You are on page 1of 32

Applied Energy 174 (2016) 37–68

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Multi-stage and multi-objective optimization for energy retrofitting


a developed hospital reference building: A new approach to assess
cost-optimality
Fabrizio Ascione a, Nicola Bianco a, Claudio De Stasio a, Gerardo Maria Mauro a,⇑, Giuseppe Peter Vanoli b
a
University of Naples Federico II, Piazzale Tecchio 80, 80125 Napoli, Italy
b
University of Sannio in Benevento, Piazza Roma 21, 82100 Benevento, Italy

h i g h l i g h t s

 The energy retrofit of hospital buildings is a complex and relevant issue.


 A hospital reference building is developed and the cost-optimal retrofit is found.
 Retrofit measures are screened via preliminary sampling and sensitivity analysis.
 Cost-optimality is assessed by multi-stage and multi-objective Pareto optimization.
 Investment cost, primary energy consumption and global cost are minimized.

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

Article history: The ‘Energy Performance of Buildings Directive’ Recast (i.e., 2010/31/EU) establishes that building energy
Received 14 January 2016 retrofit should pursue ‘‘cost-optimal levels”. However, a reliable and rigorous cost-optimal analysis is an
Received in revised form 23 March 2016 arduous and computationally-expensive issue, especially for complex buildings such as hospitals. The
Accepted 16 April 2016
paper tackles this issue by providing a novel methodology to identify robust cost-optimal energy retrofit
Available online 22 April 2016
solutions. Multi-stage and multi-objective (Pareto) optimization is performed with the aim of minimizing
the computational burden required to achieve reliable outcomes. The methodology combines EnergyPlus
Keywords:
and MATLABÒ and includes two optimization stages, preceded by a preliminary energy investigation that
Building energy retrofit
Cost-optimal analysis
performs Latin hypercube sampling and sensitivity analysis. The preliminary investigation and the first
Reference building optimization stage, which runs a genetic algorithm, aim at detecting efficient energy retrofit measures
Pareto genetic algorithm (ERMs) to reduce thermal energy demand for space heating and cooling. In the second optimization stage,
Building sampling these ERMs are combined with further ERMs, addressed to improve the efficiency of energy systems and
Hospital buildings to exploit renewable energy sources. Investment cost, primary energy consumption and global cost
related to the resulting retrofit packages are investigated by means of smart exhaustive sampling.
Finally, the cost-optimal solution is identified both in presence of a limitless economic availability and
of limited budgets. The methodology is applied to a hospital reference building (RB), which represents
hospitals built in South Italy between 1991 and 2005. The RB is defined by using an original approach,
as required by the complexity of the examined building category. The achieved cost-optimal retrofit
packages imply a reduction of primary energy consumption up to 67.9 kW h/m2 a (12.2%) and of global
cost up to 2932 k€ (24.5%) with a maximum investment of 1236 k€.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction and literature review lighting the main objectives and original aspects of the study in
Section 1.1.
This section provides preliminary introductive notes as well as It is well known that the building sector is responsible for a sub-
the literature background in which the paper is framed, by high- stantial amount of energy consumption, and thus polluting emis-
sions, at European (around 40% [1]) and World level (around 32%
[2]). Furthermore, the building turn-over rate is very low, espe-
⇑ Corresponding author. cially in the industrialized countries that account for a wide part
E-mail address: gerardomaria.mauro@unina.it (G.M. Mauro). of World consumption. For instance, most members of the

http://dx.doi.org/10.1016/j.apenergy.2016.04.078
0306-2619/Ó 2016 Elsevier Ltd. All rights reserved.
38 F. Ascione et al. / Applied Energy 174 (2016) 37–68

Nomenclature

Symbols SRRC standardized rank regression coefficient (–)


