ArticlePDF Available

Factors Affecting Energy Performance of Large-Scale Office Buildings: Analysis of Benchmarking Data from New York City and Chicago


Abstract and Figures

Buildings in high-income, industrialized cities are responsible for more than 50% of global energy consumption; consequently, many developed cities have legislated energy benchmarking and disclosure policies to understand their buildings’ energy-use dynamics better. By utilizing these benchmarking data and additional information taken from 3D models, this paper presents a comprehensive analysis of large-scale office buildings located in New York and Chicago, with respect to their energy use intensity (EUI). To identify the primary factors affecting the EUI, Spearman’s correlation analysis and multiple variate regression tests were performed on office buildings over 500,000 ft2 (46,452 m2) gross floor area. The results showed the number of floors, construction year, window-to-wall ratio (WWR), and source-to-site ratio statistically significant, while morphological factors such as the relative compactness and surface-to-volume ratio showed limited relation to EUI. In New York City, the smallest EUI median was found in the buildings with 20 to 30 floors, and in Chicago, the buildings with 60 floors or more. A higher source-to-site ratio generally had lower overall EUI in both cities. Despite the high correlation, different kinds of dependency were found for window-to-wall ratio (WWR) and construction year between NYC and Chicago. These findings highlight the relative role that each building’s characteristics play concerning the EUI, depending on the particular building’s typology, scale, and the urban context.
Content may be subject to copyright.
Energies 2019, 12, 4783; doi:10.3390/en12244783
Factors Affecting Energy Performance of Large-Scale
Office Buildings: Analysis of Benchmarking Data
from New York City and Chicago
ChungYeon Won
, SangTae No
* and Qamar Alhadidi
Department of Architecture, School of Engineering, Korea University, 145 Anam-ro, Seongbuk-gu, Seoul
02841, Korea;
Department of Architecture, Korea National University of Transportation, 50 Daehak-ro, Geomdan-ri,
Daesowon-myeon, Chungju-si, Chungcheongbuk-do 27469, Korea
Department of Architecture, School of Engineering, Korea University, 145 Anam-ro, Seongbuk-gu, Seoul
02841, Korea;
* Correspondence:; Tel.: +82-43-841-5204
Received: 11 November 2019; Accepted: 12 December 2019; Published: 15 December 2019
Abstract: Buildings in high-income, industrialized cities are responsible for more than 50% of global
energy consumption; consequently, many developed cities have legislated energy benchmarking
and disclosure policies to understand their buildings’ energy-use dynamics better. By utilizing these
benchmarking data and additional information taken from 3D models, this paper presents a
comprehensive analysis of large-scale office buildings located in New York and Chicago, with
respect to their energy use intensity (EUI). To identify the primary factors affecting the EUI,
Spearman’s correlation analysis and multiple variate regression tests were performed on office
buildings over 500,000 ft
(46,452 m
) gross floor area. The results showed the number of floors,
construction year, window-to-wall ratio (WWR), and source-to-site ratio statistically significant,
while morphological factors such as the relative compactness and surface-to-volume ratio showed
limited relation to EUI. In New York City, the smallest EUI median was found in the buildings with
20 to 30 floors, and in Chicago, the buildings with 60 floors or more. A higher source-to-site ratio
generally had lower overall EUI in both cities. Despite the high correlation, different kinds of
dependency were found for window-to-wall ratio (WWR) and construction year between NYC and
Chicago. These findings highlight the relative role that each building’s characteristics play
concerning the EUI, depending on the particular building’s typology, scale, and the urban context.
Keywords: energy benchmarking; large-scale office buildings; energy disclosure policy; source
energy use intensity; site energy use intensity
1. Introduction
Several decades have passed since the building industry recognized the importance of
monitoring the actual energy use intensity of existing buildings. Greater information and data on
energy consumption of buildings enables owners, operators, and tenants to make informed energy
management decisions. Transparent, timely information can help track performance against goals,
and the collection of general statistical information about buildings’ energy use enables better policy
and program design [1]. For this critical data gathering, an increasingly popular policy, which has
been adopted in many European and US cities, is the requirement that building owners disclose their
annual energy use and benchmark it relative to other buildings [2]. New York and Chicago, the two
subject cities of this research, are among the early adopters of this policy and have accumulated data
Energies 2019, 12, 4783 2 of 16
since 2009 and 2013, respectively, for buildings over 50,000 ft² (4645 m²) [3,4]. Among the various
building typologies of such developed, high-income cities, large-scale office buildings are ubiquitous,
compose a large portion of the overall area, and are considered one of the most energy-intensive. In
Manhattan, in terms of the area sum of space taken, nearly 90% of the overall office building stock is
accounted for by large office buildings over 500,000 ft² (46,452 m²) [5]. Due to their sizeable
quantitative portion compared to various other building groups, understanding the energy-
consumption dynamics and prominent features of these large-scale office buildings are crucial.
Already many types of research are available that deep dive into what the accumulated data tell
us. The prevalence of the energy disclosure policy and worldwide benchmarking data have enabled
diverse research that has revealed attributes and patterns of building energy consumptions; a lot of
this research has focused on instituting rigorous methodologies for analyzing performance patterns
over time or how to take multiple features into account when trying to understand energy use
intensity (EUI) dynamics [6–8]. For New York City, by using a K-means clustering algorithm,
Papadopoulos et al. [6] showed energy reductions are mostly driven by office buildings, with larger,
newer, and higher-value buildings showing significant improvement in EUI between 2011 and 2016.
Gao and Malkawi [7] also utilized the clustering method to demonstrate that multidimensional
similarity can be used as a ‘performance typology’ to define building type; this approach is similar to
the traditional one-dimensional use type but is much more comprehensive. In this way, the energy
performance of the buildings considered can be more properly benchmarked. A study utilizing the
UK’s DEC (Display Energy Certificate) data from public schools conducted by UCL (University
College London) [8] also address the alternative analysis method, which compares the top-down and
bottom-up approaches. They concluded the top-down approach utilizing descriptive statistics and
artificial neural networks (ANN) presents many benefits.
Some research has confirmed widely held beliefs, such as the relation between the building’s use
type and energy use intensity in detail. It was shown that building-use typology is the key
determinant of occupant density and operating hours of a building, which inevitably becomes the
most influential factor affecting the energy use intensity of the whole. For example, even though
multifamily properties significantly outnumber offices in many cities, the office sector is the more
energy-intensive of the two, using 50 percent more energy per square foot than the multifamily
building sector in the case of NYC [5]. Within an identical floor area, the number of occupants in an
office is usually much higher than in residential buildings, hence more energy is used. In addition,
Constantine [9], who analyzed NYC’s 2010 LL84 data for commercial buildings, found that increased
operating hours and occupant density results in higher EUI, which is a predictable result and aligns
with the empirical data from the Commercial Buildings Energy Consumption Survey (CBECS).
One of the important variables included in this research, which has also been a subject of several
previous studies, is the relationship between building height and energy consumption. Godoy-
Shimizu et al. [10] studied 611 office buildings in England and Wales, concluding there was a
significant energy use increase for high-rise offices defined as 10 stories or above, compared to low
and mid-rise buildings [10]. Another study from the UK, which compared buildings with six stories
or fewer with buildings with 20 stories or more, concluded the electricity use in the high-rise
buildings was nearly two and a half times greater than in low-rise buildings, and that carbon
emissions were more than doubled when going from ‘low’ to’ high-rise’ [11]. Guthrie [12] concluded,
based on research from Hong Kong’s 20 commercial office buildings, that tall buildings generally use
more energy. Even though the results appear to be consistent that, the higher the building, the more
energy-intensive it is, these results are limited to very specific contexts and locations. For example, in
NYC and Chicago, buildings with 20 stories could be considered as low or mid-rise, depending on
the location. If we consider relative difference in defining high-rise buildings, the results from
previous research cannot be directly applied to all cities, and further investigation is required.
Meanwhile, other research related to the building morphology of large-scale office buildings
besides the height focuses on hypothetical building geometries and orientations and provides results
from model simulations [13–15]. Many studies have found a strong correlation between the shape of
a commercial building, the relative compactness, the percent glazing, and the building’s total energy
Energies 2019, 12, 4783 3 of 16
use intensity [16]. However, there is limited research discussing the subject in conjunction with the
actual geometries and data from real buildings, let alone studies that look at the relative effect
compared to other factors, such as the construction year and building systems. Even though energy
consumption is becoming one of the key factors in determining a building’s form, many other urban,
financial, and legal factors limit architects from exploring the optimal shape for energy use. When we
attempted to categorize the building geometries from these two cities, quite distinct types were
identifiable, which are quite different from the assumed models used in previous researches. For
example, circular, oval, and triangular buildings with plan aspect ratio beyond 1:3 or with courtyards
were rarely found in this specific group [15,17,18]. Therefore, to understand the morphology of
existing buildings, a different approach may be necessary; our assumptions are further discussed in
detail later in this paper.
