Design of Energy Management Strategies for a Battery-Ultracapacitor Electric Vehicle

The battery pack is the most expensive component in electric vehicles. Electric vehicles are prone to accelerated battery degradation due to the high charging/discharging cycles and high peak power demand. One solution to this issue would be increasing the battery capacity to meet the high energy requests. However, increasing the battery size is not reasonable due to the high cost and volume. An alternative solution is integrating other energy storage systems into the vehicle powertrain. The additional energy storage system highlights an energy management strategy to distribute the power among onboard energy storage systems effectively. Energy management systems incorporate different strategies classified based on their computational time, implementability in real-time, and measurable performance to be optimized. This thesis considers the case study of Chevy Spark model year 2015 with a hybrid energy storage system including battery and ultracapacitor. First, an overview of diffrent energy storage systems is presented, followed by a review of different hybrid energy storage' configurations. Second, energy management strategies are categorized into three main classifications: rule-based, optimization-based, and data-based algorithms. Third, the selected vehicle model with an embedded rule-based energy management strategy is developed in MATLAB Simulink, and battery performance is validated against available real-world data. Optimal power distribution among battery and ultracapacitor is achieved through an offline global optimal algorithm in chapter 5 in a way to improve battery life. Finally, optimal results are used as a training dataset for an online data-based energy management strategy. Results prove the strategy's effectiveness by improving battery life by an average of 16% compared to the rule-based and 12% difference from the globally optimal strategy on various driving conditions. The proposed energy management strategy provides near-optimal performance while it is real-time implementable and does not need to have beforehand knowledge of driving cycles.