Building Sustainability in Regional Energy Transition

A Narrative Synthesis of Large‑Scale Energy Demand Modelling with Socio‑Economic and Heritage Dimensions

Ambitious decarbonization targets in the European Union require robust, scalable approaches to reduce energy demand in buildings—responsible for ~40% of final energy use and ~36% of energy-related emissions. Lukas Dahlström’s dissertation advances Urban Building Energy Modelling (UBEM) through an open-data, probabilistic, and heritage-aware framework. The work integrates cadastral (GIS/LiDAR), Energy Performance Certificates (EPCs), statistical socio-economic datasets, and measured/typical meteorological year (TMY) weather, to identify representative building archetypes using multivariate cluster analysis and to propagate archetype characteristics across entire regional stocks. Probabilistic treatments of infiltration, ventilation, window-to-wall ratio, and stochastic occupancy reduce deterministic bias, while heritage classification (Stockholm Cultural-Historical Inventory, CHI) enables retrofit strategies that balance energy savings and conservation values. Applied to Gotland (urban–rural) and Södermalm (historic urban district), the framework captures the measured diversity of energy performance, achieving weighted mean percentage errors (wMPE) around ~5% on Gotland and ~1% for Södermalm’s multifamily buildings, and demonstrating renovation scenarios with 10–39% savings depending on constraints and heritage protection. The results show that EU efficiency targets are reachable for historical building stocks through differentiated, conservation-compatible measures—without compromising high-value heritage.

Reducing energy demand is a direct pathway to mitigate climate impacts. In Europe, buildings dominate energy use, with heating, hot water and cooling accounting for ~80% of residential demand; up to 75% of existing buildings are energy‑inefficient. Despite strong policy ambitions—including full decarbonization of the building stock by 2050—the scientific consensus on how to achieve large‑scale, context‑sensitive demand reductions remains unsettled. Urban Building Energy Modelling (UBEM) provides a bridge: bottom‑up, physics‑based aggregation of building performance that can forecast demand, evaluate retrofits, and test policy scenarios at scale. Dahlström’s thesis responds to critical UBEM gaps—data availability, archetype identification, probabilistic parameterization, validation, and application to historic districts—by developing an open, transferable framework that is scientifically rigorous yet practical for planners and decision-makers.

Narrative Overview of State of the Art

UBEM comprises top‑down statistical and bottom‑up physics‑based approaches, increasingly converging in hybrid models. The core challenge is data: fine-grained geometric and operational characteristics are rarely available comprehensively; measured energy is often restricted, limiting reproducibility and validation. Archetypes—representative buildings—remain central to aggregation, but deterministic segmentation (e.g., by age/type) can miss real diversity. Dahlström reviews UBEM advances and advocates multivariate cluster analysis (MCA) to identify archetypes from open data, along with probabilistic parameter modelling (notably for infiltration, occupancy, WWR, and thermostat setpoints) to better capture variability and reduce the “performance gap” between simulated and measured use. Importantly, most UBEM treats strictly urban areas; the thesis argues for regional scales that include rural stocks where energy use patterns, systems (e.g., boilers), and building forms differ, and where policy is often enacted. The work also addresses a neglected frontier: UBEM applications for historic districts, where energy efficiency must be reconciled with cultural values.

Methodological Framework

Data Ecosystem (Open and National Sources)

• Geometric & Topological: GIS building footprints, cadastral attributes, and LiDAR point clouds (LoD1 shoebox models).
• Performance & Characteristics: EPC databases (Sweden’s EPCs primarily measured rather than purely calculated), including energy by carrier, total EP (kWh/m²), construction year, heated area, ventilation/system metadata.
• Socio‑economic & Statistical: Statistics Sweden (SCB) aggregated grids (e.g., 250×250 m), population density, income, age; regional energy use by carrier and building type; living area per person.
• Weather: Measured datasets (SMHI) and TMY/EPW files; bespoke 2022 hourly compilation for Gotland validation.
• Heritage Classification: Stockholm City Museum’s Cultural‑Historical Inventory (CHI), categorizing buildings (Blue, Green, Yellow, Grey, Dashed) by heritage value.

