Guided model predictive control for connected vehicles with hybrid energy systems

The development of intelligent transportation system has immensely promoted information interaction and provided higher fuel economy potential for connected vehicles. In this paper, a novel guided predictive energy management strategy with online state of charge (SoC) planning is proposed for connected vehicles with hybrid energy systems, such as plug-in hybrid electric vehicles. Its major advantage lies in the comprehensive integration of the driving-cycle information to ameliorate the global optima of real-time control algorithm. At the upper SoC planning level, a supervised learning method based on neural network is employed to derive a reference SoC trajectory in real time; while at the lower control level of model predictive control (MPC), the power allocation or optimization is guided by the reference SoC trajectory to achieve a globally optimal solution. The main contributions of this paper include: (1) A supervised learning method for fast SoC planning is introduced and further optimized by adjusting the sample size and sampling interval, thus reducing the SoC planning error rate to less than 2.28%. (2) A guided MPC structure is constructed to achieve close-to-optimal effect in instantaneous control. Simulation results demonstrate that this guided MPC approach is able to save up to 34.73% energy consumption compared to conventional charge depleting and charge sustaining strategy under a 7-h historical bus test cycle.