Even though several previous studies have offered methodologies to improve how we analyze
the ever-increasing benchmarking data or have informed us of the use patterns of a specific region or
a typology; none have attempted to compare data from a group of buildings that share multiple
attributes, such as the program or size, from more than one city. Based on a comprehensive analysis
of the large-scale office building’s energy disclosure data from these two major US cities, this research
seeks to investigate the common factors that affect the energy use intensity in these highly developed
urban environments. In addition to the disclosure data, this research analyzes the formal attributes,
such as the compactness factors, and it introduces a new morphological variable considering the
vertical distribution of the building area, as well as a ratio between the site and source energy use
intensity (EUI).
2. Materials and Methods
2.1. Data Sources and Preparation
The data for this research are primarily from New York and Chicago’s energy benchmarking
disclosure reports from 2015 to 2018; the reported data include information from prior years of
reporting [19–21]. The disclosure data acquired from both cities’ websites include but are not limited
to the following information: building identification number (BIN), property name, address, primary
and secondary property type, gross floor area, year built, Energy Star score, weather normalized site
EUI, weather normalized source EUI, electricity use, natural gas use, district steam use, and oil fuel
use. Additional data, such as the number of floors, floor-to-area ratio (FAR), construction, and
renovation dates (latest construction dates), were added from various online resources, as indicated
in Table 1. Other physical characteristics, including the compactness factor (CF), relative compactness
(RC) [22], and morphology types, are generated from the 3D models. Window-to-wall ratios (WWR)
are calculated based on the available plans, elevations, and photos of each selected sample building.
Table 1 summarizes the data sources for each variable.
Table 1. Data source of variables for analysis.
Variable Type New York Chicago
Physical Parameters
Number of Floors,,
Gross Floor Area
Floor Area Ratio Calculated from Chicago Data
Disclosure and Portal
Compactness Factor Calculated from 3D Models1 Calculated from 3D Models 2
Relative Compactness Calculated from 3D Models1 Calculated from 3D Models 2
Window/Wall Ratio Calculated from Google Earth Pro,
Photos, Plans, Elevations
Calculated from Google Earth Pro
Photos, Plans, Elevations
Morphology Type Calculated from 3D Models1 Calculated from 3D Models 2
Consumption Data 1
Source/Site EUI NYC Energy Data Disclosure Chicago Energy Data Disclosure
Energies 2019, 12, 4783 4 of 16
Use NYC Energy Data Disclosure Chicago Energy Data Disclosure
Source Site Ratio Calculated from NYC Energy Data
Calculated from Chicago Energy Data
Year Constructed NYC Energy Data Disclosure Chicago Energy Data Disclosure
Recent Renovation Date NYC City DOB Chicago City DOB
Notes: 1 downloaded from NYC Department of City Planning (updated in 2018), open data website; 2
downloaded from The City Project by boscorelli3D (updated August in 2017).
The initial benchmarking data acquired from both cities contained information on
approximately 13,000 buildings in NYC and 7200 buildings in Chicago. This research focuses only on
large-scale buildings used primarily for commercial offices. First, the properties whose primary and
secondary use type was defined as offices with a gross floor area of 500,000 ft² (46,452 m²) or higher
were included. Other reports [5], as well as the CBEC data—public use microdata file published by
the US Department of Energy [23] have used this 500,000 ft² (46,452 m²) limit for defining the category
of ‘very large’ buildings. Next, based on the provided source and site EUI, we excluded the top and
bottom 5% outliers, which were approximately above 350 kBTU/ft² (1104 kWh/m²) or below 50
kBTU/ft² (158 kWh/m²) in case of the source EUI. These outliers included a large proportion of
abnormal space use—for example, data centers or storage—and would have affected the average
significantly. The final number of building samples that resulted in NYC was 221, and in Chicago, it
was 106, and almost all were located in the central areas (Figures 1 and 2).
Figure 1. NYC (New York City) office buildings larger than 500,000 ft² (46,452 m²).
Figure 2. Chicago office buildings larger than 500,000 ft2 (46,452 m²).
Also, the total number of reported buildings and the data from those buildings have been
increasing every year between 2014 and 2018, for both cities. To utilize the most consistent data and
maximize the number of samples, we calculated the average EUI for each building from all available
years. In addition, for each building, abnormal EUIs compared to other years were omitted since
these may have been caused by temporary high vacancy rates or construction activities. Other
Energies 2019, 12, 4783 5 of 16
consumption data and the source-to-site ratio included in this paper are also averages of the available,
years as described above. Figure 3 summarizes the process taken for preparing the data.
Figure 3. Summary of data preparation methodology and process.
2.2. Physical Variables
In this section, we describe in detail a few variables that require additional explanations. Based
on a previous literature review, compactness factor (CF) (1) and relative compactness (RC) (2) are
defined as below, and for this research [22], information is taken from the 3D models.
Building Surface Area
CF Compactness Factor = Building Volume (1)
 
Vol ume/Surf ace Area Bui ldi ng Surface Area Refer ence Cube
RC Rel a t i ve Compact nes s = =
Volume/ Surface Area Reference Cube Surface Area Bui ldi ng
Even though previous studies present a strong correlation between this compactness aspect of a
building and its energy consumption, those researches looked at buildings that are relatively small
in scale and low in height [16]. For large-scale, high-rise buildings, the shapes concerning how the
area is vertically distributed could have additional advantages or disadvantages for energy
consumption. Hence, for this research, we have added a variable that categorizes the buildings into
• Collection of benchmarking disclosure data for New York and Chicago reported between
the years 2015 and 2018 (data for 2014-2017).
Raw Data
• Office as the primary and secondary property type.
• Gross Floor Area to 500,000 ft² (46,452 m²) or higher.
• Exclude abnormal data, top and bottom 5% (Source EUI above 350 or below 50).
• Eliminate Samples with missing values for Weather Normalized Source EUI and Weather
Normalized Site EUI.
Data Filtering Process
• Number of floors;
• Floor Area Ratio (FAR);
• Initial Construction Date;
• Latest Major Renovation Date.
Additional Data Set 1 (Websites and City's Building Department)
• Building Height;
• Building Exterior Surface Area;
• Morphology Type;
• Compactness Factor (CF);
• Relative Compactness (RC);
• Window to Wall Ratio (WWR).
Additional Data Set 2 (Calculated from 3D Models, Online Resources)
• Combine all data in Excel Sheet;
• Unify units and measurements;
• Average Consumption Data from 2014 to 2017;
• Export data to SPSS for processing and analyzing.
Cleaning and Exporting Data
Energies 2019, 12, 4783 6 of 16
five types representing different forms of vertical distribution found in specific urban locations, as
shown in Figure 4.
T1. Horizontal
T2. Vertical
Type 1
T3. Tower on
Podium T4. Stepping
T5. Vertical
Type 2
Overall volume is
more horizontally
dispersed L > H
medium slender
ratio, L < H < 2L
Podium volume
is more than 30%
of the overall
volume with
height not
exceeding 120 ft
reduces from
the lower levels
to higher levels
volume, high
slender ratio, 2L
< H
Figure 4. Morphology type: types defined to categorize vertical space and volume distribution.
In addition, the source-to-site ratio is another variable introduced in this research which will
inform the relative proportion of energy-source types used for each building. Source energy converts
energy use, considering the primary energy, which accounts for the raw fuel that is burned to create
heat and electricity, such as natural gas or fuel oil. The conversion factor currently used for grid-
purchased electricity in the US is 2.8 compared to 1.05 for gas [24]. Due to the limited data available
regarding the actual MEP systems, we believe this information will inform certain patterns in
conjunction with the quantitative-use data for each energy resource’s consumption.
2.3. Descriptive Data
Table 2 lists a summary of the processed data, which include 221 buildings from NYC and 106
from Chicago. NYC’s median site EUI is 275 (kWh/m²), and its source EUI is 670 (kWh/m²); Chicago’s
median site EUI is 232 (kWh/m²), and its source EUI is 585 (kWh/m²). Compared with the citywide
medians for all offices, these numbers are approximately 10% higher in NYC and almost identical in
Chicago [19], which implies no strong correlation between the building floor area and energy use
intensity. One compelling aspect when we compare the two cities is that Chicago’s EUIs are lower
than NYC’s (Figure 5), and this is somewhat unexpected, considering the degree-days. Chicago has
a slightly colder climate, with a lower average temperature in winter and similarly in summer.
