Adaptation of Kriging in daylight modeling for energy simulation

Research output: Contribution to journalArticle

Abstract

Kriging is a group of geo-statistical techniques used to interpolate the value of a random field at an unobserved location from discrete observed values at nearby locations. Kriging has been very popular in oceanic research used for predicting the variance of geographical conditions. However, it has been difficult to find a similar application in the built environment. This paper aims to uncover possibilities of using Kriging for built environment studies, specifically focusing on indoor daylight modeling. Currently, most energy simulation tools use algorithms that are based on Daylight Factor (DF) [1] to capture the complexity of daylight behavior. However, DF has limitations in its ability to fully account for realistic daylight conditions. Thus, efficiently integrating the light domain with thermal calculations remains a major hurdle for building energy simulation tools [2]. For this reason, the paper aims to find possibilities of using Kriging for daylight modeling. Kriging will use sample points to predict illuminance values of nearby locations. Kriging will considerably reduce the number of points needed to map indoor light intensity profiles. Furthermore, Kriging will be able to provide a more seamless daylight model for energy simulation tools. This proposed method can produce light distribution, such as illuminance (lux), while obtaining more realistic results than DF and at the same time reduce computational expense through statistical methods. The paper first investigates accuracy to determine whether Kriging modeling can interpolate the profile of illuminance from a select sample of data points. To validate the daylight model used in this paper, real indoor illuminance levels were measured for different days and compared with the daylight simulation model. Further, this paper tests Kriging predictions and compares them with both advanced physics-based light simulation and energy simulation to determine possibilities and limitations of the proposed method. In comparison with RADIANCE, this paper found that the proposed Kriging model was able to predict better than the energy simulation daylight model. The difference between Kriging and RADIANCE was 25.52 lux. The Kriging model showed great promise since it was able to cut the full RADIANCE simulation time in half and proved to be much closer to RADIANCE predictions than other methods.

Original languageEnglish (US)
Pages (from-to)479-496
Number of pages18
JournalEnergy and Buildings
Volume111
DOIs
StatePublished - Jan 1 2016

Fingerprint

Daylight simulation
Statistical methods
Physics

Keywords

  • Building energy
  • Daylight
  • EnergyPlus
  • Kriging
  • RADIANCE
  • Simulation

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Building and Construction
  • Mechanical Engineering
  • Electrical and Electronic Engineering

Cite this

Adaptation of Kriging in daylight modeling for energy simulation. / Yi, Yun Kyu.

In: Energy and Buildings, Vol. 111, 01.01.2016, p. 479-496.

Research output: Contribution to journalArticle

@article{2f57545eca244b82ad7dab8cdd027c85,
title = "Adaptation of Kriging in daylight modeling for energy simulation",
abstract = "Kriging is a group of geo-statistical techniques used to interpolate the value of a random field at an unobserved location from discrete observed values at nearby locations. Kriging has been very popular in oceanic research used for predicting the variance of geographical conditions. However, it has been difficult to find a similar application in the built environment. This paper aims to uncover possibilities of using Kriging for built environment studies, specifically focusing on indoor daylight modeling. Currently, most energy simulation tools use algorithms that are based on Daylight Factor (DF) [1] to capture the complexity of daylight behavior. However, DF has limitations in its ability to fully account for realistic daylight conditions. Thus, efficiently integrating the light domain with thermal calculations remains a major hurdle for building energy simulation tools [2]. For this reason, the paper aims to find possibilities of using Kriging for daylight modeling. Kriging will use sample points to predict illuminance values of nearby locations. Kriging will considerably reduce the number of points needed to map indoor light intensity profiles. Furthermore, Kriging will be able to provide a more seamless daylight model for energy simulation tools. This proposed method can produce light distribution, such as illuminance (lux), while obtaining more realistic results than DF and at the same time reduce computational expense through statistical methods. The paper first investigates accuracy to determine whether Kriging modeling can interpolate the profile of illuminance from a select sample of data points. To validate the daylight model used in this paper, real indoor illuminance levels were measured for different days and compared with the daylight simulation model. Further, this paper tests Kriging predictions and compares them with both advanced physics-based light simulation and energy simulation to determine possibilities and limitations of the proposed method. In comparison with RADIANCE, this paper found that the proposed Kriging model was able to predict better than the energy simulation daylight model. The difference between Kriging and RADIANCE was 25.52 lux. The Kriging model showed great promise since it was able to cut the full RADIANCE simulation time in half and proved to be much closer to RADIANCE predictions than other methods.",
keywords = "Building energy, Daylight, EnergyPlus, Kriging, RADIANCE, Simulation",
author = "Yi, {Yun Kyu}",
year = "2016",
month = "1",
day = "1",
doi = "10.1016/j.enbuild.2015.11.036",
language = "English (US)",
volume = "111",
pages = "479--496",
journal = "Energy and Buildings",
issn = "0378-7788",
publisher = "Elsevier BV",

