The state of health (SOH) of a Li-ion battery is a critical parameter that determines the remaining capacity and the remaining lifetime of the battery. In this paper, we propose SambaMixer a novel structured state space model (SSM) for predicting the state of health of Li-ion batteries. The proposed SSM is based on the MambaMixer architecture, which is designed to handle multi-variate time signals. We evaluate our model on the NASA battery discharge dataset and show that our model outperforms the state-of-the-art on this dataset. We further introduce a novel anchor-based resampling method which ensures time signals are of the expected length while also serving as augmentation technique. Finally, we condition prediction on the sample time and the cycle time difference using positional encodings to improve the performance of our model and to learn recuperation effects. Our results proof that our model is able to predict the SOH of Li-ion batteries with high accuracy and robustness.
Battery state of health (SoH) is a vital indicator of its performance and longevity, making accurate SoH predictions crucial for effectively managing and maintaining battery-powered systems. Deep learning (DL) frameworks enhance SoH prediction accuracy by learning complex patterns from large datasets. Transformer architecture is particularly beneficial for processing sequential data, capturing temporal dependencies within battery data, and improving SoH estimation reliability. In this research, we introduce a novel dynamic model that integrates nonlinear auto-regressive with exogenous (NARXs) inputs into a transformer encoder-decoder architecture. This model aims to predict long-term battery SoH and end-of-life (EoL), focusing on multiple cycles and the dynamic behavior of SoH changes. We evaluated our model using a unique half-half experiment, showing that it reduces the mean absolute percentage error in long-term predictions to 1.4%. This method enables accurate battery health and EoL predictions in practical applications, showing robustness against measurement noise.