CN-122019715-A - Intelligent power transformation monitoring log filling method based on knowledge enhancement and voice interaction

CN122019715ACN 122019715 ACN122019715 ACN 122019715ACN-122019715-A

Abstract

The invention provides an intelligent report method for a power transformation monitoring log based on knowledge enhancement and voice interaction, and relates to the technical field of power monitoring logs. The method comprises the steps of generating knowledge enhancement representation for monitoring events by constructing a power transformation monitoring field knowledge graph, and combining voice recognition, intention recognition and slot extraction to realize event positioning, automatic log field generation and incremental revision under the voice interaction of a monitoring operator. The system forms monitoring log entries to be confirmed in multiple rounds of voice interaction, files the monitoring log entries after confirmation by an attendant, and performs alignment storage on semantic information of monitoring events, voice interaction text and final log entries. And comprehensively evaluating the incremental contribution degree of the newly added logs based on semantic differences, statistical distribution differences, consistency scores and novelty scores, and adaptively updating the knowledge base in the transformer monitoring field, so that the log filling efficiency and the completeness and timeliness of the knowledge base are improved. The problem that the sustainable self-learning intelligent filling function cannot be realized is solved.

Inventors

JIN CHAO
CUI HAO
SHANG YONG
JIA RUIHENG
ZENG LINLONG
YANG DONG
JIN JING
FANG SANPING
ZHENG XUEYANG
TAN AOSHUANG
ZHANG YANYAN

Assignees

国网湖北省电力有限公司襄阳供电公司

Dates

Publication Date: 20260512
Application Date: 20260123

Claims (10)

