iTransformer Network Based Approach for Accurate Remaining Useful Life Prediction in Lithium-Ion Batteries

Electric vehicles (EVs) are paving the way toward a sustainable future by reducing carbon footprints and gaining widespread global acceptance; predicting the EV battery’s remaining useful life (RUL) is crucial. As the Li-ion batteries degrade and lose lithium and active material, managing the battery’s state of health and charge is necessary to prevent it from reaching its End of Life (EOL). This paper explores a unique application of the iTransformer neural network for RUL prediction using multi-channel charging profiles. This implementation adopts multivariate forecasting, taking advantage of mapping the high-dimensional features to low-dimensional spaces by using the inverted self-attention mechanism to learn the distinctions and interactions between the time-series data. The iTransformer is applied to two popular open-source Li-ion battery datasets: NASA and CALCE. This technique is compared against existing techniques like the Long Short-Term Memory (LSTM) and Vanilla Transformer. The proposed iTransformer improves the root mean square error (RMSE) by 44.62% and 29% for the NASA and CALCE datasets, respectively, compared to the next best-evaluated methods.