PERFORMANCE BASED SPATIALLY CONSTRAINED CLUSTERING FOR BUILDING CLIMATE ZONING

Abstract

Climate Zoning for Buildings (CZB) plays a pivotal role in optimizing energy efficiency by tailoring strategies to localized climatic conditions. However, Kazakhstan’s existing CZB framework is based on simplistic climate indicators and is affected by inaccuracies and does not align well with actual building energy use. To address this gap, this thesis proposes a novel performance-based CZB methodology that integrates spatially constrained clustering algorithms (SKATER, REDCAP, and AZP) with building energy simulation (BES) data for Kazakhstan’s diverse climate regions. By enforcing contiguity in clustering, the approach ensures each climate zone is a geographically coherent region reflecting both climatic similarity and building energy performance. Clustering quality was evaluated with multiple indices – Adjusted Rand Index (ARI), Calinski–Harabasz Index (CHI), Davies–Bouldin Index (DBI), and a Mean Overlap Value (MOV) – to rigorously benchmark the new zones against the conventional scheme. Using a comprehensive climatic dataset (10 variables across 94 weather stations) and BES results for representative residential buildings, the performance-based CZB achieved markedly more homogeneous and distinct zones. Compared to the present map, the optimized zones revealed up to an 80% decrease in overlapping areas and a 50–70% increase in within-zone climate/energy consistency. These outcomes show the strength of the method by exceeding conventional k-means and official zoning techniques. This study is the first in Kazakhstan to combine spatial algorithms with building performance, and the results show that such methods have the potential to provide more accurate and actionable climate zone maps. S patially constrained CZB can help building designers, urban planners, and government officials in Kazakhstan create energy-efficiency policies and climate-responsive design rules with more confidence in their regional relevance.