Avaliação de estratégias de gerenciamento de energia em veículos de grande porte via programação dinâmica

Abstract

Due to economic, environmental, political and social pressings, the automotive industry faces a new challenge in the development of the future fleet of vehicles, namely a severe improvement in fuel efficiency. One of the promising alternatives is the implementation of alternative propulsion systems such as the hybrid electric propulsion, which is becoming viable not only for personal vehicles, but also for heavy duty applications. One of the critical tasks during the design of such a vehicle is the implementation of an algorithm capable of providing the synergy between both energy converters (internal combustion engine and electric motor), their respective energy storage systems and other subsystems present in the vehicle, such as brakes and auxiliaries. Simultaneously, the program, which is called the Energy Management Strategy, has the objective of improving fuel economy at its maximum potential. Given the unpredictability of real drive cycles and the limitations of the vehicle processor, online strategies are not able to achieve the ideal fuel economy of a specified road. On the other hand, the algorithm can be tuned in a way that its usage for determined paths ensures gains much closer to the global optimum. One of the tools that can be employed to assist the tuning procedure of an online EMS is the Dynamic Programming, which is an offline global optimization algorithm that can be used not only to determine the fuel economy limit of a given path, but also to show patterns that can be followed by the designer to improve fuel consumption. In this project the dynamic programming was implemented as a Matlab function and coupled with a Simulink model provided by IKA of a research truck developed in cooperation with DAF and the Eindhoven University in the Netherlands. To test its efficacy and to determine the influence of both optimization and physical parameters over its results, some test drive cycles were created and optimized and their results analysed. Furthermore, three real elevation profiles were also optimized. Finally, a sample online EMS based on patterns extracted from the DP optimizations was developed to illustrate the application of this powerful tool.