Estimation of Battery State of Charge for UAV Using Adaptive Extended Kalman Filter

This paper proposes an algorithm to estimate a battery state of charge(SoC) for an UAV using an adaptive extended Kalman filter. Shepherd model is applied to represent the relationship between the battery voltage and the SoC. The parameters of the Shepherd model is identified through a nonlinear optimization method. In order to verify the proposed approach, both an open data set named the Panasonic 18650PF and flight data set are utilized. The performance of the proposed approach is evaluated based on the error between the measured SoC and the estimation of SoC, in comparison to the extended Kalman filter. Then, three sorites of flight data set performed to check an endurance of the UAV are selected in order to verify the proposed system. In comparison to the performance of the extended Kalman filter, the proposed adaptive extended Kalman filter shows better SoC estimation accuracy over both the Panasonic 18650PF data sets and the flight data sets.