Energy 121 (2017) 606e615

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Development of a new methodology to optimize building life cycle

cost, environmental impacts, and occupant satisfaction
Ehsan Mostavi a, Somayeh Asadi b, *, Djamel Boussaa c
a
Department of Architectural Engineering, Pennsylvania State University, 104 Engineering, Unit A, University Park, PA, 16802, USA
b
Department of Architectural Engineering, Pennsylvania State University, 104 Engineering, Unit A, University Park, PA 16802, USA
c
Department of Architecture and Urban Planning, Qatar University, P.O. Box 2713, Doha, Qatar

a r t i c l e i n f o a b s t r a c t

Article history: Thermal comfort and occupant thermal satisfaction are critical aspects in the indoor environment quality
Received 1 September 2016 assessment and have received considerable attention by designers and building occupants. Improper
Received in revised form indoor temperature not only decreases the level of occupant thermal satisfaction, but also has serious
18 November 2016
health related consequences. Despite the importance of occupant thermal satisfaction that has been
Accepted 9 January 2017
Available online 16 January 2017
vastly emphasized, studies incorporating occupants' satisfaction during the design process are very
limited. Therefore, this study aims to develop a multi-objective design optimization model to minimize
life cycle cost and life cycle emission, and maximize occupant satisfaction level in a typical commercial
Keywords:
Occupant satisfaction
building. To solve the multi-objective optimization problem, a Harmony Search based algorithm is
Utility-theory developed and employed. Moreover, to identify the level of design thermal satisfaction, a novel utility-
Multi-objective optimization theory based thermal comfort index is defined and calculated. A small office building is selected as a case
Building energy performance study to analyze four different designs which are identified as optimum solutions. To determine the
Building envelope optimum designs, the satisfaction level of all the design combinations having cost and emissions similar
to previously distinguished optimum solutions are compared and best designs are identified.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction the relationship between satisfaction and different parameters

Thermal comfort and occupant thermal satisfaction are
considered as critical aspects in the assessment of indoor envi-
ronment quality and have received considerable attention by de-
signers and building occupants. Thermal comfort is defined as the
order at which occupants are having no intention to modify their
environment [1]. Improper indoor temperature not only decreases
the level of occupant thermal satisfaction, but also has serious
health related consequences [1,2,3]. It is also seen that thermal
comfort and in particular temperature set, have a significant impact
on building energy consumption level [4]. Desire to understand the
extensive influence of occupants on building energy performance
has initiated large number of studies to focus on them as an
important determinative subject [5,6].
Several studies have been conducted to investigate the rela-
tionship between occupants' working environment perception and
their efficiency [7,8]. Moreover, studies were conducted to identify
* Corresponding author.
E-mail address: asadi@engr.psu.edu (S. Asadi).
including temperature, humidity, air velocity, and radiant temper-
ature [4,9]. In a study performed by Varjo et al. [10] on satisfaction
and performance of office occupants, a high decrement in working
efficiency was observed among people who were working in offices
with low thermal comfort.
As mentioned, various factors contribute to thermal comfort and
satisfaction. Among these factors, temperature play a critical role in
occupants' perception [8]. In a study performed by Huizenga et al.
[7], it was seen that having personal control on indoor environ-
mental conditions significantly increases the level of occupant
satisfaction. Other study conducted by Hancock et al. [11], showed
that people who are exposed to low indoor environmental quality
may have different symptoms including short term memory and
performance reduction due to the improper working condition. In a
study performed by Li et al. [12], a multi-objective optimization
model aiming to minimize the costs and maximize the thermal
comfort was developed. They showed that the costs and savings are
highly dependent on the occupants' requirements.
In addition, in recent decades, environmental problems, espe-
cially greenhouse gas (GHG) emissions and global warming, have
enforced designers to estimate the level of environmental emission

