Study On Remaining Useful Life Predicition Methods Of Lithium-ion Battery Based On Data Driven

With the increasing use of the number of lithium-ion battery began to appear aging phenomenon,security issues brought by the aging become the hidden dangers of lithium-ion battery development.Through the implementation of lithium-ion battery?Remaining Useful Life,RUL?accurately prediction,which can help to reduce the accident to improve safety,reliability of the lithium ion battery.In order to carry on the research of the RUL prediction method of lithium-ion battery,this topic choose SVM model as prediction method,which method has mature,learning and generalization ability.And aim at the problems of the existing prediction process such as the single variable and the precision of prediction,a further improvement method for SVM prediction model should be proposed.Based on this,this article on the content of the research.Firstly,based on the analysis of the factors affecting the capacity of lithium-ion battery and combined with the charging and discharging mode of the existing battery,explored the effects of different cathode materials,discharge rate,ambient temperature,battery temperature rise and cycle times on the capacity by designing of charging and discharging experiments for three element lithium ion battery and LiFePO₄ battery.The experimental results showed that the three element lithium ion battery is more sensitive to temperature and discharge rate,under the high discharge rate,discharge capacity was not monotone.The low temperature performance of LiFePO₄ battery was poor,and the capacity of the lithium iron phosphate battery was not reversible.Secondly,a new method based on phase space reconstruction and SVM was proposed to predict the RUL of lithium-ion battery.In order to verified the applicability and feasibility of the method for different life length samples,the two part of the data was used to predict.First,for the sake of reducing the data error caused by the experimental results,the capacity data of three element lithium ion battery No.1 and No.2 in the experiment were pretrested.second,by using NASA PCoE four types of battery capacity data to train and predict to verify the feasibility and applicability of the method to small sample;last,the capacity data of three element lithium ion battery No.1 and No.2 after the pretreatment were trained and predicted.The experimental results showed that the designed method can better fitted the capacity of the lithium-ion battery and predicted the capacity of the battery.Thirdly,based on the proposed phase space reconstruction and SVM,established a quantum particle swarm optimization phase space reconstruction-SVM lithium-ion battery RUL prediction method,which combined with the quantum particle swarm optimization algorithm to optimize the parameters of SVM model,effectively improve the SVM parameters selection problem in the prediction process and increased the prediction precision of the battery RUL.In the end,according to the experimental data of two batteries,the phase space reconstruction-SVM method and the quantum particle swarm optimization phase space reconstruction-SVM method are used to predict and compare with the traditional method.Analysis showed that the proposed two methods of RUL prediction relative error were maintained at about±10%;among them,the proposed quantum particle swarm optimization phase space reconstruction-SVM in prediction capacity and RUL prediction was better than the other two methods.