*3.3. Step 3: Model Parameter-Update Estimation*

The third step of the approach should not be mistaken as the parameter identification described in step 1. The first parameter identification of a battery requires an experimental setup, i.e., testing the cell in a battery cycler and temperature chamber. However, for identifying the parameters of partially (in operation) or completely aged cells (at EoL), the option of removing the cell from its application and testing it does not necessarily exist. Instead, the sensor measurements from the BMS and knowledge of how the battery has been used (charge/discharge cycles and environmental data) are the basis for the parametrization of the battery DT.

The procedures used to estimate the model parameters and states primarily limit the model usability. From the existing parameter identification techniques for ECM and DFN models, what role do they play when identifying the parameters of a battery DT that continuously needs to evolve as the battery is aging? Ultimately, the parameter estimation procedure that can track model parameters evolution as the cell ages is ideal for battery DT.

The battery DT parameter-update estimation procedure cannot entirely rely on the existing identification methods (Equation (1)) because, with the battery-DT-workflow, a cell in operation cannot be dismounted in order to perform experiments. Although the first step of accurate model development involves estimating parameters using experimental datasets and validation datasets, in practice, many parameters are constantly changing. For battery DT the parameter-update estimation needs to be repeated during operation after a certain number of cycles (N) or time (t). The input data from the BMS is not necessarily retrieved continuously in real-time. N and t will differ based on the battery application (EV, grid storage) and its usage practices. We leave the evaluation of optimal values of N and t for future studies, but an apparent range of N as evaluated from [60,61] is 500–1000 cycles, after which a significant change in electrochemical model parameters is observed. Parameter update for DTs during battery operation and at EoL includes but is not limited to the following methods:


Looking at the state of art **parameter** estimation algorithms, the gaps between the currently used battery models and the proposed battery DT are as follows; (1) Availability of cycling data to the battery model; (2) Model parameter-update method that does not entirely rely on experimental inputs, but instead on the charge/discharge characteristics and environmental data. Nevertheless, the **state** estimation algorithms would inherently remain the same in both battery model and DT.
