Research On Semantic Parsing Technology In Power Control System

With the expansion of power grid dispatching control scale,power-related enterprises accumulate a large number of data,especially text data.In the trend of power intelligent construction and artificial intelligence technology,how to use these data effectively becomes the common concern of industry and academia.Early,the information extraction systems often rely on manual extraction rules,and are less adaptable and scalable.In recent years,machine learning methods have been widely concerned in the field of information extraction,especially the research on machine learning based on statistics.In the actual power dispatching control process,the system stores a large number of dispatching operation instructions,monitoring operation content,switch trip processing content,monitoring abnormal records,on-site exception processing content,such as text form of record information.Here,these words are called as the power regulation text.These texts are mainly in the form of the Chinese short text which is mixed with the date,the numeral,the letter and the special symbol,which is rich in the equipment running state,the overhaul process,the operation record and so on.However,because the power text has many ambiguities,difficult to cut,professional and strong characteristics,the above information has not been fully utilized.Based on the semantic analysis of the text used in the power dispatching control system,the computer can transform the text into the analytic form for computer.Aiming at the text in the process of power dispatching control,this paper presents a semantic analytic framework for power dispatching control system,and designs a semantic analytic scheme based on the text in power dispatch control field.Firstly,According to the characteristics of various kinds of operation texts in power,a special corpus for power control system including sample library,physical environment database and terminology database are established.The lexical standard of power regulation text and the dictionary of power regulation are established,including basic word dictionary and compound word dictionary for the segmentation of power regulation text.The paper designs the annotation set including POS tagging,semantic component annotation,semantic model and so on.The experimental results show that the domain dictionary established in this paper can effectively improve the segmentation accuracy of power regulation text compared with the general Chinese dictionary.Secondly,this paper studies the common Chinese text semantic analysis methods,tools and open source platform,and carries on the experiment contrast analysis based on the electricity control text,and explores the possibility of their application in the electric power dispatch control system.Combined with CRF,LSTM and other machine learning algorithms we parse semantics of the text and take the effective information extraction and integration;Combining with the form characteristics of computer communication data and the semantic analysis demand of power dispatching control system,the paper designs the semantic analytic processing result form of convenient access and processing based on JSON.Finally,The text semantic parsing framework is given.Combined with a custom domain dictionary and corpus retraining model,based on the JAVA programming language based on open source program development,and write program will organize the output results are extracted to generate a computer can identify common format,periodic can use results were obtained.In this paper,the semantic analysis and the knowledge of power domain are integrated,and the semantic analytical framework for power dispatching control system is established.Integrating semantic analysis and knowledge of technology and power domain,this paper establishes a semantic analytic framework for power dispatching control system.The experiment shows that the proposed semantic analytic framework is feasible and effective in the power control text.