Estimating Lithium-Ion Battery Health Parameters Using Deep Learning Techniques

Prognostics and health management (PHM) application is a viable strategy to proactively manage and reduce the risks of failure. PHM has been widely utilized in various applications, including battery systems. Battery capacity is a critical parameter that requires careful assessment to ensure safe operation and informed decision-making. Quantifiable data from the battery management system, such as the current (I), voltage (V), and temperature (T) profiles, which fluctuate with battery aging, are used to anticipate failure. Using this information, neural networks may learn the relationship between capacity and charging profiles. In the present work, a framework, utilizing multi-channel convolutional neural networks (CNN), long short-term memory (LSTM), and multi-channel CNN-LSTM is proposed to accurately predict the status of health that incorporates V, I, and T profiles. The proposed multi-channel CNN-LSTM strategy outperforms the other two techniques, achieving a mean absolute percentage error reduction of up to 25%-58%. The potential of utilizing multi-channel data for accurate PHM of batteries can be concluded from the present work.