An intelligent detection method and system for abnormal data of an intelligent converged terminal

Abstract

The application provides an abnormal data intelligent detection method and system of an intelligent fusion terminal, and relates to the technical field of data analysis.The method comprises the following steps: acquiring multi-terminal data and power grid states in real time; setting a target terminal and determining an associated terminal set; performing inversion detection on the target terminal based on the associated set to generate a first abnormal index; constructing a double shaft and performing linkage detection to generate a second abnormal index; correcting and mutually exclusive mapping based on power grid state data to obtain a target abnormal characteristic chain; decomposing and positioning the abnormal contribution degree to generate a target terminal abnormal atlas, so as to solve the problem that the existing intelligent fusion terminal abnormal detection is easily disturbed by the power grid operation state, it is difficult to accurately distinguish the terminal itself abnormality and the power grid fluctuation, and the abnormal misjudgment and positioning difficulty are caused, and the effect of improving the detection reliability and operation efficiency is achieved through the combination of upstream and downstream measurement inversion and double shaft linkage detection, the correction of power grid state interference, the realization of abnormal accurate identification, contribution degree decomposition positioning and abnormal atlas generation.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, specifically to an intelligent method and system for detecting abnormal data in a smart fusion terminal. Background Technology

[0002] With the deepening of the construction of new power systems, distributed terminals are widely deployed in all aspects of power grid generation, grid, load, and storage, resulting in an exponential increase in the number of terminals and a dramatic expansion in the amount of data collected and interacted with. These terminals undertake critical tasks such as electrical measurement, condition monitoring, and communication relay, and their data quality directly affects power grid operation and control, fault diagnosis, and asset management decisions. However, intelligent converged terminals face complex anomaly causes in actual operation. On the one hand, the terminals themselves may experience hardware aging, firmware defects, clock drift, and communication module failures, leading to measurement deviations, data packet loss, or out-of-order messages. On the other hand, the power grid operating status and communication network environment can also introduce external disturbances, causing the data uploaded by the terminals to exhibit abnormal characteristics. Existing anomaly detection methods mostly use single-dimensional threshold judgment or statistical verification, which is difficult to effectively distinguish between terminal anomalies and conducted interference from the power grid or communication side, resulting in a large number of false alarms and missed alarms. At the same time, once anomalies are identified, the lack of a systematic tracing mechanism makes it difficult for maintenance personnel to quickly locate the root cause, resulting in low efficiency in fault handling. Summary of the Invention

[0003] This application provides an intelligent detection method and system for abnormal data of intelligent converged terminals, aiming to solve the technical problem that the existing anomaly detection of intelligent converged terminals is easily affected by the power grid operation status, making it difficult to accurately distinguish between the terminal's own anomalies and power grid fluctuations, resulting in misjudgment and difficulty in locating anomalies.

[0004] The first aspect disclosed in this application provides an intelligent detection method for abnormal data of a smart fusion terminal. The method includes: acquiring a multi-terminal dataset of a power grid distributed terminal in real time and synchronously loading power grid status data; traversing the power grid distributed terminals to set a target terminal and determining the upstream and downstream associated terminal sets of the target terminal; based on the multi-terminal dataset, performing measurement inversion anomaly detection on the target terminal according to the upstream and downstream associated terminal sets to generate a first anomaly index; constructing the acquisition time axis and reception time axis of the target terminal, and performing dual-axis linkage anomaly detection on the target terminal according to the acquisition time axis and the reception time axis to generate a second anomaly index; performing power grid anomaly interference correction and mutual exclusion discrimination mapping on the first anomaly index and the second anomaly index according to the power grid status data to obtain a target anomaly characteristic chain; and performing anomaly contribution decomposition and localization on the target anomaly characteristic chain to generate a target terminal anomaly map.

[0005] The second aspect of this application discloses an intelligent detection system for abnormal data of a smart fusion terminal. This system is used in conjunction with the aforementioned intelligent detection method for abnormal data of a smart fusion terminal. The intelligent detection system for abnormal data of a smart fusion terminal includes: a data acquisition module for real-time acquisition of multi-terminal datasets from distributed power grid terminals and synchronous loading of power grid status data; a terminal determination module for traversing the distributed power grid terminals to set a target terminal and determining the upstream and downstream associated terminal sets of the target terminal; a first anomaly detection module for performing measurement inversion anomaly detection on the target terminal based on the multi-terminal dataset and the upstream and downstream associated terminal sets, generating a first anomaly index; a second anomaly detection module for constructing the acquisition time axis and reception time axis of the target terminal, and performing dual-axis linkage anomaly detection on the target terminal based on the acquisition time axis and the reception time axis, generating a second anomaly index; a correction mapping module for performing power grid anomaly interference correction and mutual exclusion discrimination mapping on the first and second anomaly indices based on the power grid status data, obtaining a target anomaly characteristic chain; and a decomposition and localization module for decomposing and localizing the target anomaly characteristic chain based on its anomaly contribution, generating a target terminal anomaly map.

[0006] One or more technical solutions provided in this application have at least the following beneficial effects:

[0007] First, data from multiple distributed power grid terminals is collected in real time, and the current power grid operating status is simultaneously acquired. Then, target terminals are selected one by one for testing, and their upstream and downstream associated terminals are analyzed. Next, the reasonable measurement range of the target terminal is deduced using data from the associated terminals to determine if any measurement anomalies exist. Simultaneously, anomalies such as communication delays, out-of-order transmissions, and retransmissions are identified by combining the data acquisition time of the target terminal with the platform reception time. Finally, the detection results are corrected based on the power grid status, distinguishing between terminal anomalies and power grid anomalies. The contribution of each anomaly factor is further analyzed to form an anomaly localization map of the target terminal, thereby achieving accurate detection and cause analysis of abnormal data from intelligent fusion terminals.

[0008] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0009] Figure 1 This is a schematic flowchart of an intelligent data detection method for an intelligent fusion terminal provided in an embodiment of this application.

[0010] Figure 2This is a schematic diagram of the structure of an intelligent data detection system for an intelligent fusion terminal provided in an embodiment of this application.

[0011] Explanation of reference numerals in the attached figures: Data acquisition module 11, Terminal determination module 12, First anomaly detection module 13, Second anomaly detection module 14, Correction mapping module 15, Decomposition and positioning module 16. Detailed Implementation

[0012] This application provides a method and system for intelligent detection of abnormal data in smart converged terminals, which solves the technical problem that existing intelligent converged terminal anomaly detection is easily affected by the power grid operation status, making it difficult to accurately distinguish between the terminal's own anomalies and power grid fluctuations, resulting in misjudgment and difficulty in locating anomalies.

[0013] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0014] Example 1, as Figure 1 As shown in the figure, this application embodiment provides an intelligent detection method for abnormal data in a smart fusion terminal, the method comprising: Real-time acquisition of multi-terminal datasets from distributed power grid terminals, and synchronous loading of power grid status data.

