Research On Online Evaluation Of Measurement Error State Of Electronic Transformers

As the key measurement equipment in smart substation,the stability and reliability of electronic transformer is related to the safe and stable operation of smart grid.Accuracy is an important parameter to measure the measurement performance of transformer.Its long-term stability has a very important impact on the measurement of measuring equipment,relay protection and other functions.At this stage,the main research on the monitoring and evaluation technology of the error state of transformer is off-line periodic maintenance or short-term online verification,which can not achieve long-term online monitoring and evaluation.Therefore,how to accurately evaluate the operation status of transformer on-line without high-precision standard will be the focus and difficulty in this field.This paper mainly focuses on the evaluation and prediction methods of measurement error state of electronic transformer in engineering application.The main contents are as follows:1.Based on the analysis of EVT electrical physical network model and the physical characteristics of primary voltage,an on-line evaluation of EVT error state based on correlation analysis is proposed.Principal component analysis(PCA)is used to analyze the correlation of EVT measurement data,and the measurement deviation caused by a system fluctuation and the error caused by EVT anomaly are separated.A Q statistic representing the error state of transformer is constructed in the residual subspace,and the operation state of EVT is monitored by comparing the relationship between the Q statistic characteristic and its statistical threshold.The simulation results show that the proposed method can accurately detect the abnormal errors of EVT.This method can greatly improve the evaluation efficiency of the measurement performance of EVT,and can be used to evaluate the error state of EVT in engineering applications.2.In view of the possibility that the traditional principal component analysis method will produce false detection and classification,a RPCA method with stronger self-adjustability and time-varying is proposed,that is,after obtaining a new observation data,the data at the last time are updated to realize on-line monitoring of the measurement error state of transformers.The simulation results show that the state evaluation of measurement error of electronic transformer based on recursive principal component analysis is more suitable for abnormal data detection in time-varying process,and can accurately determine the abnormal phase.This method has good applicability.It can can evaluate the measurement error of electronic transformer online without standard transformer.It has a certain reference value for predicting the operation state of electronic transformer.3.Aiming at EVT error state prediction,an EVT measurement error state prediction model based on PCA-ARMA is proposed.By collecting the output data of three-phase EVT,principal component analysis(PCA)is used to extract the error characteristics of transformer output data,and the prediction of transformer error state is mapped to the prediction of the physical correlation of grid information.The time series method is used to predict the variation trend of EVT measurement error state in the next sampling time or in a future time period.The experimental results show that the average error between the predicted value and the actual value of the difference feature based on PCA-ARMA prediction model is about 5%.The preliminary exploration of the state prediction method for measurement error of electronic voltage transformer is realized.4.Machine learning algorithm is applied to the prediction of the error trend of transformer.By comparing the performance of three prediction methods,ARMA,ANN and SVR,the prediction method based on support vector machine regression is selected to predict the error state of transformer.The simulation results show that the prediction method of EVT error state based on RPCA-SVR is better than that based on RPCA-ARMA.The average prediction errors of ratio error and angular error are 4.1% and 6.2% respectively,which can accurately evaluate the real-time operation state of transformer and predict its future operation state.