Embedded Hybrid Model Predictive Control For Energy Management of Parallel Hybrid Electric Vehicles

Abstract

This thesis deals with hybrid model predictive control for energy management of parallel hybrid electric vehicles. A hybrid system is a system that shows both continuous and discrete characteristics and can be used to describe the drive system of hybrid vehicles that exhibits different dynamical behaviors depending on their powertrain mode. This study modeled the drive system of the target vehicle using a hybrid system and designed a hybrid model predictive controller that decides not only an optimal torque split but also an optimal powertrain mode with consideration of future information. The controller requires a mixed-integer quadratic program, that optimizes both continuous and discrete variables. It is an NP-hard problem, which has an exponential time complexity. For real-time implementation, this paper proposes an approximate optimization algorithm using the combinatorial integral approximation (CIA) technique instead of the global optimization method. The proposed algorithm, which decomposes the mixed-integer quadratic program into ADMM-based (Alternating direction method of multipliers) quadratic programs and linear programs, can reduce the time complexity of the problem. The optimality and computational time of the algorithm are investigated, and the proposed controller is demonstrated through the software in the loop (SIL) test and the processor in the loop (PIL) test. Simulation results show that the proposed controller achieved better fuel economy than conventional methods with fast computation time.