(a), (b), . . . , (h), (i) labels of the ERMs for TED reduction (–) U thermal transmittance of opaque components (W/m2 K)
a absorptance to solar radiation (–) Uw thermal transmittance of the windows (glass + frame)
cov percentage of roof area covered by solar collectors or PV (W/m2 K)
panels (–) Vair air flow handled by heat recovery systems (m3/h)
dGC difference in GC compare to RB (€) e efficiency of heat recovery systems (Wth/Wth)
dPEC difference in PEC compare to RB (W h/m2 a) g efficiency of boilers (Wth/Wpr)
e thermal (infrared) emissivity (–) gel electrical efficiency of CHP and CCHP systems (Wel/Wpr)
gmax maximum number of generations concerning the GA (–) gth thermal efficiency of CHP and CCHP systems (Wth/Wpr)
n number of mixed-integer variables introduced by ERMs
addressed to TED (–) Subscripts
r ratio between N and n (–) cool referred to space cooling
s population size concerning the GA (–) el referred to electrical power
t thickness of thermal insulation layer (m) heat referred to space heating
x0 vector of design variables of the 1st optimization stage pr referred to primary power
(–) roof referred to the roof
x00 vector of design variables of the 2nd optimization stage th referred to thermal power
(–) tot referred to the total demand for space conditioning
COP coefficient of performance of heat pumps (Wth/Wel) (heating + cooling)
EER energy efficiency ratio of chillers (Wth/Wel) walls referred to the external vertical walls
EERass energy efficiency ratio of chillers absorption chillers RB base configuration of the reference building (baseline)
(Wth/Wth)
F0 objective functions of the 1st optimization stage: Acronyms
F = [TEDheat, TEDcool] (–) BPO building performance optimization
F00 objective functions of the 2nd optimization stage: BPS building performance simulation
F = [IC, dPEC, dGC] (–)
CCHP combined cooling, heating and power
GC global cost (€)
CHP combined heating and power
IC investment cost (€)
DHW district hot water
N number of cases included in the building sample
ERM energy retrofit measure
(i.e., sample size) (–)
GA genetic algorithm
PEC primary energy consumption per unit of conditioned
HVAC heating, ventilating and air conditioning
area and year (W h/m2 a)
LHS latin hypercube sampling
P power (W)
PV photovoltaic
S building sample (–)
RB reference building
SHGC solar heat gain coefficient (–)
RES renewable energy source
Tsol transmittance to solar radiation (–)
SA sensitivity analysis
Tvis transmittance to visible radiation (–)
TED thermal energy demand for space conditioning
(W h/m2 a)

European Union (EU) extend their stock by less than 1% per year savings by minimizing, at the same time, building lifecycle costs.
[3]. Therefore, building energy retrofit is one of the most powerful Nevertheless, this kind of analysis requires high computational
weapons that our generation owns to fight in name of environment efforts because numerous retrofit scenarios have to be simulated
protection and sustainability. Nevertheless, it is a complex issue by means of reliable BPS (building performance simulation) tools,
that involves two divergent perspectives: which run time-costly dynamic simulations [7], in order to achieve
a robust cost-optimal solution. Because of this computational
 the public one, which aims at reducing energy consumption and issue, the cost-optimal analysis cannot be applied to every build-
polluting emissions, ing, and therefore it can be limited to reference buildings (RBs)
 the private one, which aims at achieving economic benefits [4]. as established in [6]. The RBs have to represent the national build-
ing stock, thereby covering any building category, which is defined
In order to find a compromise between these two perspectives, as a cluster of buildings with the same climatic conditions (loca-
the Energy Performance of Buildings Directive 2010/31/EU (EPBD tion), construction type and end-use classification [6]. The cost-
Recast) [5] prescribes the cost-optimal analysis to address building optimal energy solutions are identified for the RBs, and then the
energy retrofit. The objective is minimizing the global cost related outcomes can be extended to most buildings of the represented
to energy uses over building lifecycle, as detailed in the Delegated categories. Based on the source and type of collected data used
Regulation (EU) No. 244/2012 [6]. More in detail, the cost-optimal to define the RBs, there are the following RB models [8,9]:
analysis requires to compare the global cost (GC) and the primary
energy consumption (PEC) for different packages of proper energy  example RB: ideal building model, defined on the basis of
measures. The outcome is a cost-optimal curve that depicts the experts’ inquiries and assumptions;
value of GC in function of PEC for all packages. The minimum point  real RB: real existing building with average characteristics
on the curve identifies the cost-optimal solution. This is a powerful concerning a specific building category;
concept that ensures the best trade-off between the public and  theoretical RB: ideal building model defined by processing
the private perspectives, because it yields high potential energy statistical data.
F. Ascione et al. / Applied Energy 174 (2016) 37–68 39

