Performance Degradation Characteristics Extraction of Lithium-ion Batteries Based on Convolutional Neural Networks

Currently, lithium-ion batteries, with their high energy density and memoryless effect, are comprehensively applied to the fields of aerospace, new energy vehicles, and mobile terminal devices. However, the performance degradation caused by capacity loss will weaken the efficiency of lithium-ion batteries, and even cause property damage and casualties in serious cases. Battery capacity is a physical quantity that reflects the remaining useful life of the battery, but collecting the capacity data of battery directly is costly and has a long cycle. Therefore, this paper proposes a Performance Degradation Characteristics (PDCs) extraction method to find PDCs that can indirectly and effectively reflect the decline of battery capacity. First, a characteristic extraction method based on Convolutional Neural Network (CNN) is proposed to extract PDCs from original battery data, which can automatically learn the characteristic representation, and possesses high extraction efficiency and strong generalization ability. Second, Nine PDCs, including seven Measurable Characteristics (MCs) and two Computational Characteristics (CCs), are extracted to describe the battery capacity decline validly. Finally, The feasibility and effectiveness of the extraction method are verified by using the nine PDCs to predict the remaining using life of lithium-ion batteries based on a series of CNN models.