to close to zero. The joint reconstruction and scoring module is used to jointly reconstruct the fused representation and calculate the anomaly score based on the reconstruction error and other auxiliary indicators. A physical consistency verification module is used to perform a physical consistency verification on the anomaly score based on the physical mechanism of the power grid. The physical consistency verification module is configured to perform a secondary verification on the anomaly score or anomaly judgment result based on the physical mechanism of the power grid. The alarm and explanation output module is used to generate abnormal alarms and output interpretable analysis results based on the verified results; The alarm and explanation output module is configured to output one or more of the following: the specific device or area where the anomaly occurred, the severity level of the anomaly, the main contributing modality of the anomaly, the anomaly propagation path, the specific physical rules violated, and a natural language explanation text describing the anomaly. In some embodiments, the system is communicatively connected to a power Internet of Things (IoT) platform, which provides the system with electrical topology information and multi-source multimodal data, and receives anomaly alarms and interpretability analysis results output by the system. The platform is deployed in any one or more of the following locations: cloud master station, edge computing node, or station-side intelligent terminal, to support centralized, distributed, or edge anomaly monitoring architecture.
[0016] Compared with existing technologies, this invention provides a power Internet of Things (IoT) anomaly monitoring system and method based on topological constraint modal characterization, which has the following beneficial effects: This power IoT anomaly monitoring system and method based on topology-constrained modal characterization improves the detection capability of linked anomalies and cascading effects by setting topology constraint modeling and explicitly utilizing electrical topology structure for neighborhood consistency learning. It achieves dynamic weighting through quality feature-driven credibility assessment, significantly reducing false alarms caused by occlusion, drift, and packet loss. It can still be trained under conditions of scarce anomaly samples, facilitating its application in real power grid scenarios. The introduction of auditable power mechanism constraints enhances theoretical interpretability and engineering credibility. Simultaneously, it constructs a two-dimensional anomaly profile to achieve differentiated alarms for physical faults, sensor / communication anomalies, and topology configuration anomalies, improving responsiveness. Attached Figure Description
[0017] Figure 1 This is a schematic diagram of the electrical topology of the power Internet of Things according to the present invention.
[0018] Figure 2 This is a schematic diagram of the overall process of the power Internet of Things anomaly monitoring method based on topological constraint modal characterization of the present invention.
[0019] Figure 3 This is a schematic diagram of the topological constraint neighborhood consistency representation learning structure of the present invention.
[0020] Figure 4 This is a schematic diagram of the modal reliability assessment and quality perception fusion structure of the present invention.
[0021] Figure 5 This is a schematic diagram illustrating the two-dimensional anomaly profile and the distinction between anomaly types in this invention. Detailed Implementation
[0022] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments and accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0023] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.
[0024] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.
[0025] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.
[0026] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.
[0027] Please see Figure 1-5 In this implementation plan, the power Internet of Things anomaly monitoring method based on topology-constrained modal characterization includes the following steps: Step 1: Obtain electrical topology information of the target area and construct a topology graph G=(V,E), while simultaneously obtaining multimodal data of each node in the topology graph within a preset time window; Multimodal data includes time-series measurement data, equipment operation data, text data, and environmental and condition sensor data, among which: Time-series measurement data include voltage, current, active power, reactive power, and frequency; Equipment operation data includes visible light images and equipment operating sounds; Text data includes operation logs, alarm records, and maintenance records; Environmental and condition sensor data include temperature, humidity, smoke concentration, and vibration levels; Step 2: Preprocess and time-align the multimodal data to construct a node-1 window and multimodal samples; The device node mapping table is used to map primary devices or monitored objects within the target area to a set of topology nodes V. The node_id is a unique node-level identifier used to associate with multimodal data and log / alarm data, as shown in Table 1 below: Table 1
[0028] The connection relationship table is used to describe the electrical connection relationships between nodes to form the edge set E, and includes a switch-type device on / off status field to determine the validity of the edge, as shown in Table 2 below: Table 2
[0029] Connection relationship determination rules: R1: When switch_state is Closed, add the connection relationship corresponding to (from_node_id, to_node_id) to the edge set E; R2: When switch_state is Open, do not add it to the edge set E (or keep the edge but reset its weight to 0, maintaining consistency throughout the text); R3: If weight_param exists, use it as the edge weight; otherwise, assign the default weight to the valid edge. The multimodal data is preferably associated with node_id and timestamp as the key, and the platform can provide it through the historical database query interface or the real-time subscription interface. The following are examples of data fields for each subclass, containing at least one or more of these fields, as detailed in Table 3: Table 3-1 Time Series Measurement Data
[0030] Table 3-2 Image / Infrared Data
