An improved data-driven approach for NMC battery remaining useful life prediction 

The transition to sustainable electric mobility requires significant advances in the accurate prediction of NMC battery capacity, a critical aspect in ensuring the reliability and durability of electric vehicles. In this study, a hybrid version of Convolutional Neural Network (CNN), Temporal Convolutional Network (TCN) and Deep Neural Network (DNN), to accurately predict the battery’s capacity, named CNN-TCN-DNN, is designed. There are two layers in the proposed learning framework: CNN and TCN-DNN. The CNN layer is employed to extract features from the original battery datasets, and TCN-DNN is used to generate the final estimated capacity by leveraging the characteristics extracted by CNN. The proposed technique uses battery’s voltage, charge/discharge currents under ambient temperature to estimate capacity. An experimental validation was carried out, based on the NMC battery data sets, using two topologies: serial and parallel. Finally, the effectiveness of the proposed methods are proven by a comparative study with other advanced techniques. The results show that our approaches outperforms other state-of-the-art techniques in terms of prediction accuracy.