CN113314106A - Electric power information query and regulation function calling method based on voice and intention recognition - Google Patents
Electric power information query and regulation function calling method based on voice and intention recognition Download PDFInfo
- Publication number
- CN113314106A CN113314106A CN202110546414.4A CN202110546414A CN113314106A CN 113314106 A CN113314106 A CN 113314106A CN 202110546414 A CN202110546414 A CN 202110546414A CN 113314106 A CN113314106 A CN 113314106A
- Authority
- CN
- China
- Prior art keywords
- voice
- regulation
- scheduling
- model
- corpus
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000006870 function Effects 0.000 claims abstract description 29
- 238000004364 calculation method Methods 0.000 claims abstract description 14
- 230000015572 biosynthetic process Effects 0.000 claims abstract description 11
- 238000003786 synthesis reaction Methods 0.000 claims abstract description 11
- 238000013135 deep learning Methods 0.000 claims abstract description 10
- 239000013598 vector Substances 0.000 claims abstract description 8
- 230000004044 response Effects 0.000 claims abstract description 7
- 238000004422 calculation algorithm Methods 0.000 claims description 15
- 230000001105 regulatory effect Effects 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 9
- 238000004458 analytical method Methods 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 8
- 238000012423 maintenance Methods 0.000 claims description 6
- 238000012790 confirmation Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000010223 real-time analysis Methods 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 abstract description 18
- 238000005206 flow analysis Methods 0.000 abstract 1
- 238000013473 artificial intelligence Methods 0.000 description 7
- 230000009286 beneficial effect Effects 0.000 description 6
- 230000001276 controlling effect Effects 0.000 description 6
- 230000003252 repetitive effect Effects 0.000 description 6
- 239000002904 solvent Substances 0.000 description 6
- 239000002184 metal Substances 0.000 description 5
- 238000011161 development Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 210000004556 brain Anatomy 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 125000004122 cyclic group Chemical group 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 239000002775 capsule Substances 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000003012 network analysis Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/06—Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice
- G10L15/063—Training
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/01—Assessment or evaluation of speech recognition systems
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/02—Feature extraction for speech recognition; Selection of recognition unit
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/08—Speech classification or search
- G10L15/10—Speech classification or search using distance or distortion measures between unknown speech and reference templates
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/08—Speech classification or search
- G10L15/18—Speech classification or search using natural language modelling
- G10L15/183—Speech classification or search using natural language modelling using context dependencies, e.g. language models
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/26—Speech to text systems
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
- G10L2015/223—Execution procedure of a spoken command
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Telephonic Communication Services (AREA)
Abstract
The method comprises the steps of calling a power information query and regulation function based on voice and intention recognition, and outputting a scheduling telephone text by combining an acoustic model and a language model based on deep learning according to a scheduling voice corpus; carrying out scheduling intention recognition in a field recognition model and an intention recognition model based on deep learning according to the regulation instruction word vector; outputting a regulation instruction matching result by the slot position identification model by using a rule method; according to the regulation and control instruction matching result, power grid load flow analysis, short-circuit current calculation and switching-on and switching-off operations are combined to perform power information inquiry, power grid automatic calculation and power automatic scheduling; and inputting the scheduling telephone text and the regulation and control instruction matching result into a speech synthesis engine, converting the scheduling telephone text and the regulation and control instruction matching result into scheduling telephone speech, and performing automatic response of a scheduling telephone. The method realizes the function call of the power information inquiry and the regulation and control system from three aspects of a voice recognition technology in the power regulation and control field, an intention recognition technology in the power regulation and control field and a multi-turn dialogue technology of the regulation and control system.
Description
Technical Field
The invention relates to the technical field of power grid regulation and control data analysis, in particular to a power information query and regulation and control function calling method based on voice and intention recognition.
Background
At present, as a core driving force of a new industrial revolution, artificial intelligence has remarkable effect in the aspects of enabling and promoting the traditional industry and hastening the emergence of new industries, and is expected to cause great revolution of economic structures. China highly pays attention to the development of the artificial intelligence industry, and with the construction and the improvement of new technologies such as a regulation cloud platform and the like, the data volume and the computing capacity of a power system are greatly improved recently, so that favorable conditions are created for developing the technical research of artificial intelligence robots in the field of power regulation.
