School of Reliability and Systems Engineering, Beihang University, Beijing, China.
Science and Technology on Reliability and Environmental Engineering Laboratory, Beijing, China.
PLoS One. 2018 Jul 6;13(7):e0200169. doi: 10.1371/journal.pone.0200169. eCollection 2018.
Estimation of remaining capacity is essential for ensuring the safety and reliability of lithium-ion batteries. In actual operation, batteries are seldom fully discharged. For a constant current-constant voltage charging mode, the incomplete discharging process affects not only the initial state but also processed variables of the subsequent charging profile, thereby mainly limiting the applications of many feature-based capacity estimation methods which rely on a whole cycling process. Since the charging information of the constant voltage profile can be completely saved whether the battery is fully discharged or not, a geometrical feature of the constant voltage charging profile is extracted to be a new aging feature of lithium-ion batteries under the incomplete discharging situation in this work. By introducing the quantum computing theory into the classical machine learning technique, an integrated quantum particle swarm optimization-based support vector regression estimation framework, as well as its application to characterize the relationship between extracted feature and battery remaining capacity, are presented and illustrated in detail. With the lithium-ion battery data provided by NASA, experiment and comparison results demonstrate the effectiveness, accuracy, and superiority of the proposed battery capacity estimation framework for the not entirely discharged condition.
剩余容量估计对于确保锂离子电池的安全性和可靠性至关重要。在实际操作中，电池很少被完全放电。对于恒流-恒压充电模式，不完全放电过程不仅会影响初始状态，还会影响后续充电曲线的处理变量，从而主要限制了许多基于特征的容量估计方法的应用，这些方法依赖于整个循环过程。由于恒压曲线的充电信息无论电池是否完全放电都可以完全保存，因此在这项工作中，针对不完全放电情况，从恒压充电曲线中提取一个几何特征作为锂离子电池的新老化特征。本工作通过将量子计算理论引入经典机器学习技术，提出并详细说明了一种集成的量子粒子群优化支持向量回归估计框架及其在提取特征与电池剩余容量之间关系的应用。通过使用美国宇航局提供的锂离子电池数据，实验和比较结果证明了所提出的电池容量估计框架对于不完全放电情况的有效性、准确性和优越性。
