Prognostics of Lithium ion battery using functional principal component analysis

摘要

Lithium ion batteries are widely used for energy storage. Its capacity degradation modeling and cycle to failure estimation has become very significant. Repeated capacity measurements over the whole life of batteries, i.e., the longitudinal data, are investigated to understand the degradation process and the cycle to failure behavior. This paper proposes a method for prognostics of lithium ion batteries based on the functional principal component analysis. The observed degradation signal is decomposed into the mean function and variance-covariance function. The local quadratic smoothing method is used to estimate the mean function. Functional principal components of the variance-covariance function are represented and modeled through eigenfunctions, which are further approximated and estimated using a combination of B-Splines. For the battery capacity prognostics, three eigenfunctions explained 99.98% of the total variation. Capacity prediction and cycle to failure distribution are also analyzed and evaluated based on the proposed method.