Adaptive Linear Time-Varying Parameter-Varying Modeling of Lithium-Ion Batteries Considering Aging Phenomenon

Achieving a balance between accuracy and complexity remains a significant challenge for lithium batteries. The accuracy of the model is usually positively correlated with the complexity of the model. A simple model cannot capture more influencing factors, as simpler models often fail to capture the full range of influencing factors. This article presents an enhanced battery model that extends the traditional equivalent circuit model (ECM) by integrating a set of exogenous independent parameters, encompassing both internal and external battery characteristics, into the system matrix via a polynomial relationship. This results in a time-varying system matrix that allows for a comprehensive representation of battery parameters and effective data fusion while simplifying the model. The model also addresses aging and other time-varying phenomena through time-varying exogenous functions, thus simplifying the management of aging effects. Recursive least squares algorithms with an adaptive forgetting factor handle time-varying parameters and aging effects, ensuring rapid adaptation to data changes while preserving high performance and accuracy. Through simulation and aging battery experiments, the mean absolute error and root mean square error of the proposed scheme can be controlled within 13.6 and 16.0 mV, which proves its accuracy and superiority, and has great potential in aging battery applications.