Analysis And Research Of EMS Alarm Information Based On Deep Learning Technology

With the development of smart grid in our country,the application of smart substation is more and more extensive.The level and quantity of intelligent secondary equipment in substation have been greatly improved.Smart secondary equipment plays an important role in the safe and stable operation of power system.Therefore,more timely and accurate maintenance of secondary equipment is needed.The discovery of secondary equipment defects in intelligent substations is mainly determined by the alarm signals generated automatically by the equipment in the Energy Management System(EMS),but the alarm signals is vast,and contains a lot of interference and useless signals.It is very difficult for operators to find alarm signals that truly represent secondary equipment defect.It is urgent to find an intelligent and efficient alarm information analysis and screening method to help operators find secondary equipment defect alarm signals accurately.In recent years,the rise of natural language processing(NLP)theory and the development of deep learning(DL)theory make it possible to analyze and process alarm information intelligently.On the basis of intensive study of the characteristics of EMS alarm information,combined with the actual equipment operation and maintenance situation and the historical data of alarm information,this paper proposes an alarm information analysis method based on deep learning technology,which mainly includes two steps:artificial experience screening method and deep learning screening method.It can realize the intelligent and accurate screening of massive alarm signals,and we used the actual data to verify the feasibility of this method.The following work has been accomplished:(1)This paper introduces the background and significance of this research.introduces the research status of EMS alarm information analysis technology and deep learning natural language processing technology in detail,and summarizes the progressiveness of deep learning technology.(2)Efficient artificial experience screening method is proposed.The characteristics and generation mechanism of the original alarm signals are introduced,and the principle of artificial experience screening method is introduced in detail.Artificial experience screening method includes field division,keyword selection and screening scheme.According to the different types of alarm signals,the screening scheme can be divided into keyword screening method,date-time comparison method and frequency statistics method.(3)A deep learning screening method based on Long Short-Term Memory(LSTM)is proposed and its principle is introduced in detail.The general process of deep learning screening method is introduced,and highlights the principle of deep learning technology,including artificial neural network,cyclic neural network and LSTM model.Finally,the text vectoring technology is briefly introduced.(4)The EMS alarm information analysis method based on deep learning technology is validated by experiments using actual alarm information data.Using Microsoft SQL Server database to process the artificial experience screening method,some defect warning signals and suspected defect warning signals obtained by keyword screening method are obtained.Then the results of artificial experience screening method are used as input data of deep learning screening method.Proper pretreatment and text vectoring is carried out,LSTM classifier is trained and validated,and the best parameters are searched by repeated experiments with MATLAB software.Then the evaluation index of screening accuracy is obtained.Finally,a comparison with the traditional machine learning classification method and performance evaluation is carried out,which proves that the method proposed in this paper has a higher accuracy of comprehensive classification,a lower rate of missed selection rate and miselection rate.The screening system established can be used in practical application and is feasible.(5)Finally,the full text of the paper is summarized,and the future research is prospected.