CN-122021644-A - Automatic power grid accident reporting method and system based on AI semantic understanding and event chain tracking

CN122021644ACN 122021644 ACN122021644 ACN 122021644ACN-122021644-A

Abstract

The invention provides an automatic power grid accident reporting method and system based on AI semantic understanding and event chain tracking, and relates to the technical field of intelligent power grid monitoring. The method comprises the steps of constructing an event chain model through event time alignment and equipment topological association, deducing causal relation among events to generate an electrical law causal graph, identifying root cause events by combining node ingress, egress and confidence weights, and automatically judging and generating an accident report text and an evidence packet according to accident severity and confidence threshold. The invention can realize intelligent diagnosis and automatic report of the electrical accident, and improves analysis accuracy and response efficiency. The method solves the problem that cross-system time alignment and automatic report of power grid accidents based on causal constraint cannot be realized under the condition of multi-source heterogeneous data in the prior art.

Inventors

JIA RUIHENG
TAN AOSHUANG
CHEN BO
JIN CHAO
ZENG LINLONG
YANG DONG
SHANG YONG
LIU SISI
ZHANG YANYAN
CUI HAO
HU ZHIGANG

Assignees

国网湖北省电力有限公司襄阳供电公司

Dates

Publication Date: 20260512
Application Date: 20260127

Claims (10)

