This dissertation develops algorithms and frameworks to obtain the optimal design and control solutions for a non-linear dynamic system in a computationally efficient manner. These methods and their advantages are dem...
详细信息
This dissertation develops algorithms and frameworks to obtain the optimal design and control solutions for a non-linear dynamic system in a computationally efficient manner. These methods and their advantages are demonstrated by applying them to a Plug-in Hybrid Electric Vehicle (PHEV) powertrain's optimal design and supervisory control. Since a PHEV draws energy from the electric grid it is important to consider these interactions in optimal design and control decisions of the PHEV. At the same time the battery size significantly affects the amount of grid energy transferred to propulsion and consequently the on-road power management decisions. Thus, we develop and apply algorithms capable of highlighting the optimal PHEV battery size and control decisions that result in a synergistic interaction with the electric grid. First, we develop a Dynamic Programming (DP) based optimal control algorithm capable of evaluating optimal on-road power management for a series PHEV. This algorithm was based on a backward looking implementation of the PHEV powertrain's dynamic model. Such an implementation of the DP algorithm avoided the need to interpolate the value function or enforce constraints through penalty functions, thereby alleviating crucial computational concerns. The performance of two supervisory control strategies for series PHEVs was compared using this algorithm. For a series PHEV, the results show that a charge deplete and sustain approach is comparable in performance (in $ costs) to the optimal strategy in most cases (esp. when gasoline is more expensive per mile than electricity). Then, we extend this algorithm to consider optimal charging on the electric grid. This extension was made possible by understanding and utilizing the conditions at the boundaries of the optimal charging and driving problems, and the computational attractiveness offered by the above DP algorithm. The results showed the tradeoffs between optimal charging and power management decisions
暂无评论