Research On Time Domain Compensation Of CVT Transient Transmission Characteristics

With the increasing complexity of the grid structure and the energy attributes of the access nodes,the frequency components contained in the fault voltage in the power grid are also more complicated.The implementation of broadband transient voltage monitoring is of great significance for analyzing the cause of power grid faults and guiding insulation design,but it puts forward higher requirements for voltage monitoring devices.Capacitor Voltage Transformer(CVT)substation has a relatively high volume,but it can only measure power frequency signals.If the time-domain compensation method of CVT transient transfer characteristics can improve its ability to sense fault voltage,and then realize the monitoring of transient voltage without major changes to the power grid hardware,it will have important theoretical research and engineering application value.Combined with the analysis of CVT transient transmission characteristics,a time-domain compensation method for CVT transient transmission characteristics based on feature quantity calculation and classification compensation is proposed.Based on the high-frequency oscillation characteristics of the lightning impulse response,the wavelet characteristic in time domain tw is designed,and then for the CVT under linear and non-linear conditions,the timedomain regression compensation method considering the influence of iteration error and the neural network compensation method based on sub-band preprocessing are respectively given.The transfer characteristic curves obtained based on the time-domain regression model method,simulation method,and frequency sweep method are compared,and a bidirectional modeling method is proposed.The normalized amplitude of the transfer characteristic curve based on the time-domain regression model at the resonance frequency point of 420 Hz is proposed.The ratio error of transformation is reduced from 3 to 0.172,which improves the accuracy of the transfer characteristic curve based on the time-domain regression model.The parameters of the time-domain compensation system for CVT transient transfer characteristics are analyzed.Based on the variation characteristics of wavelet characteristic in time domain,head time at primary side,and half-peak time at primary side,it is found that wavelet characteristic in time domain is not easily affected by head time at the primary side and will not fluctuates sharply.Since the wavelet time-domain characteristic quantities of the operational impulse response and lightning impulse response are at the ms level and μs level respectively in the test,100μs is determined as the threshold of the wavelet characteristic in time domain.The estimated peak voltage data of CVT operating impulse voltage is easily affected by the historical error of the rising edge.Therefore,this paper proposes a comprehensive error evaluation index based on the iterative error coefficient analysis to facilitate the optimization of the robust coefficient.Based on the optimized robust coefficients,the peak error of the compensated waveform is reduced from 0.15 k V to 0.09 k V.While the frequency components contained in the high-frequency oscillation signal of the lightning impulse response have little influence on the primary waveform,this paper constructs a Dropout-BPNN neural network compensation model based on the filtered lightning impulse response.The results show that the dual hidden layer structure neural network with the Dropout layer can meet the fitting requirements of the neural network,and the Dropout layer reduces the ratio of the mean square error of the test data to the mean square error of the training data from 490% to 122%.Therefore,the lightning impulse compensation model based on the Dropout-BPNN neural network has a good generalization ability.Based on the field test of the substation,the lightning impulse voltage data with the same head time and different peak values are obtained,and the wavelet characteristic in time domain are all within 10μs,and then the secondary side response data is preprocessed for compensation.The results show that the head time errors of the compensated waveforms are within 3.6μs,and the peak errors are within 12.4%,which are in good agreement with the expected waveform and verifies the effectiveness of the pre-established time-domain compensation system of CVT transient characteristics.