A Characterized Method for the Real-Time Compensation of Power System Measurement Transducers

This paper presents a real-time compensation method for the improvement of the frequency behavior of measurement transducers already operating or to be installed in transmission and distribution grids. The developed technique relies on the identification of an infinite impulse response digital filter with complex frequency response equal to the inverse of the transducer, whose parameters are evaluated through a hybrid scheme based on the combination of a stochastic and a deterministic procedure. Attention is focused on the algorithm identification capability, on the sensitivity to the weighting array introduced in the optimization cost function, and on the propagation uncertainty associated with the algorithm input quantities. Such investigation has been carried out on a circuit model of a resistive divider thought for medium-voltage (MV) measurements. To estimate the uncertainty associated with the identified filter frequency behavior, the Monte Carlo method has been implemented. The overall improvement varies with the frequency; at 10 kHz, the improvement for both the ratio as well as phase error is of two orders of magnitude. The uncertainty, estimated with 10 500 draws, associated with the filter is lower or equal to the input ones all over the 10 Hz-50 kHz frequency range. The algorithm is finally applied to a laboratory resistive capacitive divider, which is the low-voltage stage of an MV divider. The obtained improvement is of two orders of magnitude for the ratio and phase errors over all the considered frequency range.