State-of-Health Estimation of Lithium Titanate Batteries Under Locomotive Conditions With Meta-Learning Based on Long Short-Term Memory Network

Accurate state-of-health (SOH) estimation of lithium titanate batteries as power sources is essential for the safe operation of locomotives. The existing achievements few focus on the SOH of lithium titanate batteries under locomotive operating conditions. This article explores a novel meta-learning based on a long short-term memory (LSTM) network to estimate the SOH of lithium titanate batteries under locomotive varying conditions. First, the simulated operating condition of locomotives is designed. Then, among the various health indicators (HIs) that affect the accuracy of SOH estimation, three key indicators are selected. Further, the meta-learning based on LSTM (Meta-LSTM) network is proposed, where the meta-learning performance is independent of the similarity between the pretrained and target data. Experimental results reveal that the selected HIs are suitable for characterizing the decay of lithium titanate batteries. Based on the CALCE public dataset, designed 5C full charge–discharge experimental data and simulated locomotive operating condition data, the results show that the accuracy of SOH estimation is enhanced based on Meta-LSTM, especially in small sample scenarios, when the training set to test set ratio is 1:9 and 2:8, the root mean square error (RMSE) is 2.33% and 3.58%, respectively, demonstrating the effectiveness of Meta-LSTM for lithium titanate batteries under locomotive conditions.