Harnessing CNN-DNC and CNN-LSTM-DNC Architectures for Enhanced Lithium-Ion Remaining Useful Life Prediction

This research explores the application of the Differentiable Neural Computer (DNC), a Neural Turing Machine model, in predicting the Remaining Useful Life (RUL) of Lithium-Ion Batteries, especially within Renewable Energy Systems and electric vehicle domains. The introduction of two novel models, CNN-DNC and CNN-LSTM-DNC, marks a significant advancement in this task. An extensive evaluation on a dataset comprising 124 Lithium-Ion Battery cells highlights the CNN-DNC’s exemplary performance, delivering an MAE of 80.133 cycles and a loss value of 0.0037, outperforming the CNN- LSTM-DNC which reported an MAE of 99.028 cycles and a loss value of 0.0064. Despite utilizing 95.2% fewer parameters and achieving a 19.1% improvement in prediction accuracy, the CNN-DNC demands more training time compared to the CNN- LSTM-DNC. This comprehensive study not only highlights the potential of integrating DNC for precise RUL predictions but also sets a pathway for future research. The emphasis on enhancing these promising models and testing their practical applicability aims to bolster the reliability and efficiency of Lithium-Ion Batteries in various critical applications.