Content area
Abstract
Battery systems have been ubiquitously adopted in commercial, residential and military applications. Due to the increasing demands for longer operating life and inevitable battery imbalances, active management is critical. However, limited management can be conducted in traditional battery systems because of the lack of flexibility. To overcome this issue, dynamically reconfigurable battery ( DRB) systems have been proposed and validated. Unfortunately, majority of current research frameworks focus on topology design with traditional battery management systems (BMS), lacking a theoretical foundation to investigate the design and operational optimization. Furthermore, before DRB systems can be applied, it is essential to figure out how the systematic ability can be released to the fullest potential. Based on previous research work as well as experience obtained from practice, this problem can be broken down into three phases and solved accordingly, namely modeling, estimation, and optimization. Modeling refers to behavior modeling of batteries and DRB systems have been used to support load profiles under specific environmental conditions. Estimation mainly deals with state estimations especially state of charge (SOC) estimation. Optimizations are then conducted based on modeling and estimation results to prolong operating life while keeping balanced cell states.
In this dissertation, we focus on addressing the fundamental problems in DRB systems by considering the aforementioned three phases in a comprehensive way. Contributions of this work include: 1) An enhanced circuit-based single battery model is established to accurately describe the performances of battery cells, which can be extended to capture multi-cell battery behaviors; 2) A novel voltage based SOC estimation algorithm is proposed and validated to conduct estimation in a real-time fashion. The complexity is considerably lower than sophisticated algorithms such as the neural network, whereas the accuracy is much higher than Coulomb counting; 3) Optimization is performed to extend operating time of the entire system, where the problem is formulated as a shortest path problem in a directed acyclic graph (DAG) and solved by a dynamic programming based algorithm. The results indicate the operating time is extended by 33% while the degree of cell imbalance is kept under a reasonable boundary.
