Research On Prediction Algorithm Of Lithium Battery State Of Health

Pure electric vehicles have now been widely popularized.As the main power source of lithium batteries,researches on how to reduce the use cost of lithium batteries and extend the service life of lithium batteries are getting more and more attention.As the number of times of use increases,the actual capacity of the lithium battery will decrease,and the state of health(SOH)of the lithium battery will also decrease.The attenuation of the actual capacity will inevitably affect the measurement of the state of charge and remaining service life of the lithium battery.Accurately predict the practical significance of lithium battery SOH,on the one hand,it can avoid the safety hazards caused by the actual capacity of the car when driving is too low,on the other hand,it can avoid the waste of funds and environmental pollution caused by the premature replacement of the battery.Therefore,this article has launched a research on how to accurately predict the SOH of lithium batteries.In the algorithm design,this paper chooses Back Propagation(BP)neural network and Radial Basis function(RBF)neural network as the models for predicting the SOH of lithium batteries.Since the prediction accuracy of these two kinds of neural networks are more dependent on the setting of initial parameters,and the randomness of the initial weights and thresholds of the BP neural network,the prediction results are easy to fall into the local optimum,so it is proposed to use Cuckoo Search(Cuckoo Search,CS)algorithm improves two kinds of neural networks.Designed CS-BP neural network and CS-RBF neural network.At the same time,this paper uses the Mean Absolute Error(MAE),the Mean Absolute Percentage Error(MAPE),the Root Mean Square Error(RMSE)and the fitness(R2)as the evaluation indicators of the model prediction effect.In the simulation analysis,the lithium battery charge-discharge cycle experimental data provided by the National Aeronautics and Space Administration Ames Research Center is used as a sample.Before the simulation,it is necessary to perform feature extraction for the SOH aging of the lithium battery.The correlation between the characteristic parameters and the SOH of the lithium battery was analyzed by the grey relational analysis method.Then the principal component dimensionality reduction is performed on these parameters,and the cumulative contribution rate of principal component two is more than 90%.Therefore,principal component one and principal component two are selected as the input of the prediction model.The simulation experiment was performed under the same set of lithium battery data,and the MAE,MAPS,RMSE,and R2 of the CS-BP neural network were 0.1950,0.3319,0.2658,0.9847,respectively,and the MAE,MAPE,RMSE,and R2 of the CS-RBF neural network were 0.1605,0.2678,0.1928,0.9913,respectively.It can be seen that the prediction accuracy of CS-RBF neural network is higher than that of CS-BP neural network.The same results can be obtained by applying the two neural networks to other lithium battery data with different working environments,which verifies that the CS-RBF neural network has good generalization.This paper uses the data collected during the charging process to predict the SOH of the lithium battery during the discharge process,which can effectively avoid safety accidents caused by the failure of the lithium battery to be replaced in time,and avoid the waste of funds and environmental pollution caused by premature replacement.