A parameter identification method of lithium ion battery electrochemical model based on combination of classifier and heuristic algorithm

Abstract

Parameters of lithium-ion electrochemical battery model have a great impact on the simulation accuracy, so their accurate identification plays an important role in terms of battery characteristic simulation and health management. Currently, global optimization algorithm is a common method for lithium-ion battery parameter identification, however this kind of method may lead to local optimization, which fails to get accurate identification results. In the search range of the global optimization algorithm, there are certain parameter vectors that may cause the battery model to not converge. Such parameters reduce the computing efficiency seriously. This work proposes a new parameter identification method for lithium-ion battery electrochemical model, which combines machine learning based classifier with improved particle swarm optimization algorithm. The classifier is used to filter the parameter vectors in the swarm generated by improved particle swarm optimization algorithm that may make the battery model fail to converge. The classification accuracy is up to over 90 %. Validation results indicate that the battery model with identified parameters obtained by the developed method has acceptable simulation accuracy, and the terminal voltage simulation errors are within 24.6 mV. Also, the parameter identification method can significantly improve the efficiency of parameter identification.