Pre‑processing and Calibration

A major contribution is robust, transparent pre‑processing:

• Spatial alignment and de‑duplication: Addresses EPCs issued to addresses rather than buildings; resolves multi‑address MFH duplicates; reconciles property‑level EPCs to building objects.
• LiDAR filtering and height inference: Median roof points within polygons; outlier correction; calibration to EPC indoor area and number of floors via correction factors (height) and (footprint-to-heated area), minimising RMSE (Table 4.1).
• Living area and occupancy estimation: MFH “loss factor” (≈0.85) for common areas; mixed‑use ground floors excluded from residential living area; validation against SCB living area and occupants shows ≤1% error (Gotland SFH) and ~0.3% error (Södermalm MFH) once mixed‑use accounted.

Archetype Identification via MCA (k‑means/k‑means++ with Replicates)

• Parameters (Table 4.2): Construction year, building volume, EP, socio‑economic (income, age), population density, heating system shares (% DH/boilers/electric/heat pumps).
• Normalization: Min‑max scaling for Euclidean metric consistency.
• Model selection: Dual validity metrics—RMSSTD (root‑mean‑square standard deviation) and a bespoke “distinctness” combining representativeness and spread—applied in an elbow‑style analysis to choose optimal (e.g., SFH in Uppsala).

Archetype Propagation to Non‑EPC Buildings

A nearest‑neighbour (KD‑tree) algorithm assigns archetypes to buildings lacking EPCs, leveraging geographic clustering of archetype characteristics. Validation identifies high accuracy for MFH (~94%), lower for rural SFH where EPC coverage is sparse—highlighting the dependency on data completeness and the need for continued expansion of open performance datasets.

Probabilistic Parameter Modelling

• Infiltration (ACH): Log‑normal sampling with mean ACH and ACH, adjusted by type, age, height, urban/rural context.
• Ventilation: Probabilistic assignment of natural/exhaust/HRV based on archetype distributions; HRV efficiency set at ~75%.
• Occupancy: Rescaled Markov chains (RMC) from Swedish time‑use data produce three‑state hourly patterns (“away/sleeping/awake”), with calibrated internal gains and household electricity/hot water intensities typical for Sweden.
• Other: WWR sampled from normal distribution (mean ≈0.21, SD ≈0.03); heated basements assigned by archetype probability and modelled as separate zones.

Simulation Engine and Modelling Choices

• Shoebox models (LoD1): One thermal zone per building (two if heated basement) for computational tractability with minor accuracy trade-off.
• Adjacency: Party walls simulated as adiabatic; simple internal shading (blinds/curtains) triggered by temperature/irradiance thresholds; external shading not modelled given high runtime cost and minimal district‑scale benefit.
• Weather and runs: EnergyPlus (9.4.0) via Python (Eppy), running hourly simulations with normalised weather (for EPC comparability) and measured weather (for statistical validation).

Case Studies and Key Results

Gotland (Urban–Rural Regional UBEM)

Archetypes: 8 SFH and 7 MFH archetypes reflecting distinct construction eras, volumes, EP distributions, and heating systems (Table 5.1). Spatial maps (Figures 5.3–5.4) reveal geographically coherent clusters (e.g., boilers prevalent in rural areas, DH concentrated in urban centres).

Validation:

• EPC alignment: Cumulative EP distributions of simulations vs EPCs show similar shapes; overestimation occurs primarily in mid-to-high EP segments. Weighted metrics indicate average differences of ~5–8% for SFH and ~4–5% for MFH (Table 5.2).
• Statistical (2022) validation: Using measured 2022 weather and SCB energy by carrier/type, total energy was overestimated by ~7% for SFH and ~27% for MFH; living area and occupants matched closely for SFH (≤1–4% error) but were ~10–15% high for MFH, likely due to mixed-use and EPC coverage artefacts (Table 5.5).
• Hourly energy profiles: Electricity dominates total demand when including household uses; space heating energy roughly balanced among boilers, DH, and electric heating, with boilers comparably significant outside dense urban contexts (Figure 5.8).

Interpretation: The framework captures heterogeneity and aggregate behaviour well, with MFH overestimation attributable to living area/occupancy estimation and system assumptions—later addressed/improved in the Södermalm study.