http://dx.doi.org/10.1016/j.energy.2017.01.049
0360-5442/© 2017 Elsevier Ltd. All rights reserved.
 E. Mostavi et al. / Energy 121 (2017) 606e615
of their design and reduce their environmental contribution
[13,14,15]. Moreover, customer expectations regarding the project
budget imposes higher pressure on designers and decision makers
to reduce the project costs. To have a better understanding about
environmental impacts of different projects, Life Cycle Assessment
(LCA) has been extensively implemented during the last decades
[16,17]. LCA includes accumulating of all environmentally relevant
streams inventory associated with production processes, trans-
portation, and demolition of a product [18]. Accordingly, a signifi-
cant body of studies have been conducted to identify the optimum
designs with minimum Life Cycle Emission (LCE) and Life Cycle Cost
(LCC) [19]. Studies have implemented multiple strategies to provide
designers with information regarding the level of environmental
impacts of their design. In an innovative approach, Basbagill et al.
[20] presented a methodology to provide designers with LCA data
of different designs in Building Information Model (BIM). This
method integrates BIM with an energy simulation model to identify
the LCE of each design during the operation stage of the building
lifetime. Emissions associated with other stages of design (pre-
operation and maintenance/replacement) were calculated accord-
ing to different databases such as SimaPro and CostLab. Although,
the developed framework was well suited, lack of a unique sys-
tematic information exchange process was a barrier to expand the
application of such methods [21].
However, during the design process, we should consider mul-
tiple, and usually competitive, objectives such as reduction of en-
ergy consumption, financial costs and environmental impacts. This
makes the design as a challenging multi-objective optimization
problem. In a very rough classification, the optimum search
methods can be categorized into two classes: exact algorithms and
heuristic search. The exact algorithms identify a concrete answer
for the optimization problem. However, the heuristic search algo-
rithms do not guarantee to find the best answer and there is always
a probability to miss the local optimums, but they provide results
with reasonable accuracy [22]. The heuristic search algorithms are
problem dependent and should be adjusted for different problems.
In a more advanced form, the metaheuristic search algorithms are
highly problem-independent and can be applied to various types of
optimization problems. Metaheuristic algorithms provide set of
strategies to develop heuristic optimization [23].
Several studies have employed different metaheuristic algo-
rithms such as Genetic Algorithm (GA), Particle Swarm Optimiza-
tion (PSO) and Harmony Search (HS) to solve their multi-objective
optimization problems [24e27]. Rapone and Saro [28] employed a
PSO based algorithm to solve a single-objective optimization
problem of identifying the best façade for a highly glazed building.
They identified the optimum percentage and type of glazing and
depth shadings to minimize the level of carbon emissions of an
office building. In a study performed by Asadi et al. [29], a multi-
objective optimization model using GA and artificial neural
network was employed to determine the best retrofit strategies
quantitatively. The proposed model determines the tradeoff be-
tween retrofit cost, energy consumption, and hours of occupant
discomfort. Building envelope and selection of proper materials for
building façade have always received significant attention [30].
Fesanghari et al. [31] employed a HS based optimization algorithm
to determine the best combination of building envelope to mini-
mize the LCC and LCE of the project. In another study conducted by
Ascione et al. [32], a GA method was employed to determine the
best mix of renewable energy for a residential building. The
objective of this study was to minimize the primary energy demand
and investment cost. In addition to selecting the best materials,
some studies have focused on identifying the optimum window to
wall ratio and windows geometry [33]. In another study, Ruiz et al.
[34] utilized a Tabu Search algorithm to identify the proper
607
materials and HVAC system with lower investment cost and energy
demand for residential buildings. Tabu search has similar behaviors
as GA, however it has better performance in local searches, while
GA has a better global performance.
As it was explained so far, multiple heuristic algorithms have
been successfully employed in different optimization problems.
Each of these algorithms uses different strategies to solve specific
range of problems, but not all of them. For example, PSO generates a
semi-random movement of population of birds (particles) to
identify the location of the optimum solution through the search
environment; while, GA utilizes the genetic recombination of
random generated parents to attain new generations. Different
characteristics of HS algorithm, such as using single-point random
search with an improving memory, has empowered the method to
identify the local optimums with better precision and fewer
mathematical requirements. Moreover, comparing the exploration
and focusing ability of mentioned algorithms, HS algorithms shows
better operation.
Targeting to involve the end user satisfaction into design pro-
cess, the objective of the current study is to identify the optimum
designs having lowest LCC and LCE as well as highest thermal
satisfaction. To attain this objective, the effect of different con-
struction materials in various building components including
external and internal wall system, glazing system, floors, roof and
ceiling on the LCC, LCE and occupant satisfaction were investigated.
To solve the multi-objective optimization problem, a HS based al-
gorithm is developed and employed.
2. Methodology
2.1. Design variables
To investigate the effect of different building construction ma-
terials on defined objective functions, a database including 65
different materials was inserted to the optimization code. Table 1
shows variables included in the database. Different building com-
ponents including walls, floors, ceilings, glazing system, doors were
considered in this study. This database includes physical and
thermal properties of materials as well as associated environmental
emissions to calculate LCE and LCC. Table 2 shows number of layers
and the range of variables considered for each element. For
instance, living room external walls are considered to have 4
different layers and the materials that can be assigned to the 4th
layer are 19, 20, and 21 which are gypsum boards with different
thicknesses.
2.2. Objective functions
Current study proposes a multi-objective optimization process
to minimize the LCC and LCE as well as maximize the occupants'
thermal comfort (TC). Since several variables affect building energy
performance [35], the process of identifying the optimum design is
time consuming. Therefore, it is necessary to use a heuristic algo-
rithm to accelerate the process of identifying a proper approximate
result. The general procedure of heuristic optimization algorithms
are similar, however, the main search engines and their ability in
identifying the local optimums are different [36].
Fig. 1 shows a schematic view of the developed optimization
code. As it can be seen in this figure, the process of optimization
includes three steps: simulation, evaluation, and improvement
[37,38]. The simulation step includes running the energy simulation
software and determining the magnitudes of energy consumption,
zone temperatures, and environmental emissions. EnergyPlus V8.4
[39] which is a powerful and reliable energy simulation software,
and has the highest utilization share between the energy
608 E. Mostavi et al. / Energy 121 (2017) 606e615

Table 1
Variables of the optimization process.

Ref. No Material Ref. No Material Ref. No Material

1 Wood wall 23 Slate or tile shingle 45 Blanket insulation 3.500

2 Vinyl wall 24 Wood shingle 46 Blanket insulation-600
3 Cellular polyurethane 100 25 Metal surface 47 Aluminum window
4 Cellular polyurethane-200 26 Built-up roofing 48 Clear 3- mm
5 Cellular polyurethane 300 27 Membrane 49 Clear 2.5-mm
6 Cellular polyurethane 3.500 28 Concrete roof tile 50 Clear 6-mm
7 Cellular polyurethane 400 29 Carpet 51 Clear 12-mm
8 Cellular polyurethane 500 30 Acoustic tile 52 Low e clear 3-mm
9 Cellular polyurethane 5.500 31 Wood subfloor 53 Low e clear 3-mm rev
10 Cellular polyurethane 600 32 Steel door 54 Low e clear 6-mm
11 Concrete 50-mm 33 Mineral wool insulation 900 55 Low e clear 6-mm rev
12 Concrete 100-mm 34 Fiberglass insulation board 5.500 56 Low e clear 6-mm rev
13 Concrete 150-mm 35 Rigid insulation fiberglass 1.500 57 Gray 3-mm
14 Concrete 200-mm 36 Rigid insulation fiberglass 200 58 Gray 6-mm
15 Plywood 10-mm 37 Rigid insulation fiberglass 2.500 59 Gray 12-mm
16 Plywood 13-mm 38 Rigid insulation fiberglass-300 60 Ref a clear low 6-mm
17 Plywood 16-mm 39 Rigid insulation fiberglass-3.500 61 Ref a clear mid 6-mm
18 Concrete block 40 Fiberglass insulation board 600 62 Ref a clear high 6-mm
19 Gypsum board 16-mm 41 Fiberglass insulation board 8.800 63 Ref b clear low 6-mm
20 Gypsum board 13-mm 42 Fiberglass insulation board 1000 64 Ref b clear low 6-mm
21 Gypsum board 9.5-mm 43 Fiberglass insulation board 11.500 65 Ref b clear high 6-mm
22 Asphalt shingle 44 Fiberglass insulation board 1300

