A NOVEL APPROACH FOR CLIMATE CLASSIFICATION USING AGGLOMERATIVE HIERARCHICAL CLUSTERING

Abstract

Climate classification plays a significant role in the development of building codes and standards. It guides the design of building envelope and systems by considering their location’s climate conditions. ASHRAE standard 169, as well as established climate classification systems such as Köppen and Trewartha, employ meteorological parameters like temperature, humidity, solar radiation, and precipitation for such classifications. However, the above approaches often fall short in acknowledging the relation between climatic conditions and the energy consumption of buildings, a critical consideration for comprehensive energy efficiency assessments. This research employs clustering techniques to group cities into different climate zones based on the number of similar days. Two days are considered similar when the absolute difference between their meteorological parameters falls within a specified threshold range. A similarity matrix for the given set of cities is created using three key meteorological parameters: mean daily temperature, mean daily relative humidity, and mean daily solar radiation. This matrix, indicating the number of similar days between each pair of cities based on these meteorological characteristics, is used to cluster cities into distinct climate zones. To assess the quality of clustering, the simulated annual energy consumption of a standard building model for each city is used. The analysis focuses on three annual energy parameters: sensible cooling, latent cooling, and heating. The quality of the clustering is evaluated using the silhouette score method, which uses annual energy consumption data. The silhouette score method considers both inter- and intra-cluster distances, with the best value being 1, the worst -1, and values near 0 indicating overlapping clusters. The application of the proposed methodology to 786 U.S. cities' meteorological datasets has shown that the clustering, evaluated using silhouette score, computed across a set of threshold values (7 °C for daily mean temperature, 45 % for daily mean relative humidity, and 35 Wh/m² for daily mean solar radiation), has given a better clustering over the prevalent ASHRAE Standard 169 classification. The clustering achieves a better score (0.113) than ASHRAE Standard 169 (0.054).