Remaining useful life prediction based on denoising technique and deep neural network for lithium-ion capacitors

Abstract

Lithium-ion capacitor is a hybrid electrochemical energy storage device which combines the merits of lithium-ion battery and electric double-layer capacitor. It is of great importance to monitor the real capacity to evaluate failures of lithium-ion capacitors. Remaining Useful Life (RUL), which is referred to remaining cycle number before reaching its End of Life (EOL) threshold, is a key part in the prognostics and health management and an important indicator of the depletion capacity of lithium-ion capacitor. In this paper, we propose a hybrid neural network which combine with the convolutional neural network and Bidirectional Long Short-Term Memory Network (Bi-LSTM), the data will be used to train this model. Finally, the verifications among different prediction horizons and other methods are discussed. According to the experimental and analysis results, the proposed approach has high reliability and prediction accuracy, which can be applied to battery monitoring and prognostics, as well as generalized to other prognostic applications. • A method for RUL prediction of lithium-ion capacitor was proposed. • CEEMDAN-ICA-thresholding method was proposed to remove noisy signal. • Utilized a hybrid neural network of CNN and BiLSTM for RUL prediction. • The efficiency of the method is verified under different parameters.