Power reserve predictive control strategy for hybrid electric vehicle using recognition-based long short-term memory network

Abstract

For series hybrid electric vehicle (SHEV), due to the limitation of operating characteristic of engine, it is challenging to design an efficient power control strategy that can ensure instantaneous response of engine-generator set (EGS) to the sudden increase in demand power. In this paper, a power reserve predictive control strategy for SHEV is proposed. Firstly, a driving pattern recognition-based long short-term memory network is developed to predict future demand power. An improvement in the accuracy of this prediction is achieved by using a new time-series changing structure. Secondly, a novel method to regulate operation points of engine is presented, in which the instantaneous power output of engine can be responded to meet suddenly increased demand power by pre-regulating operating points according to the predicted power. Thirdly, the coordinative optimization for speed regulation of engine and power flow of vehicle is formulated in model predictive control framework considering multiple constraints of SHEV. Finally, the performance of the proposed strategy is both validated in simulation and hardware-in-loop test. The results show that the expected output power of EGS can be ensured, and the fuel economy can be improved by 9.16% and 5.91% over rule-based strategy, under the two test driving cycles, respectively.