Sales of electric vehicles (EVs) increases 70% energy year thanks to emission free, quiet and higher energy effi- ciency compared with internal combustion engine (ICE) vehicles. However, EVs generally have only around 22% driving ranges compared with Internal combustion engine vehicle (ICEVs) with a similar price range. Running out of the EV battery State of charge (SoC) while driving gives the same inconvenience as vehicle breakdown. Energy consumption of EVs directly impacts on the required battery capacity. A larger battery pack increases the curve weight and thus causes lower fuel economy, a lower performance and a higher cost.
This thesis introduces a novel system-level framework that implements an accurate EV power consumption model and derives minimum-energy EV driving speed profile to extend EV driving range without additional cost. This framework is inspired from the system-level low power design for the computing system. We first imple- ments a hybrid power modeling methodology combining a vehicle physics with motor-loss model with empirical data. Then, we validate the accuracy of the hybrid model with a target EV. To insure model fidelity, we fabri- cate a lightweight custom EV, perform extensive measurement, and derive model coefficients using multivariable regression analysis.
This thesis introduce derivation of the minimum-energy driving speed profile over time for a given driv- ing mission defined by the route, time constraint and payload. We first formulates an optimization problem that minimizes the total energy consumption to complete the given driving mission. Then, we use the dynamic pro- gramming algorithm to derive a minimum-energy speed profile considering dynamically varied road slopes. The proposed idea gives additional energy saving up to 15.6% compared with constant vehicle speed driving without any vehicle modification. Also, we show a possibility of the online EV driving optimization by introducing var- ious solution tracking methods. Runtime of the proposed scaling algorithm is 34% and 50% of runtime of the binary search algorithm and greedy algorithm with a initial value derivation, respectively. We verify the optimal driving results with real test drives on various road slopes and analyze the impact of vehicle propulsion system and power consumption model accuracy on the optimization results.