Experiments on battery capacity estimation

Modern life heavily relies on the continuous and stable power supply. All kinds of devices and systems are driven by batteries. The behavior of battery has direct impact on the operation and performance of those devices and systems. Thus, the knowledge of state of health and remaining useful life will facilitate the proper use and management of batteries. In this study, battery capacity is estimated with selected machine learning algorithms. Three strategies for using the training data are proposed. Experiments were carried out with the data from Lithium-ion batteries undergoing accelerated aging process through repeated charge and discharge cycles. The preliminary results demonstrate the feasibility of these machine learning approaches.