[0015] In one embodiment, the power grid distributed terminal consists of multiple intelligent fusion terminals, which are deployed at different distributed nodes of the power grid, including but not limited to transformer substations, feeder nodes, branch line nodes, distributed power source access nodes, and user-side metering nodes. The system establishes data communication connections with each intelligent fusion terminal according to a preset acquisition cycle or a real-time triggering mechanism, continuously acquiring data uploaded by each intelligent fusion terminal, such as voltage, current, active power, reactive power, power factor, electrical energy, frequency, switch status, alarm information, equipment operating status, data acquisition time, data reception time, and communication message status. This data is then uniformly organized according to terminal number, node location, data type, and timestamp to form a multi-terminal dataset. Simultaneously with acquiring the multi-terminal dataset, the system synchronously loads current power grid status data, including power grid topology connections, line operating status, switch on / off status, load change information, power flow distribution information, fault alarm information, communication network status, and dispatch control information. The system aligns multi-terminal datasets with power grid status data according to a time window, and performs missing data completion, anomaly marking, and time synchronization processing on data from different sources and sampling frequencies. This ensures that data from all smart fusion terminals can be analyzed under the same time reference. Missing data completion can be achieved through interpolation of adjacent time points or compensation using historical data from the same period. Anomaly marking can be performed through threshold exceedance detection or historical statistical deviation analysis. Time synchronization processing uses a unified clock source to calibrate the acquisition and reception times of data uploaded by each smart fusion terminal and aligns them according to a unified time window. Through these processes, basic data reflecting the real-time operation of multiple smart fusion terminals and the corresponding power grid operating environment can be obtained, providing data support for subsequent target terminal selection, upstream and downstream correlation analysis, and anomaly detection.

[0016] The target terminal is set by traversing the distributed terminals of the power grid, and the upstream and downstream associated terminal sets of the target terminal are determined.

[0017] In one embodiment, after acquiring multi-terminal datasets and grid status data, the system establishes a terminal list based on the multiple connected smart converged terminals, including terminal number, installation location, affiliated transformer area or feeder, access line, and topology node number. The system sequentially reads the smart converged terminals in the terminal list according to a preset terminal polling strategy and sets the currently read smart converged terminal as the target terminal, enabling each smart converged terminal to participate in anomaly detection as a target object. After determining the target terminal, the system calls upon the grid topology connection relationship and terminal installation location relationship to identify adjacent and associated nodes of the target terminal in the electrical connection path. Specifically, the system traces back to the power supply side along the feeder, branch line, or transformer area power supply path where the target terminal is located to identify smart converged terminals with upstream power supply, upstream aggregation, or upstream measurement constraint relationships with the target terminal, as upstream associated terminals. Simultaneously, it extends along the power supply path where the target terminal is located to the load side, branch side, or user side to identify smart converged terminals supplied by or allocated by the target terminal's downstream power supply or with downstream measurement constraint relationships, as downstream associated terminals. The system merges the identified upstream and downstream associated terminals and records the topological distance, connection direction, line impedance, switch status, branch to which each associated terminal belongs, and measurement type matching relationship between the target terminal and the target terminal, forming a set of upstream and downstream associated terminals of the target terminal. This set is used to perform measurement inversion and anomaly comparison of the target terminal using data from adjacent and related intelligent fusion terminals.

[0018] Furthermore, traversing the power grid distributed terminal to set the target terminal includes: Obtain the terminal polling strategy; perform a traversal polling of the distributed terminals of the power grid according to the terminal polling strategy to obtain the target terminal.

[0019] Preferably, the system pre-generates a terminal polling strategy based on the importance level information and historical anomaly frequency records of the distributed terminals in the power grid. It then reads information from all connected smart converged terminals, using this terminal list as the list of terminals to be polled. The terminals are sorted according to their power supply direction in the power grid topology, starting from the power source side and proceeding sequentially to the load side. A polling pointer is then established for the list of terminals to be polled, initially pointing to the first smart converged terminal in the list. When the start time of the polling cycle is reached, the system reads the smart converged terminal currently pointed to by the polling pointer and determines whether it is online and whether it has valid measurement data in the most recent acquisition cycle. If the smart converged terminal is online and has valid measurement data, it is set as the target terminal, and its terminal number, topology node number, and installation location are extracted as basic information for subsequent anomaly detection. After completing the detection task for the target terminal, the system moves the polling pointer to the next smart converged terminal in the list of terminals to be polled. If the intelligent fusion terminal currently pointed to by the polling pointer is offline, or if there is no valid measurement data in the most recent acquisition cycle, it will not be set as the target terminal. Instead, the skip flag, skip time, and skip reason of the intelligent fusion terminal will be recorded, and the polling pointer will be moved to the next intelligent fusion terminal. The system traverses the list of terminals to be polled in the above manner. When the polling pointer moves to the end of the list, it returns to the beginning of the list and enters the next round of polling. By combining this fixed cycle, topological order, and polling pointer, the system can continuously and stably determine the target terminal to be detected from multiple intelligent fusion terminals, thereby improving the targeting and efficiency of anomaly detection.

[0020] Furthermore, the terminal polling strategy includes: Based on the importance level information of the power grid distributed terminals, an initial polling decision is generated; based on the historical anomaly frequency records of the power grid distributed terminals, the initial polling decision is dynamically adjusted to generate the terminal polling strategy.

[0021] Optionally, the system first reads the importance level information of each smart converged terminal. This importance level information is pre-determined according to the importance of the power grid node where the terminal is located and divided into three levels: Level 1, Level 2, and Level 3. Level 1 terminals correspond to main feeders, distribution area master meters, or key branch nodes; Level 2 terminals correspond to ordinary branch nodes; and Level 3 terminals correspond to end-user side nodes. The system sets different initial polling periods for Level 1, Level 2, and Level 3 terminals. For example, the initial polling period for Level 1 terminals is 1 minute, for Level 2 terminals it is 3 minutes, and for Level 3 terminals it is 5 minutes. The system generates initial polling decisions in the order of priority for Level 1 terminals, followed by Level 2 terminals, and lastly for Level 3 terminals. Subsequently, the system reads the historical anomaly frequency records of each smart converged terminal within the most recent preset number of days and compares the historical anomaly count with a preset frequency threshold. When the number of historical anomalies in a smart fusion terminal within the most recent preset number of days exceeds the anomaly threshold, the system shortens the terminal's polling cycle by one level. Specifically, third-level terminals are polled according to the second-level terminal cycle, second-level terminals are polled according to the first-level terminal cycle, and the first-level terminals remain unchanged. When the number of historical anomalies does not exceed the anomaly threshold, the initial polling cycle remains unchanged. The system then rearranges the polling order of each smart fusion terminal according to the adjusted polling cycle, prioritizing terminals with shorter polling cycles in the detection queue, ultimately generating a terminal polling strategy. This strategy includes the terminal number, polling cycle, polling priority, next polling time, and polling order for each smart fusion terminal. This strategy is used for subsequent traversal polling of distributed power grid terminals, thereby improving the rationality of anomaly detection resource allocation and the timeliness of detection.

