Remaining mileage estimation for electric vehicles based on dual extended Kalman filter and eXtreme gradient boosting
Zhiqiang Han, Zeyu Chen, Yang Zhou, Zilu Zhang, Bo Zhang
Abstract
Precisely estimating the remaining mileage of electric vehicles is highly important for vehicle control and battery recharging determinations. Remaining mileage estimation (RME) is a technique difficulty in practice since it is impacted by many factors, including the battery state of charge (SOC), state of health (SOH), ambient temperature, and traffic condition, etc. In this study, an online RME method is proposed based on dual extended Kalman filter (DEKF) and extreme gradient boosting (XGB) algorithms. Firstly, the battery SOC and SOH are co-estimated based on DEKF with considering the impacts of ambient temperature. Secondly, the current traffic condition are analyzed by using a historical data segement, and then the energy consumpation rate is predicted by XGB algorithm. The XGB algorithm's accuracy under the varying length of data segment is analyzed for determining the proper algorithm parameters. The presented method is evaluated by a simulation study. The results under several typical driving cycles indicate that the precise RME can be achieved with the maximum error less than 1.2%. The method is expected to be useful in providing credible mileage estimation in electric vehiecle applications. • The remaining mileage estimation is conducted using DEKF and XGB algorithms • The impacts of SOC, SOH, temperature and driving condition are considered • The energy consumption rate is online estimated by using a short period of data • The presented method can obtain a precise estimation with error less than 1.2%