The State of Charge Predication of Lithium-Ion Battery Using Contrastive Learning

The State of Charge (SOC) plays a crucial role as an indicator of the current energy level in lithium-ion batteries. However, obtaining the precise value of SOC is challenging due to it being a hidden state quantity. Existing neural network models commonly employ an end-to-end prediction paradigm for SOC estimation, which fails to fully exploit the rich information present in the time-series battery data. To address this limitation, this paper developed a new SOC prediction method utilizing contrastive learning named CLDMM. The proposed approach utilizes data augmentation, multi-scale encoder, and multi-layer perceptrons to learn latent representations, which are subsequently employed for downstream predictive tasks. The Panasonic NCR18650PF dataset is used to evaluate the performance of the proposed method, and the results of experiments demonstrate that CLDMM outperforms the baseline methods and achieves an average mean absolute error (MAE) of 0.73%, and an average maximum error (MAX) of 2.54%.