During the 4 years of data collection, Chicago had approximately 1100 °C longer heating degree days
(HDD) and 340 °C shorter cooling degree days (CDD) [25]. To verify the assumptions, we performed
energy simulation with a generic building mass of 1000 m² by utilizing the Prototype Building Models
provided by the US Department of Energy. The results indicated Chicago was using 7% more energy.
Even though detailed analysis is required for an accurate comparison, it is an industry-wide
understanding that heating requires more energy than cooling. In the US, previous studies have
shown a typical central air conditioner is about four times more energy-efficient than a typical furnace
or boiler [26]. However, Chicago’s reported site and source EUI were lower than NYC’s, which
suggests the superior energy efficiency of buildings in Chicago.
Energies 2019, 12, 4783 7 of 16
Table 2. Descriptive statistics of NYC and Chicago.
NYC Variables Descriptive Statistics
Variables Min Max Mean Median
Statistic Std. Error
Number of Floors 5 102 33.64 34.00 13.080 0.606 0.164
Gross Floor Area (m²) 46,475 337,859 93,666 78,540 49,583 1.821 0.164
Floor Area Ratio (%) 2.88 48.08 20.42 20.27 6.11 0.816 0.164
Compactness Factor (CF) 0.072 0.262 0.124 0.121 0.026 1.399 0.164
Relative Compactness (RC) 0.313 0.948 0.75 0.74 0.09 –0.552 0.164
Window Wall Ratio (%) 15 76 41.40 41.00 13.554 0.189 0.164
Source Site Ratio 1.170 5.590 2.41 2.40 0.42 –0.020 0.164
Electricity Intensity (kWh/m²) 0.00 309.46 148.47 144.49 48.62 0.916 0.164
Natural Gas Use (kWh/m² 0.00 700.33 16.31 0.05 55.81 8.988 0.164
District Steam Use (kWh/m²) 0.00 251.08 81.72 69.60 64.51 0.561 0.164
Site EUI (kWh/m²) 118.91 730.04 291.87 275.51 84.46 1.020 0.164
Source EUI (kWh/m²) 327.55 1453.24 685.93 670.71 161.91 0.693 0.164
Chicago Variables Descriptive Statistics
Variables Min Max Mean Median
Statistic Std.
Number of Floors 6 110 34.99 36.00 15.117 1.379 0.235
Gross Floor Area (m²) 46,574 416,514 107,030 89,151 64,586 2.432 0.235
Floor Area Ratio (%) 3.48 61.93 22.67 22.19 10.79 0.824 0.235
Compactness Factor (CF) 0.068 0.194 0.113 0.109 0.024 0.948 0.235
Relative Compactness (RC) 0.447 0.975 0.78 0.80 0.11 –0.499 0.235
Window Wall Ratio 19 74 44.74 43.00 14.996 0.314 0.235
Source Site Ratio 1.442 3.142 2.63 2.67 0.52 –0.381 0.235
Electricity Intensity (kWh/ft²) 52.74 318.59 165.12 160.10 49.47 0.543 0.235
Natural Gas Use (kBtu/ft²) 0.00 228.83 51.36 4.54 0.235 0.00 228.83
District Steam Use (kBtu/ft²) No use reported for the selected samples
Site EUI (kWh/m²) 137.44 451.66 243.74 232.89 62.27 0.628 0.235
Source EUI (kWh/m²) 270.93 1194.22 619.37 585.04 135.41 1.049 0.235
Other significant differences between the two cities were the proportion of the primary energy
sources used. For example, in Chicago, most of the buildings in the selected samples used electricity
as the primary energy source, reaching almost 70% of the overall consumption and approximately
20% gas; no district steam use was reported. Meanwhile, in NYC, more than 30% of the consumption
relied on the district’s steam, electricity was a little over 50%, but there was very limited natural gas
use (Figure 6). New York City’s district steam system of Con Edison is the largest district heating
system in the Western world and has been serving Manhattan’s large-scale commercial and
residential buildings since 1882 [26]. This may be an essential fact for understanding the
abovementioned unexpected dynamic of the EUI. District steam systems utilize pipes for
distribution, resulting in a significant energy loss in the system, lowering the efficiency of any such
scheme. In modern-day district heating, these schemes have high operating and maintenance costs,
make it difficult to connect to end-users (high/low-pressure interface), and have high thermal losses
[27]. Also, utilizing district steam for heating and other purposes generally limits other choices for
the building’s systems that could have higher efficiency. Lastly, due to the relatively small amount
reported, we have excluded the fuel oil uses from our analysis.
Other morphological data, such as the area, floor area ratio, compactness factor, relative
compactness, and window-to-wall ratio, were found to have a similar median, with less than 10%
difference between the two cities.
It should be noted that benchmarking data generally do not include detailed energy end-use
breakdown (e.g., water heating, lighting, cooling, and heating) information and have inherent
limitations for explaining the exact composition of the total consumptions. However, where possible,
additional explanations are discussed.
Energies 2019, 12, 4783 8 of 16
2.4. Method: Multiple Regression Analysis and Spearman Correlation
First, considering the nonparametric distribution of the data, we carried out Spearman’s rank
correlation analysis, to identify any underlying relations between the variables. Second, multiple
regression analysis [28] was carried out, using IBM SPSS Statistics 25 to identify the most significant
common factors affecting the energy consumption of the sampled large-scale office buildings in each
city. The common factors were then evaluated in detail, utilizing scatter and box plot diagrams to
determine whether the trends reveal any congruity and to investigate further the potential cause of
the patterns found in the regression models.
The ‘Statistical Package for the Social Sciences’ (SPSS) is a package of statistic programs from
IBM for solving research problems by means of analysis, hypothesis testing, geospatial analysis, and
predictive analytics [29].
Figure 5. Site and Source energy use intensity comparison between NYC and Chicago.
Figure 6. Electricity, natural gas and district steam use comparison between NYC and Chicago.
3. Results and Discussion
3.1. Correlation Results between Variables
Table 3 presents the result of Spearman’s correlation analysis between the variables. Fairly many
variables show strong statistical correlations, of which some reveal unexpected patterns while others
were obvious and expected. Considering our research focus, we have highlighted only the factors
that are commonly significant in both cities. Overall, number of floors and construction year were the
two factors that most related to other variables. Starting in the early 20th century, we can assume
Energies 2019, 12, 4783 9 of 16
that, as the years progressed, there were taller buildings built with higher FAR and WWR, and the
results below confirm this relationship. One obvious pattern to note is the relationship between the
construction year and compactness factor. In both cities, the negative numbers signify that recently
constructed buildings are more compact in their forms, with smaller surface-area-to-volume ratio,
which reflects the modern trends of maximizing efficiency for the limited available real estate. In
addition, the connections between site-to-source ratio, number of floors, and construction year are
worth further attention. The construction year and source-to-site ratio have a strong positive relation,
which means the newer buildings have a higher source-to-site ratio. The higher the rate, the more the
building’s relative reliance on electricity as its primary fuel increases. Especially in Chicago, the
significance level appears to be exceptionally high compared to NYC. Moreover, the ratio positively
relates to the number of floors in Chicago, while in NYC, a negative relation is found. This indicates
that, in Chicago, taller buildings rely more on electricity, while in NYC, taller buildings rely on other
energy sources. These findings relate to the earlier discussion in the statistical description section of
this paper, and it is evident that NYC’s district steam system is the primary cause of these results.
Table 3. Spearman’s rank correlation coefficients between variables.
New York City Morphology
Floor Area
Ratio (%)
Wall Ratio
Factor (CF)
(RC) Relative
Source Site
Number of Floors 0.386** 0.362** 0.115 0.567** 0.104 0.125 0.450** 0.229**
Morphology Type 0.300** 0.128 0.127 0.183** 0.044 0.272** 0.039
Year Constructed 0.234** 0.165* 0.262** 0.132* 0.035 0.151*
Latest Construction Date
0.051 0.029 0.113 0.098 0.064
Floor Area Ratio (%) 0.046 0.100 0.169* 0.071
Window Wall Ratio (WWR)
0.032 0.043 0.060
(CF) Compactness Factor
0.465** 0.025
(RC) Relative Compactness
*. Correlation is significant at the 0.05 level (2-tailed), **. Correlation is significant at the 0.01 level (2-tailed).
Chicago Morphology
Floor Area
Ratio (%)
Wall Ratio
Factor (CF)
(RC) Relative
Source Site
Number of Floors 0.327** 0.459** 0.119 0.639** 0.036 0.397** 0.086 0.233*
Morphology Type 0.138 0.206* 0.169 0.002 0.002 0.184 0.113
Year Constructed 0.027 0.460** 0.307** 0.411** 0.265** 0.727**
Latest Construction Date
0.011 0.121 0.068 0.394** 0.154
Floor Area Ratio (%) 0.133 0.229* 0.253** 0.352**
Window Wall Ratio (WWR)
0.098 0.289** 0.281**
(CF) Compactness Factor
0.476** 0.144
(RC) Relative Compactness
*. Correlation is significant at the 0.05 level (2-tailed), **. Correlation is significant at the 0.01 level (2-tailed).
