Enhancing State-of-Health Estimation Through Deep Architecture Modification of LSTM Networks

An accurate assessment of the State-of-Health(SoH) of Li-ion batteries is crucial in ensuring their reliability, safety, and longevity. However, traditional SoH estimation methods often struggle to identify the complex degradation patterns that batteries exhibit over their lifespan. To address this issue, an innovative approach that involves the modification of the architecture for deeper learning and comprehensive utilization of battery capacity, cycle, and SoH data based on Long Short-Term Memory(LSTM) networks was proposed. Through meticulous data preprocessing, we will bridge the gap between raw data and meaningful insights, thus facilitating a transformative shift in batteiy health assessment. Our model adeptly captures the cumulative effects of repeated charge-discharge cycles, ensuring accurate predictions over an extended battery lifespan. The proposed model showed significant improvement by using the NASA battery aging dataset, resulting in 39.1% and 69.35 % accuracy when employing 50 % and 70 % training data, respectively. The observed exceptional accuracy highlights the effectiveness of our approach by addressing the complexities of battery degradation.