RMDA: A Regressive Multiple-Source Domain Adaption Approach for Early Prediction of Lithium-ion Battery Lifetime

Early prediction of battery lifetime is crucial for manufacturers to sort and group lithium-ion batteries (LIBs). This article introduces a regressive multiple-source domain adaption (RMDA) method for accurate early lifetime prediction with multiple LIB datasets. First, a common feature extractor is used to extract common features from different domains. Then, the domain-specific extractors followed by a maximum mean discrepancy loss are utilized to reduce the shift of source domains. Next, the domain-specific regressors are adopted to predict the lifetime of the source domains and the target domain. The effectiveness of the proposed RMDA method is validated on three LIB datasets with different chemistries and operating conditions.