Real-time Long Horizon Model Predictive Control of a Plug-in Hybrid Vehicle Power-Split Utilizing Trip Preview

Given a forecast of speed and load demands during a trip, a hybrid powertrain power-split Trajectory Optimization Problem (TOP) can be solved to optimize fuel consumption. This can be done on desktop to set performance benchmarks; however, it has been believed that the TOP could not be solved in real-time and is not a realizable controller. As such, several approximations of the TOP have been made in the interest of obtaining a real-time near-optimal controller, for example, Equivalent Consumption Minimization Strategies (ECMS) and their adaptive counterparts. These strategies decide on the power-split by, at each sampled time instant, minimizing a Horizon-0 (without predicting forward in time) composite function of fuel consumption and equivalent battery energy. The fuel economy that results from these strategies is highly sensitive to the calibration of the associated equivalence factor, and furthermore, must be chosen differently for different drive cycles. This paper presents a strategy for solving the TOP in real-time, i.e., as an Economic Model Predictive Controller (MPC) with horizon length sufficiently long to cover the entire trip. Unlike ECMS, this MPC is arguably calibration-free. Simulation results demonstrate the performance and robustness of the MPC by comparing the fuel consumption improvements to a rule-based Charge Deplete Charge Sustain (CDCS) strategy under both a perfect forecast assumption and a simple forecasting scheme using driving patterns in the California Household Travel Survey.