Table 2
Building parameters and number of layers in each parameter.

Element Layer Variable Range

Store floor 1 X1 11, 12, 13, 14

Living floor 1 X2 11, 12, 13, 14
2 X3 29, 30, 31
Roof 1 X4 22, 23, 24, 25, 26, 27, 28
2 X5 15, 16, 17
Interior walls 1 X6 15, 16, 17
2 X7 34e44
3 X8 15, 16, 17
living room ex walls 1 X9 1, 2
2 X10 15, 16, 17
3 X11 34e44
4 X12 19, 20, 21
Store ex walls 1 X13 1, 2
2 X14 15, 16, 17
Windows 1 X15 48e65
Air gap X16 3 mm, 6 mm, 8 mm
2 X15 48e65
Store ceiling 1 X17 19, 20, 21
Living room ceiling 1 X18 34e44
2 X19 19, 20, 21
Attic ceiling 1 X20 19, 20, 21
Gable 1 X21 15, 16, 17
Doors 1 X22 32

simulation programs [38], is used in this study. Second step in-

cludes the calculation of objective functions; LCC, LCE, and thermal
comfort index. Calculation of objective functions is discussed in
details in next sections. The magnitudes calculated in second step Fig. 1. The optimization process.
are transferred to third step to evaluate, improve and generate in-
puts for next simulations.
value of electricity consumption, calculated by energy simulation
In order to employ the HS algorithm in solving the mentioned
software for the whole building lifetime is added to the LCC.
optimization problem, a C# code is developed which is capable of
Following equation is used to calculate the LCC:
generating and executing EnergyPlus input file (.IDF file), reading
the EnergyPlus output files (.CSV file), and comparing the obtained
results with previous iteration. The logic of HS algorithm in pro- LCC ¼ Initial Cost þ P: V:m=r þ P: V:e (1)
ducing new input files has been discussed in next sections.
where: P: V:m=r is the present value of maintenance and replace-
2.3. LCC and LCE ment of materials, and P: V:e is the present value of energy
consumed to operate the building. To calculate the present values,
To calculate the LCC of each design, the present value of mate- building life time is assumed to be 50 years and the energy discount
rials and maintenance costs is calculated. Additionally, the present rate is considered to be 3% [40].
 E. Mostavi et al. / Energy 121 (2017) 606e615
To calculate the LCE of the building, all phases of building life
including the per-use phase (material extraction and production),
use phase (maintenance and replacement), and end of life phase
(demolition) are considered in this study. Equation (2) is used to
calculate the LCE of different designs.
LCE ¼ EPre Use þ EUse þ EEnd of Life þ EOperation
Where EPre Use ; EUse ; and EEnd of Life are the amount of emission
during the pre-use, use, and demolition stages, and the EOperation is
the amount of GHGs emitted during the operating stage of the
building life. It should be noted that the EUse includes just the
emission for maintenance and replacement of materials, however
EOperation includes all the emissions during building operation and
service period. SimaPro database is used to identify the level of
emission of each stage. To calculate the LCE of the building, total
amount of GHG emissions of each design including CO2, CO, CH4,
and N2O are calculated and the total Global Warming Potential
(GWP) of each design is identified according to the Global Warming
Factors (GWF). Following equation is used to calculate the emission
(E) of each phase in the current study:
X
n
E¼ GHGi  GWFi
i¼1
Materials which have emission of gases with higher GWF may
have more environmental impacts than materials with higher
emissions but lower GWF. Table 3 shows the major GHGs and their
GWPs.
The calculated LCE includes aggregated amounts of emission of
each material during the material production, transportation,
maintenance and replacement, as well as demolish/transportation
which are extracted through the LCA data bases. It should be noted
that this study did not consider the environmental impacts asso-
ciated with transportation of materials and products.
2.4. Thermal comfort indices
Thermal comfort is described as: “the condition of mind which
expresses the satisfaction with the thermal environment” [42].
Thermal comfort assessment methods have been investigated and
developed by various researchers. In a study conducted by De
Freitas & Grigorieva [43], they identified about 162 thermal comfort
indices that can be used for different purposes. However, these
indices are different based on their variables, application, methods,
interpretation and accuracy.
According to ASHRAE 55, thermal comfort zone is the condition
where 80% of slightly active occupants feel thermally satisfied [42].
It is expected that the heat strain experienced by different in-
dividuals is a function of exposed heat stress. Accordingly, different
Heat Stress Indexes (HSI) have been developed to estimate the level
of exposure to heat stress [44]. A critical constraint of all these
studies for a normal body function is the deep body temperature of
37  C (normal body situation) [44]. Based on the method of
calculation of index, the developed HSIs can be divided into three
categories: indices that are calculated by solving the heat balance
equations, indices that are focusing on the subjective strain, and
Table 3
GWFs of gases considered in calculation of LCE [41].
20 years
Carbon Dioxide CO2 1 1
Carbon Monoxide CO
Methane CH4 56
Nitrous Oxide N2O 280
609
indices based on environmental measurements [44].
As most of the famous HSIs such as wet-bulb dry temperature
[45], modified discomfort index [46], predicted mean vote (PMV)
[47] and discomfort index [48] are focusing on environmental
measurements (wet-bulb temperature, humidity, wind velocity),
they cannot reflect the occupants' perception clearly. For instance,
(2) the PMV and Predicted Percentage of Dissatisfied (PPD) are calcu-
lated based on different parameters such as clothing factor, clothing
insulation, metabolic rate, vapor pressure, etc., but they do not
consider the occupants' preferences. On the other hand, in order to
determine the optimum design solutions, there is a need to have a
solid function for thermal comfort which can be used by optimi-
zation program. Hence, this study employs the utility theory to
identify the occupant satisfaction. The results of the proposed
method (based on the subjective studies) are compared with PMV
method (based on the heat transfer balance) for the selected case
study environment.
2.5. Predicted mean vote method (PMV)
This method is developed by Fanger [47] and approved as an ISO
(International Organization for Standardization) [49] standard. It
(3) uses thermal scales of 3 (cold) to 3 (hot) to describe the zone
thermal situation. In order to develop the mathematical model,
statistical studies which were conducted on the large number of
questioners in different climate regions are used. Equation (4)
shows the model developed by Fanger [47] to determine the PMV:
h i h
PMV ¼ 0:303e0:036 M þ 0:028  ðM  WÞ
h i
 3:96E8 fcl ðtcl þ 273Þ4  ðtr þ 273Þ4  fcl hc ðtcl  ta Þ
 3:05½5:73  0:007 ðM  WÞ  pa
 0:42½ðM  WÞ  58:15  0:0173Mð5:87  paÞ
i
 0:0014Mð34  ta Þ
(4)
where:fcl shows clothing factor, M shows metabolic rate equals to
115 W/m2, W shows external work which assumed to be zero, tcl , tr ,
and ta show the clothing surface temperature, mean radiant tem-
perature, and air temperature, hc shows heat transfer coefficient
through convection, and pa shows the air vapor pressure.
Clothing factor is a function of the cloths insulation and usually
considered to be between 1.1 and 1.2 according to the summer or
winter clothing. Clothing surface temperature (tcl Þ is calculated
based on the metabolic rate, clothing insulation, air temperature
and air vapor pressure using the following equation:
tcl ¼ 35:7  0:0275 ðM  WÞ  Rcl ½ðM  WÞ  3:05  ½5:73
 0:007ðM  WÞ  pa  0:42  ½ðM  WÞ  58:15
 0:0173Mð5:87  paÞ  0:0014Mð34  ta Þ
(5)
where Rcl shows the clothing thermal insulation which is a function
of clothing insolation (Icl [clo]):
Rcl ¼ 0:155 Icl (6)
Convection transfer coefficient (hc ) is a function of air velocity
100 years
(V) and calculated using equation (7):
2
21
hc ¼ 12:1 ðVÞ0:5 (7)
310
Using the calculated PMV, PPD is calculated as following:
610 E. Mostavi et al. / Energy 121 (2017) 606e615