Recently, various studies and projects aimed at collecting data the building level. Recently, Karmellos et al. [32] proposed a bi-
on the existing building stocks in order to develop national RBs objective approach to identify the optimal prioritization of building
[10–13], especially for the residential sector. Definitely, the theo- energy efficiency measures, based on primary energy consumption
retical RB yields the highest reliability in representing a building and investment cost criteria. Conversely, Chantrelle et al. [25] con-
category because it is generated ‘ad hoc’ to gather the category sidered three objective functions and developed a multi-criteria
average characteristics. As aforementioned, once defined a reliable tool for optimizing building retrofit, with the aim of minimizing
RB, this can be subjected to the cost-optimal analysis to address energy consumption, retrofit costs and CO2-eq emissions. In the
the energy retrofit of the represented category. However, the same vein, in various worthy studies, Echenagucia et al. [26], Dia-
robust assessment of cost-optimality is very time costly even when kaki et al. [27] and Delgarm et al. [33] applied a three-objective
only one building is investigated. In this regard, proper BPO (build- approach in order to design energy-efficient buildings, by minimiz-
ing performance optimization) algorithms can be used to reduce ing the energy demands for heating, cooling and lighting.
the required computational time by cutting the domain of explored Finally, as aforementioned, a very interesting application of BPO
scenarios, because these are properly selected by the optimization algorithms is supporting the robust cost-optimal analysis of build-
logic. Several approaches for BPO are available [14,15], among ing energy retrofit, since it implies significant computational time
which derivative-free methods [16] are the most popular because savings. In this regard, only few studies [17,21] proposed thorough
a differentiable objective function generally does not exist in BPO methodologies to couple the cost-optimal analysis with BPO algo-
problems. These methods perform simulation-based optimization rithms, but concerning small or medium, simple residential build-
routines by coupling numerical optimization algorithms with BPS ings. However, this coupling is particularly worthwhile when big
tools [14,15]. Thus, they consist of iterative procedures that deter- and complex buildings are examined because, in this instance, reli-
mine the progressive improvement of the solution until a conver- able energy simulations require high running times and a huge
gence criterion is satisfied. As highlighted in the admirable review domain of energy retrofit scenarios must be explored [38]. Thus,
of Nguyen et al. [15], stochastic population-based algorithms are an exhaustive investigation of all scenarios is prohibitive thereby
the dominant numerical methods in BPO, because they provide a hindering the robust assessment of cost-optimality if standard
good suboptimal solution in a reasonable computational time approaches are used. The implementation of BPO algorithms tack-
(the ‘true optimum’ is generally unknown [17]). They employ arti- les this issue by properly reducing the explored domain without
ficial intelligence techniques, among which genetic algorithms prejudicing the detection of the actual cost-optimal solution. Def-
(GAs), cultural algorithms, evolutionary programming, differential initely, this can allow a rigorous cost-optimal retrofit design also
evolution, particle swarm optimization, ant or bee colony algo- for the mentioned big and complex buildings. Among these, the
rithms, intelligent water drop and hybrid methods [15]. Among proposed study focuses on hospital buildings because they are dif-
these, the most frequently used methods for BPO are the GAs, as fuse, wide and energy-intensive structures, which present the
shown by a number of previous studies [17–26]. These allow to highest energy consumption per unit of floor area in the building
handle a high number of design variables and are particularly suit- sector [39]. Therefore, their proper energy retrofit can ensure sig-