Spoken language understanding is an important component of a human-machine dialog system, while intent recognition is a subtask and crucial in spoken language understanding. The accuracy of intent recognition is directly related to the performance of semantic slot filling and aids in the construction of subsequent dialog systems. In the prior art, the intention recognition method mainly comprises a semantic recognition method based on a rule template and a classification algorithm using statistical characteristics. The intent recognition method based on the rule template generally needs to artificially construct the rule template and classify the user intent text according to the category information. The statistical feature classification-based method needs to extract key features of the text, such as characters, word features, N-grams and the like, and then realizes intent classification by training a classifier. In consideration of the difficulty of intention recognition in a man-machine conversation system, the traditional machine learning method cannot understand deep semantic information of user words, and the research of an intention recognition method based on a deep neural network is promoted. With the development of the deep neural network, more and more students apply word vectors, convolutional neural network networks, cyclic neural networks, variable long-term memory networks, gated cyclic units, attention mechanisms and capsule networks to the intention recognition task, and compared with the traditional machine learning method, the deep learning model has the advantages that the recognition performance is greatly improved, and the single intention recognition and the multiple intention recognition are changed into more practical achievements.
With the rapid development of power systems, the market-oriented work of power is promoted, the integration characteristics of large power grids become more and more obvious, and the coupling relationship between different elements in the systems and the external environment is continuously enhanced. The types of power grid resources are continuously enriched, the new energy ratio is continuously improved, new control targets such as the improvement of the new energy consumption level are continuously evolved, and the uncertainty of the power grid state is further enhanced. A series of deep development changes of the power system and the regulation and control service enable the complexity of the scheduling control strategy and the regulation to be continuously improved, and higher requirements are provided for automation and intellectualization of the regulation and control service. The power dispatching control center is a 'command brain' integrating high-value data, analysis rules, expert experience and calculation decisions, the existing regulation and control mode mainly takes manual experience analysis as a main mode, a dispatcher needs to perform experience knowledge correlation on massive and diverse data and scheme models, more repetitive 'human brain labor' exists, and the efficiency is lower. Therefore, intelligent regulation is realized, and the working strength of regulation personnel is reduced.
Through the above analysis, in the prior art, the problems and defects in the real-time analysis of the regulation and control service are as follows:
(1) the existing multi-turn conversation technology has the problems of low intelligent degree, large research and development workload and the like.
(2) The existing power dispatching regulation and control mode mainly adopts manual experience analysis, and dispatching personnel need to perform experience knowledge correlation on massive diverse data and scheme models, so that more repetitive 'human-brain labor' is realized, and the efficiency is lower.
Disclosure of Invention
In order to solve the defects in the prior art, the invention aims to provide a power information query and regulation function calling method based on voice and intention recognition, and the power information query and regulation function calling is realized from three aspects of a voice recognition technology in the power regulation field, an intention recognition technology in the power regulation field and a multi-turn dialogue technology of a regulation system.
The invention adopts the following technical scheme.
The electric power information query and regulation function calling method based on voice and intention recognition comprises the following steps:
step 1, collecting scheduling voice corpora and constructing a scheduling voice corpus;
step 2, using the scheduling voice corpus as input data and inputting the input data into the trained acoustic model and language model; jointly outputting a dispatching telephone text by the trained acoustic model and the language model;
the trained acoustic model and language model are trained by adopting a deep learning algorithm by taking a scheduling voice corpus as a training set;
step 3, collecting a regulation and control instruction corpus, and vectorizing regulation and control instruction corpus words to construct a regulation and control instruction corpus according to the obtained regulation and control instruction word vectors;
step 4, taking the vector of the regulation instruction word as input data, and respectively inputting the input data into the trained field recognition model and the trained intention recognition model;
outputting a field matching result by a field recognition model, and outputting an intention recognition result by an intention recognition model according to the field matching result;
the method comprises the following steps that a trained field recognition model and an intention recognition model are trained by adopting a deep learning algorithm by taking a control instruction corpus as a training set;
step 5, taking the intention recognition result as input data of the slot position recognition model, and outputting a regulation and control instruction matching result by the slot position recognition model;
the slot position identification model extracts semantic tags from the scheduling instructions by using a rule analysis algorithm;
step 6, according to the regulation and control instruction matching result, defining a scheduling instruction to prepare for carrying out power scheduling; determining the running state of the power grid by combining power grid load flow calculation, short-circuit current calculation and switching-on and switching-off operations; meanwhile, according to the matching result of the regulation and control instruction, the power information query, the automatic power grid calculation and the automatic power dispatching response operation are executed;
and 7, inputting the scheduling telephone text and the regulation and control instruction matching result into a voice synthesis engine, converting the scheduling telephone text and the regulation and control instruction matching result into scheduling telephone voice through the voice synthesis engine, and automatically responding to the scheduling telephone by using the scheduling telephone voice.