3.2. Multiple Regression Analysis
Given the multiple independent variables, we used a multiple linear regression test in SPSS, to
determine the most influential factors affecting the site and source energy use intensity (dependent).
Tables 4 and 5 present the results, which include several variables significant at or above a 95%
confidence level. In both cities, the number of floors, window-to-wall ratio, and source-to-site ratio
were the most significant factors for site EUI. Year constructed was another common factor, with a
confidence level at above 90%. However, the significance level was not as strong for the source EUI.
Even though electricity, gas, and district steam use showed the highest significance, the above factors
are parts that compose the overall EUI and are apparent results.
Energies 2019, 12, 4783 10 of 16
Table 4. NYC multiple regression analysis result.
Weather Normalized Site EUI Weather Normalized Source EUI
R = 0.872 R2 = 0.761 R = 0.799 R2 = 0.639
Coefficients t Sig
Coefficients t Sig.
Number of Floors –0.232 1.552 0.063 * –0.289 0.452 0.056*
Grouped by Number of Floors 0.302 –1.869 0.012 ** 0.343 –1.923 0.019 **
Morphology –0.026 2.523 0.530 –0.036 2.366 0.460
Window Wall Ratio 0.062 –0.629 0.013 ** 0.054 –0.740 0.237
Gross Floor Area 0.106 1.643 0.065 0.165 1.186 0.018 **
Year Constructed –0.089 1.855 0.048 ** –0.070 2.383 0.195
Latest Construction Date 0.051 –1.988 0.188 0.044 –1.300 0.347
Floor Area Ratio (%) –0.038 1.321 0.427 –0.037 0.943 0.530
(CF) Compactness Factor 0.045 –0.796 0.488 0.054 –0.629 0.494
(RC) Relative Compactness 0.050 0.695 0.425 0.067 0.685 0.378
Source Site Ratio –0.251 0.799 0.000 *** 0.176 0.884 0.001 ***
Electricity Intensity 0.439 –5.775 0.000 *** 0.636 3.34 0.000 ***
Natural Gas Use 0.463 11.460 0.000 *** 0.309 13.71 0.000 ***
District Steam Use 0.522 12.047 0.000 *** 0.465 6.633 0.000 ***
*** Significant at the 99% confidence level, ** significant at the 95% confidence level, * significant at
the 90% confidence level.
Table 5. Chicago multiple regression analysis result.
Weather Normalized Site EUI Weather Normalized Source EUI
R = 0.919 R2 = 0.845 R = 0.902 R2 = 0.814
Coefficients t Sig.
Coefficients t Sig
Number of Floors 0.544 –0.188 0.014 ** 1.389 0.158 0.027 **
Grouped by Number of Floors –6.770 1.074 0.018 ** –16.445 1.193 0.059 *
Morphology 1.500 –1.173 0.374 1.639 –1.238 0.649
Window Wall Ratio –3.057 0.742 0.009 ** –5.114 0.353 0.117
Gross Floor Area –8.201 × 105 –1.575 0.072 0 –1.145 0.059 *
Year Constructed 0.140 –1.109 0.054 * 0.058 –1.203 0.675
Latest Construction Date 0.079 0.935 0.320 0.094 0.168 0.536
Floor Area Ratio (%) 0.248 0.675 0.365 0.042 0.347 0.937
(CF) Compactness Factor –41.743 0.748 0.739 –284.712 0.055 0.302
(RC) Relative Compactness –72.377 –0.182 0.009 –179.293 –0.541 0.023
Source Site Ratio –97.220 –1.517 0.005 ** –50.714 –1.633 0.009 **
Electricity Intensity 1.094 –7.810 0.005 ** 3.046 –1.770 0.005 **
** significant at the 95% confidence level, * significant at the 90% confidence level.
Interestingly, variables such as the compactness factor (CF) and relative compactness (RC),
which were found to be the most significant factors in some studies utilizing benchmarking data from
the low-rise school buildings in the UK [8], did not appear significant for this specific group. Overall
floor area and floor area ratio were not found to be significant factors. In addition, the latest
construction date, which indicates the date of the latest major renovations made to each building,
showed no significance, while the initial built year indicated strong relations with the energy
consumption in both cities.
The primary goal of this research was to investigate the prevailing common indicators in two
different cities; to this end, further investigation was carried out for the four factors identified above.
3.2.1. Number of Floors and Energy Use Intensity
Due to the unprecedented increase in high-rise building construction in recent years, a few
studies have examined the relationship between building height and energy consumption. However,
limited research exists for the cities where buildings with 20 floors or above are prevalent. Godoy-
Shimizu et al. [10] observed energy use and office building heights for buildings in England and
Wales, where they concluded energy use intensifies as the height increases, due to the greater
Energies 2019, 12, 4783 11 of 16
exposure of taller buildings to lower temperatures, stronger winds, and more solar gains. Other
related studies utilizing data from Display Energy Certificates (DEC) and neighborhood density
around London also conclude higher buildings are more energy-intensive [11]. However, high-rise
building heights in European cities are significantly lower if compared with other continents; the
above research categorizes buildings with 10–20 floors as high-rise, which could be categorized as
low-rise buildings in cities like New York and Chicago. Another study based on a hypothetical
parametric model, which analyzes the individual energy use of the different urban components,
suggests energy intensity changes profoundly with height. Rather than a linear relationship, this
study identified that the optimal number of floors are found to be in the range of 7–27, depending on
population and building lifetime [30].
The regression result for the number of floors in this research shows a robust significance
compared to other variables. In Figure 7, the box and whisker plot diagrams present the energy use
intensity per group (buildings grouped based on ten-floor intervals) for ease of comparison. In NYC,
the lowest median EUI was found for buildings between 20 and 30 floors, and In Chicago, the group
above 60 floors. In both cities, EUI for the buildings with less than 20 floors indicated relatively higher
EUI than buildings of the taller groups. Due to the limited sample numbers in some groups, it may
be premature to generalize the results. However, in contrast to the previous research, the building
height and energy use intensity did not show a linear relationship for large-scale, super-tall
commercial buildings in these highly developed cities. Even though statistical studies have limits in
explaining causes in full detail, this result potentially relates to the distribution of mechanical
systems. It is common for a single technical floor to support between 15–20 floors, either above or
below its location. Technical floor spacing beyond these parameters will increase losses from friction
and gravity forces, which diminishes energy performance and gives rise to ongoing operation and
maintenance issues [31]. Such concentrated mechanical floors are hard to incorporate in lower-height
buildings; hence, these systems get scattered on the roof or in basements, reducing the efficiency of
distribution. Also, the taller building groups include more buildings constructed in recent years, so
more highly efficient systems are incorporated.
Figure 7. Site and source energy use intensity per number of floors.
3.2.2. Construction, Renovation Years, and Energy Use Intensity
Due to the lower glazing ratio and thermal masses created by relatively thick masonry walls,
some studies indicate that the pre-war buildings are generally more energy-efficient than recently
constructed buildings [9]. These construction attributes are similar in both cities. However, in
Chicago, the buildings constructed and renovated since 1980 clearly showed a decline in their site
EUI, even though the source EUI conversion still shows an increase, while in NYC, our result
coincided with previous findings (Figures 8 and 9). This finding warrants attention, as the differences
between the two cities may inform us of the positive factors that affect use intensity. As described in
Energies 2019, 12, 4783 12 of 16
Section 2.3, New York City’s district steam system is the largest in the Western world, serving the
majority of the selected samples in this research [28]. Meanwhile, of the buildings in Chicago’s
sample, approximately 70% rely on electricity, and the rest rely on natural gas, as their fuel source.
Including high-efficiency electrical and gas-powered heat pumps, technological advances for heating
and cooling systems in recent decades could be the primary cause of this result. Due to the high
reliance on district steam, especially for heating, these recent innovative systems may not have been
utilized fully in NYC. Still, in the case of Chicago, the actual site EUI has benefitted from these
advances over time. Currently, the site-to-source conversion factor for grid-purchased electricity in
the US is 2.8, which is exceptionally high compared to other resources. However, many cleaner
options are being considered to substitute for coal, which is the primary energy source used to create
the majority of the electricity, and the site EUI for the electricity-dependent buildings will possibly
get lower in the future. Based on this result, we could argue the installation of high-efficiency
electrically dependent systems over time may result in better performing buildings than by using
district systems.
