Energy management strategies for hybrid diesel vehicles by dynamic planning embedded in real-world driving emission model

Abstract

The optimization of energy management strategy for hybrid vehicles is often based on engine steady performance data and the standard driving cycle conditions in the laboratory. However, these methods cannot fully capture the vehicle’s dynamic characteristics under real-world driving conditions. This study uses a BP-Adaboost algorithm combined with a transfer learning strategy to construct a learning model of real-world driving emissions based on several real-world driving emission tests of a hybrid diesel light truck. The real-world driving emission model is then embedded into the dynamic planning algorithm using a bi-variate interpolation algorithm on the state-space plane. Accordingly, the optimal engine and motor torque control under real-world driving conditions is determined. It is found that the energy management strategies balancing the CO₂ and NO_x emissions for the hybrid diesel light truck can obtain a good NO_x emission benefit while slightly sacrificing the CO₂ emission benefit, and the trade-off consideration between energy consumption, pollutant emissions, and state-of-charge maintenance leads to a better overall social and economic benefit.