Södermalm (Historic Urban District with Heritage-Aware Retrofits)

Archetypes: 8 MFH archetypes (and small SFH share), spanning 18th–20th centuries, with varied EPs and systems; archetypes show strong alignment with CHI classes (Table 5.4), demonstrating that MCA captures heritage-relevant features.

Calibration & Validation:

• Living area/occupants: Including mixed-use corrections yields ~0.3% deviation from SCB values—substantially improved over Gotland MFH.
• EPC alignment: Weighted mean EP difference and wMPE around ~1% for MFH (Table 5.2), indicating excellent archetype representativeness.

Renovation Scenarios (Table 4.5; Table 5.6–5.7; Figure 5.10):

• Least Changes: Non‑invasive measures (airtightness, thermostat setbacks, internal shading).
• Balanced (by CHI class): Conservation‑compatible envelope upgrades (e.g., attic/roof insulation, internal wall insulation for Yellow/Green, renovated original windows, HRV for select classes).
• Optimal: Efficiency‑maximised measures unconstrained by heritage (external wall insulation, replacement windows, deeper floor/basement insulation, HRV).

Outcomes:

• Average EP reductions for MFH: Reference 154 kWh/m² → Least Changes 138 → Balanced 114 → Optimal 79 kWh/m² (Table 5.6).
• All buildings: Savings of ~10.5% (least), ~20.7% (balanced), ~38.7% (optimal).
• Worst-performing 43% (WPB): Disproportionate impact—~53% of total savings accrue in WPB for balanced scenario.
• Historic subset (Blue/Green/Yellow): In balanced scenario, historic buildings deliver ~85% of total savings, with modest reductions for Blue (≈4%) reflecting conservation constraints; optimal scenario yields large savings (~44–51%) but risks heritage values.
• 3D visualizations (Figure 5.9): EP heatmaps highlight spatial concentrations of high demand and reveal block-scale retrofit leverage.

Conclusion from Case Work: EU energy‑efficiency targets are achievable for historical stocks through differentiated, conservation‑compatible strategies. High‑value Blue buildings contribute little to aggregate savings (due to small share and strict constraints); major gains lie in Green/Yellow stocks and worst performers.

Discussion

Methodological Advances:

• Open‑data UBEM: A reproducible, national‑scale approach using public datasets addresses the field’s validation and transferability gaps.
• MCA Archetypes: k‑means/k‑means++ with dual metrics delivers robust clusters; geographic coherence confirms real‑world representativeness; a corrigendum demonstrates method resilience even when correcting population density calculations.
• Probabilistic Modelling: Stochastic infiltration and occupancy are essential to narrow performance gaps at stock scale; deterministic parameters alone mask variability and propagate bias.

Limitations and Sensitivities:

• Data Coverage: EPC sparsity for SFH in rural areas constrains archetype propagation accuracy; mixed‑use misclassification affects MFH living area/occupancy estimates.
• Model Scope: LoD1 shoeboxes and simplified shading are justified for speed but may understate microclimatic effects; heritage-compatible HVAC replacements were not modelled exhaustively.
• Performance Gap Residuals: Overestimation in some archetypes suggests further calibration opportunities and sensitivity analyses (e.g., thermostat behaviour, appliance loads, occupancy mobility).

Strategic Insights:

• Urban–Rural Integration: Regional UBEM is vital; rural boilers and envelope characteristics require tailored policies distinct from district‑heated urban cores.
• Heritage‑Aware Planning: Differentiated strategies—“do the right things in the right buildings”—can balance conservation and efficiency, focusing on WPB and mid‑value historic stocks for maximal impact.
• Beyond Efficiency—Toward Sufficiency: The thesis acknowledges rebound risks and the need for sufficiency-oriented scenarios (behavioural change, space sharing, adaptive occupancy), which UBEM can integrate via socio‑economic and mobility modules.

Dahlström’s dissertation delivers a scalable, open, and probabilistic UBEM framework that accurately models large, heterogeneous building stocks and integrates heritage values into retrofit decision-making. Across Gotland and Södermalm, the approach captures the measured diversity of energy performance, validates aggregated outcomes, and demonstrates that substantial savings (10–39%) are attainable—even with conservation constraints—by prioritizing worst performers and mid‑value historic buildings. The work addresses major UBEM gaps (data, archetypes, probabilistic parameters, validation, historic applications) and sets the stage for future directions: richer sensitivity analyses, extended non‑residential coverage, integrated mobility/energy-system coupling, and sufficiency‑oriented scenario testing. For policymakers and practitioners, the central message is pragmatic and actionable: large‑scale energy planning can be both evidence‑based and culturally responsible.