Figure 8. Site and source energy use intensity per construction year.
Figure 9. Site and source energy use intensity per latest construction date including the major
3.2.3. Window-to-Wall Ratio and Energy Use Intensity
Energies 2019, 12, 4783 13 of 16
The balance between glazing and opaque areas alone has an impact on many aspects of the
energy balance; it influences solar gain (and thus energy use for heating and cooling) and heat loss
(mainly affecting energy use for heating), but it also impacts daylight availability (with implications
on energy use for artificial light) [32]. A study by Ballarini et al. [33] on retrofitted office buildings
enveloped with reduced glazing area showed significant improvements in the thermal performance
and comfort; however, a daylighting reduction occurred, with a consequent higher electricity
demand for lighting (36%). Consequently, many previous types of research have suggested the
optimum ratio depends on climate but ranges from 30% to 50% in the case of continental climate
locations similar to NYC and Chicago [34]. In both cities, the median WWR was approximately 40%,
even though many buildings had a much higher or lower percentage. Our regression model indicated
a significant relationship between the EUI and WWR. Without detailed information on the thermal
values or the orientation of each façade, the ratio itself cannot provide an accurate evaluation of the
effects. However, as presented in Figure 10, we can see the trends in NYC and Chicago were quite
different. The site EUI increased in NYC, while Chicago showed a decrease with higher WWR. The
recently constructed buildings have a higher WWR in general (Figure 11); these results imply that
Chicago’s newly built buildings are much more energy-efficient, even overcoming the disadvantages
of the higher glazing ratio. Again, we speculate that the cause of the difference is due to the limits of
the district steam widely used in NYC, while Chicago has been installing high-performance systems,
taking advantage of technological advances in heating and cooling systems.
Figure 10. Site and source energy use intensity per window-to-wall ratio.
Figure 11. Relation between Construction Year and Window to Wall Ratio.
3.2.4. Source–Site Ratio and Energy Use Intensity
In order to assess the relative efficiencies of buildings with varying proportions of primary and
secondary energy consumption, the industry recommends converting these two types of energy into
equivalent units of raw fuel consumed, to generate one unit of energy consumed on-site. Primary
Energies 2019, 12, 4783 14 of 16
energy is the raw fuel that is burned to create heat and electricity, such as natural gas or fuel oil used
in on-site generation. Secondary energy is the energy product (heat or electricity) created from a raw
fuel, such as electricity purchased from the grid or heat received from a district steam system [35].
Site energy is the amount of heat and electricity consumed by a building, as reflected in utility
bills, whereas source energy represents the total amount of raw fuel that is required to operate the
building. It incorporates all transmission, delivery, and production losses. By considering all energy
use, the data provide a complete assessment of the energy efficiency of a building. As noted by many,
accounting for the source energy is vital for giving an accurate picture of a building’s energy
consumption [24]. However, the majority of previous research utilizing the benchmarking data has
not addressed the relations of the two energy-use types [6,9]. In this research, we included both site
and source EUI, as well as the ratio in between as one of the test variables.
Electricity from the grid, in most countries, is still considered to be one of the least-efficient fuel
types; in the US, the conversion factor from site to the source is 2.8 compared to natural gas, at 1.05,
and steam or hot water at 1.2. Therefore, a higher source-to-site ratio implies the building’s primary
systems heavily relying on electricity rather than other resources, such as gas and district steam.
Based on the benchmarking data, we calculated the site-to-source ratio for the selected samples. For
both New York and Chicago, this ratio indicated the highest significance in predicting the source and
site EUI. However, the trends from the two cities were not uniform, as shown in Figure 12. In NYC,
a higher ratio means both the source and site EUI decreases, while in Chicago, a higher ratio means
the source EUI increases but the site EUI decreases. In addition, in NYC, many buildings’ ratios are
concentrated between 2–3, while the ratios for buildings in Chicago are concentrated above 3. From
the results, it is clear that the more the building relies on electricity, the less the overall site EUI
compared to other energy resources. Chicago’s high dependency on electricity compared to NYC also
warrants attention; this could be the primary cause for the lower EUIs.
Figure 12. Site and source energy use intensity per source site ratio.
4. Conclusions
This research evaluated the source and site EUI of 327 large-scale office buildings with a gross
floor area exceeding 500,000 ft² (46,452 m²), located in New York City and Chicago. Statistical tests
were carried out to identify the distinctive trends and predictors for energy consumption.
The analysis results presented in this work identify four statistically significant key common
factors affecting the building energy use intensity: number of floors, construction year, window-to-
wall ratio, and source-to-site ratio.
Contrary to the findings of many previous studies that found the relation between building
height and energy consumption to be linear, in NYC, the lowest EUI median was found in buildings
with 20–30 floors, and in Chicago, the buildings with more than 60 floors. For construction year and
Energies 2019, 12, 4783 15 of 16
window-to-wall ratio, our results also go beyond previous research, showing that, for example, in
Chicago, the newer buildings with higher WWR used less energy. The ratio between the source and
site energy use intensity was one of the unique variables we have introduced to this research, and
with higher ratios, the EUI declined in both cities. Additional, comprehensive analysis is required;
however, the above findings imply that Chicago’s electrical, gas-driven systems could be much more
energy efficient than NYC’s district steam-based systems, outweighing other adverse morphological
and climate factors of large-scale office buildings.
The presented results in this paper should be interpreted as observations of general trends from
the reported benchmarking data and have inherent limitations for identifying exact causes. Therefore,
further work is certainly under consideration to disentangle these complexities through theoretical
simulations and models that consider more-specific building systems, as well as urban contexts.
Author Contributions: Conceptualization, C.W. and S.N.; data curation, Q.A.; formal analysis, C.W. and Q.A.;
investigation, C.W. and S.N.; methodology, C.W. and Q.A.; supervision, C.W.; validation, C.W.
Funding: This research was funded by the School of Engineering, Korea University, grant number K180911.
Acknowledgments: The author would like to sincerely thank the anonymous reviewers for providing helpful
comments. In addition, thanks are due to the Energies’ staff and editors for handling the paper.
Conflicts of Interest: The authors declare no conflicts of interest.
1. Managan, K.; Layke, J.; Araya, M.; NeslerJen, C. Driving Transformation to Energy Efficient Buildings: Policies
and Actions, 2nd ed.; Institute for Building Efficiency: Washington, DC, USA, 2012.
2. Palmer, K.; Walls, M. Using information to close the energy efficiency gap: A review of benchmarking and
disclosure ordinances. Energy Effic. 2017, 10, 673–691.
3. City, N.Y. Greater, Greener Buildings Plan. Available online: (accessed on 25 September 2019).
4. Chicago, C.O. Chicago Energy Benchmarking Homepage. Available online:
(accessed on 25 September 2019).
5. Council, U.G. New York City’s Energy and Water Use 2014 and 2015 Report; Urban Green Council: New York,
NY, USA, 2017.
6. Papadopoulos, S.; Bonczak, B.; Kontokosta, C.E. Pattern recognition in building energy performance over
time using energy benchmarking data. Appl. Energy 2018, 221, 576–586.
7. Gao, X.; Malkawi, A. A new methodology for building energy performance benchmarking: An approach
based on intelligent clustering algorithm. Energy Build. 2014, 84, 607–616.
8. Hong, S.M.; Paterson, G.; Burman, E.; Steadman, P.; Mumovic, D. A comparative study of benchmarking
approaches for non-domestic buildings: Part 1—Top-down approach. Int. J. Sustain. Built Environ. 2013, 2,
9. Kontokosta, C.E. A market-specific methodology for a commercial building energy performance index. J.
Real Estate Financ. Econ. 2015, 51, 288–316.
10. Godoy-Shimizu, D.; Steadman, P.; Hamilton, I.; Donn, M.; Evans, S.; Moreno, G.; Shayesteh, H. Energy use
and height in office buildings. Build. Res. Inform. 2018, 46, 845–863.
11. Steadman, P. Energy ‘High-Rise Buildings: Energy and Density’ Research Project Results; UCL: London,
UK, 2017.
12. Guthrie, A. Tall buildings sustainability from the bottom up. In Proceedings of the CTBUH 8th World
Congress, “Tall & Green: Typology for a Sustainable Urban Future”, Dubai, March, 3–5 March 2018; pp. 3–5.
13. Khamma, T.R.; Boubekri, M. Statistical analysis of Impact of Building Morphology and Orientation on its
Energy Performance. J. Eng. 2017, 5, 15–25.
14. Catalina, T.; Virgone, J.; Iordache, V. Study on the impact of the building form on the energy consumption.
In Proceedings of the Building Simulation 2011: 12th Conference of International Building Performance
Simulation Association, Sydney, 14–16 November 2011; pp. 1726–1729.