able when a huge solution domain is explored, such as in BPO nificant reductions of energy consumption and polluting emissions
problems. Furthermore, they permit to perform multi-objective at large scale [40]. However, the assessment of energy performance
optimization, which is more proper for BPO because different con- [41–43] and retrofit potentials [44,45] for hospital buildings is very
flicting objectives subsist [27]. In this case, the optimization out- arduous because of the substantial size, energy-intensity and
come is not a single solution but a set of non-dominated heterogeneity. Hospitals host very different functions, being a kind
solutions, which form the so-called Pareto front [28]. The designer of small cities, so that some thermal zones (e.g., surgery blocks,
makes the last decision according to the specific weights that he intensive care units) need a rigid control of indoor conditions,
gives to all in-field objectives [29,30]. Normally, the objective func- while many others (e.g., offices, generic wards) can be compared
tions of multi-objective BPO concern the minimization of energy to typical destinations of the tertiary sector. Furthermore, both
consumption, economic expenditure, indoor discomfort or space heating and cooling demands are generally very significant,
environmental impact. Among these, two [17–25,31,32] or three due to the need of high ventilation rates [40], sometimes also 15
[26–28,33] goals are generally investigated, because a higher num- air changes per hour, and to the strict requirements for microcli-
ber would imply an excessive computational complexity and diffi- matic control [46], which is fundamental for patients’ health
culty to interpret the outcomes. In this regard, Wright et al. [18] [47,48]. Also the demand of district hot water (DHW) is substan-
applied a bi-objective GA to find the optimal trade-off between tial, as underlined by Bujack [49], and the same occurs for electric-
energy cost and occupant thermal comfort in building thermal ity demand because of the wide presence of electrical equipment
design. In the same vein, Magnier and Haghighat [19] combined for care purposes, as well as of the high need of artificial lighting
a bi-objective GA with TRNSYSÒ [34]-based artificial neural net- [50]. Finally, the energy behavior of hospitals is extremely com-
works to optimize building design by minimizing the energy plex, and thus the effective design of energy retrofit requires a
demand and the absolute value of predicted mean vote (PMV). through and careful approach, which must consider all levers
Hamdy et al. [20] developed an original bi-objective approach that affecting energy performance, namely the building thermal envel-
couples a GA with the BPS tool IDA-ICE [35] in order to design low- ope, the energy systems and the exploitation of renewable energy
emission and cost-effective dwellings by minimizing investment sources (RESs). Nevertheless, current scientific literature in matter
costs and equivalent carbon dioxide (CO2-eq) emissions. Likewise, of energy retrofit of hospitals does not provide worthy studies that
Fesanghary et al. [31] proposed a novel method to design low- investigated all mentioned levers by exploring the huge domain of
emission and energy-efficient buildings by minimizing lifecycle possible retrofit scenarios. Indeed, some studies focused on the
cost and CO2-eq emissions. Ascione et al. [22–24] developed a bi- energy rehabilitation of the building envelope [46], others on the
objective framework based on the coupling between EnergyPlus efficiency of HVAC systems [40,44,51,52], others on the optimiza-
[36], employed as BPS tool, and MATLAB [37], employed to run a tion of primary energy systems especially by means of the imple-
GA. This framework was used by the authors in different studies mentation of combined heating power (CHP) and combined
in order to optimize the envelope thermal design [22], the control cooling heating power (CCHP) systems [39,53–57]. This knowledge
of HVAC (heating, ventilating and air conditioning) systems [23] gap needs to be filled for a diffuse effective energy retrofit of the
and the mix of renewable energy source (RES) systems [24] at hospital building stock.
40 F. Ascione et al. / Applied Energy 174 (2016) 37–68

