SambaMixer: State of Health Prediction of Li-ion Batteries using Mamba State Space Models
José Ignacio Olalde-Verano, Sascha Kirch, Clara Pérez-Molina, Sergio Martin
2024-11-04
Summary
This paper introduces SambaMixer, a new model for predicting the state of health (SOH) of lithium-ion batteries. It uses advanced techniques to analyze battery performance and helps determine how much longer a battery can last.
What's the problem?
Understanding the health of lithium-ion batteries is crucial because it affects how well they work and how long they will last. Current methods for predicting battery health often rely on complex calculations that can be inaccurate. Additionally, they usually require extra data, which can complicate the process.
What's the solution?
SambaMixer is a structured state space model that improves the prediction of battery health by using a new approach called anchor-based resampling. This method ensures that the data used for predictions is consistent and helps improve accuracy. The model also incorporates positional encodings to better understand how battery performance changes over time. The authors tested SambaMixer on a NASA dataset and found that it outperformed existing models in accuracy and reliability.
Why it matters?
This research is important because it provides a more accurate way to assess the health of lithium-ion batteries, which are used in many devices like smartphones, laptops, and electric vehicles. By improving battery monitoring, this model can help extend battery life and enhance safety, making it valuable for consumers and manufacturers alike.
Abstract
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.