Multi-cell battery systems have been pervasively adopted as power supplies in industrial, commercial, and residential applications. Traditionally, battery systems consist of a large number of single cells interconnected by fixed topology to fulfill the requirements on voltage, current, capacity, and power. However, various cell unbalances introduced in manufacture and operational processes make battery systems suffer several major system-level deficiencies, including low energy efficiency, safety, second-use, and reliability issues. Therefore, dynamically reconfigurable battery (DRB) systems have gained popularity in recent years to achieve a higher energy efficiency and to prolong cycle life of battery systems by overcoming the adverse effects caused by cell unbalances. Unfortunately, current research on DRB systems mainly focuses on topology design with traditional management methods, lacking a theoretical foundation to study the performance optimization. In this paper, we propose a theoretical framework to optimize the DRB system performance by holistically considering various system design trade-offs on dynamical reconfiguration of the cell topology. Extensive numerical and experimental studies indicate the effectiveness and efficiency of the proposed optimization framework. Furthermore, the proposed framework can characterize the DRB system behaviors in a quantitative fashion, which provides design insights for such systems. A real-world case study has been provided to gain in-depth understanding of the proposed DRB systems.