MPC-based energy management with adaptive Markov-chain prediction for a dual-mode hybrid electric vehicle

The energy management strategy is an important part of a hybrid electrical vehicle design. It is used to improve fuel economy and to sustain a proper battery state of charge by controlling the power components while satisfying various constraints and driving demands. However, achieving an optimal control performance is challenging due to the nonlinearities of the hybrid powertrain, the time varying constraints, and the dilemma in which controller complexity and real-time capability are generally conflicting objectives. In this paper, a real-time capable cascaded control strategy is proposed for a dual-mode hybrid electric vehicle that considers nonlinearities of the system and complies with all time-varying constraints. The strategy consists of a supervisory controller based on a non-linear model predictive control (MPC) with a long sampling time interval and a coordinating controller based on linear model predictive control with a short sampling time interval to deal with different dynamics of the system. Additionally, a novel data based methodology using adaptive Markov chains to predict future load demand is introduced. The predictive future information is used to improve controller performance. The proposed strategy is implemented on a real test-bed and experimental trials using unknown driving cycles are conducted. The results demonstrate the validity of the proposed approach and show that fuel economy is significantly improved compared with other methods.