Preferably, the first and second electrodes are formed of a metal,
in the step 1, noise reduction processing and feature extraction are carried out on the collected scheduling voice corpus, and the obtained feature sequence is stored in a scheduling voice corpus; the scheduling voice corpus comprises a general corpus and an electric power professional corpus, and further comprises a standard voice signal for matching degree testing.
Preferably, the first and second electrodes are formed of a metal,
the step 2 comprises the following steps:
step 2.1, using the scheduling voice corpus as input data and inputting the input data into the trained acoustic model and language model;
step 2.2, the trained acoustic model carries out noise processing and acoustic feature extraction on the scheduling voice corpus to output an acoustic feature sequence; obtaining the matching degree of the acoustic feature sequence, namely the score of an acoustic model, by using a standard voice signal;
step 2.3, the trained language model carries out noise processing and semantic feature extraction on the scheduling voice corpus to output a semantic feature sequence; obtaining the matching degree of the semantic feature sequence, namely the language model score by using a standard voice signal;
step 2.4, taking the feature sequence with the highest comprehensive score of the acoustic model and the language model as a dispatching telephone feature sequence;
and 2.5, performing language decoding on the scheduling telephone feature sequence based on the electric power professional corpus to obtain a scheduling telephone text.
Preferably, the first and second electrodes are formed of a metal,
in step 2.2, standard voice signals are used as matching degree test input data and input into the trained acoustic model, and the acoustic model outputs a standard acoustic feature sequence; and calculating the matching degree of the acoustic feature sequence and the standard acoustic feature sequence based on a distance algorithm.
Step 2.3, standard voice signals are used as matching degree test input data and input into the trained language model, and the language model outputs a standard semantic feature sequence; and calculating the matching degree of the semantic feature sequence and the standard semantic feature sequence based on a distance algorithm.
Preferably, the first and second electrodes are formed of a metal,
step 6, integrating and summarizing equipment tide, power grid events, operation modes, risk early warning, equipment maintenance, accident plans and meteorological information, and inquiring and watching various electric power information in the same interface; the interface supports the dragging of the information module, and the currently required information can be customized and displayed according to the requirements.
Further, step 6 further comprises: and the functions of report self-customization, picture self-generation and telephone self-response are realized by interacting with a regulation knowledge base.
Further, step 6 further comprises: and according to the switching operation knowledge graph and the disconnecting link position image recognition result, the intelligent ticket forming, the automatic execution of the operation ticket and the confirmation of the disconnecting link position of the operation ticket are realized.
Further, the picture self-generation is to integrate and summarize the equipment tide, the power grid event, the operation mode, the risk early warning, the equipment maintenance, the accident plan and the meteorological information, and check various information in one interface.
Preferably, the first and second electrodes are formed of a metal,
the step 7 comprises the following steps: acquiring a scheduling telephone text and a regulation and control instruction matching result; through setting the characteristic parameters of the synthesized voice, converting the text of the dispatching telephone and the matching result of the regulating instruction into voice by using a voice synthesis engine; the synthesized voice is broadcasted to the regulating personnel.
Compared with the prior art, the invention has the beneficial effects that:
and developing the application of an artificial intelligence technology in regulating and controlling the operation of the streamlined business based on the voice and intention recognition, and taking the intention recognition as a core module of a multi-turn conversation scene for understanding the intention of a dispatcher, converting the real intention of the dispatcher into a machine language and transmitting the machine language to a computer. The method faces to the actual scene, accurately identifies the information intention, realizes the flow intelligent execution of the regulation and control operation business, particularly the function calling of the power information inquiry and regulation and control system, lightens the flow business burden of the regulation and control professional, improves the working efficiency and the safety level, and ensures the safe and stable operation of the power grid.
