Date

2022-12-26

Author

Canlı, İlkim

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

165
views

158
downloads

Cite This

Daylight illumination has been an essential consideration during design for architects throughout history. Daylight is a crucial design component for long-term sustainability that influences the visual and thermal comfort of the occupants and energy usage in buildings. Utilizing daylighting effectively reduces the energy required for artificial lighting and the indoor thermal loads of spaces. However, dense urban areas prevent daylight from reaching buildings. Each surrounding building acts as a shadow element obstructing the building's access to natural light. Therefore, analyzing daylight illuminances and understanding the building design characteristics and urban form parameters that affect daylight illuminance is unavoidable for sustainable building design. Simulations are one of the most preferred tools to analyze the level of illuminance in building designers. Simulations require detailed modeling knowledge and expertise to get precise results. Also, daylighting simulations performed at an urban scale take much computational time. In contrast, machine learning (ML) models enable designers to analyze daylighting levels with less computational time and detailed knowledge. This study aims to develop a method to predict hourly indoor daylighting illuminances in an urban context using ML models. For the development of the method, three different ML models (multi-layer perceptrons, random forest, extreme gradient boosting) were developed, and their performance results were compared. The ML model with the highest performance accuracy was selected as the final model. The developed method helps designers/ architects to analyze hourly indoor daylight illuminances in an urban context. The developed methodology also calculates how much daylight-dependent electric lighting is used in buildings by analyzing hourly indoor daylighting illuminances on an urban scale. The proposed methodology enables the integration of indoor daylighting analysis with the electric lighting energy consumption calculation based on real-time estimation of daylight illuminances. Residential units in the Bahçelievler neighborhood in Ankara were simulated using various design factors, and the simulation results were utilized for training and evaluating the machine learning models. The proposed model can enhance the usage of machine learning in architectural design stages to analyze daylight illuminances, accordingly, forecast the artificial electric load in buildings, and help designers integrate daylight into the buildings.

Subject Keywords

Daylighting, Daylighting Illuminance Prediction, Machine Learning Models, Urban Scale Daylighting Analysis

URI

https://hdl.handle.net/11511/101289

Collections

Graduate School of Natural and Applied Sciences, Thesis