Hybrid Electric Vehicle Downshifting Strategy Based on Stochastic Dynamic Programming During Regenerative Braking Process

Regenerative braking is the most important way to save energy for hybrid electric vehicles (HEVs) and electric vehicles (EVs). Wherein, for HEVs/EVs equipped with stepped automatic transmissions, downshifting is an effective method to improve recovery energy by adjusting motor work points. However, it is difficult to design a proper downshifting strategy due to complicated traffic conditions and unknown drivers’ intention. Therefore, this paper proposes a downshifting strategy using cluster-based stochastic dynamic programing (SDP). First, driving conditions are clustered using K-means algorithm based on a large quantity of urban traffic historical data. Then, static Markov chains are built to describe the transition of the future braking torque demand for each cluster. Next, a four-dimensional SDP optimization problem for downshifting is formulated and solved through Bellman iteration. Finally, support vector machine is adopted to identify the driving conditions online, based on which SDP results are used to give the downshifting command. Simulations and controller-in-loop tests are carried out, and the results show that regenerative braking could recover more energy by SDP-based strategy than no downshifting and rule-based strategy.