Attention Based Mechanism for State-of-Health Estimation of Li-Ion Batteries

To extend the lifespan of Lithium-ion (Li-ion) batteries and optimize their performance, an efficient State-of-Health (SoH) estimation technique is required. Deep learning techniques are capable of handling large datasets and are helpful in long-term predictions. This research proposes a hybridized deep learning architecture that consists of a Convolutional Neural Network (CNN), an attention mechanism, and a Long Short-Term Memory (LSTM) network for precise and efficient SoH prediction. The Pearson correlation coefficient (PCC) is adopted in data pre-processing to choose features that exhibit a strong correlation with each other. The selected parameters are then normalized within the range of 0 to 1 to make the training process fast. Following pre-processing, the CNN-Attention-LSTM model receives all these features as input. Performance indices elucidate a value of mean absolute error (MAE), mean absolute percentage error (MAPE), and root-mean-square error (RMSE) as 0.0039, 0.0059, and 0.0042 respectively and indicate that the proposed model is capable of estimating effective and reliable SoH for Li-ion batteries.