Nonlinear battery modeling using continuous-time system identification methods and non-uniformly sampled data

A battery model identification approach, based on non-uniformly sampled data, aiming to reflect the nonlinear dynamic behavior of a lithium-ion cell is presented in this work. To accurately predict the voltage response, the underlying model should reproduce the fast and slow dynamics of the battery cell. Therefore direct identification from non-uniformly sampled measurement data based on continuous-time model identification is applied. To take into account the nonlinear behavior of the battery, local linear model partitioning for the state of charge is performed. The resulting dynamic battery model is able to accurately predict the system response. With a parameter conversion to physically interpretable parameters, based on an equivalent circuit model, the parameter variance among similar cells and the temperature dependency of the model identification are investigated as well as the parameter characteristics over time. All results are based on non-uniformly sampled input output measurement data of three identical lithium-ion power cells.