Intelligent voice recognition and conversion method and system for power dispatching

Abstract

The invention provides an intelligent voice recognition conversion method and system for power dispatching, and the method comprises the steps: collecting voice data in a power dispatching process inreal time, and storing the voice data through an audio file; constructing a speech-text matching model library, and training an RNNT speech recognition model based on the model library; converting theaudio file into a text file by using the trained RNNT speech recognition model; optimizing the text file by using a distributed word vector tool, and filtering repeated texts; extracting key information related to report content from the optimized text file by utilizing the trained BP neural network to form structured data conforming to the dispatching log specification of the power grid dispatching management system; the problems that the dispatching telephone information can only be manually recorded by a dispatcher, the efficiency is low, and repeated confirmation is needed for many timesdue to the call quality problem are effectively solved, the working efficiency is improved, and the dispatcher can handle the abnormality and fault of the power grid equipment timely and accurately.

Description

technical field [0001] The disclosure belongs to the technical field of voice recognition, and in particular relates to an intelligent voice recognition conversion method and system for electric power dispatching. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art. [0003] As the nerve center of large power grid operation management and control, the power dispatching system relies on the rapid development of modern network information technology and tends to be more intelligent and digital; due to the increasing power consumption, the number of substations and various user stations (including electric railway Traction stations, important industrial users, etc.) have increased sharply; at the same time, the dispatching agencies at the upper and lower levels, and the dispatching calls between the dispatching agencies and the operation and maintenance p...

AI Technical Summary

Problems solved by technology

[0006] (1) In the case of relatively tight time for dispatching and receiving calls, the recording of information and repeated confirmation will increase the duration of calls and affect the rapid and accurate delivery of dispatching information

[0007] (2) After receiving dispatch calls, dispatchers need to manually record and analyze and extract the information, and cannot interact with the power grid dispatch management system (OMS system) for data interaction, and the work efficiency is relatively low

The mainstream representative of traditional continuous speech recognition technology is GMM-HMM, but its recognition rate is far from meeting the needs of commercial applications; practical results show that the power dispatching speech recognition performance using DNN-HMM is significantly better than the GMM-HMM recognition framework, but due to Due to the technical limitations of the HMM framework itself, there is still room for improvement in training time and accuracy

Images