1.1. Objective and originality of the investigation investigated in this study, because it allows a detailed and flexible
definition of building geometry and thermal zones by employing
As previously shown, the literature review about the optimiza- graphical interface programs, such as DesignBuilder [58] (i.e., the
tion of building energy retrofit shows that the robust assessment of one here adopted). Because of its high reliability, EnergyPlus
cost-optimal solutions is arduous because requires huge computa- is the most used dynamic BPS tool for BPO [15]. MATLABÒ is
tional times (order of magnitude of days and even weeks), espe- employed as processing and optimization engine, because of its
cially for big and complex buildings, e.g., hospitals. Nevertheless, flexibility and high computational capability that allows to
these buildings are generally highly energy-intensive, and there- perform Latin hypercube sampling (LHS), sensitivity analysis (SA)
fore their energy retrofit is crucial to achieve substantial energy and optimization algorithms. In this regard, it was used in a
savings at large-scale. In this regard, the current body of knowl- number of previous studies concerning the analysis and optimiza-
edge does not show: tion of building energy performance, e.g., [4,17,20–24,59,60]. Fur-
thermore, through the developed coupling function, MATLABÒ
I. Thorough methodologies that provide a robust assessment can launch EnergyPlus simulations as well as handle text-based
of cost-optimal solutions for big and complex buildings by EnergyPlus inputs and outputs. Thus, the coupling of these two
exploring the huge domain of possible retrofit scenarios in programs allows to run automatically a huge set of dynamic energy
a reasonable computational time. simulations that are managed by the optimization or processing
II. Worthy studies that optimize the retrofit design of hospital algorithms written in MATLABÒ environment [4,17,22–24,59,60].
buildings by carefully investigating all levers affecting Definitely, this coupling is not a novelty, but it represents a power-
energy performance, namely the characteristics of building ful tool that is here employed in a totally new framework.
thermal envelope and energy systems, as well as the The methodology is suitable for any building, but its application
exploitation of RESs. to reference buildings (RBs) is particularly worthwhile because, in
III. Models of RBs concerning the Italian hospital building stock. this case, the achieved cost-optimal solutions are valid for most
members of the represented building category, as argued in Sec-
The objective of this study is addressing these knowledge gaps tion 1. Therefore, a systematic implementation of the methodology
by proposing a novel methodology that is, then, employed to iden- requires the preliminary definition of a RB that represents an
tify the cost-optimal energy retrofit of a developed Italian hospital established category (as done in Section 3). Then, the procedure
RB, by considering all levers affecting energy performance. for assessing cost-optimal energy retrofit solutions is performed
The methodology provides a robust cost-optimal analysis and by following the three main stages reported below and outlined
requires a reasonable computational time even for big and com- in Fig. 1:
plex buildings. It performs simulation-based multi-stage and
multi-objective (Pareto) optimization, combined with Latin hyper- (1) Preliminary analysis of energy retrofit measures (ERMs) for
cube sampling (LHS) and sensitivity analysis (SA). In particular, it the reduction of thermal energy demand for space condi-
employs the coupling between EnergyPlus [36] and MATLABÒ tioning (TED).
[37] and presents two optimization stages, preceded by a prelimi- (2) First optimization stage: Bi-objective optimization of ERMs
nary analysis of building energy performance that adopts LHS and for the reduction of TED.
SA. The first optimization stage runs a multi-objective GA, while (3) Second optimization stage: Tri-objective optimization of the
the second one is based on a smart exhaustive sampling. Invest- whole energy retrofit and assessment of cost-optimality.
ment cost, primary energy requests and global costs are investi-
gated for a huge number of energy retrofit scenarios and, finally, This multi-stage framework allows to largely reduce the com-
the cost-optimal solution is identified both in presence of a limit- putational times compared to holistic approaches (see Sections
less economic availability and of limited budgets. Therefore, the 2.3), by ensuring, at the same time, the rigorous exploration of a
methodology presents a novel framework that combines different huge domain of energy retrofit scenarios. The following subsec-
tools, i.e., LHS, SA, GA, smart exhaustive sampling, which are, in tions detail the three methodology stages.
general, used separately, in order to fill the first mentioned knowl-
edge gap (I). 2.1. Preliminary analysis
Furthermore, in order to fill the other two knowledge gaps (II
and III), the methodology is applied to a RB that represents hospital The baseline energy performance of the RB is assessed in terms
buildings built in South Italy between 1991 and 2005. The RB is of thermal energy demand for space heating, space cooling, and
defined by using an original approach, as required by the complex- district hot water (DHW), respectively, and of electricity demand
ity of the examined building category. for fans, pumps, lighting and equipment (direct electric uses).
Then, the impact of proper ERMs addressed to the reduction of
thermal energy demand for space conditioning (TED) is investi-
2. Methodology: cost-optimal analysis of building energy gated by means of Latin hypercube sampling (LHS) and sensitivity
retrofit by multi-stage and multi-objective optimization analysis (SA).
The RB is modeled in EnergyPlus by exploiting the graphical
The methodology presents a multi-stage framework that allows interface DesignBuilder to define the building geometry and subdi-
to minimize the computational burden required to identify reliable vision in thermal zones. Most notably, EnergyPlus model requires
cost-optimal energy retrofit solutions for any building. Most nota- to set:
bly, it tackles the crucial and outstanding issue concerning the
robust assessment of cost-optimality for big and complex buildings  the thermo-physical characteristics of the building envelope;
with a reasonable computational time. It combines EnergyPlus and  the profiles of building use for each thermal zone in terms of
MATLABÒ, which communicate via a coupling function developed hourly schedules of: occupancy, people activity, ventilation
in MATLABÒ environment. EnergyPlus is used as BPS tool – in order need, lighting power density, electrical equipment power
to run dynamic energy simulations – because of its high accuracy density, DHW demand;
in predicting building energy performance. Furthermore, it is  the operation of HVAC systems by assigning hourly schedules of
particularly proper for complex multi-zone buildings, as those set-point temperatures;
F. Ascione et al. / Applied Energy 174 (2016) 37–68 41