15. Raji, B.; Tenpierik, M.; Van den Dobbelsteen, A. Early-stage design considerations for the energy-efficiency
of high-rise office buildings. Sustainability 2017, 9, 623.
Energies 2019, 12, 4783 16 of 16
16. Ourghi, R.; Al-Anzi, A.; Krarti, M. A simplified analysis method to predict the impact of shape on annual
energy use for office buildings. Energy Convers. Manag. 2007, 48, 300–305.
17. Elotefy, H.; Abdelmagid, K.S.S.; Morghany, E.; Ahmed, T.M.F. Energy-efficient Tall buildings design
strategies: A holistic approach. Energy Procedia 2015, 74, 1358–1369.
18. Ling, C.S.; Ahmad, M.H.; Ossen, D.R. The effect of geometric shape and building orientation on minimising
solar insolation on high-rise buildings in hot humid climate. J. Constr. Dev. Ctries. 2007, 12, 27–38.
19. City of Chicago; 2017 Chicago Energy Benchmarking Report. Available online:
benchmarking/Chicago_Energy_Benchmarking_Reports_Data.html (accessed on 30 November 2019).
20. City, N.Y. LL84 Data Disclosure & Reports. Available online: (accessed on 28 October 2019).
21. Chicago, C.O. Chicago Energy Benchmarking Results, Analysis, & Building Data. Available online:
benchmarking/Chicago_Energy_Benchmarking_Reports_Data.html (accessed on 28 October 2019).
22. Mahdavi, A.; Gurtekin, B. Shapes, numbers, perception: Aspects and dimensions of the design-
performance space. In Proceedings of the 6th International Conference Design and Decision Support
Systems in Architecture, Ellecom, The Netherlands, 1 January 2002; pp. 291–300.
23. CBECS. User’s Guide to the 2012 CBECS Public Use Microdata; CBECS: Washington, DC, USA, 2016.
24. Ueno, K.; Straube, J. BSD-151: Understanding Primary/Source and Site Energy. Available online:
(accessed on 28 October 2019).
25. EPA. Portfolio Manager Degree Days Calculator. Available online: (accessed on 27 October 2019).
26. Sivak, M. Air conditioning versus heating: Climate control is more energy demanding in Minneapolis than
in Miami. Environ. Res. Lett. 2013, 8, 014050.
27. Millar, M.A.; Burnside, N.M.; Yu, Z. District heating challenges for the UK. Energies 2019, 12, 310.
28. Multiple Regression Analysis Using SPSS Statistics. Available online:
tutorials/multiple-regression-using-spss-statistics.php (accessed on 21 October 2019).
29. Landau, S.; Everitt, B.S. A Handbook of Statistical Analyses Using SPSS; Chapman and Hall/CRC: Boca Raton,
FL, USA, 2003.
30. Burton, C. Designing High Performance MEP Systems for Supertall Buildings: A Review of Challenges and
Opportunities. Int. J. High Rise Build. 2017, 6, 301–306.
31. Chiesa, G.; Acquaviva, A.; Grosso, M.; Bottaccioli, L.; Floridia, M.; Pristeri, E.; Sanna, E.M. Parametric
Optimization of Window-to-Wall Ratio for Passive Buildings Adopting A Scripting Methodology to
Dynamic-Energy Simulation. Sustainability 2019, 11, 3078.
32. Goia, F. Search for the optimal window-to-wall ratio in office buildings in different European climates and
the implications on total energy saving potential. Solar Energy 2016, 132, 467–492.
33. Ballarini, I.; De Luca, G.; Paragamyan, A.; Pellegrino, A.; Corrado, V. Transformation of an office building
into a nearly zero energy building (nZEB): Implications for thermal and visual comfort and energy
performance. Energies 2019, 12, 895.
34. Agency, E.P. Energy Star PortFolio Manager. Available online:
manager/understand-metrics/difference (accessed on 12 September 2019).
35. Meng, T.; Hsu, D.; Han, A. Estimating energy savings from benchmarking policies in New York City.
Energy 2017, 133, 415–423.
© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (
... This phenomenon is referred to as negative transfer [218]. More precisely, within the context of building energy modeling, energy consumption depends greatly on several contextual factors, such as the building typology (residential, industrial or commercial), shape, size and age [226]. Combining energy data from very disparate source buildings is consequently counterproductive and will adversely impact the target performance in the target building. ...
... More precisely, for the task of building energy modeling, energy consumption depends greatly on several building-specific contextual characteristics, namely typology (residential, industrial, or commercial), shape, size and age [226]. Hence, buildings have varying probability distributions, over different characteristics. ...
Predictive modeling of energy consumption in buildings is essential for intelligent control and efficient planning of energy networks. One way to perform predictive modeling is through machine learning approaches. Alongside their good performance, these approaches are time efficient and facilitates the integration of buildings into smart environments. However, accurate machine learning models rely heavily on collecting relevant building operational data in a sufficient amount, notably when deep learning is used. In the field of buildings energy, historical data are not available for training, such is the case in newly built or newly renovated buildings. Moreover, it is common to verify the energy efficiency of buildings before construction or renovation. For such cases, only a contextual description about the future building and its design is available. The goal of this dissertation is to address the predictive modeling tasks of building energy consumption when no historical data are available for the given target building. To that end, existing data collected from multiple different source buildings are leveraged. This is increasingly relevant with the growth of open data initiatives in various sectors, namely building energy. The main idea is to transfer knowledge across building models. There is little research at the intersection of building energy modeling and knowledge transfer. An important challenge arises when dealing with multi-source data, since large domain shift may exist between different sources and also between each source and the target. As a contribution, a two-fold query-adaptive methodology is developed for cross-building predictive modeling. The first process recommends relevant training data to a target building solely by using a minimal contextual description on it (metadata). Contextual descriptions are provided as user queries. To enable a task-specific recommendation, a deep similarity learning framework is used. The second process trains multiple predictive models based on recommended training data. These models are combined together using an ensemble learning framework to ensure a robust performance. The implementation of the proposed methodology is based on microservices. Logically independent workflows are modeled as microservices with single purposes and separate data sources. Building metadata and time series data collected from multiple sources are integrated into an unified ontology-based view. Experimental evaluation of the predictive model factory validates the effectiveness and the applicability for the use case of building energy modeling. Moreover, because of its generic design, the methodology for query-adaptive cross-domain predictive modeling can be re-used for a diverse range of use cases in different fields.
... Awareness about the amount of actual energy consumption in existing buildings straightly affects the energy efficiency, since the building owners/users compare their energy consumption with that of peer buildings. Landlords, operators, and occupants can manage their consumption more wisely and make smart decisions if they have accurate knowledge of the energy consumption data in buildings [8]. Energy benchmarking is an effective tool to evaluate and compare the energy efficiency of buildings because systematically evaluating the building's energy performance, helps to determine over-standard (more than the benchmark) consumption and identify inefficient buildings [9]. ...
Public buildings such as schools have a great potential for energy efficiency as well as improving their performance to reduce CO2 emissions. Schools with huge building areas, a large number of users, and various equipment use more energy compared with other building types. Top-Down energy benchmarking is a policy tool that detects inefficient buildings in a community, so it has a fundamental role in energy efficiency management on a large scale. This study aims to evaluate end-use energy benchmarks for secondary schools based on a Top-Down statistical method. In this regard, the real monthly energy consumption from 2016 to 2019 in 31 school buildings were analyzed. The standards of 9, 171, and 180 kWh/m²/yr were determined as electricity, thermal, and total energy benchmarks respectively. In addition, based on the Spearman test correlation analysis, 14 parameters that may affect the energy consumption were analyzed; among them, the building useful area and volume had the highest impact on electricity consumption; however, population density and building volume had the highest impact on the thermal consumption. The developed benchmarking methodology presents a band of benchmarks instead of a single-value mostly found in the literature.
... This phenomenon is referred to as negative transfer [11]. More precisely, within the context of building energy modeling, energy consumption depends greatly on several contextual factors, such as the building's use type (residential, industrial, or commercial), shape, size and age [12]. Combining energy data from very disparate source buildings is consequently counterproductive and will adversely impact the target performance in the target building. ...
Full-text available
Predictive modeling in buildings is a key task for the optimal management of building energy. Relevant building operational data are a prerequisite for such task, notably when deep learning is used. However, building operational data are not always available, such is the case in newly built, newly renovated, or even not yet built buildings. To address this problem, we propose a deep similarity learning approach to recommend relevant training data to a target building solely by using a minimal contextual description on it. Contextual descriptions are modeled as user queries. We further propose to ensemble most used machine learning algorithms in the context of predictive modeling. This contributes to the genericity of the proposed methodology. Experimental evaluations show that our methodology offers a generic methodology for cross-building predictive modeling and achieves good generalization performance.
