Cross-Domain State-of-Charge Estimation of Li-Ion Batteries Based on Source Domain Selection and Attention Mechanism

Abstract

In the domain of lithium-ion (Li-ion) battery state-of-charge (SOC) estimation, deep neural network models commonly assume a congruent distribution between training and testing data. Nonetheless, this assumption often proves inadequate in real-world scenarios, due to variations in environmental temperature, aging levels, and operational conditions. To tackle this challenge, this study proposes a novel approach centered around a deep transfer network, incorporating source domain selection and an attention mechanism, for the task of cross-domain SOC estimation. This approach leverages a transfer network to extract bidirectional temporal features and accentuate salient information within sequences. The selection of an appropriate source domain for pretraining is contingent upon establishing domain similarity between the source and target domains. Experimental results demonstrate that the proposed method excels at feature extraction from Libs sequence data, yielding enhanced performance even when confronted with limited data.