一种预测锂电池剩余寿命的改进粒子滤波算法

粒子滤波算法本身存在着粒子退化问题,对于衰减趋势变化剧烈的模型,难以获得精确的预测结果,限制了算法的适用范围。针对以上问题对粒子滤波进行改进,通过引入粒子群优化算法中的粒子更新机制,优化粒子的全局位置信息,进而重新分配各粒子权重,降低了重采样阶段粒子重置的比例,改善了算法固有的粒子退化现象,达到改进算法、提升算法预测性能的目的;同时,为验证算法的实际效果,以马里兰大学先进寿命周期工程中心(CALCE)发布的锂电池容量实验数据集为基础,分别使用传统粒子滤波算法与改进的算法进行剩余寿命预测仿真。经过对比发现:改进算法误差下降33.6%,可获得更为精确的预测结果,重采样率下降18.3%,粒子退化问题得到改善。

An Improved Particle Filter Algorithm for Lithium-Ion Battery Remaining Useful Life Prediction

The capacity degradation of Lithium battery is influenced by excessive factors and the mechanism is complex, and the remaining useful life (RUL) is very difficult to predict. The particle filter (PF) is one of the main stream methods in the current RUL prediction research because of its excellent non-linear non-Gaussian characteristics. However, the PF has a problem of particle degeneration in nature, which weakens the precision especially when the model has a dramatic trend of changes. In order to overcome the above problems, the PF is improved by introducing the renewal philosophy of particle swarm optimism (PSO) to reassign particle weight by optimizing the global position of particle. The simulation results reveal that compared with the original PF, a more precise prediction of the RUL can be obtained with the 33.6% reduction of error, and an alleviation of particle degeneration is reached for the 18.3% reduction of resampling rate.