CN-121999783-A - Dispatching voice voiceprint recognition and online analysis system and method based on power grid dispatching

CN121999783ACN 121999783 ACN121999783 ACN 121999783ACN-121999783-A

Abstract

The invention relates to the technical field of power grid dispatching voice recognition and data analysis, in particular to a dispatching voice voiceprint recognition and online analysis system and method based on power grid dispatching. The method comprises the steps of receiving and identifying double-track voice real-time information acquired by a voice acquisition module through a voice identification module in the system, receiving an identification result of the voice identification module and incoming call information of an incoming call source positioning module through an information analysis module in the system, carrying out big data analysis on the identification result and the incoming call information, carrying out association application with dispatch maintenance log information, and outputting an association analysis report or a visual result. The invention realizes the efficient identification, data association and safe storage of the power grid dispatching call, and improves the dispatching identity verification accuracy, the work coordination efficiency and the accident analysis convenience.

Inventors

CHENG YAN
ZHANG YUANHAO
Gulinesa Ibrahim
YU HONGWEI
WANG LU
WANG PENG
WANG JINKUN
MA WEIJUN
Feng Tenghui
HU TIANJIE

Assignees

国网新疆电力有限公司哈密供电公司

Dates

Publication Date: 20260508
Application Date: 20260115

Claims (10)

1. The dispatching voice voiceprint recognition and online analysis system based on power grid dispatching is characterized by comprising a voice acquisition module, an incoming call source positioning module, a voice recognition module and an information analysis module; The voice acquisition module is used for acquiring double-track voice real-time information and transmitting the double-track voice real-time information to the voice recognition module; The incoming call source positioning module is used for extracting incoming call information and transmitting the incoming call information to the information analysis module; The voice recognition module is used for receiving the double-track voice real-time information transmitted by the voice acquisition module, recognizing the double-track voice real-time information and transmitting a recognition result to the information analysis module; the information analysis module is used for receiving the recognition result transmitted by the voice recognition module and the incoming call information transmitted by the incoming call source positioning module, analyzing big data of the recognition result and the incoming call information, and carrying out association application with the scheduled maintenance log information.
2. The system for identifying and analyzing the voice and the voiceprint of the dispatching voice based on the power grid dispatching according to claim 1, wherein the incoming call source positioning module comprises an incoming call source positioning capturing sub-module and an incoming call source positioning identifying sub-module; The incoming call source positioning capturing submodule is used for synchronously capturing the display content of the client from the display port of the dispatching telephone management system; The incoming call source positioning and identifying sub-module is used for sensing and displaying the change of the content in the specific interested area through a motion detection algorithm, and the optical character identifying module is used for identifying the change content and outputting number information and line information of the call carried by the dispatching system.
3. The system for identifying and analyzing the voice and the voiceprint of the dispatching voice based on the dispatching of the power grid according to claim 1 or 2, wherein the voice identification module comprises a preprocessing unit, a feature extraction unit, a model training unit and a model optimizing unit; the preprocessing unit is used for denoising the double-track voice real-time information by adopting a noise reduction method based on deep learning; the feature extraction unit is used for extracting the acoustic features of the noise-reduced voice signals to obtain acoustic feature representations containing time step descriptions; The model training unit is used for calculating time step attention energy by using a transcription attention mechanism, converting the time step attention energy into attention weights, forming recognition result probability sequences, revising the sequences by using the technical term acoustic feature weights provided by the information analysis module to obtain a voice recognition result, wherein the weights are obtained by calculating historical scheduling data and are integrated with supervised learning parameters in real time; the model optimizing unit is used for introducing transfer learning, antagonism generation network and self-adaptive voice recognition algorithm, and enhancing model generalization capability and robustness.
4. The system for identifying and analyzing the voice and the voiceprint of the dispatching voice based on the dispatching of the power grid according to claim 1 or 2, wherein the information analysis module is used for classifying and storing the identification result and the incoming call information when carrying out big data analysis, and the classification dimension comprises the conversation time and the dispatching service type.
5. The system for identifying and analyzing the voice voiceprint of the dispatching voice based on the dispatching of the power grid according to claim 1 or 2, wherein when the information analysis module is associated with dispatching maintenance log information, key information related to the log in the identification result is automatically marked, and the key information comprises a device name and a dispatching instruction key word and is used for rapidly associating corresponding dispatching maintenance log entries.
6. The system for identifying and analyzing the voice and the voiceprint of the dispatching voice based on the dispatching of the power grid according to claim 1 or 2, further comprising a data encryption module, wherein the data encryption module is used for encrypting, storing and transmitting the dual-track voice real-time information, the incoming call information, the identification result and the associated analysis data.
7. A dispatching voice voiceprint recognition and online analysis method based on power grid dispatching is characterized by comprising the following steps: Step S1, providing a dispatching voice voiceprint recognition and online analysis system based on power grid dispatching, wherein the dispatching voice voiceprint recognition and online analysis system based on power grid dispatching is the dispatching voice voiceprint recognition and online analysis system based on power grid dispatching of any one of claims 1 to 6; s2, acquiring double-track voice real-time information through a voice acquisition module in the system; s3, extracting incoming call information through an incoming call source positioning module in the system; s4, receiving and identifying the double-track voice real-time information acquired by the voice acquisition module through a voice identification module in the system; And S5, receiving the recognition result of the voice recognition module and the incoming call information of the incoming call source positioning module through an information analysis module in the system, analyzing big data of the recognition result and the incoming call information, performing association application with the scheduled maintenance log information, and outputting an association analysis report or a visual result.
8. The method for identifying and analyzing the voice voiceprint of the dispatching voice based on the dispatching of the power grid as set forth in claim 7, wherein the step S3 specifically includes: s3.1, synchronously capturing the display content of the client from a display port of the dispatch telephone management system by a call source positioning capturing sub-module; And S3.2, the change of the content in the specific region of interest is perceived and displayed by the call source positioning and identifying sub-module through a motion detection algorithm, the change content is identified by the optical character identifying module, the number information and the line information of the call carried by the dispatching system are output, and the call positioning is completed.
9. The method for identifying and analyzing voice voiceprints on schedule based on power grid scheduling according to claim 7 or 8, wherein step S4 specifically comprises: S4.1, modeling typical noise in a modulation environment, and carrying out noise reduction processing on the acquired dual-track voice real-time information by adopting a noise reduction method based on deep learning; S4.2, extracting acoustic features of the noise-reduced voice signals to obtain acoustic feature representations containing time step descriptions; S4.3, calculating time step attention energy by using a transcription attention mechanism, converting the time step attention energy into attention weight, forming recognition result probability sequencing, revising and sequencing by using the technical term acoustic feature weight provided by the information analysis module to obtain a voice recognition result, wherein the weight is obtained by calculating historical scheduling data and is integrated with supervised learning parameters in real time; and S4.4, introducing transfer learning, antagonizing a generation network and a self-adaptive voice recognition algorithm, and enhancing the generalization capability and robustness of the model.
10. The method for identifying and analyzing the voice and the voiceprint of the dispatching voice based on the dispatching of the power grid according to claim 7 or 8, wherein in the step S5, the information analysis module classifies and stores the identification result and the incoming call information when analyzing big data, and the classification dimension comprises the conversation time and the dispatching service type.

