Optimisation of light shelf for daylight-thermal balance and climate change impact in a hot and humid climate using machine learning

Abd. Salam, Nur Nasuha 2026. Optimisation of light shelf for daylight-thermal balance and climate change impact in a hot and humid climate using machine learning. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF - Accepted Post-Print Version Download (18MB) | Preview
	PDF (Cardiff University Electronic Publication Form) - Supplemental Material Restricted to Repository staff only Download (211kB)

Abstract

The global push for sustainable buildings aligns with the United Nations' Sustainable Development Goal (SDG) 2030 and net-zero energy building adoption to combat climate change and enhance efficiency. In Malaysia, environmental sustainability is a priority under the Construction Industry Transformation Programme (CITP), which aims to reduce carbon dioxide (CO₂) emissions from construction by 4 million metric tons annually. Residential and commercial buildings are a critical focus, accounting for 15% of the nation's energy use. Terrace housing, the largest residential type at 41% of the housing stock, is central to achieving Malaysia's carbon reduction goals. However, in the equatorial climate, sustainable design faces the challenge of balancing daylighting and thermal performance. Conventional solutions, like large overhangs, reduce solar heat gain but limit daylight, increasing the need for artificial lighting and reducing energy efficiency. This research addresses the challenges of balancing daylight and energy performance in equatorial terrace housing. It aims to optimise light shelf and facade designs to maximise daylight utilisation, minimise energy use, and ensure resilience to future climate. The research had three phases: assessing current and future climate impacts, conducting sensitivity analyses on eleven design variables, and implementing multi-objective optimisation using artificial neural network based surrogate modelling. The study found critical design factors influencing daylight and energy performance, though their rankings differed for each goal. Notably, future climate had minimal impact on these parameter rankings. The optimal solutions from multi-objective optimisation increased acceptable useful daylight illuminance by 13.4% and reduced cooling energy use intensity by 50% under current conditions. This performance was largely maintained under future climate, with only a 0.5% rise in cooling EUI due to higher outdoor temperatures. Analysis showed that incorporating a light shelf requires a smaller window-to-wall ratio, enhancing daylight even in deep floor plans. While multi-objective optimisation achieves optimal building performance, it is computationally intensive, taking about 400 hours per iteration. To address this, the research used artificial neural network-based surrogate models as proxies, demonstrating high accuracy and reducing simulation time by 87.6%. Statistical analysis confirmed no significant differences between predicted and simulated optimisation results, with the surrogate models producing comparable Pareto fronts. Applying the integrated workflow to multiple design variants for a large housing project 4 took 154.2 hours versus 1248 hours for the conventional method, a substantial enhancement in efficiency. This research developed an artificial neural network-based multi-objective optimisation (ANN-MOO) framework to improve daylight and thermal performance for terrace housing in hot, humid climates. The approach enables faster exploration of sustainable design options during early design, considering climate adaptability. The findings contribute to Malaysia's net-zero energy and carbon reduction goals, providing scalable solutions for sustainable urban development. The framework's adaptability to diverse urban housing scenarios demonstrates its potential to advance sustainable building practices worldwide.

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Schools > Architecture
Date of First Compliant Deposit:	5 May 2026
Last Modified:	06 May 2026 09:44
URI:	https://orca.cardiff.ac.uk/id/eprint/186788

Actions (repository staff only)

Edit Item