1. The intelligent transformer monitoring log reporting method based on knowledge enhancement and voice interaction is characterized by comprising the following steps of: The method comprises the steps of obtaining power transformation monitoring operation data and preprocessing the power transformation monitoring operation data to obtain preprocessed power transformation monitoring operation data, wherein the power transformation monitoring operation data comprises alarm data, telemetry data, operation mode data, maintenance plan data, equipment defect record data, history monitoring logs and voice data corresponding to monitoring telephone records; Aggregating the preprocessed substation monitoring operation data into monitoring events according to a preset time window and equipment identification, and associating the monitoring events with corresponding context information to obtain a monitoring event semantic set; constructing a knowledge base in the field of power transformation monitoring; generating corresponding knowledge enhancement representations for each monitoring event in the monitoring event semantic set according to the power transformation monitoring field knowledge base to obtain a knowledge enhancement monitoring event representation set; Receiving voice input of a monitoring attendant, performing voice recognition to obtain a voice transcription text, and performing intention recognition on the voice transcription text to obtain a log-related intention; determining an update target event semantic context according to the monitoring event semantic set and the monitoring event semantic set based on the log-related intent; generating monitoring log candidate items based on the knowledge enhancement monitoring event representation set, and carrying out field correction on the monitoring log candidate items according to the increment constraint in the semantic context of the update target event to obtain a monitoring log candidate item set; Outputting monitoring log candidate items to a monitoring attendant through voice broadcasting based on the monitoring log candidate item set, and supplementing and revising the monitoring log candidate items based on multiple rounds of voice interaction to obtain monitoring log items to be confirmed; Under the condition that the monitoring operator confirms the monitoring log entry to be confirmed, archiving the monitoring log entry to be confirmed to obtain a final monitoring log entry; And carrying out alignment storage on the semantic information of the monitoring event, the voice interaction text information and the information of the final monitoring log entry corresponding to the final monitoring log entry, and updating the power transformation monitoring domain knowledge base based on an alignment storage result to obtain an updated power transformation monitoring domain knowledge base.
2. The intelligent power transformation monitoring log reporting method based on knowledge enhancement and voice interaction according to claim 1, wherein the steps of obtaining power transformation monitoring operation data and preprocessing the power transformation monitoring operation data to obtain preprocessed power transformation monitoring operation data comprise: carrying out format standardization processing and field completion processing on alarm data, telemetry data, operation mode data, maintenance planning data, equipment defect record data and history monitoring logs in the substation monitoring operation data set so as to enable time fields, equipment identification fields and event type fields of various data to be kept consistent semantically and obtain a structured operation data set; Time alignment is carried out on time fields in the structured running data set, and normalization processing is carried out on time stamps according to preset time granularity, so that a running data set after time alignment is obtained; Performing voice recognition on voice data corresponding to the monitoring telephone record in the power transformation monitoring operation data set to obtain a monitoring telephone voice transfer text, and associating the monitoring telephone voice transfer text with time stamp information and equipment identification information to obtain a monitoring telephone text data set; And merging the time-aligned operation data set with the monitoring telephone text data set, and outputting the operation data set according to a unified data format to obtain the preprocessed substation monitoring operation data.
3. The intelligent transformer monitoring log reporting method based on knowledge enhancement and voice interaction according to claim 1, wherein the aggregating the preprocessed transformer monitoring operation data into monitoring events and associating the monitoring events with corresponding context information according to a preset time window and a device identifier to obtain a monitoring event semantic set comprises: Performing time division on the preprocessed substation monitoring operation data according to a preset time window, and performing preliminary aggregation on alarm data, telemetry data, operation mode data, maintenance plan data, equipment defect record data and monitoring phone text data which are in the same time window and have the same equipment identifier to obtain a candidate event data set; Performing event boundary judgment and type recognition on candidate event data sets under the same time window and the same equipment identifier based on a preset event triggering rule to obtain single monitoring events, and generating unique event identifier, event start-stop time and event type information for each monitoring event to obtain a monitoring event basic set; According to the event identification and the event start-stop time of the monitoring event basic set, operation mode data, maintenance plan data, equipment defect record data and history monitoring logs which are related to the current monitoring event are retrieved from the preprocessed power transformation monitoring operation data, and the retrieved related data are used as context information to be associated with the corresponding monitoring event to obtain a monitoring event expansion set; Based on the monitoring event expansion set, each monitoring event and associated context information are subjected to structural representation, and a monitoring event semantic set is obtained.
4. The intelligent report method for the power transformation monitoring log based on knowledge enhancement and voice interaction according to claim 1, wherein the power transformation monitoring domain knowledge base comprises: Substation equipment topology, monitoring point tables, alarm coding and meaning, standard handling procedures, log template specifications and historical monitoring logs, constraint entities for substation equipment entities, monitoring event types, handling measures and log fields.
5. The intelligent power transformation monitoring log reporting method based on knowledge enhancement and voice interaction according to claim 1, wherein the generating a corresponding knowledge enhancement representation for each monitoring event in the monitoring event semantic set according to the power transformation monitoring domain knowledge base to obtain a knowledge enhancement monitoring event representation set comprises: Carrying out structural modeling based on a knowledge base in the power transformation monitoring field to obtain a knowledge graph representation in the power transformation monitoring field; Based on each monitoring event in the monitoring event semantic set, retrieving knowledge entities and relations corresponding to equipment identification, event types, alarm coding and treatment processes of the monitoring event from the power transformation monitoring field knowledge graph representation to obtain a candidate field knowledge sub-graph set; Screening and aggregating the candidate domain knowledge subgraphs of each monitoring event based on the candidate domain knowledge subgraphs set, and combining equipment information, event type information, treatment procedure information and log template constraint information which are closely related to the semantics of the monitoring event to obtain a domain knowledge context set; Based on the domain knowledge context set, feature fusion is carried out on semantic representations of all monitoring events in the monitoring event semantic set and corresponding domain knowledge contexts, and a knowledge enhancement monitoring event representation set is obtained.
