A computationally efficient model predictive control energy management strategy for hybrid vehicles considering driving style

In this study, a computationally efficient energy management strategy for model predictive control for hybrid vehicles considering the driving style is proposed. Driving data were collected through driver-in-the-loop simulation experiments, a certain number of feature parameters related to driving styles were analyzed, and the final feature parameters were determined through two screenings. Driving data were labeled according to unsupervised clustering, and the performance of multiple machine learning classifiers was compared under the driving style recognition task. The extra Tree, which had the highest accuracy rate, was selected as the driving style recognition model for online applications. A driving style adaptive speed predictor was designed, and the prediction accuracy of the proposed predictor was simulated and compared with that of some common predictors on different driving style data. The driving-style adaptive Pontryagin’s minimum principle for model predictive control (DSA-PMP-MPC) algorithm was designed as a real-time energy management strategy for Hybrid Electric Vehicles (HEVs). The performance of the rule-based, Pontryagin’s minimum principle (PMP), dynamic programming (DP-MPC), Pontryagin’s minimum principle for model predictive control (PMP-MPC), and the strategy proposed in this study are compared through simulation on test data of four driving styles, and it is verified that the strategy in this study can achieve good overall performance, including fuel economy and computational efficiency.