Algorithm Research On State Of Health Of Lithium Battery Based On Electrochemical Impedance Spectroscopy And Incremental Capacity Curve Features

In recent years,with the rapid growth of renewable energy demand,fossil fuels have been gradually replaced by other clean energy.The rise of electric vehicles can effectively alleviate the energy crisis and environmental problems,and people are increasingly relying on the application of lithium batteries in the electric vehicle industry.An accurate estimate of the state of health(SOH)of the battery can help users better evaluate the aging degree of the battery,so as to take timely measures to prevent the premature decay of the battery capacity which could ensure the safety of users.Therefore,accurate prediction of battery health is of great significance for the battery management system.This topic is based on the data-driven method to predict the health status of lithium batteries.The main contributions are as follows:1.With the perspective of the lithium battery health status,the data is preprocessed.Firstly,the working principle and practical application of electrochemical impedance spectroscopy(EIS)and incremental capacity(IC)curve method are introduced.Through these two methods,the high-quality health features of lithium battery can be obtained from the perspective of time domain and frequency domain.Secondly,because of the noisy curve,it is necessary to use filtering methods to reduce the noise of the incremental capacity curve.Finally,the cleaned high-quality electrochemical impedance spectrum and incremental capacity curve features are fused to build a new training dataset.2.Comprehensively explore convolutional neural network(CNN),recurrent neural network(RNN)and long short-term memory(LSTM),a lithium battery health prediction model based on CNN-LSTM is established.The research found that CNN can effectively extract the battery health features,and LSTM has the function of capturing the data timing relationship.Combining these two methods can make up for each other’s disadvantages.Under the condition of using the same CNN-LSTM model,compare the health status of lithium battery under different health features.The experimental results show that compared with single domain features,the fusion features obtained from electrochemical impedance spectroscopy and incremental capacity curve method can improve the prediction accuracy by about 20%on average,and better track the state of health of lithium batteries.3.The CNN-TLSTM network is established based on the CNN-LSTM network structure,that is,adding a 1-tanh function after the LSTM input gate to change the value range of the output value of the input gate to ensure that the important features of the input data are retained as much as possible.The introduction of improved particle swarm optimization(IPSO)algorithm can effectively solve the problem of manual parameter adjustment caused by too many parameters of neural network,which is time-consuming and not objective,so as to better reflect the correlation between lithium battery data and health features.The experimental results show that the accuracy of CNN-TLSTM model to predict the state of health of lithium battery is around 17%higher than that of CNN-LSTM model.Compared with no-optimization,traditional particle swarm optimization(PSO)and quantum particle swarm optimization(QPSO)algorithms,the prediction accuracy of the improved particle swarm optimization algorithm can be improved by 73%,61%and 11%,and the prediction effect is more obvious.