Research On State Of Health Prediction Of Lithium Ion Battery Based On ICA And SVR

With the continuous development of electric vehicles,their safety issues have received extensive attention.The health of lithium-ion batteries is an important factor in the stable operation of electric vehicles.With the use of lithium-ion batteries,complex chemical reactions occur inside them,which will cause the consumption of active materials and the continuous decline of capacity,which will gradually degrade the state of health(SOH)of lithium-ion batteries,which will affect electric vehicles in severe cases.Normal operation,so accurate monitoring of the SOH of lithium-ion batteries is very important for the safe operation of electric vehicles.First of all,this paper conducts a large number of related literature surveys around the aging feature extraction of lithiumion batteries and SOH prediction methods,and establishes a data-driven modeling method to predict SOH.On the basis of analyzing the working principle and performance characteristics of lithiumion batteries,design and build a lithium-ion battery test platform,complete battery performance test experiments,and analyze the impact of different cycle test conditions on the SOH of lithium-ion batteries,and predict the SOH of subsequent lithium-ion batteries Research provides basic data.Secondly,in view of the small sample data problem of lithium-ion battery SOH with few available data and difficult extraction of characteristic parameters,the Incremental Capacity Analysis(ICA)method is used to analyze the attenuation of lithium battery SOH and study the capacity under different working conditions.The change trend of the incremental curve with SOH.The wavelet packet transform analysis method is used to filter the noise of the capacity increment curve,and multiple characteristic parameters such as peak height,peak position and peak area are extracted,and the gray correlation analysis method is used to verify the correlation between the characteristic parameters and SOH Effectiveness.The principal component analysis(PCA)is used to integrate the original information,extract hidden factors,eliminate redundant information,and improve generalization performance.Then,using the characteristic parameters processed by ICA and PCA as the input and SOH as the output,the Support Vector Regression(SVR)algorithm suitable for small sample data prediction is adopted to establish the lithium-ion battery SOH prediction model.By analyzing the parameter optimization algorithm of SVR,a grid search cross-validation method suitable for small sample data optimization is selected to optimize the penalty factor and kernel function parameters of SVR.The prediction model was validated with test set data,and good prediction accuracy was achieved.At the same time,a SOH prediction model without PCA processing was established,and the influence of redundant information on the prediction accuracy was analyzed and compared.Finally,in order to further solve the problem of the decrease in SVR prediction accuracy caused by the regeneration of the SOH capacity of lithium-ion batteries under the condition of small samples,an adaptive integrated learning algorithm(Adaptive Boosting,Ada Boost)is proposed to improve the SVR.Continuous weighting changes to solve the problem of poor prediction accuracy of SOH capacity regeneration samples.The Ada Boost-SVR model was verified with the test set data.The results proved that the lithium-ion battery SOH prediction model based on Ada Boost-SVR improved the global prediction ability of the SOH with nonlinear attenuation state under the condition of small samples,and obtained a higher precision SOH forecast results.