Experimental Analysis, Modelling and Optimisation of Large Scale Lithium-ion Batteries

Globally, there is a substantial increase in the use of large batteries. One key application of these is grid-connected batteries. To maximise revenue, it is important that the full capacity of the battery is used. To achieve this, an operator must be confident of the capability of a battery at any given time.

For grid-connected batteries consisting of upwards of tens of thousands of cells, this can be challenging. This is because the cells all behave slightly differently due to manufacturing tolerances, with effects combining to give the observed output behaviour. Two main model-based approaches for these large batteries are observed in literature - a cell model for each cell, or modelling the battery with a single cell model. The former results in high computation demand while the latter is less accurate, ignoring the cell behaviour.

This thesis investigates the means for maximising the useable capacity of a large battery, through the consideration of cell level behaviour. This behaviour is demonstrated experimentally using a 2MW/1MWh system, with the behaviour examined through rigorous testing at the lab scale. A new model is proposed which considers nine cell models to represent a large battery to achieve accurate estimation. This is validated and shown to be computationally efficient while giving cell-level detail. A method is then explored to identify model parameters for cells within a large battery, which is successful at parameterising a model to match a physical system. It is demonstrated using 1584 cells in-situ, showing the variance of capacity and impedance in a large battery. The model and parameter identification method are then combined to produce an online state estimator to estimate key battery metrics in real-time, considering cell behaviour. The observations and methodologies presented thus far are then applied to consider maximising the capacity of a grid-connected battery. It is shown through simulation that choosing an appropriate SoC at which to perform cell voltage equalisation can increase the capacity, and that this is dependant on the cell parameter variance. Additionally, it is shown that rearranging modules in a large system based on temperature can further improve the system capacity. This can be as high as 5\% depending on the capacity variance.