1. An automatic power grid accident reporting method based on AI semantic understanding and event chain tracking is characterized by comprising the following steps: acquiring multi-source data of a power grid, and carrying out semantic analysis on the multi-source data to obtain event types, time clues and equipment entity information; Normalizing and hierarchically mapping the event types, the time clues and the equipment entity information according to the equipment vocabulary and the topology knowledge to obtain an initial event model data set; Performing drift estimation and alignment processing on time information of different sources in the initial event model dataset based on a synchronous anchor point in the multi-source data, and establishing a confidence interval for the aligned time information to obtain a time constraint event set aligned by a cross-system time scale; according to the time constraint event centralization time sequence and the topological association relation between the electrical equipment, carrying out event sequencing and aggregation to obtain an event chain model; carrying out causal inference on the association between event chain models according to time constraint and topology constraint to obtain an electrical law causal graph; Determining root cause analysis result data according to the electrical law causal graph; And automatically triggering and judging the root cause analysis result data according to the severity of the accident and the confidence coefficient threshold to obtain an accident report text and a corresponding evidence packet.
2. The automatic reporting method of power grid accidents based on AI semantic understanding and event chain tracking according to claim 1, wherein the steps of obtaining multi-source data of a power grid and performing semantic analysis on the multi-source data to obtain event types, time cues and equipment entity information comprise: Acquiring multi-source data from operation monitoring, protection wave recording, operation management and field communication channels, and executing format recognition, noise filtering and time tag extraction processing on the multi-source data to obtain a preprocessed multi-source data set; Performing word segmentation, part-of-speech tagging and named entity recognition on text data in the preprocessed multi-source data set, and extracting alarm tags, time stamps and measurement sequence features of signal data in the multi-source data set to obtain unified data representation with semantics and signal features; performing semantic classification on the unified data representation based on a semantic rule base and contextual features to identify semantic elements comprising event type candidates, time cue segments, and device entity designations; Based on the semantic elements, performing co-index normalization and numbering mapping processing on the equipment entity names to obtain a structured event semantic element set; And obtaining event types, time clues and equipment entity information according to the event semantic element set.
3. The automatic reporting method of power grid accidents based on AI semantic understanding and event chain tracking according to claim 1, wherein the normalizing and hierarchical mapping process is performed on the event type, the time clue and the equipment entity information according to the equipment vocabulary and the topology knowledge, so as to obtain an initial event model data set, and the method comprises the following steps: Invoking a preset equipment vocabulary and a power grid topology knowledge base, and carrying out vocabulary matching and node positioning on equipment entities involved in the event semantic element set to obtain an event element set; Unified normalization processing is carried out on different named or numbered forms of the same equipment in the event element set, and a standardized equipment entity mapping table is obtained; Based on a standardized equipment entity mapping table, combining primary and secondary system structures of a power grid, determining a subordinate relation, a connection relation and an action path between equipment to obtain an electrical hierarchical structure model; Binding event types, time clues and equipment entity information according to the associated equipment nodes according to the electrical hierarchical structure model to obtain a structured event record set after hierarchical mapping; carrying out unique index and number management on event relations in the structured event record set to obtain an initial event model data set; the expression of the electrical hierarchy model is: ; Wherein, the A comprehensive representation matrix for the electrical hierarchy model; The total number of the equipment nodes in the power grid; Is a node And node The coefficient of electrical communication relationship between the two, And (3) with Respectively nodes And A device class encoding value of (2); Is a node And node Shortest path distance in topology; Is the average hierarchical scale constant of the network structure; Is a first exponential term describing the electrical impact attenuation relationship between topological layers.
4. The automatic reporting method of power grid accidents based on AI semantic understanding and event chain tracking according to claim 1, wherein the drift estimation and alignment processing are performed on time information of different sources in the initial event model dataset based on a synchronization anchor point in multi-source data, and a confidence interval is established on the aligned time information to obtain a time constraint event set aligned by a cross-system time scale, and the method comprises the following steps: Extracting time attribute information corresponding to each data source from the initial event model data set, and grouping and sorting the time attribute information according to the data source type to obtain a source-separated time attribute set; Analyzing public event identification and trigger signals of the source-separated time attribute set to identify synchronous anchor points which can be aligned among different sources, and obtaining a time anchor point set containing a synchronous anchor point matching relationship; Calculating the time drift amount and drift trend of each data source relative to a reference time source according to the time anchor point set to obtain a cross-source time drift parameter set; correcting time attributes in the initial event model data set by utilizing the cross-source time drift parameter set, and adjusting and aligning all event time by taking a synchronous anchor point as a reference to obtain an event model data set with aligned time marks; establishing a confidence interval for the time attribute of each event in the event model data set to obtain a time calibration event set with confidence information; And carrying out unified formatting and index management on the event time and the confidence coefficient in the time calibration event set to obtain a time constraint event set after cross-system time mark alignment.
5. The automatic reporting method of power grid accidents based on AI semantic understanding and event chain tracking according to claim 1, wherein the expression of the event chain model is: ; Wherein, the An associated intensity matrix of the event chain model; constraining the number of events in the event set for a time; Is event And events Connection relation index on equipment topology; Is event And events Time sequential symbols of (a) Is event And (3) with Time difference between; a time decay constant that is a chain of events; and a second index term is used for describing trigger impact attenuation caused by event increase along with time interval.
6. The automatic reporting method of power grid accidents based on AI semantic understanding and event chain tracking according to claim 1, wherein the causal inference is performed on the association between event chain models according to time constraint and topology constraint to obtain an electrical law causal graph, comprising: Analyzing the time sequence relation of the event nodes in the event chain model and the topology node information of the affiliated equipment to obtain an event chain input set containing the event nodes and constraint information; identifying a front-to-back trigger relationship of the event chain input set based on the time constraint; determining a logic communication relation between adjacent equipment nodes in the event chain input set based on the topology constraint; obtaining an event association pair set according to the front-back trigger relationship and the logic communication relationship; Calculating causal association strength and directivity indexes of each pair of events in the event association pair set to obtain a candidate causal relationship set with causal direction labels; correcting the candidate causal relation set by using a time confidence interval and a topological path constraint to obtain an effective causal relation set; and based on the effective causal relation set, constructing an electrical law causal graph by taking event nodes as vertexes and effective causal relations as directed edges.
7. The automatic reporting method of power grid accidents based on AI semantic understanding and event chain tracking according to claim 6, wherein the calculating causal association strength and directivity index of each pair of events in the event association pair set to obtain a candidate causal relationship set with causal direction labels comprises: Reading the time interval, the event type label and the associated equipment node information of each pair of events according to the event associated pair set to obtain an event associated feature set for causal analysis; Calculating time sequence probability based on event triggering time difference, sampling period and time line sequence, and correcting time occurrence sequence indexes of event association feature sets according to confidence interval weights to obtain sequence measurement results; According to the event association feature set and the electrical hierarchy model, calculating a topological influence index of an upstream event on a downstream event to obtain a topological influence result; Performing normalization synthesis on the sequencing measurement result and the topology influence degree result according to a preset weight model to obtain an event causal relation set; And judging the causal direction according to the time sequence relation and the causal strength threshold, and marking a directional identifier in the event causal relation set to obtain a candidate causal relation set with causal direction marks.
8. The automatic reporting method of power grid accidents based on AI semantic understanding and event chain tracking according to claim 1, wherein determining root cause analysis result data according to the electrical law causal graph comprises: the method comprises the steps of reading event nodes and causal side attribute information in an electrical law causal graph to extract node ingress, egress, event types and time confidence weights, and obtaining a causal node feature set; traversing event links along a causal direction based on the causal node feature set, and identifying terminal event groups which are not triggered by other events on time and topology paths to obtain a candidate root cause event set; Based on the candidate root cause event set, calculating influence degree and overall confidence score of each candidate event according to the position of the event in the electrical topology, the number of related events, the causal transfer length and the time confidence weight, and obtaining root cause event scoring set with confidence scores; The method comprises the steps of screening highest scoring events according to confidence threshold and influence weight based on Yu Genyin event evaluation sets to obtain deterministic root cause event sets; based on the deterministic root cause event set, the root cause analysis result data is obtained by combining the associated event chain, time information and topology node information.
9. The automatic reporting method of power grid accidents based on AI semantic understanding and event chain tracking according to claim 1, wherein the automatic triggering judgment is performed on the root cause analysis result data according to the severity of the accidents and a confidence coefficient threshold, so as to obtain an accident reporting text and a corresponding evidence packet, and the automatic reporting method comprises the following steps: analyzing accident types, severity levels and confidence indexes in root cause analysis result data to obtain an accident analysis input set; Based on the accident analysis input set, comparing according to the accident type and the influence range with a severity threshold value table, determining whether the report level is reached, and obtaining a preliminary accident judgment result; based on the preliminary accident judgment result, comparing the root cause event confidence coefficient with a preset confidence coefficient threshold to obtain an effective report event set; judging whether a triggering condition is met based on the effective reporting event set, if so, calling an accident reporting template generation module and locking a related evidence index path to obtain a triggering state identifier and a corresponding evidence packet index set; Based on the triggering state identification and the evidence package index set, embedding the accident type, severity, time, root cause description and evidence path information into a preset text template to obtain an accident report text and a corresponding evidence package.
10. An automatic power grid accident reporting system based on AI semantic understanding and event chain tracking is characterized by comprising: the multi-source data acquisition module is used for acquiring multi-source data of a power grid and carrying out semantic analysis on the multi-source data to obtain event types, time clues and equipment entity information; the device vocabulary matching and hierarchy mapping module is used for carrying out normalization and hierarchy mapping processing on the event types, the time clues and the device entity information according to the device vocabulary and topology knowledge to obtain an initial event model data set; The time alignment and confidence modeling module is used for carrying out drift estimation and alignment processing on time information of different sources in the initial event model data set based on a synchronous anchor point in the multi-source data, and establishing a confidence interval for the aligned time information to obtain a time constraint event set aligned across system time marks; the event chain construction module is used for carrying out event sequencing and aggregation according to the time constraint event centralization time sequence and the topological association relation between the electrical equipment to obtain an event chain model; The causal inference module is used for performing causal inference on the association between the event chain models according to the time constraint and the topology constraint to obtain an electrical law causal graph; The root cause analysis module is used for determining root cause analysis result data according to the electrical law causal graph; And the automatic report generation module is used for automatically triggering and judging the root cause analysis result data according to the severity of the accident and the confidence coefficient threshold to obtain an accident report text and a corresponding evidence packet.

