Abstarct: ABSTRACT Mohammad Taghi Nodeh, M.S., Sharood University of Technology, February 2013. Predictive control strategy for power management in parallel hybrid electric vehicle(PHEV). In this thesis we describe the modeling of hierarchical control and implementation of nonlinear model predictive control for parallel hybrid electric vehicles(PHEV). Three levels characterize the hierarchical form: supervisor, coordinator, and local, although this thesis only develops the supervisory level control strategy. The PHEV model consists of an internal combustion engine (ICE), battery-electric-drive, coupling device and differential, and vehicle dynamics. For various (velocity and road grade) driving profiles such as a US06 profiles, a supervisory level controller is computed and detailed for the solution of the associated PHEV power management problem. Specifically, the supervisory controller decides on the (optimal) power flows among the subsystems, i.e., the modes of operation and the power split between the ICE and the battery-electric-drive to achieve optimal or near-optimal performance, e.g., the trade-off between power usages, desired velocity tracking, battery charge sustaining and drivability constraints for each of the driving profiles. Solution of this problem requires an underlying mathematical power flow model that captures pertinent physical properties of the subsystems and that presupposes the given hierarchical control structure. The supervisory problem solution yields power profiles that are to be tracked by the local/decentralized controllers of the subsystems. The developed power flow model is amenable to recent advances in hybrid optimal control theory. In hybrid optimal control, different modes of operation as well as classical control inputs are utilized. For the PHEV there are two modes of operation (motoring or generating) determined by the mode of operation of the electric-drive as a motor or generator. For practicability in computing and implementing the optimal (power flow) control technique in real time, a nonlinear model predictive control strategy is also adopted for computing sub-optimal solution to the power management problem.