[0031] Table 3-3 Sound Data
[0032] Table 3-4 Text Log / Alarm Data
[0033] Table 3-5 Environmental and Condition Sensor Data
[0034] Step 3: Encode each modal data to obtain modal representations, and perform neighborhood aggregation on the topology graph to obtain topological constraint representations; Topological constraint representation is obtained by performing neighborhood aggregation on the node representation in the topological graph G. Neighborhood aggregation considers at least the information of adjacent nodes and the information of the node itself, and sets the adjacency weight and on / off attributes according to the electrical connection relationship. Step 4: Extract the quality features of each modality and calculate the modality confidence coefficient. Based on the modality confidence coefficient, perform quality perception fusion on the modality representations to obtain the fused representation. Modal quality characteristics include at least one or more of the following: missing rate, saturation / truncation rate, signal-to-noise ratio, outlier ratio, short-term stability, and time alignment bias. The modal confidence coefficient is output by the confidence assessment network based on the modal quality characteristics, and its value ranges from [0,1]. The quality perception fusion uses modal scoring and modal credibility to jointly determine modal weights, and performs a weighted summation of the modal representations to obtain the fused representation, where the lower the modal credibility, the smaller the weight corresponding to the modality. Step 5: Based on topological constraint representation and fusion representation, jointly reconstruct the modal data and features, calculate the reconstruction error and generate a physical deviation score; Step 6: Based on the consistency of topological neighborhood and the consistency of power mechanism, obtain the consistency deviation index; The consistency check of power mechanism includes: Calculate at least one of the following based on measurement data and topology constraints: state estimation residual, power flow trend consistency deviation, and power balance deviation; Step 7: Construct an anomaly profile based on the physical deviation score and consistency deviation index, determine the threshold, output anomaly alarms, and at least distinguish between physical fault anomalies, sensor / communication anomalies, and topology configuration anomalies. The distinction between exception types should include at least the following: When both the physical deviation score and the consistency deviation index exceed the threshold, it is judged as a physical fault anomaly; when the physical deviation score is mainly contributed by low confidence modes and the consistency deviation index does not exceed the threshold, it is judged as a sensor / communication anomaly; when the consistency deviation index exceeds the threshold but the physical deviation score does not increase significantly, it is judged as a topology configuration anomaly.
[0035] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.
[0036] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for anomaly monitoring in the power Internet of Things based on topological constraint modal characterization, characterized in that, Includes the following steps: Step 1: Obtain electrical topology information of the target area and construct a topology graph G=(V,E), and at the same time obtain multimodal data of each node of the topology graph within a preset time window; Step 2: Preprocess and time-align the multimodal data to construct a node-1 window and multimodal samples; Step 3: Encode each modal data to obtain modal representations, and perform neighborhood aggregation on the topology graph to obtain topological constraint representations; Step 4: Extract the quality features of each modality and calculate the modality confidence coefficient. Based on the modality confidence coefficient, perform quality perception fusion on the modality representations to obtain the fused representation. Step 5: Based on the topological constraint representation and fusion representation, jointly reconstruct the modal data and features, calculate the reconstruction error, and generate a physical deviation score; Step 6: Based on the consistency of topological neighborhood and the consistency of power mechanism, obtain the consistency deviation index; Step 7: Construct an anomaly profile based on the physical deviation score and consistency deviation index, determine the threshold, output an anomaly alarm, and at least distinguish between physical fault anomalies, sensor / communication anomalies, and topology configuration anomalies.
2. The power Internet of Things anomaly monitoring method based on topological constraint modal characterization according to claim 1, characterized in that, The multimodal data includes time-series measurement data, equipment operation data, text data, and environmental and status sensor data, among which: The time-series measurement data includes voltage, current, active power, reactive power, and frequency; The device operation data includes visible light images and device operation sounds; The text data includes operation logs, alarm records, and maintenance records; The environmental and condition sensor data includes temperature, humidity, smoke concentration, and vibration level.
3. The power Internet of Things anomaly monitoring method based on topological constraint modal characterization according to claim 1, characterized in that, The topological constraint representation is obtained by performing neighborhood aggregation on the node representation in the topological graph G. The neighborhood aggregation considers at least the information of adjacent nodes and the information of the node itself, and sets the adjacency weight and on / off attributes according to the electrical connection relationship.
4. The power Internet of Things anomaly monitoring method based on topological constraint modal characterization according to claim 1, characterized in that, The modal quality characteristics include at least one or more of the following: missing rate, saturation / truncation rate, signal-to-noise ratio, outlier ratio, short-term stability, and time alignment bias. The modal confidence coefficient is output by the confidence evaluation network based on the modal quality characteristics, and its value range is [0,1].
