Aging model development based on multidisciplinary parameters for lithium‐ion batteries. (29th December 2019)

• Record Type: Journal Article
• Authors: Garg, Akhil; Shaosen, Su; Gao, Liang; Peng, Xiongbin; Baredar, Prashant
• Is Part Of: International journal of energy research. Volume 44:Number 4(2020)
• Page Start: 2801
• Page End: 2818
• Publication Date: 2019-12-29
• Subjects: battery aging model -- battery management system -- diffusion coefficient -- energy conversion -- genetic programming; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Power resources -- Research -- Periodicals; 621.042
• DOI: 10.1002/er.5096
• Languages: English
• ISSNs: 0363-907X

Abstract

Summary: In this paper, a method composed of state of health (SOH) testing experiments and artificial intelligence simulation is proposed to carry out the study on the change of battery characteristic during its operation and generate mathematical models for the prediction of aging behaviour of battery. An experiment comprising of multidisciplinary parameters‐based SOH detection is conducted to study the battery aging characteristics from several aspects (ie, electrochemistry, electric, thermal behaviour and mechanics). In total, 200 sets of data (corresponding 200 charging/discharging cycles) are collected from the experiment. The data obtained from the first 150 cycles are employed in generation of the models. The result of sensitivity analysis based on the obtained genetic programming models shows that it is better to apply voltage value at the end of charging step, charging time and cycle number to predict the operational performance of the battery. The average predicted accuracy of model (without stress) is 94.52%, whereas the average predicted accuracy of model (with stress effect) is 99.42%. The proposed models could be useful for defining the optimised charging strategy, fault diagnosis and spent batteries disposal strategies. Abstract : The state of health of lithium‐ion battery reflects the performance of hybrid energy systems and electric vehicles. A battery aging model which can monitor and predict the status of batteries is vital for optimising the optimised charging strategy, fault diagnosis and spent batteries disposal strategies. Experiments comprising of multidisciplinary parameters based state of health detection were conducted. Multidisciplinary parameters comprise of parameters from several aspects such as chemical (number of Li‐ions), electrochemical (cyclic voltametry, diffusion coefficient), electrical (internal resistance), thermal (temperature) and mechanical (stack stress). In addition, considering the aging effects, it was found that the battery capacity was largely depended on battery voltage and charging time. Based on the validation results, the average predicted accuracy of model (without stress effect) was 94.52%, whereas the average predicted accuracy of model (measuring under specific stress value) was 99.42%. It can be thus concluded that the stack stress should be considered as an input parameter during the formulation process of battery aging model. Moreover, several future research directions are summarised in the end.