An optimized Relevance Vector Machine with incremental learning strategy for lithium-ion battery remaining useful life estimation

In most industrial fields, it needs to evaluate the performance degradation and remaining useful life (RUL) of lithium-ion battery. With the uncertainty representation of the RUL, the Relevance Vector Machine (RVM) becomes an effective approach in lithium-ion battery RUL prognostics. But, the small sample size and low precision of multi-step prediction will bring district to RUL prediction for sparse RVM algorithm. With the on-line monitoring data updating, the dynamic training ability and online algorithm are necessary to improve the prediction precision for battery RUL model. Moreover, the operating efficiency and computing complexity are needed for on-line and real-time processing. A simple and effective on-line training strategy is introduced for RVM algorithm to realize high prediction performance. An incremental optimized RVM algorithm is proposed to achieve efficient online training for model updating. Furthermore, with the on-line training strategy, the prediction precision can increase for battery RUL estimation. Using proposed method, we carry out experiments with NASA battery data and the results show that our method has excellent performance on predicting the RUL of lithium-ion battery.

[1]  Kai Goebel,et al.  Modeling Li-ion Battery Capacity Depletion in a Particle Filtering Framework , 2009 .

[2]  David J. C. MacKay,et al.  Bayesian Interpolation , 1992, Neural Computation.

[3]  Liu Datong,et al.  A review: Prognostics and health management , 2010 .

[4]  Kai Goebel,et al.  Comparison of prognostic algorithms for estimating remaining useful life of batteries , 2009 .

[5]  James Theiler,et al.  Accurate On-line Support Vector Regression , 2003, Neural Computation.

[6]  Enrico Zio,et al.  Particle filtering prognostic estimation of the remaining useful life of nonlinear components , 2011, Reliab. Eng. Syst. Saf..

[7]  George Eastman House,et al.  Sparse Bayesian Learning and the Relevan e Ve tor Ma hine , 2001 .

[8]  Bo-Suk Yang,et al.  Intelligent prognostics for battery health monitoring based on sample entropy , 2011, Expert Syst. Appl..

[9]  Bhaskar Saha,et al.  An Adaptive Recurrent Neural Network for Remaining Useful Life Prediction of Lithium-ion Batteries , 2010 .

[10]  Roger A. Dougal,et al.  Dynamic lithium-ion battery model for system simulation , 2002 .

[11]  Stefan Rüping,et al.  Incremental Learning with Support Vector Machines , 2001, ICDM.

[12]  A. Thakker,et al.  Health monitoring algorithms for space application batteries , 2008, 2008 International Conference on Prognostics and Health Management.

[13]  Liu Qiao,et al.  Automotive battery management systems , 2008, 2008 IEEE AUTOTESTCON.

[14]  Michael Osterman,et al.  Prognostics of lithium-ion batteries based on DempsterShafer theory and the Bayesian Monte Carlo me , 2011 .

[15]  Michael E. Tipping Sparse Bayesian Learning and the Relevance Vector Machine , 2001, J. Mach. Learn. Res..

[16]  Jay Lee,et al.  A review on prognostics and health monitoring of Li-ion battery , 2011 .

[17]  Huan Liu,et al.  Handling concept drifts in incremental learning with support vector machines , 1999, KDD '99.