Time-Efficient Stochastic Model Predictive Energy Management for a Plug-In Hybrid Electric Bus With an Adaptive Reference State-of-Charge Advisory

In order to develop a practicality oriented low-cost energy management controller for a plug-in hybrid electric bus, besides minimizing energy consumption, algorithmic time efficiency should be put great attention so as to substantially lower the requirement of the controller hardware. This paper first compares two forecasting methods including a Markov chain model and an artificial hack propagation neural network based on real driving cycles, showcasing significant superiority of the Markov chain especially in computational efficiency. Moreover, an adaptive reference state-of-charge (SOC) advisement, which is tuned iteratively by taking advantage of speed forecasts in each prediction horizon, is provided with the aim of guiding the battery to discharge reasonably. Then, the Markov chain-based model predictive control is conducted and compared with a linear SOC reference model. Moreover, numerous influencing factors of the computational efficiency, including the prediction horizon length, the sampling width of the optimal power sequence, and the discretization size of state/control variables for solving the dynamic programming problem, are systematically investigated. The results show that the proposed reference SOC advisory is superior to the linear model. We further introduce several ways of accelerating the operational efficiency for the model predictive controller. Comparisons with common dynamic programming and charge-depleting and charge-sustaining solutions are also carried out to show the improved performance of the proposed control approach.