Search

Parasitic capacitance has a considerable impact on conductive electromagnetic interference (EMI), especially on the common mode (CM) noise. In vehicular applications, such as more electric aircraft and electrical vehicles, the value of CM stray capacitance can vary during the drive cycle, due to some change in environmental, and operating conditions, resulting in fluctuations in EMI level. Designing EMI filters according to potentially overestimated CM capacitance results in additional weight, volume, and cost which should be avoided. In this paper, for the first time, the parasitic capacitance relevant to CM noise is estimated in real-time using unscented Kalman filter (UKF). The UKF is employed due to its ability in dealing with nonlinearity and stochastic nature of the EMI as well as the measurement noises. The proposed method is tested on a typical dc–dc buck converter in different operating conditions. The effectiveness of the proposed method is tested using some hardware-in-the-loop experiments using dSPACE1104 development platform. [ABSTRACT FROM AUTHOR]

Online accurate estimation of supercapacitor state-of-health (SoH) and state-of-energy (SoE) is essential to achieve efficient energy management and real-time condition monitoring in electric vehicle (EV) applications. In this article, for the first time, unscented Kalman filter (UKF) is used for online parameter and state estimation of the supercapacitor. In the proposed method, a nonlinear state-space model of the supercapacitor is developed, which takes the capacitance variation and self-discharge effects into account. The observability of the considered model is analytically confirmed using a graphical approach. The SoH and SoE are then estimated based on the supercapacitor online identified model with the designed UKF. The proposed method provides better estimation accuracy over Kalman filter (KF) and extended KF algorithms since the linearization errors during the filtering process are avoided. The effectiveness of the proposed approach is demonstrated through several experiments on a laboratory testbed. An overall estimation error below 0.5% is achieved with the proposed method. In addition, hardware-in-the-loop experiments are conducted and real-time feasibility of the proposed method is guaranteed. [ABSTRACT FROM AUTHOR]