基于CPSO-RVM的锂电池剩余寿命预测方法

Approach for Lithium-ion Battery RUL Prognostics Based on CPSO-RVM

针对锂电池健康状况测量数据中经常伴随着各种类型及大小的噪声,本文提出了一种基于小波去噪和混沌粒子群-相关向量机的锂电池剩余寿命预测方法。执行小波二次降噪,削弱测量数据中的大噪声信号及消除测量数据中的小噪声信号,从而提取原始数据;将经混沌粒子群算法优化的相关向量机算法用于估计锂电池各个充放电周期健康状况的变化轨迹,并预测锂电池的剩余寿命。基于美国国家航空航天局提供的锂电池测量数据,对提出的方法进行了有效性验证。

On account of the measured battery state of health （SOH） data are often subject to different levels of noise pollution, a battery remaining useful life （RUL） prognostics approach is presented based on wavelet denoising and CPSO-RVM in the paper. Wavelet denoising is performed twice with different thresholds in order to weaken the strong noise and remove the weak noise. Relevance vector machine （RVM） improved by chaos particle swarm optimization （CPSO） algorithm is utilized to estimate the trend of battery SOH variation trajectory and predict the battery RUL based on the denoised data. RUL prognostic experiments using battery data provided by NASA are conducted and the effectiveness of the presented approach is validated.