How to parameterise an equivalent-circuit empirical battery model from time-domain data

Empirical battery models have many uses, including pack design, charge estimation, and cell selection. An accurate model requires accurate parameters, however there are a range of subtleties in parameter fitting that are often missed in the literature. This work presents a best-practice guide for parameterisation, covering design of appropriate experiments; using numerical optimisers to fit model parameters to the experimental data; and building a suitable interpolation of the fitted model parameters, to get the required parameter-functions for a model. Key novelties include the importance of using timescales in place of capacitances for numerical stability; the design of a multi-step parameterisation process where model complexity is gradually increased to give more accurate parameter estimates; and a combined numerical and experimental design, so that the right experiments, data, and fitting methods are chosen for the target application. Our method is presented as a general recipe, so that users can easily choose the right setup for their needs. Scripts are provided which implement the method and reproduce the provided examples. Additionally, a full set of experimental data, model scripts, and fitted parameters are provided, both for testing the described method and for re-use in modelling studies.