6. The intelligent transformer monitoring log reporting method based on knowledge enhancement and voice interaction of claim 1, wherein the log-related intent comprises: The log-filling initiating intention, the information supplementing intention, and the log-confirming and modifying intention.
7. The intelligent transformer monitoring log reporting method based on knowledge enhancement and voice interaction according to claim 1, wherein the determining the update target event semantic context according to the monitoring event semantic set and the monitoring event semantic set based on the log-related intent comprises: Carrying out intention recognition and slot extraction on the voice transcription text to obtain log related intention and slot information corresponding to the log related intention, and obtaining a log interaction request carrying the intention and the slot information, wherein the slot information comprises event positioning slots, field updating slots and confirmation state slots; Based on the event positioning slot information, searching monitoring events matched with equipment identifications, time ranges and event types related in the voice transcription text in the monitoring event semantic set to obtain a candidate target monitoring event set; When the log related intention is an information supplementing intention or a log confirming and modifying intention, selecting a monitoring event which is most matched with the current log interaction context from the candidate target monitoring event set as a target monitoring event according to event indicating information and time sequence information in a voice transfer text to obtain the target monitoring event; Acquiring knowledge enhancement monitoring event representations corresponding to target monitoring events according to a target monitoring event and knowledge enhancement monitoring event representation set, and combining historical voice interaction information of current conversation turns to acquire a target event semantic context; And updating the event attribute, the log field constraint and the semantic label in the semantic context of the target event based on the field update slot and the increment constraint information in the confirmation state slot to obtain the semantic context of the updated target event.
8. The intelligent transformer monitoring log reporting method based on knowledge enhancement and voice interaction according to claim 1, wherein the generating monitoring log candidate entries based on the knowledge enhancement monitoring event representation set, and performing field correction on the monitoring log candidate entries according to the incremental constraint in the update target event semantic context, to obtain a monitoring log candidate entry set, comprises: selecting a knowledge enhancement monitoring event representation corresponding to the semantic context of the updated target event from the semantic context of the updated target event and the knowledge enhancement monitoring event representation set to obtain a target knowledge enhancement monitoring event representation; based on target knowledge enhancement monitoring event representation, automatically generating log title, occurrence time, ending time, equipment name, equipment position, event type, reason analysis and disposal measure log fields by combining with transformation monitoring log template specification to obtain initial monitoring log candidate entries; According to the increment constraint information in the semantic context of the update target event, carrying out item-by-item correction and filling on the corresponding fields in the initial monitoring log candidate entries to obtain the monitoring log candidate entries after field correction; Determining monitoring log candidate entries according to the field-corrected monitoring log candidate entries, the confirmation state slot information and the conversation history constraint information in the semantic context of the update target event; based on the monitoring log candidate entries, outputting a plurality of log versions which can be selected or confirmed by a monitoring attendant in the current turn according to a preset ordering rule, and obtaining a monitoring log candidate entry set.
9. The intelligent transformer monitoring log reporting method based on knowledge enhancement and voice interaction according to claim 1, wherein the outputting the monitoring log candidate items to a monitoring attendant through voice broadcasting based on the monitoring log candidate item set, and supplementing and revising the monitoring log candidate items based on multiple rounds of voice interaction to obtain monitoring log items to be confirmed comprises the following steps: Selecting one or more monitoring log candidate items corresponding to the semantic context of the update target event based on the monitoring log candidate item set, and abstracting key field contents of the selected monitoring log candidate items to obtain abstract information of the monitoring log candidate items; Based on the summary information of the monitoring log candidate items, performing voice broadcasting to a monitoring attendant in a voice synthesis mode, and synchronously displaying the corresponding content of the monitoring log candidate items on a visual interface to obtain a display result of the monitoring log candidate items; Receiving voice feedback given by a monitoring attendant after listening to voice broadcasting and checking the display result of the candidate entries of the monitoring log, performing voice recognition and intention recognition on the voice feedback, extracting incremental information related to log supplement and field revision, and obtaining a voice incremental instruction; Based on the voice increment instruction, performing supplementing, modifying or deleting operation on corresponding fields in the display result of the monitoring log candidate items by using the semantic context of the update target event to obtain the monitoring log candidate items updated in the current interaction round; Judging whether the voice feedback of the monitoring attendant contains a continuous modification intention for the monitoring log candidate item updated by the current interaction round, if so, taking the monitoring log candidate item updated by the current interaction round as a new basic item, returning to continue voice broadcasting and voice feedback processing, and if not, taking the monitoring log candidate item updated by the current interaction round as the final output of the current interaction, thereby obtaining the monitoring log item to be confirmed.
10. The intelligent report method for the power transformation monitoring log based on knowledge enhancement and voice interaction according to claim 1, wherein the updated power transformation monitoring field knowledge base has the following expression: ; Wherein, the Scoring the comprehensive incremental contribution of the newly added knowledge item; vector coding for monitoring event representation for knowledge enhancement constructed based on the new event; Vector coding for the existing knowledge enhancement monitoring event representation in the current knowledge base most similar to the new event; a feature transformation function for performing nonlinear semantic mapping on the monitor event vector representation; squaring Euclidean distance of the new event and the nearest event in the knowledge base in a semantic space; Counting the discrete probability distribution obtained by the log field corresponding to the new monitoring event; The method comprises the steps of obtaining a reference probability distribution which is obtained by statistics of a set of similar or similar events which are retrieved from a current knowledge base and are similar to the new event; Kullback-Leibler divergence for the new event probability distribution relative to the knowledge base reference distribution; Scoring for integrated consistency; the new event is scored for effective novelty brought on the knowledge graph structure or semantic combination, For coefficients that balance weights between semantic and statistical difference terms and logical consistency and novelty terms, To control the balance coefficient of the relative importance of the euclidean distance term and the KL divergence term, Is a trade-off coefficient between the overall consistency score and the effective novelty score.