The invention has the beneficial effects that the power grid regulates and controls the transactional work intelligent execution technology, realizes the power information query, search and function call based on voice, and realizes the functions of customizing reports, automatically composing pictures, automatically answering typical service calls and the like. The intelligent interaction of regulation and control things can be realized through voice recognition and intention recognition, the switching operation data information is managed and controlled on line, a dispatcher is assisted to execute the whole flow, and the power grid system level warning capability and the intelligent fault handling capability are improved. The intelligent control system has the advantages that simple and repetitive control work is replaced by an artificial intelligence means, the intelligence level of application products in the control field is improved, a dispatcher can be liberated from the repetitive work, and the dispatcher can have the ability to think more work which is beneficial to the operation of a power grid.
Drawings
FIG. 1 is a flow chart of a method for calling a power information query and regulation function based on voice and intention recognition according to the present invention;
fig. 2 is a schematic diagram illustrating a power information query and a regulation and control system function call implemented by a power regulation and control field speech recognition technology, a power regulation and control field intention recognition technology, and a regulation and control system multi-turn dialogue technology in an embodiment of the present invention.
Detailed Description
The present application is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present application is not limited thereby.
As shown in fig. 1, the method for calling the power information query and regulation function based on voice and intention recognition includes:
step 1, collecting a scheduling voice corpus and constructing a scheduling voice corpus.
In particular, the amount of the solvent to be used,
in the step 1, noise reduction processing and feature extraction are carried out on the collected scheduling voice corpus, and the obtained feature sequence is stored in a scheduling voice corpus; the scheduling voice corpus comprises a general corpus and an electric power professional corpus, and further comprises a standard voice signal for matching degree testing.
Step 2, using the scheduling voice corpus as input data and inputting the input data into the trained acoustic model and language model; jointly outputting a dispatching telephone text by the trained acoustic model and the language model;
the trained acoustic model and language model are trained by adopting a deep learning algorithm by taking a scheduling voice corpus as a training set.
In particular, the amount of the solvent to be used,
the step 2 comprises the following steps:
and 2.1, taking the scheduling voice corpus as input data and inputting the scheduling voice corpus into the trained acoustic model and language model.
Step 2.2, the trained acoustic model carries out noise processing and acoustic feature extraction on the scheduling voice corpus to output an acoustic feature sequence; and obtaining the matching degree of the acoustic feature sequence, namely the acoustic model score by using the standard voice signal.
In particular, the amount of the solvent to be used,
in step 2.2, standard voice signals are used as matching degree test input data and input into the trained acoustic model, and the acoustic model outputs a standard acoustic feature sequence; and calculating the matching degree of the acoustic feature sequence and the standard acoustic feature sequence based on a distance algorithm.
Step 2.3, the trained language model carries out noise processing and semantic feature extraction on the scheduling voice corpus to output a semantic feature sequence; and obtaining the matching degree of the semantic feature sequence, namely the language model score by using the standard voice signal.
In particular, the amount of the solvent to be used,
step 2.3, standard voice signals are used as matching degree test input data and input into the trained language model, and the language model outputs a standard semantic feature sequence; and calculating the matching degree of the semantic feature sequence and the standard semantic feature sequence based on a distance algorithm.
And 2.4, taking the feature sequence with the highest comprehensive score of the acoustic model and the language model as a dispatching telephone feature sequence.
And 2.5, performing language decoding on the scheduling telephone feature sequence based on the electric power professional corpus to obtain a scheduling telephone text.
And 3, collecting the control instruction corpus, and vectorizing the control instruction corpus words to construct a control instruction corpus according to the obtained control instruction word vectors.
Step 4, taking the vector of the regulation instruction word as input data, and respectively inputting the input data into the trained field recognition model and the trained intention recognition model;
outputting a field matching result by a field recognition model, and outputting an intention recognition result by an intention recognition model according to the field matching result;
the trained field recognition model and the trained intention recognition model are trained by adopting a deep learning algorithm by taking a control instruction corpus as a training set.
Step 5, taking the intention recognition result as input data of the slot position recognition model, and outputting a regulation and control instruction matching result by the slot position recognition model;
and the slot position identification model extracts semantic tags from the scheduling instructions by using a rule analysis algorithm.
Step 6, according to the regulation and control instruction matching result, defining a scheduling instruction to prepare for carrying out power scheduling; determining the running state of the power grid by combining power grid load flow calculation, short-circuit current calculation and switching-on and switching-off operations; and simultaneously, according to the regulation and control instruction matching result, performing power information query, automatic power grid calculation and automatic power dispatching response operation.