implementation and it is not sensitive to initial condition values.

PPD ¼ 100  95 e½ð0:3353PMV þ0:2178PMV Þ
4 2
(8)
To have a comparison between the proposed thermal comfort
index and PMV which is one of the most famous thermal comfort
indexes, different PPDs for multiple temperatures have been
calculated. To calculate these magnitudes, the air velocity of 0.1 m/
s, metabolic rate of 1.1 met (metabolic rate of office spaces),
clothing of 0.5 clo (normal inside clothing) and humidity of 50% are
assumed.
2.6. Utility theory thermal comfort index (TCI)
The present study implemented the utility theory to identify
occupants' thermal utility function and calculate occupants' Ther-
mal Comfort Index (TCI). Initially, the concept of utility and deter-
mining the uncertain gambling value was proposed by Bernoulli
[50]. Later, the concept expanded into the decision making prob-
lems to evaluate decision makers' attitude toward an accident. In
this study, the following formula, suggested by Thurston [51] has
been used to identify occupants' thermal utility function.
h Ti
UðTÞ ¼ A  B  eR (9)
Where T is room temperature ( C), A and B are scaling factors,
and R is the risk tolerance. Since scaling factors are functions of R,
the first step of determining the utility function is to calculate R
according to the following equation:
h i
Highest T
CEðcertainty equivalentÞ ¼ LN 0:5  e R þ 0:5  e R
Lowest T
 ðRÞ
(10)
h Lowest T Highest T
i similar to office working schedule. The total area of this building is
A ¼ e ÷ e RT  e RT
Lowest
RT (11)
Additionally, as this algorithm does not utilize the derivatives in the
search process, it is a powerful tool for solving multi-modal prob-
lems. Fig. 2 shows a schematic of the HS algorithm. As it can be seen
in this figure, the process of finding the optimum solution in this
algorithm includes generating random initial values for variables,
evaluating the results of these inputs, save the results in the Har-
mony Memory (HM), improving the solutions, evaluation of new
inputs, and replacing the worst result with new one if the results
are better than worst result. This loop should be continued to reach
the maximum iteration numbers. The improvement of the solution
is directed by two factors of Harmony Memory Considering Rate
(HMCR) and Pitch Adjusting Rate (PAR). HMCR is the probability
that each variable will be selected from current HM, and obviously
1-HMCR is the probability that each variable will be selected
randomly. As the HM includes multiple number of solutions,
choosing whether the sources of selected variable in HMCR should
be modified or utilized as it is in HM, will take place by PAR
probability [54].
2.8. Case study
To illustrate the function of the developed optimization process,
a case study of a small office building in State College, Pennsylvania,
USA is investigated and the optimum design of building envelope is
determined. Fig. 3 shows a plan view of the case study. As it can be
seen in this figure, this office has two windows facing south and
east. This office building, having a single conditioned zone and two
unconditioned zones, has a single speed electric heat pump DX air
heating coil (EnergyPlus object name: Coil:Heating:DX:Single-
Speed) with the heat capacity of 9000 W, coefficient of perfor-
mance (COP) of 3.0 and the flow rate of 0.45 m3/s to maintain the
air temperature at 24  C. The HVAC system schedule is set to be