[0022] Based on the multi-terminal dataset, the target terminal is subjected to measurement inversion anomaly detection according to the upstream and downstream related terminal set, and a first anomaly index is generated.

[0023] In one embodiment, based on a multi-terminal dataset composed of multiple intelligent fusion terminals, real-time measurement data of the target terminal and its upstream and downstream associated terminals are extracted. Combined with the electrical connection relationships between the terminals, a measurement relationship between the target terminal and its upstream and downstream associated terminals is established. Subsequently, using the measurement results of the upstream and downstream associated terminals, the theoretical measurement range of the target terminal under the current power grid operating state is derived in reverse, and residual analysis is performed between this theoretical result and the actual data collected by the target terminal. When there is a significant deviation between the actual measurement data and the theoretical measurement range, it is determined that the target terminal may have a measurement anomaly, and a corresponding first anomaly index is generated based on the anomaly state, thereby achieving the identification of anomalies in the target terminal's measurement data.

[0024] Furthermore, based on the multi-terminal dataset, the target terminal is subjected to measurement inversion anomaly detection according to the upstream and downstream related terminal set to generate a first anomaly indicator, including: Based on the multi-terminal dataset, target measurement data of the target terminal and upstream and downstream measurement data of the upstream and downstream associated terminal set are extracted; a measurement conservation constraint relationship is constructed between the target terminal and the upstream and downstream associated terminal set; based on the upstream and downstream measurement data, the theoretical measurement interval of the target terminal is inverted according to the measurement conservation constraint relationship; residual detection is performed between the theoretical measurement interval and the target measurement data to generate the first abnormal index.

[0025] Preferably, the system uses the current detection time as a benchmark, selects a preset time length, such as 5 minutes, as the detection time window, and extracts the target measurement data of the target terminal within this time window from the multi-terminal dataset. This includes target active power, target reactive power, target current, target voltage, and target electrical energy. Simultaneously, it extracts the upstream and downstream measurement data of each associated terminal in the upstream and downstream associated terminal sets within the same time window, aligns them according to the acquisition timestamp, and forms a group of associated measurement data at the same moment. Subsequently, the system establishes measurement conservation constraints based on the connection relationship of the target terminal in the power grid topology. Specifically, the upstream associated terminal of the target terminal is taken as the input side, and the downstream associated terminal of the target terminal is taken as the output side. The correspondence between the input and output measurement values ​​is determined according to the power flow direction, and a preset line loss coefficient is introduced. For example, for active power measurement, the constraint is that the theoretical active power of the target terminal should be within the allowable fluctuation range of the upstream active power minus the downstream active power and line loss; for electrical energy measurement, the constraint is that the theoretical electrical energy of the target terminal should satisfy the balance relationship between the upstream input electrical energy, the downstream output electrical energy, and the line loss. Then, the upstream and downstream measurement data are substituted into the measurement conservation constraint relationship to calculate the theoretical measurement value of the target terminal at each acquisition time, and a theoretical measurement interval is generated according to the preset allowable error. This preset allowable error can be set as a preset proportion of the theoretical measurement value, such as 5%. That is, the theoretical measurement interval is the range of theoretical measurement value fluctuating by 5% above and below, thus obtaining the theoretical measurement interval of the target terminal at each acquisition time, including the theoretical active power interval, theoretical reactive power interval, theoretical current interval, theoretical voltage interval, and theoretical electrical energy interval, etc. Finally, the target measurement data actually transmitted by the target terminal is compared item by item with the corresponding theoretical measurement interval. If the target measurement data is within the theoretical measurement interval, the residual for that measurement item is recorded as 0; if the target measurement data is higher than the upper limit of the theoretical measurement interval, the difference between the target measurement data and the upper limit of the interval is taken as the residual; if the target measurement data is lower than the lower limit of the theoretical measurement interval, the difference between the lower limit of the interval and the target measurement data is taken as the residual. The system statistically analyzes the residual magnitude, residual duration, and residual duration for each measurement item within the detection time window, and generates a first anomaly index based on the residual results. This first anomaly index includes the abnormal measurement item, maximum residual value, average residual value, anomaly duration, and anomaly level, used to characterize whether the target terminal has measurement inversion anomalies. Through the above process, it is possible to accurately identify whether the measurement data of the target terminal deviates from the expected range, promptly detect potential measurement anomalies, thereby improving the accuracy and reliability of data acquisition and providing effective support for subsequent anomaly diagnosis and power grid operation and maintenance.

[0026] Furthermore, based on the upstream and downstream measurement data, the theoretical measurement interval of the target terminal is inverted according to the measurement conservation constraint relationship, including: Mapping characteristics are mined between the historical dataset of the power grid and the historical measurement dataset of the target terminal to construct a power grid state measurement mapping model; the power grid state data is input into the power grid state measurement mapping model to obtain the measurement mapping interval; the reliability of the theoretical measurement interval is evaluated based on the measurement mapping interval to obtain the theoretical measurement reliability; if the theoretical measurement reliability is less than the predetermined reliability, the generation process of the theoretical measurement interval is backtracked to establish a theoretical measurement backtracking block; the theoretical measurement interval is corrected and updated based on the theoretical measurement backtracking block.

[0027] Optionally, the system first acquires historical power grid datasets and historical measurement datasets for the target terminal. The historical power grid dataset contains long-term power grid operating status information, such as voltage, current, active power, reactive power, and power grid topology. The historical measurement dataset for the target terminal includes electrical measurement data for the target terminal within a historical time period. The system performs feature mining on both datasets to identify the correlation between power grid operation and target terminal measurements. This feature mining can be based on Gaussian process regression, time-series correlation regression, machine learning regression, etc. Taking Gaussian process regression as an example, the historical power grid dataset and historical measurement dataset are aligned by timestamps to form a training sample set. A radial basis function kernel is selected as the covariance function, and initial hyperparameters are set, such as signal variance set to 1 and feature length scale set to 0.5. The training sample set is then input into the Gaussian process regression framework. The kernel function hyperparameters are iteratively optimized by maximizing the logarithmic marginal likelihood function using a conjugate gradient optimization algorithm with a maximum of 200 iterations and a convergence threshold of 1×10⁻⁶. -5 The process continues until convergence, resulting in a trained posterior Gaussian process model, which serves as the power grid state measurement mapping model. This model implicitly learns the nonlinear mapping law and noise distribution characteristics between power grid state characteristics and terminal measurement values. In the application phase, real-time acquired power grid state data is input into the constructed model. The model outputs the predicted mean and standard deviation of theoretical measurements such as voltage, current, active power, and reactive power. Based on the posterior distribution properties of the Gaussian process, a confidence interval is constructed by adding or subtracting twice the predicted standard deviation from the predicted mean. This interval represents the measurement mapping interval under the current power grid state, reflecting the expected range of target terminal measurement data under a given power grid state.