In particular, the amount of the solvent to be used,
step 6, integrating and summarizing equipment tide, power grid events, operation modes, risk early warning, equipment maintenance, accident plans and meteorological information, and inquiring and watching various electric power information in the same interface; the interface supports the dragging of the information module, and the currently required information can be customized and displayed according to the requirements.
Further, step 6 further comprises: and the functions of report self-customization, picture self-generation and telephone self-response are realized by interacting with a regulation knowledge base.
The picture self-generation is to integrate and summarize the equipment tide, the power grid event, the operation mode, the risk early warning, the equipment maintenance, the accident plan and the meteorological information, and check various information in one interface.
Further, step 6 further comprises: and according to the switching operation knowledge graph and the disconnecting link position image recognition result, the intelligent ticket forming, the automatic execution of the operation ticket and the confirmation of the disconnecting link position of the operation ticket are realized.
And 7, inputting the scheduling telephone text and the regulation and control instruction matching result into a voice synthesis engine, converting the scheduling telephone text and the regulation and control instruction matching result into scheduling telephone voice through the voice synthesis engine, and automatically responding to the scheduling telephone by using the scheduling telephone voice.
In particular, the amount of the solvent to be used,
the step 7 comprises the following steps: acquiring a scheduling telephone text and a regulation and control instruction matching result; through setting the characteristic parameters of the synthesized voice, converting the text of the dispatching telephone and the matching result of the regulating instruction into voice by using a voice synthesis engine; the synthesized voice is broadcasted to the regulating personnel.
In the preferred embodiment of the invention, the power information query and the regulation and control system function call are realized from three aspects of a voice recognition technology in the power regulation and control field, an intention recognition technology in the power regulation and control field and a multi-turn conversation technology of the regulation and control system, as shown in fig. 2. Firstly, a corpus of a general small sample set and an electric power professional lexicon is constructed, and an acoustic model and a language model are trained based on deep learning. And obtaining a characteristic sequence by noise processing and characteristic extraction of the dispatching voice telephone, matching the characteristic sequence with the characteristic sequence in the training library after voice decoding, and synthesizing the characteristic sequence with the highest score according to the acoustic model and the language model to obtain an output dispatching telephone text. Secondly, a field identification model and an intention identification model of the dispatching instruction are respectively established by adopting a deep learning method based on the linguistic data in the instruction category of the regulating and controlling field, in view of the fact that the regulating and controlling system is difficult to understand the dispatching instruction once in practical application, a large number of slot position identification models are established by using a rule method, the real intention of a regulating and controlling person is determined by matching with the slot position of the dispatching instruction, and then the electric power information inquiry, automatic dispatching, automatic calculation and other operations are carried out by combining with the services of electric network analysis, calculation, operation and the like in the intelligent regulating and controlling system. And the instruction reply information generated by the regulation and control system converts the text into voice through the voice synthesis engine and broadcasts the voice to the regulation and control personnel.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When used in whole or in part, can be implemented in a computer program product that includes one or more computer instructions. When loaded or executed on a computer, cause the flow or functions according to embodiments of the invention to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, the computer instructions may be transmitted from one website site, computer, server, or data center to another website site, computer, server, or data center via wire (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL), or wireless (e.g., infrared, wireless, microwave, etc.)). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that includes one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.
Compared with the prior art, the invention has the beneficial effects that:
and developing the application of an artificial intelligence technology in regulating and controlling the operation of the streamlined business based on the voice and intention recognition, and taking the intention recognition as a core module of a multi-turn conversation scene for understanding the intention of a dispatcher, converting the real intention of the dispatcher into a machine language and transmitting the machine language to a computer. The method faces to the actual scene, accurately identifies the information intention, realizes the flow intelligent execution of the regulation and control operation business, particularly the function calling of the power information inquiry and regulation and control system, lightens the flow business burden of the regulation and control professional, improves the working efficiency and the safety level, and ensures the safe and stable operation of the power grid.
