State of health estimation of lithium-ion batteries based on TiDE

Given the widespread use of lithium-ion batteries, accurately forecasting their State of Health (SOH) is crucial for ensuring the secure and reliable operation of equipment. The local capacity regeneration during battery degradation can undermine prediction accuracy. This research introduces an MDT-FOA+TiDE approach for predicting SOH in lithium-ion batteries using multidimensional time series data. Initially, capacity and various feature data from the degradation process were collected and health indicators were derived. Features with high correlation to capacity were integrated into multidimensional time series. Subsequently, the Time-series Dense Encoder (TiDE) model was employed for training and prediction, while the Fox Optimization Algorithm (FOA) was used for model optimization. Experimental results using the NASA dataset demonstrate that the proposed approach outperforms SVR and LSTM models that utilize univariate data and MDT-TiDE model without FOA optimization.