Prognostics Of Li-ion Batteries Using Swarm Intelligence Optimization Particle Filtering

Lithium-ion batteries have played a vital role in all walks of life.Sudden failure or damage can cause significant loss to property even to personal safety.The regular maintenance formerly has not satisfied the requirements of stability and reliability to lithium-ion batteries,so the concept of predictive maintenance is drawing more attention.As the core technology of the predictive maintenance,remaining life prediction has been a very hot research topic.Remaining life prediction can predict the residual life of batteries before their life are ended.That provides much more time for the maintenance personnel to make property maintenance strategy.Particle filter algorithm has been widely used in the study of lithium batteries remaining life prediction.In order to improve the defect of particle filter used in the lithium-ion batteries' residual life prediction,the following work has been done in this paper.The precision estimation of parameters is not high when the particle filter is used in the prediction of lithium-ion batteries' residual life because of the loss of effective samples caused by the particle degradation.The inaccurate parameters finally lead to the decline of the accuracy of prediction.This paper introduces the swarm intelligence optimization algorithms to the particle filter aim at the inaccuracy problem of prediction.Using the intelligent optimization strategies of swarm intelligence optimization algorithms to optimize the particles of particle filter in order to drive them to get closer to the true state and increase the amount of the effective particles.So that the degradation problem is solved and the estimation of parameters is increased either.Finally the inaccuracy problem of prediction is improved effectively.Firstly this paper integrate the particle swarm optimization algorithm into the particle filter algorithm.Using the updating method of location and velocity of the particle swarm optimization algorithm to optimize the particles sampled in the particle filter.The particles optimized have better locations and bigger weights and are more effective than that are not been optimized.Secondly this paper introduce the artificial fish swarm algorithm into the particle filter.Using the foraging behavior and cluster behavior to actuate the particles closer to the high likelihood area so that the estimation performance of particle filter is improved.In order to validate the two algorithms proposed,the data of lithium-ion collected by the University of Maryland is used.The simulation results show that the proposed algorithms are more effective than the standard particle filter algorithm when they are used to predict the residual life of lithium-ion batteries.