Advancing operational characterisation of dwelling stock energy models: integrating climate, occupancy profiles and heating behaviour for Ireland

Extended abstract 7-265-26

Reference Dwellings (RDs) underpin Dwelling Stock Energy Models (DSEMs) used to inform national retrofit and decarbonisation strategies. In Ireland, recent work has produced statistically robust and geometrically validated RDs derived from large-scale Energy Performance Certificate (EPC) datasets. However, while these archetypes are physically representative in terms of form, envelope and systems, they remain operationally static. Deterministic climate assumptions, fixed occupancy schedules and uniform heating setpoints limit their ability to reflect real-world variation, reducing confidence in stock-level estimates of energy demand and retrofit performance.

This study addresses that operational gap by enriching the 20 three-dimensional thermophysical Irish RDs developed by Li et al. (2026), which represent approximately 85% of the national dwelling stock. Climatic realism is introduced through assignment of regionally representative Test Reference Year (TRY) files from the TRANSLATE dataset, replacing the single-location assumptions embedded in conventional EPC methodologies. Occupancy presence and temporal variability are captured using a Bayesian Neural Network-based Probabilistic Occupancy Model (POM). To enable integration into dynamic simulation, POM outputs are pre-processed into structured behavioural profiles that preserve temporal dynamics while providing interpretable occupancy and activity patterns at dwelling level. These structured profiles are subsequently mapped to dwelling typologies to derive population-level operational representations suitable for stock-scale modelling. Heating preferences and behavioural practices are characterised using nationally representative data from SEAI’s Behavioural Energy and Travel Tracker (BETT), providing empirical thermostat setpoints and zone-specific heating patterns.

To coherently integrate occupancy and heating behaviour, a Human-in-the-Loop (HITL) control framework is adopted. An adaptive deep-setback strategy links probabilistic vacancy and arrival patterns to heating control logic, enabling dynamic alignment between comfort provision and occupant presence. This results in a set of simulation-ready operational schedules suitable for Dynamic Simulation Modelling (DSM) at archetype and stock level.

The framework advances Irish DSEM practice by completing the operational dimension of RD characterisation, enabling behaviourally realistic stock modelling. Beyond Ireland, the methodology provides a transferable pathway for jurisdictions with large-scale EPC and behavioural datasets to enhance modelling realism, narrow the performance gap, and strengthen evidence-based retrofit and demand-side management policy in low-carbon energy systems.

Read the full abstract.

Download presentation.