Research On Remaining Useful Life Prediction Of Lithium Ion Batteries

With the rapid development of complex systems such as aerospace,their complexity is constantly improving.However,the factors that affect their normal operation are also increasing,that is mean,the probability of system go wrong or functional failures is increasing,the cost of maintenance and protection is getting more and more expensive.Therefore,the technology of prognostics and health management gains more and more attention and applications,remaining useful life prediction is the key to PHM technology.It is necessary to predict the remining useful life of lithium-ion batteries,because it is the core unit of many complex systems.Firstly,the paper introduced the working principle of lithium-ion batteries,briefly outlined the basic characteristics of lithium-ion batteries and the factors of battery capacity degradation.Secondly,the RUL prediction of lithium-ion batteries was carried out by using Autoregressive integrated Moving Average(ARIMA)model and Regularized Particle Filter(RPF).Based on the analysis of these two algorithms,according to RPF algorithm over-dependent empirical degradation model and data poor-adaptability,a hybrid prognostics algorithm is proposed based on ARIMA model and RPF algorithm.Compared with PF and RPF prediction algorithms,ARIMA-RPF hybrid algorithm can effectively predict the RUL in the medium and late stages of lithium-ion batteries degradation;also can gives the uncertainty representation of prediction result.In view of the inaccurate problem of RUL prediction by the traditional data-driven algorithm in the early stage of battery degradation,a method for predicting lithium-ion battery RUL based on Long-Short Term Memory(LSTM)model is proposed?Through LSTM model learning lithium-ion batteries capacity degradation path,effectively predict the RUL of lithium-ion batteries in different stages(early,medium and late).In order to verify the prediction accuracy of this algorithm,compared with the ARIMA and ARIMA-RPF prediction algorithm,the results show that LSTM model not only has high prediction accuracy in the medium and late stage of battery degradation,but also has good prediction accuracy in the early stage.