Deep feature extraction in lifetime prognostics of lithium-ion batteries: Advances, challenges and perspectives

The wide application of lithium-ion batteries makes their lifecycle prognosis a challenging and hot topic in the battery management research field. Feature extraction is a key step for the lifetime prognostics of lithium-ion batteries, which takes a significant effect on the accuracy of performance prognosis. More relevant and useful feature inputs would certainly bring more accurate predictions. While, the deep learning technology has great advantages in feature extraction. Being of practical feasibility, the combination of battery management and deep learning technology, e.g. deep feature extraction in battery lifetime prognostics, promise it a wide application in the future. To fully understand the deep feature extraction in lifetime prognostics of lithium-ion batteries, existing investigations on it are summarized, analyzed and concluded in the current review. The sources and searching methods of the literature are introduced first. Commonly used deep learning methods and their variants are reviewed for the feature extraction. Their applications in lithium-ion battery prognostics are then introduced in detail. On this basis, the existing problems in this research field are investigated, the challenges are analyzed and summarized, and the future research works are proposed finally.