Lithium-Ion Battery Remaining Useful Life Prediction Based on Hybrid Model

Accurate prediction of the remaining useful life (RUL) is a key function for ensuring safety and stability lithium-ion batteries. To solve capacity regeneration model adaptability under different working conditions, hybrid RUL based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) bi-directional gated recurrent unit (BiGRU) proposed. CEEMDAN used to divide into intrinsic functions (IMFs) reduce impact regeneration. In addition, an improved grey wolf optimizer (IGOW) proposed maintain reliability BiGRU network. The diversity initial population in GWO algorithm was using chaotic tent mapping. An control factor dynamic weight are adopted accelerate convergence speed algorithm. Finally, experiments conducted verify battery performance training data conditions. results indicate that method can achieve MAE less than 4% only 30% set, which verified CALCE NASA data.