Research Of Remaining Life Prediction Method For Battery Based On Small Sample

With varieties of characteristics of high specific energy,high voltage,wide temperature range,low self-discharge rate,long life,and high security,lithium-ion battery has been widely used in telephone,notebook computer and electric vehicle,etc,and extends to the fields of military communications,navigation,aviation,and spaceflight gradually.The prediction of lithium-ion battery's remaining service life has been an important content of PHM technique.based on the analyze of predicting outcomes by RUL,managing the system device excellently,it can improve the availability and reliability of system or device,meanwhile reduce even avoid heavy loss caused by fault.In practical engineering,with the limitations of engineering specifications,the number of samples is small,so the small sample problem has caused the public's concern,and it has been a hot area of research that predicting lithium-ion battery's remaining service life under the condition of small sample.SVM has been proved to be an effective method,and this method is also based on limited sample information,looking for a middle ground between model complexity and learning ability,it has unique benefit on solving small sample,non-linear and high-dimension problems.Grey theory also has unique benefit on solving small sample problem.In this paper,aiming at the prediction of lithium-ion battery's remaining service life,adopting the lithium-ion battery acceleration degradation testing data tested by University of Maryland,we carry out the following work:1.The research of SVR's fitting problems and LSSVR's fitting problems,especially under the condition of treatment factor increasing,the simulation velocity of LSSVR is faster than SVR,which has some advantages.The research of short-term prediction of Gray theory GM(1,1),secondly,studying the application of metastasis discrete gray model MDGM(1,1)on lithium-ion battery's remaining service life based on Gray theory GM(1,1)and DGM(1,1)models.The research of combining the fitting algorithm and prediction algorithm,predicting lithium-ion battery's remaining service life with small sample.2.Based on the Microsoft Visual Studio 2010 development platform,studying hybrid programming method using MATLAB and Visual Studio,achieving software implementation.It has proved that LSSVR and metastasis gray model can evaluate lithium-ion battery's remaining service life with small sample exactly through experiment.