Abstract: To solve the key problem of charging current setting in optimal charging, ant colony algorithm optimizing core parameters of regression support vector machine are put forward and used in optimal charging current forecast. Core parameters C and g of SVR are in the form of node values in the ACO. This paper uses errors under the condition of cross validation as the objective function to update pheromone concentration, to make the SVR optimal performance. According to the lithium ion battery charging data, the optimal charging current ACO-SVR model is established. The results show that the ACO-SVR model shows less optimization time and better prediction accuracy. Through the theoretical analysis and experimental results, it attests that this method is practical and effective.

Key words: ant colony algorithm, support vector machine, parameter optimization, lithium ion batteries, optimal charging

摘要： 为解决锂离子电池最优充电中电流设定的关键问题，提出蚁群算法（ACO）优化回归型支持向量机（SVR）核心参数，并将蚁群优化的回归型支持向量机（ACO-SVR）用于最优充电电流的预测。SVR核心参数[C]和[g]以节点值的形式在蚁群系统中体现，以交叉验证意义下误差作为目标函数更新节点信息素浓度，经过有限次迭代得到最优[C]和[g]值，使SVR性能最优。根据锂离子电池实测充电数据建立了ACO-SVR最优充电电流模型，结果表明ACO-SVR模型具有较少的寻优时间和较好的预测精度，通过理论分析和实验数据验证了该方法具有一定的实用性和有效性。

关键词: 蚁群算法, 支持向量机, 参数优化, 锂离子电池, 最优充电