Energy management strategies based on dynamic programming for applications with energy storing capacity
DIRECTORS: Alfred Rufer, Philippe Barrade UNIVERSITY: EPFL (École Polytechnique Fédérale de Lausanne)
Nowadays, the energy storage systems are being incorporated in many applications. For example, they are enabling a better integration of renewables. Apart from the stationary applications, these devices are also presented in mobility applications. The electric vehicle is a good example of this kind of applications. Energy storage systems have some limitations, such as safety and lifetime, which are being addressed during the last years. Nevertheless, these are not the only challenges that need to be faced, as their optimal rating and exploitation are also critical.
In fact, when an energy storage system is introduced in any application, two main issues must be solved. On the one hand, the energy storage system must be rated in order to satisfy the application requirements, mainly the power and energy requirements. On the other hand, it must be optimally managed in order to make the most of the installed capacity. Moreover, these two aspects are strongly coupled. In this thesis, it is proposed to address this problem, opening this coupling and solving the management problem first (assuming a given energy rating).
Between the different possible energy management strategies, this work is focused on a rule based control strategy which is optimized by implementing a Dynamic Programming based optimization technique. For that purpose, an implementation methodology is proposed for a systematic development and implementation of these optimized control strategies, valid for deterministic and stochastic applications. The cost function proposed for the optimization technique is based on the stock management theory. In addition, a new representation of stochastic applications is also proposed, which relates the energy requirements of an application with their probabilities of occurrence. The proposed methodology has been applied to a vertical transport application with energy storing capacity. The proposed control strategy has been tested first in simulation and then experimentally validated in a full-scale elevator with a supercapacitors based energy storage tank. In addition, a non-optimized rule based control strategy has also been analyzed, developed, implemented and compared to the optimized control strategy. These results have validated the proposed methodology and demonstrated that the optimization techniques based on Dynamic Programming are well-suited for energy management applications, and they achieve an optimal behavior of the system.