Fig. 1. Framework of the proposed methodology for the cost-optimal analysis of building energy retrofit by multi-stage and multi-objective optimization.

 the type of HVAC systems, in terms of characteristics of the current energy performance, building peculiarities and best prac-
heating and cooling terminals as well as of the distribution tice. A variable is associated to each ERM. It can be continuous or
network. discrete, depending on the number (limitless or limited) of investi-
gated options for that ERM. For instance, if the thermal insulation
It should be noted that the primary heating/cooling energy sys- is considered as ERM, the associated variable can be the insulation
tems are not modeled because this stage aims to calculate thermal layer thickness and it can be assumed as continuous. Otherwise, if
energy demand and not primary energy consumption, which is the installation of a heat recovery system is considered as ERM, the
assessed later by means of MATLABÒ post-process (see Section 2.3). associated variable can be discrete, so that it assumes: (a) the value
After modeling the RB, an EnergyPlus simulation is run by using a 0 if the measure is absent, (b) 1 if this is present. Finally, n mixed-
proper weather data file. The hourly values of thermal energy integer variables are introduced and a continuous or discrete range
demand for space heating, space cooling, DHW and of electricity of variability is assigned to each of them, thereby defining the sam-
demand for direct electric uses are achieved as simulation outputs. ple space to be explored in order to investigate all ERMs’ combina-
Hence, the annual values of TED for space heating (TEDheat [W h/ tions. Hence, LHS is employed within a Monte Carlo framework in
m2 a]) and for space cooling (TEDcool [W h/m2 a]) per unit of condi- order to generate a representative sample S of N cases that repre-
tioned area are calculated. The sum of TEDheat and TEDcool provides sent different ERMs’ combinations. For each case, an EnergyPlus
the total thermal energy demand for space conditioning, denoted building model has been built by means of the developed
as TEDtot [W h/m2 a]. MATLABÒ coupling function. The sampling framework is similar
After the energy assessment of the RB baseline, a set of n ERMs to the one used by the authors in [4] for investigating a building
for the reduction of TEDheat and/or TEDcool are identified based on stock, and implies high uniformity and coverage in the sample
42 F. Ascione et al. / Applied Energy 174 (2016) 37–68

