Remaining useful life prediction for lithium-ion batteries using a quantum particle swarm optimization-based particle filter

摘要

Lithium ion batteries are commonly used in portable electronics, spacecraft and electric vehicles due to many relative advantages. As a critical component, its performance is crucial to reliability of the whole system. An accurate prediction of battery life is important to avoid system breakdown. The health status of a lithium ion battery is usually called state-of-health (STH) which depends on the battery's charge capacity which degrades over time. When the charge capacity reaches a threshold value, the battery is considered failed. When STH can be accurately predicted, major breakdowns can be averted. Two types of approaches are available for STH prediction for lithium ion batteries: model- based and data-driven. Model-based approaches use the degradation model to describe the physical nature of degradation. The data-driven approach uses large degradation data to determine the degradation mechanism. The recurrent neural network and autoregressive moving average model are two data driven models used for predicting remaining useful life (RUL) of lithium ion batteries. Liu et al. (Ref. 1) developed a fusion prognostic framework based on the autoregressive model and regularized particle filter (RPF) algorithm to improve RUL prediction. The particle filter (PF) is better in dealing with nonlinear and non-Gaussian systems, however has a major drawback of a particle impoverish problem. The particle swarm optimization algorithm (PSO) improves performance of PF-based approaches, but requires a large amount of computation. A new algorithm called the quantum-behaved particle swarm optimization algorithm (QPSO) performs well and does not need too much computation. QPSO enables the particles to move to optimum regions with a better capacity of searching the global optimum and faster convergence, plus fewer particles and parameters are needed for this application.