Research On Remaining Useful Life Prediction Method For Lithium-ion Battery Based On Data Driven

Lithium-ion batteries,as the main power supply or energy storage components,are widely used in electric vehicles,military equipment,aerospace,power stations and other fields.However,the performance of the battery is gradually degraded during the recycling process,which shortens the battery life or even presents a potential safety hazard;it also indirectly leads to the performance degradation or malfunction of the powered device.Therefore,the remaining useful life prediction of lithium-ion batteries is crucial.This paper mainly studies the RUL prediction method of lithium-ion battery based on data-driven;the main contents are as follows:1)In this paper,the structure and working principle of lithium-ion battery are analyzed in detail,and main performance parameters of lithium-ion battery are expounded.By carrying out independent performance testing and charge-discharge experiments on battery,we study the influence of discharge rate on the battery performance,battery aging effect on discharge voltage and the actual capacity.2)The traditional support vector regression method has difficulties in selecting parameters,and it requires a lot of degradation data samples in prediction.To solve the above problems,a RUL prediction method for lithium-ion battery based on IPSO-SVR is studied,which uses an improved particle swarm optimization algorithm to optimize the parameters of SVR model.This method is verified by using NASA data and experiment data,the results show that the proposed method is effective in the selection of parameters,improves the prediction accuracy and reduce the calculation.In addition,with less training data,this method also can achieve better prediction results.3)In practical application,it's hard to monitor battery capacity data online.Moreover,the current RUL prediction algorithms are mostly offline,which is difficult to meet the actual needs of the project.Aiming at the above issues,we attempt to use indirect prediction method.First,after fully analyzing the parameters of battery life sates,we choose the equivalent voltage drop discharge time as the indirect health index of the battery.Secondly,genetic algorithm is introduced to optimize the extreme learning machine parameters to establish an indirect life prediction model for lithium-ion battery.Finally,the correctness of the GA-ELM method is verified based on NASA data and independent experimental data,the results show,compared with the ELM method and other traditional method,this new method is accurate and effective,with high predicting speed,and its output is stable.