Concurrent multi-objective optimization of plug-in parallel HEV by a hybrid evolution algorithm

This dissertation presents a new method for solving for the optima of the Plug-in HEV's overall system parameters. Different from the existing HEV optimization approaches shown in the literature that are mainly control strategies focused, our study suggested that the powertrain sizing optimization is also a crucial factor for achieving minimum fuel consumption and emissions. To solve this multi-objective problem, the dissertation research featured a concurrent approach that simultaneously optimizes both HEV powertrain sizing parameters and control logics. The novelty is using probabilistic algorithms to attack this large-scale and nonlinear problem. Such a derivative-free approach has gained high efficiency in handling the high-order, noisy and discontinuous objective functions, and nonlinear constraints of the Plug-in HEV optimization problem.; A generic design methodology of parameterizing the optimal propulsion system for the Plug-in Parallel HEV has been developed in the course of this research. It was divided into the four stages: designing search algorithms, building the cost models, analyzing multiple constraints, and implementing the optimization. Based on the modeling of Plug-in HEV and CVT (Continuously Variable Transmission) control, we implemented two global probabilistic algorithms, Genetic Algorithm and Simulated Annealing, into a case study of a parallel hybrid medium duty Step Van. Promising results or concurrent optimization have been achieved through simulations. They reveal that both algorithms are practical and effective but have certain limitations. To further enhance the overall search performance, we designed a Hybrid Evolution Algorithm that combines the strengths and overcomes the shortcomings of Genetic Algorithm and Simulated Annealing. Such an algorithm hybridization of both techniques inherits merits of each other and further enhances the overall search speed and accuracy. The overall optimization scheme simultaneously optimizes the parameters of the propulsion system and control logics adaptively based on various driving cycles. It can handle a number of design variables and lead to near-global optima in an efficient convergence path and with minimal computational resources.