Online Estimation of Lithium Polymer Batteries State-of-Charge Using Particle Filter-Based Data Fusion With Multimodels Approach

Affiliation auteurs!!!! Error affiliation !!!!
TitreOnline Estimation of Lithium Polymer Batteries State-of-Charge Using Particle Filter-Based Data Fusion With Multimodels Approach
Type de publicationJournal Article
Year of Publication2016
AuteursZhou D, Zhang K, Ravey A, Gao F, Miraoui A
JournalIEEE TRANSACTIONS ON INDUSTRY APPLICATIONS
Volume52
Pagination2582-2595
Date PublishedMAY-JUN
Type of ArticleArticle; Proceedings Paper
ISSN0093-9994
Mots-clésMultimodels data fusion, particle filter (PF), state-of-charge (SOC), weighted average
Résumé

In this paper, a robust model-based battery state-of-charge (SOC) estimating algorithm is proposed with a novel approach based on combination of multimodels data-fusion technique and particle filter (PF). The proposed method is particularly adapted for SOC estimation under real-time conditions and the presence of measurement noise. In this innovative approach, multiple battery models have been used in order to accurately estimate a battery SOC. During the estimation process, the measured battery terminal voltage is compared with the multiple battery models output to generate individual residual, which is then used to calculate the weight of estimated value from each battery model. This weight, which represents the accuracy of observation equation of each battery model, is inversely proportional to the residual. The estimated SOC values from different models are then fused and the weights of estimated values from each battery model are adjusted dynamically using PF and weighted average methodology, in order to calculate the final SOC estimation of the battery. For each proposed battery model, the corresponding parameter-tuning strategies are also presented. In addition, the proposed method has been validated by experimental results. The results demonstrate that the proposed multimodels-based algorithm can be implemented effectively for real-time application, and achieve better accuracy than single model-based methods.

DOI10.1109/TIA.2016.2524438