State-of-Health Prediction of Lithium-Ion Batteries Using Exponential Smoothing Transformer With Seasonal and Growth Embedding

Abstract

In the world of modern energy, Lithium-Ion batteries reign supreme, offering rechargeability, sustainability, and long-term energy storage. However, their lifespan is not infinite, calling for accurate prediction of remaining life under various conditions. Deep learning shines in this domain, with the Transformer architecture blossoming as a powerful tool for time series forecasting. This research dives into data collection, processing, model design, training, and evaluation, making key methodological contributions to battery life prediction. Notably, the SGEformer model, a Transformer enhanced with growth and seasonal embedding, emerges as a groundbreaking innovation. Comparing SGEformer to ETSformer, Informer, Reformer, Transformer, and LSTM reveals its unique strengths. With an impressive MSE score of 0.000117, SGEformer establishes itself as a highly effective tool for battery life prediction, highlighting the value of growth and seasonal embedding in boosting accuracy. This research propels the state-of-the-art Lithium-Ion battery state-of-health prediction, offering a robust methodological foundation for precise and reliable forecasts.