Research On SOH And SOC Prediction Of Lithium Ion Battery Based On Neural Network

As the core component of electric vehicles,lithium-ion battery is the key research field of new energy vehicles.The state-of-health(SOH)of batteries reflects the aging degree of the battery,and the state-of-charge(SOC)of batteries reflects the remaining battery power.Accurate battery SOH prediction and SOC prediction are the premise to ensure the reliability of the battery.The key problem of SOH and SOC prediction is to improve the prediction accuracy and realize real-time online prediction.In this paper,the improved neural network method is used to predict battery SOH and SOC with high accuracy online.Firstly,the experimental data of lithium-ion batteries are pretreated.The data of batteries during the whole life span are analyzed.The correlation between SOH and the health factors,such as average temperature,maximum temperature,equal temperature rise discharge time,average voltage,et,are analyzed.In order to reduce the model error,the first exponential smoothing method is used to smooth the battery SOH data,the wavelet threshold denoising method is used to de-noise the battery valley bottom voltage data,and the outliers detection based on LOF and the data outliers repair based on cubic spline interpolation are used to eliminate the average temperature,maximum temperature,equal temperature rise,discharge time,and steep drop of the batteries.In order to reduce the number of neurons in the neural network and reduce the complexity of the model,the principal component analysis method is adopted to remove the redundant information and correlation among the eight health factors.Finally,three principal components are used as the input of the subsequent neural network.Then the SOH prediction model of lithium-ion batteries is established and verified.The coupling NARX neural network model for battery SOH prediction is established and verified,which preliminarily proves the applicability of NARX network for battery SOH prediction.The NARX neural network model is optimized by Bayesian regularization algorithm,and the optimized neural network model is verified by real battery data.It is proved that the BR algorithm can improve the prediction accuracy of the model.Particle Swarm Optimization is used to optimize the improved NARX neural network model,and the optimized network model is verified with real battery data,which proves that the Particle Swarm Optimization algorithm can improve the prediction accuracy and the generalization ability of the model.Finally,the prediction errors of improved NARX neural network,Elman neural network and RBF neural network are compared,and it is verified that the optimized NARX neural network has the best prediction effect on battery SOH.Then the SOC prediction model of lithium-ion batteries is established and verified.The BP neural network model of battery SOC prediction is established and verified.The Levenberg-Marquardt algorithm is used to optimize the BP neural network model,and the optimized neural network model is verified,which proves that the Levenberg-Marquardt algorithm can reduce the prediction error of the model and accelerate the convergence speed.Genetic algorithm is used to optimize the optimized BP neural network model,and optimized neural network model is verified,which proves that Genetic algorithm can reduce the prediction error and improve the generalization ability of the model.Finally,the prediction errors of improved BP neural network,Elman neural network and RBF neural network are compared,and the prediction accuracy of the optimized BP neural network for battery SOC is verified to be the highest.Finally,the prediction system for SOH and SOC of lithium-ion batteries is developed.The prediction system includes three functions: monitoring data display,battery state prediction and battery abnormal alarm.Using the optimized NARX and BP neural network algorithm,the program block diagram and display interface of the above functions are designed.Then battery monitoring data display,state prediction display and abnormal state alert functions are tested by using lithium-ion batteries’ experimental data.It is verified that the battery state prediction system can realize the real-time display of battery parameters,SOH,SOC and alarm prompt of abnormal state.