Estimation of Lithium-ion Battery Health State Using Partial Charge-Discharge Data

Data-driven methods are techniques for estimating the health state of lithium-ion batteries with the widest scope of application, yet the difficulty lies in obtaining complete charge/discharge process data in actual scenarios. In order to increase efficiency in both data utilization and health state prediction, efficient indirect health features of lithium-ion cells are exploited from charge/discharge data based on a time series method. These features stem from partial data within specific voltage ranges, and they are subsequently imported into a Long Short-Term Memory neural network model. The parameters of the model are automatically tuned by the Pelican Optimization Algorithm. This approach achieves high-accuracy health state estimation on different batteries. The work reveals potential for lithium-ion battery health state prediction with fragmented data.