IEEE Access (Jan 2017)

Capacity Prognostics of Lithium-Ion Batteries using EMD Denoising and Multiple Kernel RVM

  • Chaolong Zhang,
  • Yigang He,
  • Lifeng Yuan,
  • Sheng Xiang

DOI
https://doi.org/10.1109/ACCESS.2017.2716353
Journal volume & issue
Vol. 5
pp. 12061 – 12070

Abstract

Read online

Lithium-ion batteries are crucial to many types of electric equipments. Hence, lithium-ion battery capacity prognostic is significantly important, and it is yet very hard for the measured battery data that are regularly polluted by miscellaneous noises. In this paper, a battery capacity prognostic approach using the empirical mode decomposition (EMD) denoising method and multiple kernel relevance vector machine (MKRVM) approach is presented. The EMD denoising method is employed to process the measured capacity data to produce noise-free capacity data. The battery capacity prediction model using MKRVM is constructed based on the denoised capacity data. The MKRVM's kernel keeps diversity by using multiple heterogeneous kernel learning method. Meanwhile, sparse weights of basic kernel functions are yielded by using particle swarm optimization (PSO) algorithm. The measured battery capacity data are used to demonstrate the effect of EMD denoising method, and battery capacity prediction experiments reveal that the proposed MKRVM approach can predict the battery's future capacity precisely.

Keywords