h Lowest T Highest T
i
B ¼ 1÷ e RT  e RT (12)

In which CE is the certainty equivalent, equals to the average of

maximum and minimum satisfying temperatures that occupants
may consider in their evaluation, Highest T and Lowest T are the
maximum and minimum temperatures that received the highest
and lowest utility grade in the conducted survey. This study in-
cludes two sets of interviews to determine the cooling and heating
utility function. Each utility interview considers 6-degree tolerance
(18  Ce24  C and 24  Ce30  C) for occupants' utility.
To calculate the TCI, the monthly living zone temperatures are
exported from the simulation model. Then, the TCI for each design
is estimated based on the following equation:
P12
1 UðTÞ
TCI ¼  100 (13)
12

where U(T) is the utility associated with average temperature of

each month. To find the magnitude of utility between the 12
mentioned temperatures, a simple linear interpolation has been
performed.

2.7. Harmony search algorithm

Since the introduction of HS algorithm in 2001, wide variety of

researches have employed this algorithm [52,53]. The HS algorithm
is a meta-heuristic algorithm, capable of solving continuous and
discrete problems. This algorithm involves few parameters for Fig. 2. HS algorithm procedure [54].
 E. Mostavi et al. / Energy 121 (2017) 606e615 611

3. Results and discussions

3.1. Utility theory thermal comfort index

Although there are several formulas to calculate the occupants'

satisfaction level (referring to Fanger method mentioned in previ-
ous sections), but to attain the most reliable function, occupants
should be directly contacted and their satisfactory temperatures
should be monitored. It should be noted that occupants' perception
of satisfaction depends on many different physical, mental and
Fig. 3. Case study located in State College, PA, USA.
cultural parameters. Hence, to determine the thermal satisfaction
function, a utility based survey is conducted in State College,
Pennsylvania, USA.
Table 4 Since the population of State College is about 50,000, a sample
Utility function for office staff. with 96 people population can be a proper statistical representative
Temperature Risk tolerance Scale factor A Scale factor B with 10% confidence interval. Hence, in the effort of calculating the
16e24 7.657 1.5426 12.47
thermal utility function of the occupants, a total number of 108
24e31 3.554 1.1621 0.00018 interviews were conducted between graduate students and office
staffs of the Pennsylvania State University who work more than 6 h
a day in their office. Each participant was asked to grade the tem-
210 m2 including 160 m2 working area and 50 m2 store room and peratures from 1 to 5 (1 represents the least preferable and 5
utilizes electricity as the only source of energy. represents the most preferable) based on their thermal perception.

Fig. 4. Occupants' utility according to the conducted survey; blue represents cold seasons and red represents hot seasons. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)

Fig. 5. Occupants' utility according to the conducted survey.

612 E. Mostavi et al. / Energy 121 (2017) 606e615
Fig. 6. LCC vs. LCE of developed models.
Fig. 7. a) LCC vs TCI and b) LCE vs TCI.
According to the performed survey study and equations (5)e(7),
risk tolerance and scale factors for temperatures less than 24  C and
more than 24  C are calculated. Table 4 shows the magnitudes of
utility function parameters.
Using the magnitudes given in Table 4, the level of thermal
utility of occupants can be drawn from Fig. 4. As it can be seen in
this figure, occupants have positive attitude toward cooler tem-
peratures than warmer temperatures. In other words, the occupant
utility decreases faster due to the increase in temperature in
comparison with temperature reduction. Since the determined
function is based on the results of the survey, it is highly dependent
on the participants' perception, previous experiences, and cultural
values.
Fig. 5 shows the results of the proposed TCI and PMV methods.
Since the PPD is the predicted dissatisfaction, to identify the
satisfied population, 100-PPD is calculated and compared with the
Utility Theory based TCI. As it can be seen in this figure, PMV
method identifies a broader range of temperatures having satis-
faction higher than zero, while the proposed TCI is more focused.
Additionally, as the proposed TCI concentrates on the real occu-
pants, the difference between the most preferable temperatures of
the two methods can be identified.
3.2. Multi-objective optimization
This section presents the results of the optimization process. As
mentioned before, to automate the optimization process, a C# code
is developed and employed in this study. This code is capable of
generating the input file for EnergyPlus, executing the EnergyPlus,
and reading the results and inserting the modifications in input file
for the next levels of simulation. To attain the optimum solution,
this code generated 3800 different models in 3 days using a com-
puter with Intel® Core™ i5 CPU 3.3 GHz and 4.00 GB Ram.
Fig. 6 demonstrates the LCE and LCC of the simulations (blue
dots) and the Pareto front solutions (red triangles). As illustrated in
this figure, eight designs are shown to be the optimum. These de-
signs labeled as A, B, C, D, E, F, G, and H which have the least LCE and
LCC among all the simulations. Design A has the least amount of
LCC and highest LCE among the optimum solutions. On the other
hand, alternative H is having highest LCC and the least LCE.
To identify the strategies with higher levels of occupant satis-
faction, calculated TCI should be considered in decision making
 E. Mostavi et al. / Energy 121 (2017) 606e615 613

