The performance impact of beginning of life cell-to-cell variation on a lithium-ion battery system using reduced-order pack models of different resolutions is investigated. The aim is to evaluate the effects of variance, how simplified models of the entire battery pack can reproduce certain key performance indicators, such as fast charge time or extractable discharge energy, under different operating conditions and levels of parameter dispersion. The aim also seeks to identify when higher model resolution is necessary and its subsequent effect on the computational time. The parameter dispersion in the system is introduced as variations in capacity, internal resistance, and state of charge imbalances between the individual cells, with varying degrees of parameter variation and combinations. Two simulation environments were used which both utilized an equivalent circuit model consisting of two RC-branches; one for fast charging procedures with a more refined thermal representation of the system, and one for discharge simulations with a less detailed thermal model. 

The results show that the influence of cell-to-cell variation depend strongly of both the parameter being varied and the operating mode. During fast charge, resistance variation has the strongest effect on the charging time. When only one parameter is varied, the lowest resolution model predicts similar performance indicators to that of the full-scale model. However, as more parameters are allowed to vary simultaneously, the reduced models deviate increasingly from the fully modelled system. During discharge, capacity and initial state of charge have the greatest impact on the usable energy in the pack. The different cell models produce minor differences in the performance indicators regardless of the amount of variation or the parameter combinations. 

The findings indicate that reduced order models can be suitable for certain analyses, particularly when variation is limited. The gain in computational runtime is however closely related to the level of spatial representation of individual cells used to capture temperature gradients in the system. Full scale modelling becomes more important, especially during fast charge simulations, when multiple sources of heterogeneity interact together. This is valid both on a system perspective as well as on a cell-to-cell perspective. Reduced order models could however be useful during discharge scenarios, both on a system and cell-to-cell perspective, if the discharging current is considered relatively low, with only minor losses in the predicted accuracy. This thesis therefore provides guidance for selecting the appropriate model resolution in battery pack simulations depending on the operating mode, and highlights the importance of accounting for cell-to-cell variations in both system design and performance evaluation.