... In recent years, researchers have been able to perform benchmark studies by obtaining data on buildingsʹ actual energy usage from a few databases around the world. Won et al. [5] presented a comprehensive analysis of large-scale office buildings located in New York and Chicago, with respect to their energy use intensity (EUI). In New York City, the smallest EUI median was found in the buildings with 20 to 30 floors, and in Chicago, the buildings with 60 floors or more. ...
Full-text available
Although interest in and the importance of green building certification have increased, it is difficult to determine how much less energy the buildings actually consume after obtaining sustainability related certification, such as LEED in the USA, and G-SEED in Korea. Focusing on building energy evaluation, we compared and analyzed the energy evaluation items of LEED and G-SEED. In addition, from the Korean public data portal site, this study investigated the annual site energy consumption (electricity and gas) per floor area of non-residential buildings that obtained G-SEED certification in Korea. The energy consumption of non-certified general buildings was compared with the energy consumption of certified buildings. As a result of examining 84 samples of this study, non-residential buildings with G-SEED certification showed energy use intensity that was at least (35.5 to 48.9)% lower than that of uncertified buildings.
Natural and human-caused extreme events can alter residential electricity demand in urban areas and stress the electricity grid, with different types of residential electricity consumers exhibiting different consumption patterns. Residential electricity demands have been widely analyzed considering single-family consumers; however, multi-family consumption patterns remain comparatively understudied. The deployment of smart electricity meters enables the identification of single- and multi-family residential electricity consumption patterns at high temporal resolution. Using smart electricity meter data for the greater Chicago area, we compare electricity demand profiles reported by smart meters from single- and multi-family consumers in a large and diverse urban environment to understand residential electricity patterns better. Our study comprehensively analyzes the daily electricity demand profiles of these two types of residential consumers to identify peak electricity consumption times and magnitudes. Results show that the electricity demand of both residential end-users follows similar time of use patterns, and single-family users approximately double the demand of multi-family users on a per household basis. We also present predictive models of the electricity demand with socioeconomic data at the zip code level. Predictive model results show that multiple linear regression models explain up to 62% and 41% of the mean daily electricity (MDE) demand of single- and multi-family users, respectively. The median age of occupants, percent age 65 and older, mean commute time, and percent high school or higher education are statistically significant predictors of the MDE demand of single-family users, with percent high school or higher education having the highest relative importance. Similarly, median building age, percent multi-family, percent female, median age of occupants, and mean commute time are statistically significant predictors of multi-family electricity consumption, with median age of occupants having the highest relative importance. Modeling electricity demand to uncover differences between single- and multi-family residential electricity demands can assist city planners, and utility managers develop tailored demand management strategies.
Conventional building energy dynamic analysis relies on climatic and geographical statistics of buildings. A large number of studies focus on investigating the major influencing factors and the energy use with engineering methods and statistical methods to predict buildings' energy use. However, due to the lack of temporal statistics when analyzing buildings' energy use intensity (EUI), very few studies concerned the energy consumption from spatial and temporal heterogeneity simultaneously. This study extends the traditional regression model by introducing the temporal variations into the geographically weighted regression (GWR) model and developed a geographically and temporally weighted regression (GTWR) model. A case study, which is based on four years’ urban building energy consumption (from 2015 to 2018) in the city of Seattle, is used to validate the performance of the proposed regression model. Meanwhile, the ordinary least square (OLS), GWR, and temporally weighted regression (TWR) models are developed for making cross-comparison of the goodness-of-fit. The results show the GTWR and TWR models reduce the mean absolute error by 13% and 7%, respectively, relative to the GWR model. Moreover, the GTWR model has a substantial improvement of goodness-of-fit, which can be referred from the value of R2 increasing by 49% in OLS, 16% in GWR, and 6% in TWR. Two driving forces, the number of floors (NF) and energy stars (ES), are identified to have the most significant coefficients in evaluating the site EUI. The novelty of this research lies in achieving better reliability and accuracy in building energy modeling by GTWR, which provides an assisting tool for government and local authorities to make the master plan and sustainability strategies for the new buildings in the city.
Full-text available
The paper analyzes and compares the perspectives for reducing the energy consumption associated to the operation of Heating Ventilation and Air Conditioning system for climatic control of large-size non-residential buildings. Three different control strategies are considered comparing the use of boiler and heat pumps as heating systems and analyzing the use of demand-controlled ventilation, operating on the effective occupancy of the building. The control strategies are applied to two different educational buildings with shapes representative of typical educational structures. The results of the analysis show how the energy consumption can be reduced up to 70%, shifting from the actual values of the energy intensity of over 300 kWh/m2 for year to values of less than 100 kWh/m2 per year. The significance of the energy savings achieved in such different buildings has led to the identification of a possible benchmark for HVAC systems in the next future years which could help reach the environmental targets in this sector.
Full-text available
A building self-shading shape impacts substantially on the amount of direct sunlight received by the building and contributes significantly to building operational energy use, in addition to other major contributing variables, such as materials and window-to-wall ratios. Deep Learning has the potential to assist designers and engineers by efficiently predicting building energy performance. This paper assesses the applicability of two different neural networks structures, Dense Neural Network (DNN) and Convolutional Neural Network (CNN), for predicting building operational energy use with respect to building shape. The comparison between the two neural networks shows that the DNN model surpasses the CNN model in performance, simplicity, and computation time. However, image-based CNN has the benefit of utilizing architectural graphics that facilitates design communication.
Today, energy problems are becoming increasingly serious. The direct energy consumption of buildings accounts for 20% of the total energy consumption in a country. There are difficulties in continuing the mode of high energy consumption in the traditional construction industry. Therefore, the future of construction is in the development of green buildings. In the life cycle of a building, the consumption during the construction phase accounts for only approximately 20% of the total energy consumption. Most of the consumption occurs during buildings operations, such as lighting, heating, air conditioning and the running of various electrical appliances. Therefore, this paper focuses on the energy consumed during the building operation period with the aim of optimizing relevant design parameters to reduce total energy consumption. The West Twelfth Teaching Building (WTTB) of Huazhong University of Science & Technology (HUST) is used as a prototype, and Design Builder is used to establish a model of energy consumption and validate the reliability of the model based on the data obtained from the investigation. Based on this model, the study takes the perspective of energy conservation to analyze ten factors that may affect the energy consumption of the building: the heat transfer coefficient of the roof, the amount of fresh air, the heat transfer coefficient of interior walls, the heat transfer coefficient of the floors, the interior temperature, the energy efficiency ratio of the air conditioner, the thickness of the outer wall insulation, the ratio of windows to walls, the natural ventilation starting temperature and the solar heat gain coefficient of the outer windows. Those factors are then ranked according to their energy-saving potential through partial factorial design tests. The six factors with the most potential for energy savings are selected and divided into 2 groups to conduct a response surface optimization analysis of three factors at three levels. The best level of each factor and the optimal combination of all factors are obtained to reduce building energy consumption to the greatest possible extent and to provide a reference for teaching buildings and similar public buildings (PBs) in achieving the goal of “green building.”
Full-text available
Counterbalancing climate change is one of the biggest challenges for engineers around the world. One of the areas in which optimization techniques can be used to reduce energy needs, and with that the pollution derived from its production, is building design. With this study of a generic office located both in a northern country and in a temperate/Mediterranean site, we want to introduce a coding approach to dynamic energy simulation, able to suggest, from the early-design phases when the main building forms are defined, optimal configurations considering the energy needs for heating, cooling and lighting. Generally, early-design considerations of energy need reduction focus on the winter season only, in line with the current regulations; nevertheless a more holistic approach is needed to include other high consumption voices, e.g., for space cooling and lighting. The main considered design parameter is the WWR (window-to-wall ratio), even if further variables are considered in a set of parallel analyses (level of insulation, orientation, activation of low-cooling strategies including shading devices and ventilative cooling). Finally, the effect of different levels of occupancy was included in the analysis to regress results and compare the WWR with corresponding heating and cooling needs. This approach is adapted to Passivhaus design optimization, working on energy need minimisation acting on envelope design choices. The results demonstrate that it is essential to include, from the early-design configurations, a larger set of variables in order to optimize the expected energy needs on the basis of different aspects (cooling, heating, lighting, design choices). Coding is performed using Python scripting, while dynamic energy simulations are based on EnergyPlus.
