Identification of passive solar design determinants in office building envelopes in hot and humid climates using data mining techniques

Abstract

Passive solar design is an effective strategy to alleviate the energy-intensive status of the building sector. Identifying passive solar design parameters that significantly impact office buildings’ energy performance can further understand sustainable design principles and prioritize energy efficiency measures. This study proposes a holistic methodology integrating data mining techniques and parametric energy simulation to explore the critical design parameters in passive solar office building envelopes in hot and humid climates. The data mining module incorporates Extreme Gradient Boosting Decision Tree (XGBoost) and association rule mining to measure feature importance and extract strong correlations, respectively. A case study, using a typical office building in Guangzhou, China as a reference building, is conducted to demonstrate the implementation procedure and feasibility of the proposed approach. In total, 115,200 design scenarios are created and simulated in EnergyPlus software. The results of XGBoost show that the glazing system, window-to-wall ratio, and roof coating are the most critical design factors, with importance scores of 0.4858, 0.3197, and 0.1297, respectively. Similarly, based on a confidence threshold of 30% and a lift threshold of 3.0, the extracted association rules indicate that the above three factors have the strongest correlations with the energy consumption level. Findings of this study will provide practical passive solar design guidance for office buildings in hot and humid climates to achieve energy-saving targets. Also, the developed simulation-based data mining method can be applied to other building types in different climates.