A novel Gaussian model based battery state estimation approach: State-of-Energy. (1st August 2015)

• Record Type: Journal Article
• Authors: He, HongWen; Zhang, YongZhi; Xiong, Rui; Wang, Chun
• Is Part Of: Applied energy. Volume 151(2015:Aug. 01)
• Page Start: 41
• Page End: 48
• Publication Date: 2015-08-01
• Subjects: Electric vehicles -- Lithium-ion battery -- State of energy -- Gaussian model -- Akaike information criterion -- Central difference Kalman filter; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2015.04.062
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: The Gaussian model is employed to construct a novel battery model. The genetic algorithm is used to implement model parameter identification. The AIC is used to decide the best hysteresis order of the battery model. A novel battery SoE estimator is proposed and verified by two kinds of batteries. Abstract: State-of-energy (SoE) is a very important index for battery management system (BMS) used in electric vehicles (EVs), it is indispensable for ensuring safety and reliable operation of batteries. For achieving battery SoE accurately, the main work can be summarized in three aspects. (1) In considering that different kinds of batteries show different open circuit voltage behaviors, the Gaussian model is employed to construct the battery model. What is more, the genetic algorithm is employed to locate the optimal parameter for the selecting battery model. (2) To determine an optimal tradeoff between battery model complexity and prediction precision, the Akaike information criterion (AIC) is used to determine the best hysteresis order of the combined battery model. Results from a comparative analysis show that the first-order hysteresis battery model is thought of being the best based on the AIC values. (3) The central difference Kalman filter (CDKF) is used to estimate the real-time SoE and an erroneous initial SoE is considered to evaluate the robustness of the SoE estimator. Lastly, two kinds of lithium-ion batteries are used to verify the proposed SoE estimation approach. The results show that the maximum SoE estimation error is within 1% for both LiFePO4 and LiMn2 O4 battery datasets.