Online accurate state of health estimation for battery systems on real-world electric vehicles with variable driving conditions considered

The environmental sustainability stimulates the development of electric vehicles with great energy saving and emission reduction effects. State of health of the battery system in an electric vehicle is crucial to the safety of vehicle operation, charging station, and the environment. The existing techniques implemented in well-controlled experimental environments fail to learn unpredictable drivers' driving behaviors and complex road/weather conditions during actual vehicular operation. This paper investigates a novel deep-learning-enabled method to perform accurate state of health estimation for battery systems on real-world electric vehicles. Eight potential evaluation schemes depending on the stable charging stages are recapped and discussed. By fitting the correlation between battery degeneration factors and various vehicle operation parameters such as ambient temperature and mileage, an approximate battery degeneration model oriented for the real application scenarios is obtained. The variable-length-input long short-term memory network is used to learn the variable battery degeneration factors acquired from different driving stages of a yearlong dataset. The test results show that the proposed method has a better performance than other estimation methods. More significantly, based on the acquisition advantages of big-data platforms, it can be used to full-state and full-climate vehicle applications unrestricted by complex actual environments. (c) 2021 Elsevier Ltd. All rights reserved.

Automated summary

This study investigates a novel deep-learning-enabled method for accurate estimation of battery system health on real-world electric vehicles. By fitting the correlation between battery degeneration factors and various vehicle operation parameters, an approximate battery degeneration model is obtained. The proposed method demonstrates better performance compared to other estimation methods.