Research On The Early Warning System Of Current Transformer Error Testing Station

Error test station is an important part of automatic verification pipeline for current transformer, its accuracy and stability concerns the evaluation of tested object. In order to maintain the traceability of value of the station, inspection and calibration are provided by the superior metrology center periodically. However, in the interval of two calibrations,accuracy of the equipment may decrease and calibration status may not credible.Therefore, intermediate checks is necessary.In this thesis, risk warning based on the data of error detection is used as the method of intermediate checks. For the characteristic in risk warning such as finite sample,nonlinearity, uncertainty, etc. CA-FSVR(Correlation Analysis-Fuzzy Support Vector Regression) and other intelligence algorithm are combined to resolve the problem.The data used to forecast is a time series sample, which means there is a strong correlation between forecasting result and recent data,but a weak correlation between forecasting result and early data. Therefore, different weights are assigned to each sample according to the sampling time, which will effect the penalty factor of FSVR. For the interaction of multidimensional characteristics, correlation is analyzed to choose the factors which are strongly correlated with the data to be forecasted, and then prediction model can be built by using FSVR which based on the factors chosen. Furthermore, AGA(Adaptive Genetic Algorithm) is used to optimize the parameter of FSVR to achieve a optimal result.Analysis of living example shows that the prediction model used in this thesis has a better performance than others for the specific problem. In the end, the prediction model is combined with the strategy of risk assessment to forecast the risk of exception occurs of error test station.In the thesis, system of risk warning is developed based on the intelligence algorithm researched, the result of trial shows that the system can provide intermediatechecks as well as the developing trends of states from devices, which is an effective technical support to the advance inspection of error test station.