Authors: Berk Ekici, Tuğçe Kazanasmaz, Michela Turrin, M. Fatih Tasgetiren, I. Sevil Sariyildiz

This study presents a methodology to optimise daylight performance of a whole high-rise building located in a dense urban district considering a variety of parameter sets at different zones of the building. For example, daylight availability in the upper zones of the high-rise building, which are close to the sky, is different than daylight availability in the lower zones near the ground level, since the surrounding buildings cause obstruction on the facade in dense urban environments.

Here you can find the paper “A Methodology for daylight optimisation of high-rise buildings in the dense urban district using overhang length and glazing type variables with surrogate modelling”

There are four steps which are proposed in methodology. These are:

• Form finding: A parametric model of a high-rise building is generated by defining design parameters. The building is equally divided into 5 zones (zone 1-5) for the performance assessment. The floors in the middle of each zone are selected for the simulation.
• Performance evaluation: Corresponding floors are selected for daylight simulation in all zones. Spatial Daylight Autonomy (sDA), a metric for sufficient daylight illuminance, and Annual Sunlight Exposure (ASE), a metric for the potential visual discomfort owing to the direct sunlight, are considered to assess the daylight performance in each zone.
• Surrogate modelling: Uniformly generated samples are collected for each part to define the fitness function. The Dodo plug-in is used to train an artificial neural networks (ANN) as a surrogate model. 500 samples are collected for each zone to train the surrogate model. The ANN uses backpropagation and bipolar sigmoid activation functions. The predicted values (outputs) obtained from ANN are compared to outputs from Diva simulations to assess the accuracy of the ANN model.
• Optimisation: The most desirable parameter sets in each zone are discovered using a computational optimisation algorithm. The single-objective self-adaptive differential evolution (jDE) algorithm presented in J. Brest, V. Zumer, and M. S. Maucec, “Self-adaptive differential evolution algorithm in constrained real-parameter optimization,” in 2006 IEEE international conference on evolutionary computation, 2006, pp. 215-222: IEEE (and coded in C#) is used for optimisation.

LIMITATIONS: Although optimised solutions were not checked against thermal performance, higher ASE values (2331%) can be interpreted as indicating overheating. Therefore, the study can be an initial step that suggest if further testing of thermal energy consumption is required. This would most likely happen for high-rise buildings in temperate-humid climates. In this situation, the zones at varying levels of high-rise buildings will likely require different combinations of parameter sets to result in the optimally performing building.

Software & Plugins Used

Rhinoceros, Grasshopper as the main interface to generate the parametric model Grasshopper plugins:

• DIVA for daylight, glare, and performance simulations
• Dodo for Artificial Neural Network trained as surrogate model

Coding in C# (inside Grasshopper) to generate uniformly distributed random values Single-objective self-adaptive differential evolution (jDE) algorithm for optimization, algorithm presented in J. Brest, V. Zumer, and M. S. Maucec, “Self-adaptive differential evolution algorithm in constrained real-parameter optimization,” in 2006 IEEE international conference on evolutionary computation, 2006, pp. 215-222: IEEE (and coded in C#)

Paper Information

• Title: A Methodology for daylight optimisation of high-rise buildings in the dense urban district using overhang length and glazing type variables with surrogate modelling
• Author(s): Berk Ekici, Tuğçe Kazanasmaz, Michela Turrin, M. Fatih Tasgetiren, I. Sevil Sariyildiz
• Year: 2019
• Link: https://repository.tudelft.nl/islandora/object/uuid:a552e612-e334-467e-9885-c2688f5138df?collection=research
• Type: PHD paper
• ML tags: Surrogate modelling, Neural Networks
• Topic tags: Architectural Design, Building Geometry, Building Energy performance