[0028] After obtaining the measurement mapping interval, the reliability of the theoretical measurement data of the target terminal is evaluated based on this interval. Specifically, the length of the intersection interval between the theoretical measurement interval and the measurement mapping interval is calculated. This intersection interval length is divided by the length of the theoretical measurement interval to obtain the interval overlap. Then, the absolute distance between the geometric center point of the theoretical measurement interval and the geometric center point of the measurement mapping interval is divided by the length of the measurement mapping interval to obtain the deviation penalty factor. By multiplying the overlap degree by the deviation penalty factor, the theoretical measurement reliability is obtained. If the theoretical measurement reliability is lower than the preset reliability threshold, such as 0.9, it indicates that the measurement results of the target terminal may be affected by fluctuations in the power grid operating status, equipment failure, or data acquisition problems, and there is a large degree of uncertainty. At this time, the backtracking process is initiated to backtrack and analyze the generation process of the theoretical measurement interval, identify factors that may affect the reliability, such as abnormal power grid status and measurement data deviation, and classify these factors into theoretical measurement backtracking blocks. The system will readjust the parameters for generating the theoretical measurement interval based on the causes of anomalies in the backtracking block, correcting the original mapping model to ensure it reflects the actual state of the current power grid. Finally, it will recalculate the theoretical measurement interval for the target terminal based on the corrected and updated measurement interval from the theoretical measurement backtracking block, using this as the basis for subsequent anomaly detection. Through this process, the system can further improve the reliability of measurement data and eliminate the interference that power grid state fluctuations may cause to the accuracy of measurement data, thereby enhancing the accuracy and effectiveness of anomaly detection.

[0029] Construct the acquisition time axis and the reception time axis of the target terminal, and perform dual-axis linkage anomaly detection on the target terminal based on the acquisition time axis and the reception time axis to generate a second anomaly index.

[0030] In one embodiment, for the target terminal, an acquisition timeline and a reception timeline are constructed. The acquisition timeline records the time points when the target terminal acquires data, while the reception timeline records the time points when this data is transmitted from the terminal to the receiving platform. These two timelines allow for tracking the data acquisition and transmission process and identifying any inconsistencies in time. Subsequently, based on the differences between the acquisition and reception timelines, dual-axis linkage anomaly detection is performed to detect significant time offsets. Further analysis is conducted to determine if there are out-of-order data phenomena and burst characteristics on both the acquisition and reception timelines, thereby obtaining a time offset feature set, a message out-of-order feature set, and a retransmission correlation feature set. Then, the system combines these feature sets to perform communication anomaly detection on the target terminal, generating a second anomaly index to quantify communication anomalies and further aid in locating potential communication problems or terminal data transmission failures during power grid operation.

[0031] Furthermore, based on the acquisition time axis and the reception time axis, dual-axis linkage anomaly detection is performed on the target terminal to generate a second anomaly indicator, including: The acquisition time axis and the receiving time axis are subjected to time offset detection for the same data point to obtain a time offset feature set; the target terminal is subjected to message out-of-order identification based on the acquisition time axis and the receiving time axis to obtain a message out-of-order feature set; a retransmission associated feature set is identified based on the continuous acquisition time characteristic of the acquisition time axis and the burst reception time characteristic of the receiving time axis; the target terminal is subjected to communication anomaly detection based on the time offset feature set, the message out-of-order feature set, and the retransmission associated feature set to obtain the second anomaly index.

[0032] Preferably, after obtaining the acquisition timeline and the reception timeline, the system first performs time offset detection on the acquisition timeline and the reception timeline. This involves comparing the time difference of the same data points on the acquisition timeline and the reception timeline of the target terminal to check if the time difference exceeds a preset threshold. If the time difference is too large, it may be due to network latency, data transmission interruption, or equipment failure. The system marks data points with time differences exceeding the preset range as time offsets and records the time differences of all offset data points as a time offset feature set, reflecting synchronization issues between data acquisition and uploading. Subsequently, based on the order of data points on the acquisition timeline and the reception timeline, the system performs message out-of-order identification. Because data transmission may be affected by network fluctuations, data packets may arrive out of order at the receiving end. By checking the order of data points on the reception timeline, if the reception time of a data packet is earlier than the reception time of its predecessor, the system marks this phenomenon as out of order. This out-of-order phenomenon often occurs during data transmission. The system records all out-of-order phenomena and the reception time of their related data packets, forming a message out-of-order feature set. This feature set helps identify and locate sequence disorder problems in network transmission. Subsequently, based on the continuity of the acquisition time on the acquisition time axis and the burst characteristics of the reception time on the reception time axis, the system identifies whether data retransmission occurs. When some data packets fail to be uploaded at the scheduled time due to network or communication problems, multiple data packets may be received in a short period of time on the reception time axis, forming a burst in reception time. The system analyzes whether there are long time intervals in the acquisition time axis and compares them with the burst reception times of data points on the reception time axis. If a burst in reception time is found and the interval is short, the system marks these data packets as retransmissions and records them as a retransmission-related feature set. Then, the system extracts the average and maximum values ​​of the offset from the time offset feature set and calculates the time offset impact by dividing them; it extracts the average time interval of out-of-order phenomena, the frequency of out-of-order phenomena, and the cumulative time of out-of-order phenomena from the message out-of-order feature set, and obtains the message out-of-order anomaly impact by normalizing and weighting the three; it extracts the average retransmission time, the frequency of retransmission phenomena, and the cumulative retransmission time of retransmission phenomena from the retransmission-related feature set, and obtains the retransmission anomaly impact by normalizing and weighting the three. Finally, the impact of time offset, message out-of-order anomaly, and retransmission anomaly are summarized to generate a second anomaly index. This second anomaly index can accurately identify communication anomalies in the target terminal during data transmission and provide support for further fault location and power grid monitoring.

[0033] Based on the power grid status data, the first anomaly index and the second anomaly index are subjected to power grid anomaly interference correction and mutual exclusion discrimination mapping to obtain the target anomaly characteristic chain.