Full-text available
Directive 2010/31/EU promotes the refurbishment of existing buildings to change them into nearly zero-energy buildings (nZEBs). Within this framework, it is of crucial importance to guarantee the best trade-off between energy performance and indoor environmental quality (IEQ). The implications of a global refurbishment scenario on thermal and visual comfort are assessed in this paper pertaining to an existing office building. The retrofit actions applied to achieve the nZEB target consist of a combination of envelope and technical building systems refurbishment measures, involving both HVAC and lighting. Energy and comfort calculations were carried out through dynamic simulation using Energy Plus and DIVA, for the thermal and visual performance assessments, respectively. The results point out that energy retrofit actions on the building envelope would lead to significant improvements in the thermal performance, regarding both energy savings (−37% of the annual primary energy for heating) and thermal comfort. However, a daylighting reduction would occur with a consequent higher electricity demand for lighting (36%). The research presents a detailed approach applicable to further analyses aimed at optimizing the energy efficiency measures in order to reduce the imbalance between visual and thermal comfort and to ensure the best performance in both domains.
Full-text available
District heating uptake has grown with the increasing need for cleaner and more efficient energy supply. This has resulted in a rising number of new developments signing up to a district heating scheme, typically powered by Combined Heat and Power (CHP) boilers or biomass boilers with supplemental electrical or gas grid connections. These schemes have advanced rapidly in recent years, with much of the research focus targeting lower carbon technologies, improved load prediction and peak demand management. We assess the current status of District Heating Networks (DHNs) in the United Kingdom using published case studies and suggest next steps to improvement. Our findings show that the United Kingdom has good potential for uptake of district energy given the current political climate and government incentives, however significant improvements must be made to further penetrate the heating market.
Full-text available
The relationship between energy use and height is examined for a sample of 611 office buildings in England and Wales using actual annual metered consumption of electricity and fossil fuels. The buildings are of different ages; they have different construction characteristics and methods of heating and ventilation; and they include both public and commercial offices. When rising from five storeys and below to 21 storeys and above, the mean intensity of electricity and fossil fuel use increases by 137% and 42% respectively, and mean carbon emissions are more than doubled. A multivariate regression model is used to interpret the contributions of building characteristics and other factors to this result. Air-conditioning is important, but a trend of increased energy use with height is also found in naturally ventilated buildings. Newer buildings are not in general more efficient: the intensity of electricity use is greater in offices built in recent decades, without a compensating decrease in fossil fuel use. The evidence suggests it is likely – although not proven – that much of the increase in energy use with height is due to the greater exposure of taller buildings to lower temperatures, stronger winds and more solar gains.
Full-text available
Building energy consumption in developed countries accounts for 20–40% of the total energy use and about 40% of primary energy use in the U.S in 2010.Office buildings account for approximately 18% of this usage. The morphology of a building has a huge impact on its energy use, especially in office buildings due to their huge glazing areas. Designing with proper regard of climate issues leads to enhanced energy performance. This paper provides an analysis of the impact of building shapes and orientations on the energy performance across small, medium and tall office buildings for the Chicago, IL, USA location. The method is based on the analysis of simulation results obtained from energy modeling software, using Pearson Correlation and Multiple Linear Regression methods. The analysis considers six different building shapes; Rectangular (1:1, 1:1.5, 1:2), T, L and U. All of these considered shapes have identical construction and Window-wall ratios as specified in the Department of Energy (DOE) standard reference buildings. The aim is to establish a relationship between the impacts of building relative compactness (RC) on the energy performance of office buildings in three different cases: Small, Medium and Tall.
Full-text available
Decisions made at early stages of the design are of the utmost importance for the energy-efficiency of buildings. Wrong decisions and design failures related to a building’s general layout, shape, façade transparency or orientation can increase the operational energy tremendously. These failures can be avoided in advance through simple changes in the design. Using extensive parametric energy simulations by DesignBuilder, this paper investigates the impact of geometric factors for the energy-efficiency of high-rise office buildings in three climates contexts: Amsterdam (Temperate), Sydney (Sub-tropical) and Singapore (Tropical). The investigation is carried out on 12 plan shapes, 7 plan depths, 4 building orientations and discrete values for window-to-wall ratio. Among selected options, each sub-section determines the most efficient solution for different design measures and climates. The optimal design solution is the one that minimises, on an annual basis, the sum of the energy use for heating, cooling, electric lighting and fans. The results indicate that the general building design is an important issue to consider for high-rise buildings: they can influence the energy use up to 32%. For most of the geometric factors, the greatest difference between the optimal and the worst solution occurs in the sub-tropical climate, while the tropical climate is the one that shows the smallest difference. In case of the plan depth, special attention should be paid in the case of a temperate climate, as the total energy use can increase more than in other climates. Regarding energy performance, the following building geometry factors have the highest to lowest influence: building orientation, plan shape, plan depth, and window-to-wall ratio.
Full-text available
Building energy use accounted for 38 % of total US carbon dioxide (CO2) emissions in 2012, and roughly half of those emissions were attributable to the commercial building sector. A new policy that has been adopted in 15 US cities and one US county is a requirement that commercial and sometimes also multifamily residential building owners disclose their annual energy use and benchmark it relative to other buildings. We discuss these nascent policies, summaries of the data that have been collected so far, and how to evaluate whether they are having an effect on energy use and CO2 emissions. Missing or imperfect information is a contributor to the energy efficiency gap, the finding that many low-cost options for improving energy efficiency fail to be adopted. These new laws may be an important step in closing the gap in the commercial and multifamily building sectors, but careful evaluation of the programs will be essential.
In recent years, many cities have adopted energy disclosure policies to better understand how energy is consumed in the urban built environment and how energy use and carbon emissions can be reduced. The diffusion of such policies has generated large-scale streams of building energy data, creating new opportunities to develop the fundamental science of urban energy dynamics. Nevertheless, there is limited research that rigorously analyzes building energy performance patterns over time. This paper provides a comprehensive framework to analyze building energy time series data and identify buildings with similar temporal energy performance patterns. We use data from approximately 15,000 properties in New York City, covering a six-year reporting period from 2011 to 2016. After pre-processing and merging the data for each constituent year, we use an unsupervised learning algorithm to optimally cluster the energy time series and statistical tests and supervised learning methods to infer how building characteristics vary between clusters. Our results show that energy reductions in New York City are mainly driven by its commercial building stock, with larger, newer, and higher-value buildings demonstrating the largest improvements in energy intensity over the study period. Moreover, voluntary energy conservation schemes are found to be more effective in boosting energy performance of commercial properties, compared to residential buildings. Our results suggest two distinct temporal patterns of energy performance for commercial and residential buildings, characterized by energy use reductions and increases. This finding highlights the differential response to energy reporting and disclosure, and presents a more complex picture of energy use dynamics over time when compared to previous studies. In order to realize significant energy use improvements over time and reach energy and carbon reduction goals, cities need to design and implement comprehensive energy policy frameworks, bringing together information transparency and reporting with targeted mandates and incentives.
A growing number of governments have begun to implement benchmarking or energy disclosure policies. By requiring owners to measure and disclose their energy use, these policies are intended to transform the market for energy-efficient investments in existing buildings. To improve future policy efforts, two critical questions are: first, how much energy do these policies save? and second, what particular aspects of these policies are most effective? To answer these questions, this study explores how different aspects of these policies were phased-in to different groups of buildings over the first four years of the City of New York’s benchmarking ordinance. By applying a novel difference-in-differences strategy, we can causally attribute observed declines in energy consumption to specific owner behaviors and policy mechanisms. Our analysis indicates that in comparison with the control group and before the policies were implemented in 2011, total disclosure of both energy use and Energy Star together can be credited with a 6% reduction in building energy use intensity (EUI) three years later and a 14% reduction in EUI four years later. Disclosure of Energy Star scores decreased building EUI by 9% three years later and 13% four years later. These two separate findings are a consequence of the policy design and different control groups.
This paper deals with the search for the optimal window-to-wall ratio (WWR) in different European climates in relation to an office building characterized by best-available technologies for building envelope components and installations. The optimal WWR value is the one that minimizes, on an annual basis, the sum of the energy use for heating, cooling and lighting.By means of integrated thermal and lighting simulations, the optimal WWR for each of the main orientations was found in four different locations, covering the mid-latitude region (35° to 60° N), from temperate to continental climates. Moreover, the robustness of the results was also tested by means of sensitivity analyses against the efficiency of the building equipment, the efficacy of the artificial lighting and the compactness of the building.The results indicate that although there is an optimal WWR in each climate and orientation, most of the ideal values can be found in a relatively narrow range (0.30 < WWR < 0.45). Only south-oriented façades in very cold or very warm climates require WWR values outside this range. The total energy use may increase in the range of 5-25% when the worst WWR configuration is adopted, compared to when the optimal WWR is used.