The invention has the beneficial effects that the power grid regulates and controls the transactional work intelligent execution technology, realizes the power information query, search and function call based on voice, and realizes the functions of customizing reports, automatically composing pictures, automatically answering typical service calls and the like. The intelligent interaction of regulation and control things can be realized through voice recognition and intention recognition, the switching operation data information is managed and controlled on line, a dispatcher is assisted to execute the whole flow, and the power grid system level warning capability and the intelligent fault handling capability are improved. The intelligent control system has the advantages that simple and repetitive control work is replaced by an artificial intelligence means, the intelligence level of application products in the control field is improved, a dispatcher can be liberated from the repetitive work, and the dispatcher can have the ability to think more work which is beneficial to the operation of a power grid.
The present applicant has described and illustrated embodiments of the present invention in detail with reference to the accompanying drawings, but it should be understood by those skilled in the art that the above embodiments are merely preferred embodiments of the present invention, and the detailed description is only for the purpose of helping the reader to better understand the spirit of the present invention, and not for limiting the scope of the present invention, and on the contrary, any improvement or modification made based on the spirit of the present invention should fall within the scope of the present invention.
Claims (10)
1. The electric power information query and regulation function calling method based on voice and intention recognition is characterized in that,
the method comprises the following steps:
step 1, collecting scheduling voice corpora and constructing a scheduling voice corpus;
step 2, using the scheduling voice corpus as input data and inputting the input data into the trained acoustic model and language model; the trained acoustic model and language model are trained by adopting a deep learning algorithm by taking a scheduling voice corpus as a training set; jointly outputting a dispatching telephone text by the trained acoustic model and the language model;
step 3, collecting a regulation and control instruction corpus, and vectorizing regulation and control instruction corpus words to construct a regulation and control instruction corpus according to the obtained regulation and control instruction word vectors;
step 4, taking the vector of the regulation instruction word as input data, and respectively inputting the input data into the trained field recognition model and the trained intention recognition model; outputting a field matching result by a field recognition model, and outputting an intention recognition result by an intention recognition model according to the field matching result; the trained field recognition model and the trained intention recognition model are trained by adopting a deep learning algorithm by taking a control instruction corpus as a training set;
step 5, taking the intention recognition result as input data of the slot position recognition model, and outputting a regulation and control instruction matching result by the slot position recognition model; the slot position identification model extracts semantic tags from a scheduling instruction by using a rule analysis algorithm;
step 6, according to the regulation and control instruction matching result, defining a scheduling instruction to prepare for carrying out power scheduling; determining the running state of the power grid by combining power grid load flow calculation, short-circuit current calculation and switching-on and switching-off operations; meanwhile, according to the matching result of the regulation and control instruction, the power information query, the automatic power grid calculation and the automatic power dispatching response operation are executed;
and 7, inputting the scheduling telephone text and the regulation and control instruction matching result into a voice synthesis engine, converting the scheduling telephone text and the regulation and control instruction matching result into scheduling telephone voice through the voice synthesis engine, and automatically responding to the scheduling telephone by using the scheduling telephone voice.
2. The power information query and regulation function calling method based on voice and intention recognition according to claim 1,
in the step 1, noise reduction processing and feature extraction are carried out on the collected scheduling voice corpus, and the obtained feature sequence is stored in a scheduling voice corpus; the scheduling voice corpus comprises a general corpus and an electric power professional corpus, and further comprises a standard voice signal for matching degree testing.
3. The power information query and regulation function calling method based on voice and intention recognition according to claim 2,
the step 2 comprises the following steps:
step 2.1, using the scheduling voice corpus as input data and inputting the input data into the trained acoustic model and language model;
step 2.2, the trained acoustic model carries out noise processing and acoustic feature extraction on the scheduling voice corpus to output an acoustic feature sequence; obtaining the matching degree of the acoustic feature sequence, namely the score of an acoustic model, by using a standard voice signal;
step 2.3, the trained language model carries out noise processing and semantic feature extraction on the scheduling voice corpus to output a semantic feature sequence; obtaining the matching degree of the semantic feature sequence, namely the language model score by using a standard voice signal;
step 2.4, taking the feature sequence with the highest comprehensive score of the acoustic model and the language model as a dispatching telephone feature sequence;
and 2.5, performing language decoding on the scheduling telephone feature sequence based on the electric power professional corpus to obtain a scheduling telephone text.
4. The power information query and regulation function calling method based on voice and intention recognition according to claim 3,
in the step 2.2, a standard voice signal is used as matching degree test input data and is input into a trained acoustic model, and a standard acoustic feature sequence is output by the acoustic model; and calculating the matching degree of the acoustic feature sequence and the standard acoustic feature sequence based on a distance algorithm.
