Partial Discharge Localization in Power Transformers Using Invariant Extended Kalman Filter

This paper proposes a novel invariant extended Kalman filter (IEKF), recently modified version of the (EKF), to estimate partial discharge (PD) location in transformer insulation system model. An acoustic signal measurement is utilized localize PD location. Unlike conventional EKF methods, where correction term used update state linearly proportional output error, proposed algorithm independent resulting fast response with significant reduction estimation error. In contrast EKF, method can successfully mimic nonlinear dynamics and mitigate noise stochasticity. Moreover, even if model fails fully capture PD’s dynamics, IEKF will still sustain reasonable performance. contrast, EKFs easily diverge mismatch between true occurs. Experimental results are shown verify method’s performance compared published variant EKF.