Enabling accurate and fast large-scale battery simulation using only a 9-cell model with variance based parameters

For grid-connected batteries consisting of upwards of tens of thousands of cells, it can be challenging to produce an effective electrical model. Two methods are commonly seen in literature to model large packs, either represent the pack with a single cell model, or represent the pack with a cell model for every cell. The former is computationally efficient and suitable for real-time applications but lacks individual cell-level behaviour across the pack, whilst the latter offers this, due to the model size, it is unsuitable for real-time applications. Thus, a novel model is presented that can represent any size of battery pack using up to nine cell-models. A method for identifying the parameters for the nine cell models is offered, focused on ensuring the capacity limiting weakest cell is accounted for. This is verified experimentally with two lab-scale tests and a method for identifying pa-rameters for a large sample of cells using the parameter distribution is evaluated. For a 48 cell pack under 1C charge/discharge cycles, modelling all cells was identical to the proposed model, with an accuracy of >99.4 % compared to experimental testing whilst being >20 times faster to simulate. The simulation time for the proposed model to provide 10 real-time hours of data was 3.7 s, compared to modelling all 48 cells in the pack requiring 79.6 s. Finally, using the parameter distribution or variance is shown to be a viable technique to estimate the achievable pack capacity, with a 1C cycle test experimentally achieving an accuracy of >98.9 % when identifying model parameters using the variance of a sample of cells.

Automated summary

For grid-connected batteries with a large number of cells, it is challenging to create an effective electrical model. Existing methods either lack individual cell-level behavior or are computationally inefficient for real-time applications. This study presents a novel model that can represent any battery pack size using up to nine cell models. Experimental results show that the proposed model has high accuracy and is faster to simulate compared to modeling all cells. Additionally, using parameter distribution is shown to be a viable technique for estimating the achievable pack capacity.