Optimized deep learning strategy for estimation of state of charge at different C-rate with varying temperature

• Published in: Electrical engineering Vol. 105; no. 6; pp. 3853 - 3860
• Main Authors: Kumari, Pooja, Singh, Ashutosh Kumar, Kumar, Niranjan
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.12.2023; Springer Nature B.V
• Subjects: Batteries; Deep learning; Economics and Management; Electric charge; Electric vehicles; Electrical Engineering; Electrical Machines and Networks; Energy consumption; Energy Policy; Energy storage; Engineering; Error analysis; Estimation; Management systems; Mathematical analysis; Onboard equipment; Original Paper; Power Electronics; Power management; State of charge; Deep learning; Lithium ion battery; Battery management system; Electric vehicle; State of charge
• ISSN: 0948-7921; 1432-0487
• DOI: 10.1007/s00202-023-01925-0

Due to advancements in e-mobility technology, more and more people are opting to use electric vehicles. As there is a lack of a backup power supply, energy storage device is especially important for onboard systems. So, battery management systems become an important aspect for these energy storage devices. State-of-charge (SoC) estimation is a vitally significant assessment index in BMS because it is one of the most critical attributes that represents the working state of power batteries in EVs. In addition to the quick display of the remaining battery capacity to the user, accurate knowledge of SoC exerts further control over the charging/discharging process, which may be used to enhance battery life. This can be done in order to increase the longevity of the battery. According to the findings of this study, the error while estimating the SoC has been reduced to near about zero. The effect of resistance, temperature and C-rate on SoC has been also considered in this study. These findings show that the resistance is proportional to SoC below 80%, and after that, it becomes nonlinear. The SoC calculation using optimized deep learning strategy is proposed in this study. This approach helps to limit the margin of error in the calculation of SoC. It has also taken into account the occurrence of changes in SoC as a result of the changing C-rate, temperature, and resistance. Hence, this study helps to offset the negative effects of inaccurate SoC prediction.