[0034] In one embodiment, anomaly detection is first performed using acquired power grid status data to identify existing abnormal events, such as abnormal power grid operation or communication. Then, based on the identified abnormal events, the first and second anomaly indicators are corrected through impact identification, resulting in corrected first and second anomaly indicators. Next, a pre-established mutual exclusion mechanism is used to analyze the corrected first and second anomaly indicators to determine if they are mutually exclusive. For example, if both the first and second anomaly indicators exist within the same time period, but the power grid status data indicates a serious fault in the power grid during that period, the anomaly at the target terminal may be caused by a fault in the power grid itself, rather than a problem with the target terminal's data acquisition or transmission, thus identifying potential causes of the anomaly. Through these correction and discrimination processes, a target anomaly characteristic chain is ultimately formed. This chain includes all influencing factors related to the target terminal's anomaly and links them to the power grid status, providing clear traceability of the anomaly source and making subsequent fault diagnosis and handling more accurate and effective.

[0035] Furthermore, based on the power grid state data, power grid anomaly interference correction and mutual exclusion discrimination mapping are performed on the first anomaly index and the second anomaly index to obtain the target anomaly characteristic chain, including: Anomaly detection is performed on the power grid status data to obtain power grid operation anomaly events and power grid communication anomaly events; the first anomaly index is corrected based on the power grid operation anomaly events to obtain a corrected first anomaly index; the second anomaly index is corrected based on the power grid communication anomaly events to obtain a corrected second anomaly index; based on the corrected first anomaly index and the corrected second anomaly index, candidate anomaly causes are determined according to a mutual exclusion discrimination mechanism; the corrected first anomaly index, the corrected second anomaly index, and the candidate anomaly causes are chained together to obtain the target anomaly characteristic chain.

[0036] Preferably, the system first reads power grid status data based on the current detection time window and classifies the data for detection. Voltage exceeding limits, current exceeding limits, power surges, load spikes, switch changes, and line fault alarms are categorized into power grid operation status detection items. Master station communication link interruptions, communication channel congestion, concentrator forwarding delays, and batch message retransmissions are categorized into power grid communication status detection items. The system compares each of these detection items with preset status thresholds. When voltage, current, power, or switch status exceeds the corresponding threshold within the detection time window, a power grid operation anomaly event is generated, and the event occurrence time, affected lines, affected nodes, and anomaly type are recorded. When communication delay, link interruption duration, or the number of retransmitted messages exceeds the corresponding threshold, a power grid communication anomaly event is generated, and the event occurrence time, affected communication links, affected terminal range, and anomaly type are recorded. After obtaining a power grid operation anomaly event, the system matches the occurrence time, affected nodes, and target terminal topology location of the event to determine whether the first anomaly indicator is within the influence range of the power grid operation anomaly event. If the abnormal time of the first abnormal indicator coincides with the occurrence time of a power grid operation anomaly event, and the target terminal is located within the affected line or node corresponding to the power grid operation anomaly event, the system considers that there is an interference component in the first abnormal indicator caused by power grid operation fluctuations. The system deducts or downgrades the residual, abnormal duration, and abnormal level in the first abnormal indicator according to the impact intensity of the power grid operation anomaly event to obtain a corrected first abnormal indicator; if the first abnormal indicator does not match the power grid operation anomaly event in terms of both time and topology, the first abnormal indicator remains unchanged.

[0037] Upon receiving a power grid communication anomaly event, the system matches the occurrence time of the anomaly event, the affected communication link, and the communication access path of the target terminal to determine whether the second anomaly indicator falls within the impact range of the power grid communication anomaly event. If the time offset, out-of-order message, or retransmission characteristics in the second anomaly indicator occur within the time window corresponding to the power grid communication anomaly event, and the target terminal uploads data through the abnormal communication link, the system considers that the second anomaly indicator contains an interference component caused by the anomaly of the common communication link. The system corrects the second anomaly indicator based on the delay duration and retransmission count of the power grid communication anomaly event to obtain a corrected second anomaly indicator; if the second anomaly indicator does not match the power grid communication anomaly event, the second anomaly indicator remains unchanged. Subsequently, the system executes a mutual exclusion discrimination mechanism based on the corrected first and second abnormal indicators. If the corrected first abnormal indicator meets the abnormality discrimination criteria, but the corrected second abnormal indicator does not, the target terminal measurement abnormality is identified as a candidate cause of the abnormality. If the corrected first abnormal indicator does not meet the abnormality discrimination criteria, but the corrected second abnormal indicator does, the target terminal communication abnormality is identified as a candidate cause of the abnormality. If both the corrected first and second abnormal indicators meet the abnormality discrimination criteria, the cause corresponding to the higher-priority abnormal indicator is identified as the primary candidate cause of the abnormality, and the cause corresponding to the other abnormal indicator is identified as the secondary candidate cause of the abnormality. Finally, the corrected first and second abnormal indicators and the candidate causes of the abnormality are linked in a chain according to time sequence and causal relationship to generate a target abnormality characteristic chain. This target abnormality characteristic chain includes the time of abnormality occurrence, type of abnormal indicator, indicator values ​​before and after correction, associated power grid events, candidate causes of the abnormality, master-slave relationship, and scope of abnormality impact. This process can eliminate interference from abnormal power grid operation and communication link anomalies on terminal detection results, avoid misjudging abnormalities on the power grid side as abnormalities on the terminal side, thereby improving the accuracy of anomaly cause identification and the reliability of subsequent location results.

[0038] Furthermore, the mutual exclusion mechanism includes: If the corrected first anomaly indicator meets the anomaly discrimination condition, but the corrected second anomaly indicator does not meet the anomaly discrimination condition, the anomaly candidate cause is generated based on the corrected first anomaly indicator; if the corrected first anomaly indicator does not meet the anomaly discrimination condition, but the corrected second anomaly indicator meets the anomaly discrimination condition, the anomaly candidate cause is generated based on the corrected second anomaly indicator; if both the corrected first anomaly indicator and the corrected second anomaly indicator meet the anomaly discrimination condition, priority analysis is performed based on the corrected first anomaly indicator and the corrected second anomaly indicator to generate the anomaly candidate cause.

[0039] Optionally, the system pre-sets anomaly detection conditions. These conditions include: the measurement residual value of the first anomaly indicator being greater than a preset residual threshold; the residual duration being greater than a preset duration; and the time offset impact of the second anomaly indicator being greater than a preset offset threshold; the message out-of-order anomaly impact being greater than a preset out-of-order threshold; and the retransmission anomaly impact being greater than a preset retransmission threshold. After completing the power grid anomaly interference correction, the system reads the corrected first and second anomaly indicators respectively and makes judgments according to the above anomaly detection conditions.

