基于特征优化和随机森林算法的锂离子电池SOH估计

State of Health Estimation of Lithium-ion Battery Based on Feature Optimization and Random Forest Algorithm

电池健康状态(State of health,SOH)估计是电动汽车动力电池管理系统的重要功能之一。对其准确估计有利于延长锂离子电池的寿命,保障车辆安全可靠运行。面向数据驱动的锂离子电池SOH估计方法,针对以往方法存在无法平衡SOH估计精度和模型计算成本的问题,提出一种基于特征优化的解决方案。首先基于部分充电电压曲线和增量容量曲线提取多个特征,通过随机森林算法中的基尼系数计算出各个特征的重要性;然后综合考虑模型的估计准确度和所选子集的特征数量等两方面因素选择出最优特征子集;最后应用随机森林算法建立电池老化模型并估计SOH。结果表明,该SOH估计方法的平均绝对误差和均方根误差分别在0.4%和0.5%以内。该方法中的特征优化策略能够选择出最优的特征集合,结合随机森林算法后可以在实现较高SOH估计精确度的同时降低模型的计算成本。

State of health(SOH)estimation plays a significant role in the battery management system for electric vehicles.Accurate estimation of the SOH is conducive to extending the lifespan of lithium-ion batteries and ensuring vehicles'safe and reliable operation.Aiming at the problem that the previous data-driven methods cannot balance the accuracy of SOH estimation and the cost of model calculation,a solution based on feature optimization is proposed.Firstly,several features are extracted based on the partial charging voltage curve and incremental capacity curve.Moreover,the importance of each feature is calculated by the Gini coefficient in the random forest algorithm.Then,the optimal feature subset is selected by considering the estimation accuracy of the model and the feature number of the selected subset.Finally,the random forest algorithm is employed to establish the battery aging model and estimate the SOH.The results show that the mean absolute error and root mean square error of the proposed SOH estimation method are within 0.4%and 0.5%,respectively.Here,the most relevant feature set can be selected by the developed feature optimization strategy.Hence,combined with the random forest algorithm,it can achieve higher SOH estimation accuracy while reducing the calculation cost of themodel.