Premium
Models based on mechanical stress, initial stress, voltage, current, and applied stress for Li‐ion batteries during different rates of discharge
Author(s) -
Cui Xujian,
Kam Shi Khai,
Chin Christina May May,
Chen Jihong,
Babu Chitti,
Peng Xiongbin
Publication year - 2020
Publication title -
energy storage
Language(s) - English
Resource type - Journals
ISSN - 2578-4862
DOI - 10.1002/est2.126
Subject(s) - voltage , stress (linguistics) , artificial neural network , battery (electricity) , stack (abstract data type) , lithium ion battery , work (physics) , materials science , computer science , engineering , electrical engineering , power (physics) , machine learning , mechanical engineering , philosophy , linguistics , physics , quantum mechanics , programming language
The most important criteria for any energy storage system such as the Li‐ion batteries are its capacity fading or the state of health (SOH). In real time, the parameters such as voltage, current cannot be used to predict SOH because these are not taken into account the self‐discharge. This article proposes experimental combined numerical methodology for studying coupled stress‐electrochemical performance of Li‐ion batteries. The work aims to evaluate and predict the SOH of lithium‐ion batteries based on mechanical stress, number of charging cycles, and induced load. Experiments are conducted to measure data corresponding to capacity, initial stress, and applied stress. Artificial neural networks are then applied in formulation of predictive models based on initial stress, stack stress, charging voltage, and discharging voltage. A neural net was successfully trained that managed to achieve correlation coefficient (prediction accuracy) of 0.9909 for capacity and 0.7260 for cycle number. This research was able to identify an ideal network configuration, predicting cycle number, and remaining capacity of a battery after multiple charges, trained from the given data values.