[0040] When the corrected first anomaly indicator meets the anomaly discrimination criteria, and the corrected second anomaly indicator does not meet the anomaly discrimination criteria, it indicates that the data transmission process of the target terminal does not exhibit obvious communication anomalies, but there is still a continuous deviation between the actual measurement value of the target terminal and the theoretical measurement range. The system attributes this type of anomaly to an anomaly on the target terminal's measurement side, and generates candidate causes of anomalies based on the abnormal measurement items, residuals, residual durations, and anomaly levels in the corrected first anomaly indicator. These candidate causes include at least one definite cause among measurement acquisition deviation, sensor anomaly, metering module anomaly, and local acquisition distortion of the terminal, and are marked as triggered by the first anomaly indicator. When the corrected first anomaly indicator does not meet the anomaly discrimination criteria, and the corrected second anomaly indicator does meet the anomaly discrimination criteria, it indicates that the measurement data of the target terminal is within a reasonable range in terms of electrical relationships, but there is an abnormal offset, out-of-order message transmission, or concentrated retransmission phenomenon between its acquisition time axis and reception time axis. The system attributes this type of anomaly to an anomaly on the target terminal's communication side and generates candidate anomaly causes based on the corrected second anomaly indicator. These candidate causes include at least one definite cause among terminal communication delay, out-of-order message transmission, abnormal data retransmission, and unstable communication link, and are marked as triggered by the second anomaly indicator. When both the corrected first and second anomaly indicators meet the anomaly discrimination criteria, the system initiates priority analysis. Specifically, it compares the anomaly start times of the two indicators. If the anomaly start time of the second anomaly indicator is earlier than that of the first anomaly indicator, and the time interval between them is less than a preset correlation time threshold, then the communication anomaly is determined to have a higher priority than the measurement anomaly, generating a candidate anomaly cause with communication anomaly as the primary cause and measurement anomaly as a secondary manifestation. If the anomaly start time of the first anomaly indicator is earlier than that of the second anomaly indicator, and the anomaly level of the first anomaly indicator is higher than that of the second anomaly indicator, then the measurement anomaly is determined to have a higher priority than the communication anomaly, generating a candidate anomaly cause with measurement anomaly as the primary cause and communication anomaly as a secondary manifestation. If the start times of the two are the same, the system compares the corrected anomaly levels, identifying the cause corresponding to the indicator with the higher anomaly level as the primary anomaly candidate cause, and the cause corresponding to the indicator with the lower anomaly level as the secondary anomaly candidate cause. Through this mutual exclusion discrimination process, the system can clearly distinguish the primary and secondary relationships when measurement anomalies and communication anomalies occur simultaneously, reducing confusion of anomaly causes and improving the accuracy of anomaly candidate cause generation.

[0041] The abnormal contribution of the target abnormal characteristic chain is decomposed and located to generate an abnormal map of the target terminal.

[0042] In one embodiment, after obtaining the target anomaly characteristic chain, the system analyzes this chain, breaks down the various anomaly contributing objects involved, and then calculates the contribution of each anomaly contributing object to the target terminal's anomaly result, determining which factors are the main causes of the anomaly and which factors are merely auxiliary influences or accompanying phenomena. Based on this, information such as the anomaly cause, associated terminals, power grid events, and anomaly indicators is structured and organized to form a target terminal anomaly map. This target terminal anomaly map can intuitively display the source, propagation relationship, and root cause location results of the target terminal anomaly, providing a basis for subsequent maintenance personnel to quickly determine the fault location, determine the handling priority, and formulate maintenance measures.

[0043] Furthermore, the abnormal contribution decomposition and localization of the target abnormal characteristic chain are performed to generate a target terminal abnormality map, including: Based on the target anomaly characteristic chain, multiple anomaly contributing objects are identified; the contribution degree of each anomaly contributing object is analyzed on the target anomaly characteristic chain to obtain the contribution degree of multiple objects; the multiple anomaly contributing objects are divided according to the contribution degree of the multiple objects to obtain the primary cause object and the secondary cause object; the root cause of the target anomaly characteristic chain is traced according to the primary cause object and the secondary cause object to obtain the target terminal anomaly map.

[0044] Preferably, after obtaining the target anomaly characteristic chain, the system first parses the anomaly occurrence time, candidate causes, corrected first anomaly index, corrected second anomaly index, associated power grid events, and upstream and downstream relationships recorded in the target anomaly characteristic chain. Then, it breaks down the chain into multiple anomaly contributing objects according to the anomaly source. These anomaly contributing objects include the target terminal measurement module, the target terminal communication module, upstream associated terminals, downstream associated terminals, power grid operation anomaly events, and power grid communication anomaly events. The system establishes an object identifier for each anomaly contributing object and writes the first anomaly index, second anomaly index, topological correlation distance, and event time overlap related to that object into the corresponding object feature record. Subsequently, the contribution degree of each anomaly contributing object is analyzed. For the target terminal measurement module, the system calculates its contribution based on the residual, residual duration, and anomaly level in the corrected first anomaly index. For the target terminal communication module, the system calculates its contribution based on the time offset impact, message out-of-order anomaly impact, and retransmission anomaly impact in the corrected second anomaly index. For upstream and downstream associated terminals, the system calculates their contribution based on the topological distance between them and the target terminal, the degree of measurement deviation propagation, and data time consistency. For power grid operation anomalies and power grid communication anomalies, the system calculates their contribution based on the degree of matching between the event's impact range and the node where the target terminal is located, and the degree of overlap between the event's occurrence time and the target anomaly's occurrence time. The system normalizes the contribution of each anomaly contributing object, making the sum of the contribution of multiple objects equal to 1, thus obtaining the contribution of multiple objects. Next, based on the contribution levels of multiple objects, the abnormal contributing objects are ranked in order of importance. The abnormal contributing object with the highest contribution level that is greater than a preset primary cause threshold is identified as the primary cause object. This preset primary cause threshold can be set to 0.5. Abnormal contributing objects with a contribution level less than or equal to the preset primary cause threshold but greater than a preset secondary cause threshold are identified as secondary cause objects. This preset secondary cause threshold can be set to 0.2. Abnormal contributing objects with a contribution level less than or equal to the preset secondary cause threshold are identified as associated reference objects. If no abnormal contributing object with a contribution level greater than 0.5 exists, the system selects the abnormal contributing object with the highest contribution level as the primary cause object and the abnormal contributing object with the second highest contribution level as the secondary cause object.

[0045] Then, based on the primary and secondary causes of the anomaly, the system traces the root cause of the target anomaly characteristic chain. In this process, the system uses the primary cause as the starting point and traces backward along the time sequence, topological connections, and indicator triggering relationships within the target anomaly characteristic chain to determine the propagation path of the anomaly from its occurrence to its manifestation. Simultaneously, secondary causes are designated as auxiliary influencing nodes, marking their impact on the amplification of the anomaly's severity and the extension of its duration. Based on the tracing results, the system generates a target terminal anomaly map. This map includes target terminal nodes, primary cause nodes, secondary cause nodes, associated terminal nodes, anomaly indicator nodes, and power grid event nodes. Connection edges representing causal, temporal, and topological relationships are established between these nodes, and information such as contribution, anomaly level, anomaly occurrence time, duration, and anomaly type is labeled. Through this process, the system can clearly identify the main sources and auxiliary influencing factors of the target terminal anomaly, forming a visualized and traceable anomaly location result, improving the accuracy of anomaly root cause analysis and the efficiency of operation and maintenance.