Fig. 8. Three dimensional view of optimization results.

process. Fig. 7a and b respectively show the LCC vs TCI and LCE vs
TCI of simulation results. As it can be seen in these figures, although
the proposed strategies in previous step indicate the minimum cost
and environmental impacts, but in terms of occupant satisfaction
they are among the worst choices.
Fig. 8 shows the 3 dimensional cloud of the simulation results
and the frontier results. In this figures the height of the cloud il-
lustrates the level of satisfaction. Fig. 8 a shows all the simulation
results with blue points, the optimum solutions considering
emission, cost and satisfaction with red points on the cloud. Fig. 8 b
shows the frontier results of this optimization process with black
dots and the Pareto front of LCC and LCE on the horizontal surface
with red dots.
A simple geometrical method to solve this optimization problem
is to identify the two objective having the highest weighing in
decision making process (which are making the horizontal surface),
and consider all the designs located above the selected design. It is
clear that the designs located in higher altitude have the higher
occupant satisfaction level. It can be seen in Fig. 8, that all the op-
timum points considering all three objectives (cost, emission, and
satisfaction) are located on the higher surface of the cloud.
Employing this strategy, all designs having LCC and LCE similar to
the mentioned optimum solutions of A, B, C, D, E, F, G, and H (and
their vicinity with 2% fluctuation) were considered in TCI evalua-
tion level. Table 5 shows the new determined optimum solutions.
As it can be seen in this table and comparing with Fig. 7a and b, the
satisfaction levels of new solutions are improved. In order to
determine the best strategies, a proper weighting system could be
employed. The decision maker's weighting can be identified
through interviews focusing on the objectives of the optimization
problem.
4. Conclusions
Making proper decisions at the early stages of design can assist
designers to attain best sustainable (i.e., occupant satisfying and

Table 5
Optimum designs.

Strategy A0 , C0 , D0 B0 E0 , G 0 F0 H0
LCC (1000  61.4 65.9 77.7 78.8 97.5
$)
LCE (Ton 183 187 178 173 175
CO2)
TCI (%) 70.61 71.23 69.49 68.98 70.11

