Filtering of Q(t) Measurement Data for Estimating Leakage Current

Q(t) measurement has been developed as a diagnostic technique for electrical insulators in DC applications, and is widely used to compare the relative properties of polymeric samples under high voltage; however, its conventional usage is limited to analyzing the charge amount compared to the space charge phenomenon. Conversely, Q(t) measurement can also serve as a diagnostic tool for leakage current through insulation materials because of its relatively high precision; hence, this study focuses on utilizing this method as an online abnormal detector. However, first, noisy data resulting from variable circumstances remains a challenge, even when simple low-pass filters are applied. This study ensures noise reduction by assuming an ordinary differential equation model and applying Kalman filters/Rauch–Tung–Striebel (RTS) smoothers to the experimental data of a polyimide sheet sample. Filtering/smoothing reduces the noise level to less than 1/900 of the simple first-difference data when an RTS smoother is applied, according to the comparison method that utilizes linearly-approximated data. Then, fixed-lag smoothers are applied, generally obtaining more smoothed data than that of the Kalman filter. Through the filtered results and Kalman filters’ property as an adaptive low-pass filter, a direction for filter designing is proposed, which depends on ‘process noise to the 2nd power’ / ‘observation noise.’ The filtering/evaluation/designing methods utilized in this study are beneficial for more precise diagnoses and are expected to provide a foundation for online leakage current detection.