}

TY - JOUR

T1 - Adaptation of Kriging in daylight modeling for energy simulation

AU - Yi, Yun Kyu

PY - 2016/1/1

Y1 - 2016/1/1

N2 - Kriging is a group of geo-statistical techniques used to interpolate the value of a random field at an unobserved location from discrete observed values at nearby locations. Kriging has been very popular in oceanic research used for predicting the variance of geographical conditions. However, it has been difficult to find a similar application in the built environment. This paper aims to uncover possibilities of using Kriging for built environment studies, specifically focusing on indoor daylight modeling. Currently, most energy simulation tools use algorithms that are based on Daylight Factor (DF) [1] to capture the complexity of daylight behavior. However, DF has limitations in its ability to fully account for realistic daylight conditions. Thus, efficiently integrating the light domain with thermal calculations remains a major hurdle for building energy simulation tools [2]. For this reason, the paper aims to find possibilities of using Kriging for daylight modeling. Kriging will use sample points to predict illuminance values of nearby locations. Kriging will considerably reduce the number of points needed to map indoor light intensity profiles. Furthermore, Kriging will be able to provide a more seamless daylight model for energy simulation tools. This proposed method can produce light distribution, such as illuminance (lux), while obtaining more realistic results than DF and at the same time reduce computational expense through statistical methods. The paper first investigates accuracy to determine whether Kriging modeling can interpolate the profile of illuminance from a select sample of data points. To validate the daylight model used in this paper, real indoor illuminance levels were measured for different days and compared with the daylight simulation model. Further, this paper tests Kriging predictions and compares them with both advanced physics-based light simulation and energy simulation to determine possibilities and limitations of the proposed method. In comparison with RADIANCE, this paper found that the proposed Kriging model was able to predict better than the energy simulation daylight model. The difference between Kriging and RADIANCE was 25.52 lux. The Kriging model showed great promise since it was able to cut the full RADIANCE simulation time in half and proved to be much closer to RADIANCE predictions than other methods.

AB - Kriging is a group of geo-statistical techniques used to interpolate the value of a random field at an unobserved location from discrete observed values at nearby locations. Kriging has been very popular in oceanic research used for predicting the variance of geographical conditions. However, it has been difficult to find a similar application in the built environment. This paper aims to uncover possibilities of using Kriging for built environment studies, specifically focusing on indoor daylight modeling. Currently, most energy simulation tools use algorithms that are based on Daylight Factor (DF) [1] to capture the complexity of daylight behavior. However, DF has limitations in its ability to fully account for realistic daylight conditions. Thus, efficiently integrating the light domain with thermal calculations remains a major hurdle for building energy simulation tools [2]. For this reason, the paper aims to find possibilities of using Kriging for daylight modeling. Kriging will use sample points to predict illuminance values of nearby locations. Kriging will considerably reduce the number of points needed to map indoor light intensity profiles. Furthermore, Kriging will be able to provide a more seamless daylight model for energy simulation tools. This proposed method can produce light distribution, such as illuminance (lux), while obtaining more realistic results than DF and at the same time reduce computational expense through statistical methods. The paper first investigates accuracy to determine whether Kriging modeling can interpolate the profile of illuminance from a select sample of data points. To validate the daylight model used in this paper, real indoor illuminance levels were measured for different days and compared with the daylight simulation model. Further, this paper tests Kriging predictions and compares them with both advanced physics-based light simulation and energy simulation to determine possibilities and limitations of the proposed method. In comparison with RADIANCE, this paper found that the proposed Kriging model was able to predict better than the energy simulation daylight model. The difference between Kriging and RADIANCE was 25.52 lux. The Kriging model showed great promise since it was able to cut the full RADIANCE simulation time in half and proved to be much closer to RADIANCE predictions than other methods.

KW - Building energy

KW - Daylight

KW - EnergyPlus

KW - Kriging

KW - RADIANCE

KW - Simulation

UR - http://www.scopus.com/inward/record.url?scp=84950131760&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84950131760&partnerID=8YFLogxK

U2 - 10.1016/j.enbuild.2015.11.036

DO - 10.1016/j.enbuild.2015.11.036

M3 - Article

AN - SCOPUS:84950131760

VL - 111

SP - 479

EP - 496

JO - Energy and Buildings

JF - Energy and Buildings

SN - 0378-7788

ER -