Approximate optimal energy management with a high-precision vehicle speed prediction algorithm
Jian Feng, Zhiyu Han, Zhenkuo Wu, Minqing Li
Abstract
A predictive energy management strategy is proposed for plug-in hybrid electric vehicles, which adopts a new high-precision vehicle speed prediction algorithm and a real-time reference state of charge (SOC) trajectory planning method. The speed prediction algorithm includes a modified Markov model based on vehicle speed features recognized by the self-organizing map neural network method, Fourier approximation, and moving average filtering. The prediction accuracy of the algorithm, measured by the root mean square error, is improved by 32.28% and 23.33%, respectively, over that of the standard Markov model when the prediction time is 5 and 10 s. To approach the optimal management strategy for improved vehicle fuel economy, the radial basis function neural network method is utilized to correlate the 10 s predicted vehicle speed with the trajectory slope of the suboptimal SOC for obtaining the reference SOC trajectory, which is applied to the adaptive equivalent consumption minimization strategy based on SOC feedback. The simulation indicates that the fuel consumption by using the proposed energy management strategy is very close to that by the theoretical suboptimal strategy with the shooting method, which leads to a superior fuel economy among the strategies studied.