5. The power information query and regulation function calling method based on voice and intention recognition according to claim 3,
in the step 2.3, the standard voice signal is used as matching degree test input data and is input into the trained language model, and the language model outputs a standard semantic feature sequence; and calculating the matching degree of the semantic feature sequence and the standard semantic feature sequence based on a distance algorithm.
6. The power information query and regulation function calling method based on voice and intention recognition according to claim 1,
step 6, integrating and summarizing equipment tide, power grid events, operation modes, risk early warning, equipment maintenance, accident plans and meteorological information, and inquiring and watching various electric power information in the same interface; the interface supports the dragging of the information module, and the currently required information can be customized and displayed according to the requirements.
7. The real-time regulation and control business analysis method based on intention recognition of claim 6,
step 6 also includes: and the functions of report self-customization, picture self-generation and telephone self-response are realized by interacting with a regulation knowledge base.
8. The power information query and regulation function calling method based on voice and intention recognition according to claim 6,
step 6 also includes: and according to the switching operation knowledge graph and the disconnecting link position image recognition result, the intelligent ticket forming, the automatic execution of the operation ticket and the confirmation of the disconnecting link position of the operation ticket are realized.
9. The method for real-time analysis of regulatory business based on intent recognition as recited in claim 7,
the picture self-generation is to integrate and summarize equipment tide, power grid events, operation modes, risk early warning, equipment maintenance, accident plans and meteorological information, and check various information in one interface.
10. The power information query and regulation function calling method based on voice and intention recognition according to claim 1,
the step 7 comprises the following steps: acquiring a scheduling telephone text and a regulation and control instruction matching result; through setting the characteristic parameters of the synthesized voice, converting the text of the dispatching telephone and the matching result of the regulating instruction into voice by using a voice synthesis engine; the synthesized voice is broadcasted to the regulating personnel.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110546414.4A CN113314106A (en) | 2021-05-19 | 2021-05-19 | Electric power information query and regulation function calling method based on voice and intention recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110546414.4A CN113314106A (en) | 2021-05-19 | 2021-05-19 | Electric power information query and regulation function calling method based on voice and intention recognition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113314106A true CN113314106A (en) | 2021-08-27 |
Family
ID=77374016
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110546414.4A Pending CN113314106A (en) | 2021-05-19 | 2021-05-19 | Electric power information query and regulation function calling method based on voice and intention recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113314106A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113688210A (en) * | 2021-09-06 | 2021-11-23 | 北京科东电力控制***有限责任公司 | Power grid dispatching intention identification method |
| CN113705224A (en) * | 2021-09-06 | 2021-11-26 | 广东电网有限责任公司 | Voice recognition scheduling service voice interaction method and system |
| CN113948079A (en) * | 2021-09-30 | 2022-01-18 | 国网福建省电力有限公司 | Power distribution network scheduling man-machine interaction method based on artificial intelligence voice recognition |
| CN114171021A (en) * | 2021-11-29 | 2022-03-11 | 国网江苏省电力有限公司南京供电分公司 | Distribution network virtual scheduling system and method based on artificial intelligence |
| CN115567637A (en) * | 2022-08-25 | 2023-01-03 | 北京科东电力控制***有限责任公司 | Dispatching conversation real-time transcription system and method based on regulation and control voice recognition |
| CN115691496A (en) * | 2022-12-29 | 2023-02-03 | 北京国安广传网络科技有限公司 | TTS-based health management robot voice interaction module |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10303978B1 (en) * | 2018-03-26 | 2019-05-28 | Clinc, Inc. | Systems and methods for intelligently curating machine learning training data and improving machine learning model performance |
| CN110444198A (en) * | 2019-07-03 | 2019-11-12 | 平安科技（深圳）有限公司 | Search method, device, computer equipment and storage medium |
