Fig. 1: The dataset utilised in this work encompasses a wide variety of electric vehicles (EVs) with different battery capacities, chemistries, and battery management systems (BMS), capturing charging behaviour under real-world and often sub-optimal conditions.

a The geographic distribution of the charging stations from which the charge curve dataset was collected. Stations located in the Netherlands did not report cumulative energy delivered and are therefore excluded from the quantitative analyses in panels (b–e). b A diverse array of charging stations with rated connector power ranging from 50 kW to 360 kW are included in the dataset, with the primary connector types being Type 2 Combined Charging System (CCS) and CHAdeMO. c Most charging sessions analysed in this study originated from Great Britain and Germany, with fewer than 1% of sessions coming from the Netherlands. The dataset was collected from 2011 to 2024, and covers a large range of sessions with various levels of delivered energies and connector power ratings. It should be noted that these describe only the subset of data examined here and should not be taken as representative of Shell’s overall charging operations. d The distribution of estimated battery capacity across different connector power ratings. Violin plots show the data distribution (kernel density), with embedded box plots indicating the median (centre line), interquartile range (box limits), and whiskers (1.5 × interquartile range). e The percentage of sessions as a function of starting and stopping state-of-charge (SoC), as well as the average session duration of these sessions. Note that for plotting purposes, these are computed from a fraction of the dataset due to minor inconsistencies in recorded starting or stopping SoC values in some sessions.