Application of Convex Optimization Techniques for Intelligent Electric Vehicles

Abstract

Batteries are often damaged by a high peak power and a steep variation of the power since it has a relatively low power density. In order to reduce battery damage, the battery/super-capacitor (SC) hybrid energy storage system (HESS) has been utilized since the SC can act as a buffer against large magnitudes and rapid fluctuations in power. The major objective regarding the battery/SC HESS is to minimize the magnitude/variation of the battery power and the power loss. To achieve the objective, I formulate optimization problems to provide the optimal HESS power using given load operation profiles. In addition, I propose an algorithm using a barrier method and a Multiplicative Increase Additive Decrease method for providing a feasible optimal solution for energy management in HESS. The battery/SC HESS can be effectively utilized for Electric Vehicles (EVs) because high peak power or rapid charging/discharging occur frequently in driving situations. However, the optimization method proposed in the second chapter cannot be adopted for EVs because it is difficult to obtain the future driving profile in advance. To calculate the optimal power of the battery/SC without the future profiles, I propose a method for computing the reference voltage of the SC based on the characteristic of power-train and the vehicle dynamics. In addition, I formulate the real-time optimization problem that minimizes the magnitude/variation of the battery power and the power loss simultaneously. To improve the power control for the battery/SC HESS in EVs, it is necessary to know the future motor power in advance. They can be derived from the future speed/acceleration of the vehicle through the method proposed in the third chapter if the future speed/acceleration can be predicted. Fortunately, there are many prediction techniques such as car following models, path planning algorithms and model predictive schemes, which are based on results of target tracking. Therefore, the driving environments, e.g., moving objects, should be accurately estimated. To improve the multi-target estimation accuracy even if there are many false detections, I propose a robust multi-target tracking scheme based on the GMPHD filter. The proposed scheme includes the processing step of evaluating multiple states/measurements which is designed to overcome the weight under/overestimation problem. Furthermore, it includes the step of generating the birth intensity for the next iteration using the measurements not associated with any tracked states. I also show that the proposed method can be extended to nonlinear Gaussian models.