Clustering Smart Retrofit Typologies through Building’s Energy Performance Certificate (EPC) Data: A K-Means Approach to Scalable Smart Spatial Planning, A Case of Cumbria, UK.

TY - GEN

T1 - Clustering Smart Retrofit Typologies through Building’s Energy Performance Certificate (EPC) Data

T2 - A K-Means Approach to Scalable Smart Spatial Planning, A Case of Cumbria, UK.

AU - Forghaniallahabadi, Maryam

PY - 2025

Y1 - 2025

N2 - The accelerating demand for energy-efficient urban transformation has positioned building retrofit as a central strategy in achieving net-zero objectives. However, retrofit planning remains hindered by the lack of typological insight into how energy performance varies across spatial and structural contexts. This paper develops a data-driven approach to classify building retrofit typologies using K-means clustering of Energy Performance Certificate (EPC) data in Cumbria, England, United Kingdom (UK). The analysis integrates four quantitative indicators, i.e., ‘efficiency gap’, ‘CO₂ reduction potential’, ‘energy-saving potential’, and ‘total floor area’ to reveal intrinsic patterns in building performance. Results identify four distinct clusters representing differentiated retrofit priorities: (1) High-Impact Retrofit stock with large efficiency gaps and substantial CO₂ reduction potential; (2) Large Homes with moderate gains showing optimization rather than deep retrofit needs; (3) Efficient Low-Impact stock requiring minimal intervention; and (4) Transitional stock offering scalable opportunities for hybrid retrofit programs. The findings demonstrate how machine learning can translate building-level or local data into actionable intelligence for smart spatial planning. By linking micro-level building performance to meso and macro-level policy scales, the paper presents a scalable methodological framework to inform energy-efficient housing strategies and spatial coherence in smart urban development. This underscores the potential of clustering techniques in guiding evidence-based retrofit prioritization, advancing data-informed approaches to sustainable and scalable spatial planning.

AB - The accelerating demand for energy-efficient urban transformation has positioned building retrofit as a central strategy in achieving net-zero objectives. However, retrofit planning remains hindered by the lack of typological insight into how energy performance varies across spatial and structural contexts. This paper develops a data-driven approach to classify building retrofit typologies using K-means clustering of Energy Performance Certificate (EPC) data in Cumbria, England, United Kingdom (UK). The analysis integrates four quantitative indicators, i.e., ‘efficiency gap’, ‘CO₂ reduction potential’, ‘energy-saving potential’, and ‘total floor area’ to reveal intrinsic patterns in building performance. Results identify four distinct clusters representing differentiated retrofit priorities: (1) High-Impact Retrofit stock with large efficiency gaps and substantial CO₂ reduction potential; (2) Large Homes with moderate gains showing optimization rather than deep retrofit needs; (3) Efficient Low-Impact stock requiring minimal intervention; and (4) Transitional stock offering scalable opportunities for hybrid retrofit programs. The findings demonstrate how machine learning can translate building-level or local data into actionable intelligence for smart spatial planning. By linking micro-level building performance to meso and macro-level policy scales, the paper presents a scalable methodological framework to inform energy-efficient housing strategies and spatial coherence in smart urban development. This underscores the potential of clustering techniques in guiding evidence-based retrofit prioritization, advancing data-informed approaches to sustainable and scalable spatial planning.

KW - Smart spatial planning

KW - Energy Performance Certificates (EPC)

KW - Building energy typologies

KW - K-means clustering

KW - Data-driven planning

KW - Scalability

KW - Sustainable urban development

KW - Machine Learning

U2 - 10.20933/100001471

DO - 10.20933/100001471

M3 - Other contribution

PB - University of Dundee

ER -