[0046] Example 2, based on the same inventive concept as the intelligent data detection method for an intelligent fusion terminal in the foregoing examples, such as... Figure 2 As shown in the figure, this application embodiment provides an intelligent detection system for abnormal data of a smart fusion terminal. The intelligent detection system for abnormal data of a smart fusion terminal includes: Data acquisition module 11: acquires multi-terminal datasets from distributed power grid terminals in real time and synchronously loads power grid status data; Terminal determination module 12: traverses the distributed power grid terminals to set target terminals and determines the upstream and downstream associated terminal sets of the target terminal; First anomaly detection module 13: based on the multi-terminal datasets, performs measurement inversion anomaly detection on the target terminal according to the upstream and downstream associated terminal sets, and generates a first anomaly index; Second anomaly detection module 14: constructs the acquisition time axis and reception time axis of the target terminal, and performs dual-axis linkage anomaly detection on the target terminal according to the acquisition time axis and the reception time axis, and generates a second anomaly index; Correction mapping module 15: performs power grid anomaly interference correction and mutual exclusion discrimination mapping on the first anomaly index and the second anomaly index according to the power grid status data, and obtains the target anomaly characteristic chain; Decomposition and positioning module 16: performs anomaly contribution decomposition and positioning on the target anomaly characteristic chain, and generates a target terminal anomaly map.

[0047] Furthermore, the terminal determination module 12 is used to perform the following operation steps: Obtain the terminal polling strategy; perform a traversal polling of the distributed terminals of the power grid according to the terminal polling strategy to obtain the target terminal.

[0048] Furthermore, the terminal determination module 12 is used to perform the following operation steps: Based on the importance level information of the power grid distributed terminals, an initial polling decision is generated; based on the historical anomaly frequency records of the power grid distributed terminals, the initial polling decision is dynamically adjusted to generate the terminal polling strategy.

[0049] Furthermore, the first anomaly detection module 13 is used to perform the following operation steps: Based on the multi-terminal dataset, target measurement data of the target terminal and upstream and downstream measurement data of the upstream and downstream associated terminal set are extracted; a measurement conservation constraint relationship is constructed between the target terminal and the upstream and downstream associated terminal set; based on the upstream and downstream measurement data, the theoretical measurement interval of the target terminal is inverted according to the measurement conservation constraint relationship; residual detection is performed between the theoretical measurement interval and the target measurement data to generate the first abnormal index.

[0050] Furthermore, the first anomaly detection module 13 is used to perform the following operation steps: Mapping characteristics are mined between the historical dataset of the power grid and the historical measurement dataset of the target terminal to construct a power grid state measurement mapping model; the power grid state data is input into the power grid state measurement mapping model to obtain the measurement mapping interval; the reliability of the theoretical measurement interval is evaluated based on the measurement mapping interval to obtain the theoretical measurement reliability; if the theoretical measurement reliability is less than the predetermined reliability, the generation process of the theoretical measurement interval is backtracked to establish a theoretical measurement backtracking block; the theoretical measurement interval is corrected and updated based on the theoretical measurement backtracking block.

[0051] Furthermore, the second anomaly detection module 14 is used to perform the following operation steps: The acquisition time axis and the receiving time axis are subjected to time offset detection for the same data point to obtain a time offset feature set; the target terminal is subjected to message out-of-order identification based on the acquisition time axis and the receiving time axis to obtain a message out-of-order feature set; a retransmission associated feature set is identified based on the continuous acquisition time characteristic of the acquisition time axis and the burst reception time characteristic of the receiving time axis; the target terminal is subjected to communication anomaly detection based on the time offset feature set, the message out-of-order feature set, and the retransmission associated feature set to obtain the second anomaly index.

[0052] Furthermore, the correction mapping module 15 is used to perform the following operation steps: Anomaly detection is performed on the power grid status data to obtain power grid operation anomaly events and power grid communication anomaly events; the first anomaly index is corrected based on the power grid operation anomaly events to obtain a corrected first anomaly index; the second anomaly index is corrected based on the power grid communication anomaly events to obtain a corrected second anomaly index; based on the corrected first anomaly index and the corrected second anomaly index, candidate anomaly causes are determined according to a mutual exclusion discrimination mechanism; the corrected first anomaly index, the corrected second anomaly index, and the candidate anomaly causes are chained together to obtain the target anomaly characteristic chain.

[0053] Furthermore, the correction mapping module 15 is used to perform the following operation steps: If the corrected first anomaly indicator meets the anomaly discrimination condition, but the corrected second anomaly indicator does not meet the anomaly discrimination condition, the anomaly candidate cause is generated based on the corrected first anomaly indicator; if the corrected first anomaly indicator does not meet the anomaly discrimination condition, but the corrected second anomaly indicator meets the anomaly discrimination condition, the anomaly candidate cause is generated based on the corrected second anomaly indicator; if both the corrected first anomaly indicator and the corrected second anomaly indicator meet the anomaly discrimination condition, priority analysis is performed based on the corrected first anomaly indicator and the corrected second anomaly indicator to generate the anomaly candidate cause.

[0054] Furthermore, the decomposition and positioning module 16 is used to perform the following operation steps: Based on the target anomaly characteristic chain, multiple anomaly contributing objects are identified; the contribution degree of each anomaly contributing object is analyzed on the target anomaly characteristic chain to obtain the contribution degree of multiple objects; the multiple anomaly contributing objects are divided according to the contribution degree of the multiple objects to obtain the primary cause object and the secondary cause object; the root cause of the target anomaly characteristic chain is traced according to the primary cause object and the secondary cause object to obtain the target terminal anomaly map.

[0055] Through the foregoing detailed description of an intelligent data detection method for an intelligent fusion terminal, those skilled in the art can clearly understand the intelligent data detection system for an intelligent fusion terminal in this embodiment. Since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and relevant parts can be referred to the method section.

[0056] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for intelligent detection of abnormal data in a smart fusion terminal, characterized in that, The method includes: Real-time acquisition of multi-terminal datasets from distributed power grid terminals, and synchronous loading of power grid status data; Traverse the distributed terminals of the power grid to set target terminals, and determine the upstream and downstream associated terminal sets of the target terminals; Based on the multi-terminal dataset, the target terminal is subjected to measurement inversion anomaly detection according to the upstream and downstream related terminal set, and a first anomaly index is generated. Construct the acquisition time axis and the reception time axis of the target terminal, and perform dual-axis linkage anomaly detection on the target terminal based on the acquisition time axis and the reception time axis to generate a second anomaly index; Based on the power grid status data, power grid anomaly interference correction and mutual exclusion discrimination mapping are performed on the first anomaly index and the second anomaly index to obtain the target anomaly characteristic chain; The abnormal contribution of the target abnormal characteristic chain is decomposed and located to generate an abnormal map of the target terminal.

