Lithium-Ion battery remaining useful life prediction method concerning temperature-influenced charging data

Developing the remaining useful life (RUL) prediction technology for lithium-ion batteries can effectively provide information for battery maintenance and diagnosis. Although there has been some development in battery RUL prediction methods like model-based methods and data-driven methods, the influence of temperature on battery system is rarely considered. Besides, in the actual operation of the battery, the data used for RUL prediction is limited. Aiming at these problems, this paper proposed an integrated battery RUL prediction method. This method establishes a temperature correction formula for health indicators and uses neural network (NN) technique to determine the SOH of the battery. Unscented particle filter (UPF) is used with an empirical model to predict when the battery reaches the end-of-life (EOL) point. At last, the RUL value is calculated. Experiment data from the Sandia National Lab is deployed for NN training and method verification, and the results show that the proposed method has high accuracy than the conventional method.