A Computationally Efficient Predictive Energy Management Framework for PHEVs with Driving Cycle Clustering and Multi-Model MPC

This paper presents a two-level predictive Energy Management (EM) strategy for Plug-in Hybrid Electric Vehicles (PHEVs) with improved performance on computational efficiency and battery power constraint compliance. The upper level of this strategy performs clustering of predicted future driving cycle using an ordered sample clustering algorithm and calculates the optimal global SoC reference trajectory; the lower level employs a multi-model Model Predictive Control (MPC) method and Forward Dynamic Programming (FDP) algorithm to handle battery power constraint with reduced computational complexity, where 1
^st
order RC battery model and Rint model are fused in an elegant manner. Numerical simulations validate the strategy's ability to reproduce the optimal global SoC trajectory online and satisfy the battery power constraint. Moreover, the fuel consumption achieved by the proposed control strategy is very close to the global minimum achieved by offline DP.