BP Neural Network Prediction Research Of Impedance Model Parameters Of Lithium Battery

For the relatively backwardness of current testing technology and classification methods for Lithium batteries which directly affect life and reliability of batteries, this paper put forward the method of using BP neural network to predict impedance model parameters of batteries, which aims to provide theoretical basis and method for the uniformity research of battery. The impedance model of battery contains a large number of information which can represent the performance of battery, so the impedance model parameters are of great significance in the uniformity research of batteries. In this paper, electrode equivalent circuit which based on E-Movement Theory is applied to the establishing process of initial impedance model, through the referred equivalent components of the impedance spectra characteristics, combined with the measured impedance curve feature of battery, the equivalent circuit is amended, finally battery's impedance model which conforms to the internal structure and working principle of battery as well as the measured results is acquired.Analysing the correlation between the response values of battery voltage and impedance model parameters, prediction model of battery's impedance model parameters is established by the samples of data obtained in the experiments using BP neural network method. In the establishing process of the prediction model ,the trainlm, trainrp, trainscg, trainbfs and traingdx-the five kinds of fast learning methods provided by MATLAB neural network toolbox are compared and the trainlm which has the best results of training is selected to train the network. Simulation test by examining samples shows that the maximal relative error of predicted impedance model parameters of batteries by the established model and the actual measured values is only 1.23% which proves that using BP neural network to predict the battery's impedance model parameters is feasible.Finally, the BP neural network and genetic algorithm are combined in the research, where genetic algorithm is used to optimize the BP neural network.The simulation result shows that the maximal relative error of predicted values of battery's impedance model parameters and the actual measured values reduced to 0.699%, showing that GA-BP algorithm improves the prediction accuracy and proves the predictive ability of this model to battery's impedance model parameters once again.