X1 Concrete 100-mm Concrete 100-mm Concrete 150-mm Concrete 150-mm Concrete 100-mm
X2 Concrete 150-mm Concrete 200-mm Concrete 200-mm Concrete 150-mm Concrete 200-mm
X3 Wood subfloor Wood subfloor Carpet Wood subfloor Wood subfloor
X4 Membrane Membrane Membrane Wood shingle Membrane
X5 Plywood 16-mm Plywood 13-mm Plywood 13-mm Plywood 10-mm Plywood 13-mm
X6 Gypsum board 16-mm Gypsum board 9.5-mm Gypsum board 16-mm Gypsum board 9.5-mm Gypsum board 9.5-mm
X7 Rigid insulation fiberglass 1.500 Cellular polyurethane 600 Rigid insulation fiberglass 1.500 Rigid insulation fiberglass-300 Rigid insulation fiberglass
2.500
X8 Gypsum board 9.5-mm Gypsum board 9.5-mm Gypsum board 9.5-mm Gypsum board 13-mm Gypsum board 9.5-mm
X9 Wood wall Vinyl wall Wood wall Wood wall Vinyl wall
X10 Plywood 16-mm Plywood 10-mm Plywood 16-mm Plywood 10-mm Plywood 10-mm
X11 Cellular polyurethane 400 Cellular polyurethane 500 Cellular polyurethane 300 Cellular polyurethane 300 Blanket insulation-600
X12 Gypsum board 13-mm Gypsum board 16-mm Gypsum board 13-mm Gypsum board 16-mm Gypsum board 9.5-mm
X13 Vinyl wall Wood wall Vinyl wall Vinyl wall Vinyl wall
X14 Plywood 10-mm Plywood 16-mm Plywood 10-mm Plywood 10-mm Plywood 16-mm
X15 Ref a clear low 6-mm Ref a clear low 6-mm Ref b clear low 6-mm Ref b clear low 6-mm Ref a clear low 6-mm
X16 8 mm 13 mm 8 mm 13 mm 8 mm
X17 Gypsum board 16-mm Gypsum board 16-mm Gypsum board 16-mm Gypsum board 16-mm Gypsum board 16-mm
X18 Fiberglass insulation board Fiberglass insulation board Fiberglass insulation board Fiberglass insulation board Cellular polyurethane 5.500
11.500 11.500 11.500 11.500
X19 Gypsum board 13-mm Gypsum board 9.5-mm Gypsum board 13-mm Gypsum board 13-mm Gypsum board 9.5-mm
X20 Gypsum board 16-mm Gypsum board 13-mm Gypsum board 16-mm Gypsum board 16-mm Gypsum board 13-mm
X21 Plywood 10-mm Plywood 16-mm Plywood 10-mm Plywood 16-mm Plywood 10-mm
X22 Steel door Steel door Steel door Steel door Steel door
614 E. Mostavi et al. / Energy 121 (2017) 606e615
cost effective) design. In this study, a multi-objective optimization
model is developed to minimize the LCC and LCE and maximize the
occupants' thermal satisfaction of a building. The optimization
process employs the HS algorithm to determine the best design
strategy. This optimization algorithm compares LCC, LCE and TCI of
different alternatives to evaluate different design variables. To
calculate TCI, a novel utility theory based methodology has been
defined to investigate occupants' perception toward the thermal
condition of their working environment.
To illustrate the process and function of the developed model, a
case study of an office building located in State College, Pennsyl-
vania has been investigated and the best design strategies have
been identified. A series of designs with lowest LCC and LCE and
highest TCI was identified as optimum results. These optimum
solutions provide decision makers with a verity of alternatives with
lowest LCC, LCE and highest TCI and the decision maker has the
ability to utilize his own weighting system to select between the
options.
As the scope and details of building design optimization algo-
rithms expand, the outcomes become more reliable in shorter
times. However, involving different human based aspects of the
design, such as architectural perspective, into the process of design
optimization is still an obstacle and needs further investigations in
future studies.
Acknowledgement
The authors would like to acknowledge the financial support
through a grant from the Qatar National Research Foundation
(QNRF)/National Priorities Research Program (NPRP) (NPRP 7 - 143
- 2 - 070).
References
[1] Rajagopalan P, Jamei E. Thermal comfort of multiple user groups in indoor
aquatic centres. Energy Build Oct. 2015;105:129e38.
[2] Artuso P, Santiangeli A. Energy solutions for sports facilities. Int J Hydrogen
Energy Jun. 2008;33(12,):3182e7.
[3] Nematchoua MK, Tchinda R, Orosa JA. Thermal comfort and energy con-
sumption in modern versus traditional buildings in Cameroon: a
questionnaire-based statistical study. Appl Energy Feb. 2014;114:687e99.
[4] Day JK, Gunderson DE. Understanding high performance buildings: the link
between occupant knowledge of passive design systems, corresponding be-
haviors, occupant comfort and environmental satisfaction. Build Environ Nov.
2014;84:114e24.
[5] Rahimian M, Cardoso-Llach D, Iulo L. Participatory energy management in
building networks. In: Sustainable Human Building ecosystems. American
Society of Civil Engineers; 2015. p. 27e35.
[6] Ghahramani A, Zhang K, Dutta K, Yang Z, Becerik-Gerber B. Energy savings
from temperature setpoints and deadband: quantifying the influence of
building and system properties on savings. Appl Energy Mar. 2016;165:
930e42.
[7] Huizenga C, Abbaszadeh S, Zagreus L, Arens E. Air quality and thermal comfort
in office buildings: results of a large indoor environmental quality survey. In:
Proceedings of healthy buildings; 2006. p. 393e7.
[8] Tsuzuki K, Arens E, Bauman F, Wyon D. Individual thermal comfort control
with desk-mounted and floor-mounted task/ambient conditioning (TAC)
systems. 1999.
[9] Holopainen R, Tuomaala P, Hernandez P, Ha €kkinen T, Piira K, Piippo J. Comfort
assessment in the context of sustainable buildings: comparison of simplified
and detailed human thermal sensation methods. Build Environ Jan. 2014;71:
60e70.
[10] Varjo J, Hongisto V, Haapakangas A, Maula H, Koskela H, Hyo € n€
a J. Simulta-
neous effects of irrelevant speech, temperature and ventilation rate on per-
formance and satisfaction in open-plan offices. J Environ Psychol Dec.
2015;44:16e33.
[11] Hancock PA, Ross JM, Szalma JL. A meta-analysis of performance response
under thermal stressors. J Hum Factors Ergon Soc 2007;49(5):851e77.
[12] Li X, Malkawi A. Multiobjective optimization for thermal mass model pre-
dictive control in small and medium size commercial buildings under summer
weather conditions. Energy 2016;112.
[13] Arup. Sustainable building design [Online]. Available, http://www.arup.com/
Services/Sustainable_Buildings_Design.aspx (Accessed:14 Mar 2015).
[14] Mahapatra K. Energy use and CO2 emission of new residential buildings built
under specific requirements e the case of Va €xjo
€ municipality, Sweden. Appl
Energy Aug. 2015;152:31e8.
[15] Tavakkoli S, Lokare OR, Vidic RD, Khanna V. Systems-level analysis of waste
heat recovery opportunities from natural gas compressor stations in the
