A novel remaining useful life prediction method based on gated recurrent unit network optimized by tunicate swarm algorithm for lithium-ion batteries

Abstract

Lithium-ion batteries have a wide range of applications in the field of new energy vehicles with advantages including small size, high efficiency, and low pollution. Accurate prediction of the remaining useful life (RUL) of lithium-ion batteries is particularly important to reduce unintended maintenance and avoid safety incidents caused by battery aging. To improve the accuracy and robustness of the RUL prediction of lithium-ion batteries, a prediction method is proposed based on gated recurrent unit (GRU) network optimized by tunicate swarm algorithm (TSA) in this paper. First, the capacity data during the life cycle of the aging battery are extracted as the prediction feature. Then, the GRU network is used to capture the dependencies between degraded capacities for RUL prediction. The main hyperparameters in the GRU network are optimized by the TSA to maximize the prediction performance. Finally, to ensure the validity and generalizability of the proposed RUL prediction method, data sets from University of Maryland, National Aeronautics and Space Administration (NASA), and our own laboratory are selected for validation. The superiority of the proposed method is verified by comparison with other different prediction methods.