Abstract:

In order to prevent safety accidents of lithium batteries due to remaining battery life (RUL) and improve the RUL prediction accuracy of lithium batteries,the lithium battery RUL prediction model is proposed based on improved particle swarm optimization (IPSO) algorithm with the leaky integrator echo state network (LIESN).Firstly,its local and global optimization-seeking ability is improved by improving the inertia weights and learning factor update rules.Next,the LIESN network parameters are optimized by IPSO to establish the degradation prediction model.The simulation experiments are conducted using the NASA’s public experimental data of lithium batteries.The results show that IPSO-LIESN has higher prediction accuracy,stability and generalization ability compared with prediction methods such as the fusion of particle fitter and Gaussian process regression (PF-GPR),indirect health indicators and echo state network (ESN) under the same data set conditions,which can demonstrate the effectiveness of the proposed method.

Key words: lithium-ion battery, particle swarm algorithm, leaky echo state network, remaining useful life