5. The power Internet of Things anomaly monitoring method based on topological constraint modal characterization according to claim 1, characterized in that, The quality perception fusion uses modal scoring and modal credibility to jointly determine modal weights, and performs weighted summation of each modal representation to obtain the fused representation, wherein the lower the modal credibility, the smaller the corresponding weight.
6. The power Internet of Things anomaly monitoring method based on topological constraint modal characterization according to claim 1, characterized in that, The consistency verification of the power mechanism includes: The residual of state estimation, power flow trend consistency deviation, and power balance deviation are calculated based on measurement data and topology constraints.
7. The power Internet of Things anomaly monitoring method based on topological constraint modal characterization according to claim 1, characterized in that, The distinction between the anomaly types includes at least: When both the physical deviation score and the consistency deviation index exceed the threshold, it is judged as a physical fault anomaly; when the physical deviation score is mainly contributed by low confidence modes and the consistency deviation index does not exceed the threshold, it is judged as a sensor / communication anomaly; when the consistency deviation index exceeds the threshold but the physical deviation score does not increase significantly, it is judged as a topology configuration anomaly.
8. A power Internet of Things anomaly monitoring system based on topological constraint modal characterization, characterized in that, include: The topology and data acquisition module is used to acquire electrical topology relationships and multimodal data in real time, thereby solving the data silo problem and providing a complete topology view and raw data foundation for all subsequent analyses. The topology and data acquisition module is configured to acquire the connection relationships of power grid buses, branches and equipment in real time as electrical topology information, which includes telemetry data, status data, image data, environmental parameter data, communication traffic datasets and text alarm information. The preprocessing and window building module is used to preprocess multimodal data and build a sliding time window; The preprocessing and window building module includes one or more of the following: a data cleaning unit, a noise reduction unit, a missing value completion unit, and an outlier suppression unit, for performing preprocessing; and includes a time alignment unit and a window segmentation unit, for aligning multimodal data and building sliding windows according to a uniform time scale; A multimodal encoding module is used to encode the multimodal data within the sliding time window to obtain the feature representation of each modality; A topology constraint representation module is used to apply topology constraints to the feature representations of each mode based on the electrical topology information, thereby generating a topology-aware multimodal representation. The topology constraint characterization module is configured to introduce one or more of the following during the feature representation process: adjacency constraint, electrical distance constraint, and power flow direction constraint, so that the generated topology-aware multimodal characterization has the ability to perceive the physical topology of the power grid. The modal credibility assessment module is used to dynamically assess the credibility of each modality in the current scenario and obtain the credibility weight of each modality. The modal credibility assessment module is configured to dynamically quantify the relative credibility of each modality at the current moment and generate the credibility weights based on one or more of the following: the current level of interference with the modality, the probability of sensor failure, communication quality, signs of potential network attacks, and the aging state of the equipment. A quality-aware fusion module is used to perform quality-aware fusion of the multimodal representations of topology awareness according to the confidence weights of each modality, and generate a fused representation. The quality perception fusion module adopts an adaptive weighted fusion mechanism, which makes the modal with higher confidence weight occupy a larger proportion in the fusion, while the weight of the modal with lower confidence weight is reduced or even suppressed to close to zero. The joint reconstruction and scoring module is used to jointly reconstruct the fused representation and calculate the anomaly score based on the reconstruction error and other auxiliary indicators. A physical consistency verification module is used to perform a physical consistency verification on the anomaly score based on the physical mechanism of the power grid. The physical consistency verification module is configured to perform a secondary verification on the anomaly score or anomaly judgment result based on the physical mechanism of the power grid. The alarm and explanation output module is used to generate abnormal alarms and output interpretable analysis results based on the verified results; The alarm and explanation output module is configured to output one or more of the following: the specific device or area where the anomaly occurred, the severity level of the anomaly, the main contributing modality of the anomaly, the anomaly propagation path, the specific physical rules violated, and an explanation text of the anomaly in natural language.
9. The power Internet of Things anomaly monitoring system based on topological constraint modal characterization according to claim 8, characterized in that, The system is communicatively connected to a power Internet of Things (IoT) platform. The power IoT platform is used to provide the system with electrical topology information and multi-source multimodal data, and to receive the abnormal alarms and interpretability analysis results output by the system. It is deployed in any one or more of the following locations: cloud master station, edge computing node, or station-end intelligent terminal, to support centralized, distributed, or edge abnormal monitoring architecture.