Abstract
Achieving net-zero carbon emissions requires enhanced energy planning, driven by the growing penetration of renewable sources, the electrification of heating and transport, and evolving consumer behaviour. Therefore, accurate urban-scale energy analysis is essential for effective planning and policy design. However, existing literature largely overlooks user behaviour as an active and structural component of energy demand, often treating it as a secondary source of uncertainty rather than as a defining feature of the system, whereas technical specifications are modelled with great precision. As a result, models tend to capture the physical performance of systems accurately yet fail to represent the stochastic and socially driven dynamics that shape real-world demand. This study introduces a hybrid, bottom-up framework that integrates user behaviour and building performance within a unified modelling structure. Occupancy dynamics are simulated through a Markov Chain Model capturing transitions between active and inactive states, while appliance operation schedules are derived using logistic regression, both based on Time Use Survey (TUS) data, thereby enriching behavioural representation. Indoor thermal dynamics are modelled via an RC model, ensuring physical accuracy alongside computational efficiency and scalability. A key innovation is the creation of Agent Energy Consumption Units (AECUs), in which occupant behaviour, appliance usage and building characteristics are tightly coupled, supporting a granular representation of household heterogeneity. The framework is simulated at district scale and includes a sensitivity analysis of the individual user on the aggregate. Results highlight that household-type distribution is as critical as traditional technical parameters, particularly when assessing its impact on energy demand.