Description

Automatic power grid accident reporting method and system based on AI semantic understanding and event chain tracking Technical Field The invention relates to the technical field of intelligent power grid monitoring, in particular to a power grid accident automatic reporting method and system based on AI semantic understanding and event chain tracking. Background In the current monitoring and scheduling work of the power system, a plurality of information sources such as a monitoring system, a protection and fault wave recording system, an operation management system, a field communication recording and maintenance planning system and the like work in parallel, so that a large amount of structured and unstructured multi-source data is formed. When an abnormality or an accident occurs in the power grid, an operator on duty needs to respectively call records, compare signals and study and judge reasons from a plurality of systems and manually generate an accident report material. The whole flow depends on manual experience, has low working efficiency, is easily influenced by personal judgment, and causes delay and inconsistency of reporting accident information. In recent years, with the development of artificial intelligence and big data technology, the industry starts to gradually shift from alarm compression and templated report forms based on fixed rules to intelligent analysis directions based on semantic understanding and causal inference. Part of researches try to unify event expression forms of different systems by utilizing a semantic recognition technology, and introduce an event chain analysis model based on a topological logic relationship so as to realize intelligent analysis and report generation of contents such as accident reasons, action correctness, influence ranges and the like. Meanwhile, in order to meet compliance and audit requirements, the automatic power grid reporting system is evolving towards data traceability, evidence verifiability and closed-loop learning. However, the prior art still has significant shortcomings in multi-system time synchronization and causal chain reasoning. The time stamp of each system has drift and delay, the influence of time uncertainty on accident study and judgment cannot be accurately quantified, the event association mainly depends on static rules or empirical templates, physical topological constraint and time logic sequence are difficult to be considered, an automatic reporting function generally lacks a trigger strategy based on severity and confidence degree dynamic judgment, and also lacks bidirectional consistency verification between site report and system judgment, and evidence chains are scattered and can not be effectively audited. Together, these problems lead to insufficient accuracy, timeliness and traceability of the grid accident report. Disclosure of Invention In order to overcome the defects of the prior art, the invention aims to provide the automatic power grid accident reporting method and the system based on AI semantic understanding and event chain tracking, and solves the problems that the cross-system time alignment and the automatic power grid accident reporting based on causal constraint cannot be realized under the condition of multi-source heterogeneous data in the prior art. In order to achieve the above object, the present invention provides the following solutions: an automatic power grid accident reporting method based on AI semantic understanding and event chain tracking comprises the following steps: acquiring multi-source data of a power grid, and carrying out semantic analysis on the multi-source data to obtain event types, time clues and equipment entity information; Normalizing and hierarchically mapping the event types, the time clues and the equipment entity information according to the equipment vocabulary and the topology knowledge to obtain an initial event model data set; Performing drift estimation and alignment processing on time information of different sources in the initial event model dataset based on a synchronous anchor point in the multi-source data, and establishing a confidence interval for the aligned time information to obtain a time constraint event set aligned by a cross-system time scale; according to the time constraint event centralization time sequence and the topological association relation between the electrical equipment, carrying out event sequencing and aggregation to obtain an event chain model; carrying out causal inference on the association between event chain models according to time constraint and topology constraint to obtain an electrical law causal graph; Determining root cause analysis result data according to the electrical law causal graph; And automatically triggering and judging the root cause analysis result data according to the severity of the accident and the confidence coefficient threshold to obtain an accident report text and a corresponding evidence packet. An a