Energy management control design for fuel cell hybrid electric vehicles using neural networks

Abstract

The design and optimization of hybrid electric vehicle powertrains can take a great benefit from mathematical models which include auxiliary management and control strategies of the energy fluxes: the use of virtual platforms reduces the expensive and time-consuming experimental activity. In this work the authors developed an online Energy Management System (EMS) controller for a FCHEV, designed to employ the same energy management over a wide range of driving style types. The controller was designed by using neural networks (NN), which were trained with the optimal power flux distribution between a fuel cell system and a battery system that minimizes the overall equivalent energy consumption. The optimal solution was obtained by carrying out a gradient-based method minimization over eight different driving cycles, and using a dynamic lumped parameter mathematical model of a FCHEV fed by hydrogen and Li-ion batteries. A quantitative and qualitative analysis was made showing the networks performances over different type of cycles. Through this analysis, a suitable classification into two cycle categories is provided, covering most of the possible driving styles with two of the developed controllers.