Description

Dispatching voice voiceprint recognition and online analysis system and method based on power grid dispatching Technical Field The invention relates to the technical field of power grid dispatching voice recognition and data analysis, in particular to a dispatching voice voiceprint recognition and online analysis system and method based on power grid dispatching. Background With the continuous expansion of the new energy power generation scale, the fluctuation and the intermittence of the new energy power generation provide serious challenges for the power grid dispatching work. The dispatcher needs to adjust the regular power supply output more frequently and rapidly to maintain the power balance of the grid. Under the background, the possibility that a plurality of stations have problems at the same time is obviously increased, the complexity and difficulty of emergency treatment faced by a dispatcher are greatly improved, and the workload of communication coordination, instruction issuing, log recording and the like is also increased. If communication is not smooth in the dispatching process, the accuracy and timeliness of a dispatching decision are easily affected, so that the dispatching object is ensured to support the guard and go on duty, the occurrence of the condition of impossibility and displacement is avoided, and the method becomes a key link for ensuring the safe operation of the power grid. The traditional dispatch office management system and the dispatch telephone system lack effective data connection and are in a state of being self-administration. In daily work, the dispatcher can only check the qualification of the butt joint personnel by means of manual inquiry, password input and the like of the system, so as to determine whether to link with the butt joint personnel to develop power dispatching business. The method not only occupies a large amount of working time and has low verification efficiency, but also has single verification method, has certain potential safety hazard, and is easy to misjudge or cause unqualified personnel to take out the identities of other people for scheduling and contact. The risk of incomplete execution of the dispatching instruction, phase change execution and even misoperation is greatly increased, and great threat is brought to safe and stable operation of the power grid. Along with the continuous increase of power grid dispatching traffic and the continuous improvement of dispatching safety requirements, the traditional identity recognition mode cannot meet the actual demands, and a more advanced, accurate and convenient identity recognition technology is urgently needed to improve the accuracy, reliability and practicability of power dispatching voiceprint recognition identity. Meanwhile, the data to be recorded is in nonlinear growth due to the fact that the scheduling business is increasingly large, and the work bearing capacity of a scheduler is limited, so that the work intensity of the scheduler is remarkably increased, the spirit is in a highly stressed state for a long time, the problems of fatigue and overlarge pressure are easily caused, and further the work efficiency and decision quality can be affected. In the evaluation analysis process after the occurrence of the scheduled event, the related information such as the scheduled log and the scheduled recording of the current day is often required to be called. At present, a dispatch log needs to be retrieved from a dispatch work management system, a dispatch record needs to be extracted intensively by a communication staff, record information related to the log cannot be retrieved directly, and log content corresponding to the record is difficult to obtain quickly. The target records are queried from massive information, so that the workload is large, the time consumption is long, the telephone recording information cannot be efficiently utilized, and the process of accident analysis is seriously influenced. In addition, work coordination among schedulers is severely hampered by the inability of scheduling telephone recording information to synchronize in time with scheduling work management system data. When a plurality of schedulers simultaneously handle different scheduling services, an effective communication mechanism is difficult to establish, and the work connection is easy to cause errors, so that the operation time sequence is influenced. In an emergency, the problem of poor coordination may delay the accident handling time, affect the efficiency of recovering power transmission, and finally cause low execution efficiency of scheduling work. Disclosure of Invention The invention provides a dispatching voice voiceprint recognition and online analysis system and method based on power grid dispatching, which overcome the defects of the prior art and can effectively solve the problems of data rupture, low identity verification efficiency and insufficient voice recognition accuracy of t