The rollout of electric vehicles (EVs) is considered as an efficient way to enhance transportation sustainability. Understanding the charging demand of EVs is therefore necessary for future deployment of charging infrastructure and the increase of EV adoption rates. Previous studies have used either data-driven or simulation-based approaches to understand EV driver behaviours and estimate their charging demand. However, these studies are limited in representing the adaptability and learning ability of EV drivers when making charging choices. To address these challenges, we introduce an agent-based reinforcement learning (RL) framework to simulate EV drivers’ charging and routing choices and the resulting public charger usage pattern in Manchester. The model is validated using real-world charging session data. We found that charging activity is not an everyday necessity for normal intra-city EV drivers unless (1) the number of trips significantly increases their total trip distances or (2) the battery charge falls below the drivers’ psychological threshold, incentivising their range anxiety. This modelling framework can contribute to a better understanding of the adaptive charging behaviours of EV drivers and their charging demand distribution in road networks.