A framework for optimising the CO2 reduction Performance of green walls using machine learning and evolutionary algorithms

Jafar, Rawan 2025. A framework for optimising the CO2 reduction Performance of green walls using machine learning and evolutionary algorithms. PhD Thesis, Cardiff University. Item availability restricted.

	PDF - Accepted Post-Print Version Restricted to Repository staff only until 14 October 2026 due to copyright restrictions. Download (37MB)
	PDF (Cardiff University Electronic Publication Form) - Supplemental Material Restricted to Repository staff only Download (186kB)

Abstract

Abstract This study aims to investigate the effect of using green walls to enhance indoor air quality and provide a useful aid for designers and architects to integrate green walls within their designs through the development of a framework for a simulation tool that can assess designers' and architects' implementation of green walls within indoor environments, especially in office buildings. This research focuses on the phytoremediation potential of green walls, using carbon dioxide (CO2) reduction as an indicator of Indoor air quality (IAQ) and the plant's ability to perform photosynthesis, which is a fundamental process for plants. The study combines lab experiments, parametric/algorithmic performance-based predictive models, surrogate model/ machine learning, and multi-objective optimisation to design green wall systems. Indoor air quality of buildings has been recognised as a key factor affecting the health and wellbeing of occupants and their mental productivity. Various studies have suggested nature-based solutions as a feasible and practical approach to improving indoor air quality. Plants can substantially mitigate indoor air pollution only in the correct environmental conditions, specifically photosynthetically active radiation (PPFD) levels, as light plays a key role in the plant's ability to perform photosynthesis, which suggests that optimising light conditions is critical to maximise the positive impact of green walls. This dissertation uses computational methods and simulation tools to address a significant gap between lab experimental conditions and in situ studies, to provide designers and architects with scientifically evidenced tools. A digital simulation tool was developed using Honeybee sunlight simulation and lab experiments to create a light response curve for Epipremnum aureum, a species well-regarded for its indoor air purification properties, as the focus of experimental investigations. A regression model with R2 of 0.997 was derived from the light response curve. The regression model between light intensities and CO2 uptake was then used in the simulation tool to predict the CO2 reduction of the green walls in IV indoor environments. This tool was used to create a synthetic dataset, which was then employed to train a two machine-learning-based ANN surrogate model. The surrogate models enable fast prediction of CO₂ reduction performance for diverse design scenarios while mitigating high computational demands. The two developed surrogate models were to predict the CO2 reduction of the wall one when leaves were on one side of the wall with R2 of 0.972, and the other when it had leaves on both sides with R2 of 0.986. These surrogate models were integrated into the multi-objective optimisation framework as fitness functions for CO2 reduction in addition to recommended indoor daylight and quality of the view. The optimisation results showed a multiplication in the CO2 reduction throughout the year by adding vegetation on both sides of the wall and adding more than one wall in the space without compromising the quality of the recommended levels of the window view or indoor daylight. This research contributes an application-ready design tool that offers architects and designers a scalable green wall design solution for enhancing indoor air quality in office buildings. By advancing our understanding of indoor plant performance, the framework and the developed simulation tool support informed decision-making in the design process

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Schools > Architecture
Date of First Compliant Deposit:	14 October 2025
Last Modified:	14 Oct 2025 10:47
URI:	https://orca.cardiff.ac.uk/id/eprint/181641

Actions (repository staff only)

Edit Item