Description

Intelligent power transformation monitoring log filling method based on knowledge enhancement and voice interaction Technical Field The invention relates to the technical field of power monitoring logs, in particular to an intelligent transformer monitoring log reporting method based on knowledge enhancement and voice interaction. Background With the continuous improvement of the scale and automation level of the power grid, various substations are widely provided with a dispatching automation system, a centralized control system and an on-line monitoring system, which are used for monitoring a large amount of operation information such as voltage, current, switch positions, protection actions, load flow, on-line monitoring and the like in real time. At present, the recording mode of the monitoring log mainly depends on a manual operator on duty, and information such as event occurrence time, equipment name, alarm content, treatment process and result is manually recorded in a log system or an operation and maintenance management system according to monitoring picture alarm, scheduling and operation and maintenance telephone communication conditions and on-site treatment feedback. Although some systems provide a semi-automatic reporting function based on a fixed template, the alarm codes, the device names and the time fields are simply brought into the system, the main content of the log still depends on manual editing, supplementing and modifying, the workload is large, the consistency is poor, and the problems of information omission, irregular expression and the like are easy to occur. In the field of operation control of an electric power system, with the development of technologies such as artificial intelligence, big data, knowledge maps and the like, power transformation monitoring service is evolving towards the direction of intelligence, scene and man-machine cooperation. On one hand, the knowledge graph, knowledge base question and answer and other technologies are adopted to carry out structural modeling on information such as equipment topology, procedure specifications, fault cases and the like, and the knowledge graph, the knowledge base question and answer technology and the method have been applied to equipment management, fault diagnosis and operation and maintenance decision support, and provide knowledge-enhanced auxiliary analysis capability for operators. On the other hand, the technologies of voice recognition, voice synthesis, dialogue management and the like are gradually mature, and the intelligent voice assistant and the voice interaction interface are applied to trial points in the scenes of scheduling communication, information inquiry, simple command control and the like. In log management, it has also emerged to generate exploration of partial log fields from alert codes, device types and preset templates using a rules engine or simple machine learning model, and attempt to perform text analysis on the history log for report statistics and typical case mining. The overall trend is to evolve the monitoring log from a passive logging tool to support the important data assets for fault analysis, risk early warning and operational assessment. The prior art still has obvious defects in the aspect of intelligent filling of the transformer monitoring log. The existing semiautomatic filling mode is mainly based on the original field of the alarm for field copying or fixed sentence pattern splicing, lacks semantic fusion of multi-source data, and cannot automatically combine information such as an operation mode, an overhaul plan, a defect record, a history similar event and the like to generate an event description and a treatment process record with complete context logic. The monitoring telephone recording and the multi-round telephone communication contain a great deal of key information about event reasons, on-site investigation results and processing measures, but the key information is usually only archived as a voice file in the prior scheme, is not subjected to fine alignment and joint modeling with log text, and also lacks a mechanism for continuously optimizing domain knowledge and generating strategies by utilizing voice and log alignment data. In addition, the existing voice interaction application stays at the level of simple voice inquiry and command execution, and a multi-round voice interaction and knowledge enhancement reasoning integrated technical path facing the monitoring log scene is not formed, so that the problems of complex operation, easy neglected recording, non-uniform semantics, difficult self-iteration lifting of system capability and the like still exist in the monitoring log filling process. Disclosure of Invention In order to overcome the defects of the prior art, the invention aims to provide an intelligent filling method for a power transformation monitoring log based on knowledge enhancement and voice interaction, which solves the problems that the p