Descriptions of Key Figures and Tables (from the Dissertation)

• Table 4.1 (Correction Factors): Presents calibrated footprint () and height () factors used to align LiDAR-derived geometry with EPC heated area and floors; indicates LiDAR height overestimation and smaller heated shares in rural SFH.
• Table 4.2 (Clustering Parameters): Lists variables used in MCA—construction year, volume, EP, socio‑economic indicators, population density, heating system shares.
• Figure 5.1 (Error Metrics for k‑means): Shows RMSSTD and “distinctness” curves vs number of clusters; elbow points inform optimal selections (e.g., SFH ).
• Table 5.1 (Archetypes): Enumerates representative archetypes for Gotland and Södermalm with construction year, EP, size, and dominant heating system types.
• Figures 5.2–5.5 (Spatial Archetypes): Maps coloured clusters across Uppsala, Gotland, Visby, and Södermalm; reveal geographic coherence and neighbourhood homogeneity.
• Table 5.2 (Simulation vs EPC EP): Compares simulated mean EP per archetype to EPC archetypes; includes wMPE and wMAPE for aggregate accuracy (notably ~1% wMPE in Södermalm MFH).
• Figure 5.6–5.7 (EP Distributions): Presents cumulative and box‑plot comparative EP distributions showing similar shapes and spreads between simulations and EPCs.
• Table 5.5 (Gotland Energy and Occupancy): Breaks down 2022 simulated energy by carrier/end use vs SCB statistics; highlights overestimation in MFH.
• Table 4.5 (Retrofit Measures): Details conservation‑compatible and optimal measures (airtightness, thermostat control, insulation, HRV, window strategies) per CHI class.
• Tables 5.6–5.7 and Figure 5.10 (Scenario Savings): Reports EP reductions and total GWh savings under least/balanced/optimal scenarios; shows disproportionate savings in WPB and significant contributions from Green/Yellow heritage classes.
• Figure 5.9 (3D EP Visualization): A KML-rendered view of EP intensities across Södermalm buildings for reference scenario.

References

Reference list from and aligned with the dissertation’s bibliography. Numbers correspond to those cited in the thesis text.

1. European Parliament, Council of the European Union. Directive (EU) 2024/1275 on the energy performance of buildings, 2024.
2. Johari, F. Urban Building Energy Modeling for Retrofit Scenarios: Development, Calibration, Validation and Implementation for Swedish Residential Buildings. PhD thesis, 2023.
3. Papaioannou, T. The importance of open research or open science for achieving sustainable development goals, 2025.
4. Shukla, P.R. et al. IPCC AR6 WG III—Mitigation, 2022.
5. Denton, F. et al. Accelerating the transition in the context of sustainable development. In IPCC, 2022.
6. Cabeza, L.F. et al. Buildings. In IPCC AR6 WG III, 2022.
7. Swedish Energy Agency. Användning av energi—fördjupad energistatistik, 2025.
8. Neij, L. et al. Transformative evaluations of energy efficiency in buildings. Energy Research & Social Science, 2021.
9. Dixit, M.K. Life cycle embodied energy review. Renew. Sust. Energy Rev., 2017.
10. Pettersson, A. Långsiktig renoveringsstrategi, 2019.
11. Keirstead, J. et al. Urban energy system models review. Renew. Sust. Energy Rev., 2012.
    12–167. (Further references continue as numbered in the dissertation’s Bibliography section, including UBEM reviews, EPC studies, clustering methodology, Swedish regulations, LiDAR resources, SMHI datasets, heritage guidelines (EN 16883:2017), CHI sources, and sufficiency literature.)

Acknowledgment

This synthesis is based entirely on Lukas Dahlström (2025), Building Sustainability in Regional Energy Transition: Large‑Scale Energy Demand Modelling Considering Socio‑Economic Factors and Heritage Values, Acta Universitatis Upsaliensis, Digital Comprehensive Summaries, No. 2586.