United States. ACS Sustain Chem Eng 2016;4(7). p. acssuschemeng.5b01685.
[16] Kovacic I, Zoller V. Building life cycle optimization tools for early design
phases. Energy 2015;92:409e19.
[17] Tavakkoli S, Lokare OR, Vidic RD, Khanna V. Systems-level analysis of waste
heat recovery opportunities from natural gas compressor stations in the
United States. ACS Sustain Chem Eng 2016;4(7):3618e26.
[18] Hawkins TR, Singh B, Majeau-Bettez G, Strømman AH. Comparative envi-
ronmental life cycle assessment of conventional and electric vehicles. J Ind
Ecol 2013;17(1):53e64.
[19] Asadi S, Amiri SS, Mottahedi M. On the development of multi-linear regres-
sion analysis to assess energy consumption in the early stages of building
design. Energy Build Dec. 2014;85:246e55.
[20] Basbagill J, Flager F, Lepech M, Fischer M. Application of life-cycle assessment
to early stage building design for reduced embodied environmental impacts.
Build Environ Feb. 2013;60:81e92.
[21] Ramaji IJ, Memari AM. Information exchange standardization for BIM appli-
cation to multi-story modular residential buildings. In: Proceeding of the
architectural engineering 6th conference. American Society of Civil Engineers;
2015.
[22] Kim JH, Geem ZW, Loganathan GV. A new heuristic optimization algorithm:
harmony search. Simulation 2001;76:60e8.
[23] Talbi E-G. Metaheuristics from design to implementation. 2009.
[24] Carlucci S, Cattarin G, Causone F, Pagliano L. Multi-objective optimization of a
nearly zero-energy building based on thermal and visual discomfort mini-
mization using a non-dominated sorting genetic algorithm (NSGA-II). Energy
Build Oct. 2015;104:378e94.
[25] Brownlee AEI, Wright JA. Constrained, mixed-integer and multi-objective
optimisation of building designs by NSGA-II with fitness approximation.
Appl Soft Comput Aug. 2015;33:114e26.
[26] Nguyen A-T, Reiter S, Rigo P. A review on simulation-based optimization
methods applied to building performance analysis. Appl Energy Jan.
2014;113:1043e58.
[27] Delgarm N, Sajadi B, Kowsary F, Delgarm S. Multi-objective optimization of
the building energy performance: a simulation-based approach by means of
particle swarm optimization (PSO). Appl Energy May 2016;170:293e303.
[28] Rapone G, Saro O. Optimisation of curtain wall facades for office buildings by
means of PSO algorithm. Energy Build 2012;45:189e96.
[29] Asadi E, da Silva MG, Antunes CH, Dias L, Glicksman L. Multi-objective opti-
mization for building retrofit: a model using genetic algorithm and artificial
neural network and an application. Energy Build Oct. 2014;81:444e56.
[30] Evins R. A review of computational optimisation methods applied to sus-
tainable building design. Renew Sustain Energy Rev 2013;22:230e45.
[31] Fesanghary M, Asadi S, Geem ZW. Design of low-emission and energy-
efficient residential buildings using a multi-objective optimization algo-
rithm. Build Environ Mar. 2012;49:245e50.
[32] Ascione F, Bianco N, De Stasio C, Mauro GM, Vanoli GP. Simulation-based
model predictive control by the multi-objective optimization of building en-
ergy performance and thermal comfort. Energy Build Jan. 2016;111:131e44.
[33] Harmathy N, Magyar Z, Foli c R. Multi-criterion optimization of building en-
velope in the function of indoor illumination quality towards overall energy
performance improvement. Energy 2016;114:302e17.
[34] Aparicio Ruiz P, S anchez De La Flor FJ, Molina Felix JL, Salmero n Lisse
n J,
Guadix Martín J. Applying the HVAC systems in an integrated optimization
method for residential building's design. A case study in Spain. Energy Build
2016;119:74e84.
[35] Amiri SS, Mottahedi M, Asadi S. Using multiple regression analysis to develop
energy consumption indicators for commercial buildings in the U.S. Energy
Build Dec. 2015;109:209e16.
[36] Shi X, Tian Z, Chen W, Si B, Jin X. A review on building energy efficient design
optimization rom the perspective of architects. Renew Sustain Energy Rev
2016;65:872e84.
[37] Si B, Tian Z, Jin X, Zhou X, Tang P, Shi X. Performance indices and evaluation of
algorithms in building energy efficient design optimization. Energy 2016;114:
100e12.
[38] Nguyen A, Reiter S, Rigo P. A review on simulation-based optimization
methods applied to building performance analysis. Appl Energy 2014;113:
1043e58.
[39] EnergyPlus 8.4. U.S. Department of Energy Building Technologies Office; 2015.
[40] Rushing AS, Kneifel JD, Lippiatt BC. Energy price indices and discount factors
for life-cycle cost analysis, 26; April 2005. p. 67. 2014.
[41] Myhre G, Shindell D, Bre on F-M, Collins W, Fuglestvedt J, Huang J, et al.
Anthropogenic and natural radiative forcing. Clim Chang 2013 Phys Sci Basis
Contrib Work Gr I Fifth Assess Rep Intergov Panel Clim Chang 2013:659e740.
[42] ASHRAE. ANSI/ASHRAE standard 55-2010. American S. Atlanta; 2010.
[43] de Freitas CR, Grigorieva EA. A comprehensive catalogue and classification of
human thermal climate indices. Int J Biometeorol 2014;59(1):109e20.
[44] Epstein Y, Moran DS. Thermal comfort and the heat stress indices. Ind Health
2006;44(3):388e98.
[45] Wallace RF, Kriebel D, Punnett L, Wegman DH, Wenger CB, Gardner JW, et al.
The effects of continuous hot weather training on risk of exertional heat
illness. Med Sci Sports Exerc 2005;37(1):84e90.
 E. Mostavi et al. / Energy 121 (2017) 606e615
[46] Moran Y, Laor DS, Epstein A, Shapiro Y. Ovid: a modified discomfort index
(mdi) as a substitute for the wet bulb globe temperature (WBGT). Med Sci
Sport Exerc May 1998;30(5):284.
[47] Fanger PO. Thermal comfort - analysis and applications in environmental
engineering. 1970.
[48] Thom EC. The discomfort index. Weatherwise February 2014;12. 1959.
[49] ISO. International organization for standardization. 2005 [Online]. Available,
http://www.iso.org (Accessed 06 Apr 2016).
[50] Bernoulli D. Exposition of a new theory on the measurement of risk. J Econom
Soc 1738:23e6.
615
[51] Thurston DL, Locascio A. Decision theory for design economics. Eng Econ
1994:41e71. http://dx.doi.org/10.1080/00137919408903138.
[52] Yoo DG, Kim JH, Geem ZW. Overview of harmony search algorithm and its
applications in civil engineering. Evol Intell 2013;7(1):3e16.
[53] Manjarres D, Landa-Torres I, Gil-Lopez S, Del Ser J, Bilbao MN, Salcedo-Sanz S,
et al. A survey on applications of the harmony search algorithm. Eng Appl Artif
Intell Sep. 2013;26(8):1818e31.
[54] Wang X, Gao X-Z, Zenger K. An introduction to harmony search optimization
method. SpringerBriefs in Computational Intelligence; 2015.