| CN111833871A (en) * | 2020-07-07 | 2020-10-27 | 信雅达***工程股份有限公司 | Intelligent outbound system based on intention recognition and method thereof |
| CN112599124A (en) * | 2020-11-20 | 2021-04-02 | 内蒙古电力(集团)有限责任公司电力调度控制分公司 | Voice scheduling method and system for power grid scheduling |
-
2021
- 2021-05-19 CN CN202110546414.4A patent/CN113314106A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10303978B1 (en) * | 2018-03-26 | 2019-05-28 | Clinc, Inc. | Systems and methods for intelligently curating machine learning training data and improving machine learning model performance |
| CN110444198A (en) * | 2019-07-03 | 2019-11-12 | 平安科技（深圳）有限公司 | Search method, device, computer equipment and storage medium |
| CN111833871A (en) * | 2020-07-07 | 2020-10-27 | 信雅达***工程股份有限公司 | Intelligent outbound system based on intention recognition and method thereof |
| CN112599124A (en) * | 2020-11-20 | 2021-04-02 | 内蒙古电力(集团)有限责任公司电力调度控制分公司 | Voice scheduling method and system for power grid scheduling |
Non-Patent Citations (1)
| Title |
|---|
| 蔡新雷等: "基于人工智能的电网调度操作智能防误***建设及实践", 《电力大数据》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113688210A (en) * | 2021-09-06 | 2021-11-23 | 北京科东电力控制***有限责任公司 | Power grid dispatching intention identification method |
| CN113705224A (en) * | 2021-09-06 | 2021-11-26 | 广东电网有限责任公司 | Voice recognition scheduling service voice interaction method and system |
| CN113688210B (en) * | 2021-09-06 | 2024-02-09 | 北京科东电力控制***有限责任公司 | Power grid dispatching intention recognition method |
| CN113948079A (en) * | 2021-09-30 | 2022-01-18 | 国网福建省电力有限公司 | Power distribution network scheduling man-machine interaction method based on artificial intelligence voice recognition |
| CN114171021A (en) * | 2021-11-29 | 2022-03-11 | 国网江苏省电力有限公司南京供电分公司 | Distribution network virtual scheduling system and method based on artificial intelligence |
| CN115567637A (en) * | 2022-08-25 | 2023-01-03 | 北京科东电力控制***有限责任公司 | Dispatching conversation real-time transcription system and method based on regulation and control voice recognition |
| CN115691496A (en) * | 2022-12-29 | 2023-02-03 | 北京国安广传网络科技有限公司 | TTS-based health management robot voice interaction module |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113314106A (en) | Electric power information query and regulation function calling method based on voice and intention recognition | |
| CN109658928B (en) | Cloud multi-mode conversation method, device and system for home service robot | |
| CN112599124A (en) | Voice scheduling method and system for power grid scheduling | |
| CN111651609A (en) | Multi-turn dialogue method and system integrating knowledge graph and emotion supervision | |
| CN110277086A (en) | Phoneme synthesizing method, system and electronic equipment based on dispatching of power netwoks knowledge mapping | |
| CN111098312A (en) | Window government affairs service robot | |
| CN112632288A (en) | Power dispatching system and method based on knowledge graph | |
| CN109345072A (en) | Dispatching method for distribution branch line scheduling scenario | |
| US20220358292A1 (en) | Method and apparatus for recognizing entity, electronic device and storage medium | |
| CN111326153A (en) | Power grid intelligent regulation and control auxiliary robot interaction method based on semantic understanding | |
| CN111651572A (en) | Multi-domain task type dialogue system, method and terminal | |
| CN113140219A (en) | Regulation and control instruction generation method and device, electronic equipment and storage medium | |
| CN117421398A (en) | Man-machine interaction method, device, equipment and storage medium | |
| CN111046189A (en) | Modeling method of power distribution network knowledge graph model | |
| CN110428816A (en) | A kind of method and device voice cell bank training and shared | |
| CN112634698A (en) | Dispatcher training simulation system, method and device based on voice recognition | |
| KR101433506B1 (en) | Operation method of energy management system using an isolated language voice recognition | |
| CN115687934A (en) | Intention recognition method and device, computer equipment and storage medium | |
| CN109002498A (en) | Interactive method, device, equipment and storage medium | |
| CN112035648B (en) | User data processing method and device and electronic equipment | |
| CN113434663A (en) | Conference summary generation method based on edge calculation and related equipment | |
| CN113689851A (en) | Scheduling professional language understanding system and method | |
| Zhao et al. | Speech emotion recognition based on multi-task learning | |
| CN117112065B (en) | Large model plug-in calling method, device, equipment and medium | |
| CN116600053B (en) | Customer service system based on AI large language model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210827 |