Fault Diagnosis and Failure Prognostics of Lithium-ion Battery based on Least Squares Support Vector Machine and Memory Particle Filter Framework

摘要: 123456A novel data driven approach is developed for fault diagnosis and remaining useful life (RUL) prognostics lithium-ion batteries using Least Square Support Vector Machine (LS-SVM) Memory-Particle Filter (M-PF). Unlike traditional data-driven models capacity failure prognosis, which require multidimensional physical characteristics, the proposed algorithm uses only two variables: Energy Efficiency (EE), Work Temperature. The aim of this framework to improve accuracy incipient abrupt faults prognosis. First, LSSVM used generate residual signal based on fade trends Li-ion batteries. Second, adaptive threshold model several factors including input, output error, disturbance, drift parameter. tackle shortcoming a fixed threshold. Third, M-PF as new method prognostic determine Remaining Useful Life (RUL). assumption availability real-time observation historical data, where can be instead within particle filter. feasibility validated battery obtained from National Aeronautic Space Administration (NASA) Ames Prognostic Center Excellence (PCoE). experimental results show following: (1) fewer dimensions input are required compared empirical models; (2) diagnostic provides an effective way diagnosing fault; (3) predict RUL with small has high prediction accuracy; and, (4) shows that in