2. The intelligent data detection method for intelligent fusion terminals as described in claim 1, characterized in that, Based on the multi-terminal dataset, anomaly detection is performed on the target terminal using measurement inversion based on the upstream and downstream associated terminal set, generating a first anomaly indicator, including: Based on the multi-terminal dataset, extract the target measurement data of the target terminal and the upstream and downstream measurement data of the upstream and downstream associated terminal set; Construct a measurement conservation constraint relationship between the target terminal and the set of upstream and downstream associated terminals; Based on the upstream and downstream measurement data, the theoretical measurement range of the target terminal is inverted according to the measurement conservation constraint relationship; The theoretical measurement interval and the target measurement data are subjected to residual detection to generate the first anomaly index.

3. The intelligent data detection method for intelligent fusion terminals as described in claim 2, characterized in that, Based on the upstream and downstream measurement data, the theoretical measurement interval of the target terminal is inverted according to the measurement conservation constraint relationship, including: A mapping feature mining process is performed on the historical dataset of the power grid and the historical measurement dataset of the target terminal to construct a power grid state measurement mapping model. Input the power grid state data into the power grid state measurement mapping model to obtain the measurement mapping interval; The reliability of the theoretical measurement interval is evaluated based on the measurement mapping interval to obtain the theoretical measurement reliability. If the reliability of the theoretical measurement is less than the predetermined reliability, the generation process of the theoretical measurement interval is backtracked to establish a theoretical measurement backtracking block; The theoretical measurement interval is corrected and updated based on the theoretical measurement backtracking block.

4. The intelligent data detection method for intelligent fusion terminals as described in claim 1, characterized in that, Based on the acquisition time axis and the reception time axis, a dual-axis linkage anomaly detection is performed on the target terminal to generate a second anomaly indicator, including: Perform time offset detection on the acquisition time axis and the receiving time axis for the same data point to obtain a time offset feature set; Based on the acquisition time axis and the reception time axis, the target terminal is subjected to message out-of-order identification to obtain a message out-of-order feature set. Based on the continuous characteristics of the acquisition time axis and the burst characteristics of the reception time axis, the retransmission associated feature set is identified; Based on the time offset feature set, the message out-of-order feature set, and the retransmission association feature set, the target terminal is subjected to communication anomaly detection to obtain the second anomaly index.

5. The intelligent data detection method for intelligent fusion terminals as described in claim 1, characterized in that, Based on the power grid state data, power grid anomaly interference correction and mutual exclusion discrimination mapping are performed on the first anomaly index and the second anomaly index to obtain the target anomaly characteristic chain, including: Anomaly detection is performed on the power grid status data to obtain abnormal power grid operation events and abnormal power grid communication events; Based on the abnormal power grid operation event, the first abnormal indicator is identified and corrected to obtain the corrected first abnormal indicator. Based on the power grid communication anomaly event, the second anomaly index is identified and corrected to obtain the corrected second anomaly index; Based on the corrected first anomaly index and the corrected second anomaly index, candidate causes of anomalies are determined according to a mutual exclusion discrimination mechanism. The corrected first anomaly index, the corrected second anomaly index, and the candidate anomaly causes are chained together to obtain the target anomaly characteristic chain.

6. The intelligent data detection method for intelligent fusion terminals as described in claim 5, characterized in that, The mutual exclusion mechanism includes: If the corrected first abnormal indicator meets the abnormality discrimination condition, and the corrected second abnormal indicator does not meet the abnormality discrimination condition, the abnormality candidate cause is generated based on the corrected first abnormal indicator. If the corrected first abnormal indicator does not meet the abnormality discrimination condition, and the corrected second abnormal indicator meets the abnormality discrimination condition, the abnormality candidate cause is generated based on the corrected second abnormal indicator. If both the corrected first anomaly index and the corrected second anomaly index meet the anomaly discrimination condition, priority analysis is performed based on the corrected first anomaly index and the corrected second anomaly index to generate the anomaly candidate cause.

7. The intelligent data detection method for intelligent fusion terminals as described in claim 1, characterized in that, The target anomaly characteristic chain is decomposed and located based on its anomaly contribution, and an anomaly map of the target terminal is generated, including: Based on the target anomaly characteristic chain, multiple anomaly contributing objects are identified; The contribution degree of the target abnormal characteristic chain is parsed for each abnormal contribution object to obtain the contribution degree of multiple objects; Based on the contribution of the multiple objects, the multiple abnormal contribution objects are divided into the main cause object and the secondary cause object of the abnormality; Based on the primary cause object and the secondary cause object of the anomaly, the root cause of the target anomaly characteristic chain is traced to obtain the target terminal anomaly map.

8. The intelligent data detection method for intelligent fusion terminals as described in claim 1, characterized in that, Traversing the power grid distributed terminal to set the target terminal includes: Obtain the terminal polling strategy; The target terminal is obtained by traversing and polling the distributed terminals of the power grid according to the terminal polling strategy.

9. The intelligent data detection method for intelligent fusion terminals as described in claim 8, characterized in that, Obtain the terminal polling strategy, including: Based on the importance level information of the power grid distributed terminals, an initial polling decision is generated; The initial polling decision is dynamically adjusted based on the historical anomaly frequency records of the power grid distributed terminals to generate the terminal polling strategy.

10. An intelligent data anomaly detection system for a smart fusion terminal, characterized in that, The system for implementing the intelligent data detection method for the intelligent fusion terminal according to any one of claims 1-9, the system comprising: Data acquisition module: acquires multi-terminal datasets from distributed power grid terminals in real time and loads power grid status data synchronously; Terminal determination module: Iterates through the distributed terminals of the power grid to set the target terminal, and determines the upstream and downstream associated terminal sets of the target terminal; First anomaly detection module: Based on the multi-terminal dataset, the target terminal is subjected to measurement inversion anomaly detection according to the upstream and downstream related terminal set, and a first anomaly index is generated; The second anomaly detection module constructs the acquisition time axis and the reception time axis of the target terminal, and performs dual-axis linkage anomaly detection on the target terminal based on the acquisition time axis and the reception time axis to generate a second anomaly index. Correction mapping module: Based on the power grid state data, the first abnormal index and the second abnormal index are corrected for power grid abnormality interference and mutually exclusive discrimination mapping are performed to obtain the target abnormal characteristic chain; Decomposition and localization module: Decomposes and locates the anomaly contribution of the target anomaly characteristic chain to generate an anomaly map of the target terminal.

Images