Stochastic dynamic programming for the energy management of a serial hybrid electric bus

Hybrid electric vehicles require an energy management strategy to determine how the requested power is to be split up between the combustion engine and the electric energy buffer. The energy management significantly influences the overall efficiency of the vehicle. A careful design and implementation is thus crucial to achieve good fuel economy. This paper proposes a strategy based on stochastic dynamic programming with the following advantages: First, optimal performance is guaranteed with respect to the available information. Second, the resulting controller is a look-up table and the computational burden of its evaluation is very light. Third, controller synthesis is a systematic procedure allowing to trade off multiple objectives, e.g., fuel economy and drivability. Simulation results show 3-5% close to global optimal performance. During practice, a hybrid bus with the proposed energy management yields a fuel consumption reduction of 22-29% compared to a Diesel bus with comparable specifications.