space thanks to LHS stratification properties [61–64]. In order to analysis, but the outcomes of the SA are exploited to properly
ensure that S represents all considered retrofit scenarios signifi- screen the ERMs. In this regard, the best ERMs are applied to all
cantly, the sample size (N) must be carefully set. Definitely, the solutions, thus these do not introduce design variables because
minimum value of N that produces representative outcomes they are always present. Conversely, the worst ERMs are com-
depends on the number of sampled variables (n). Several studies pletely excluded. Therefore, the preliminary SA allows to reduce
concerning UA and SA of building energy performance [4,63–66] the number of design variables, thereby reducing the computa-
employed a ratio (r) between N and n included in the range 2–5. tional time required to run the GA, since the number of generations
However, Hopfe [63] highlighted that 60–80 cases are generally needed to achieve a reliable Pareto front decreases with decreasing
sufficient – independently of n – because the increment of out- variables’ number [15]. In this optimization process, the screened
comes’ reliability is very slight for higher N values. Once the sam- design variables are handled as discrete, even if in the preliminary
ple S is built, MATLABÒ automatically interrogates EnergyPlus in analysis some of them have been considered continuous. This
order to evaluate the energy performance of the N building models, allows to reduce the explored domain thereby speeding up the
i.e., retrofit scenarios, included in S. For each case, TEDheat, TEDcool GA convergence, and makes the approach more realistic because
and TEDtot are calculated, and thus the impact of the proposed limited options for the ERMs are generally available. This choice
ERMs on the reduction of TED is assessed. Histograms are is largely shared in BPO studies [17,18,20–27]. Thus, a string of ni
employed to depict the distributions of TEDheat, TEDcool and TEDtot, bits is used to encode the i-th design variable, so that this latter
thereby highlighting the potential TED savings compared to the RB can take 2ni different discrete values. These bit strings are included
baseline. in the vector x0 , whose values need to be optimized in order to min-
Finally, the SA is carried out by employing a global approach – imize F0 (x0 ). Finally, the bi-objective optimization problem has the
which is more reliable than local ones for building applications following mathematical formulation, where d0 denotes the number
[14,67] – based on the assessment of the standardized rank regres- of design variables:
sion coefficient (SRRC) [68]. This sensitivity index is used because
min F0 ðx0 Þ ¼ ½TEDheat ðx0 Þ; TEDcool ðx0 Þ
it is a measure of how much a variable input affects an output,
and can vary between 1 and 1: a positive value means that input subject to
" #
and output change in the same direction, a negative values means
0
the opposite. The use of SRRCs is widely shared in the BPS commu- x ¼ x01 ; . . . ; x0n1 ; . . . ; x0 Pd0 ; . . . ; x0Pd0 with
ð ni Þnd0 þ1 ni
i¼1 i¼1
nity [4,63–66,69–71] because this index can handle non-linear cor-  Xd0
relations between inputs and outputs as those provided by BPS 0
x0j ¼ for j ¼ 1; . . . ; n
i¼1 i
tools. In this study, the SRRCs are assessed for the n variables 1
(inputs) associated to the implementation of the ERMs for TED
The problem is solved by employing a control elitist GA, namely a
reduction in relation to TEDheat, TEDcool and TEDtot (outputs) in
variant of NSGA II [72], which is run by means of the automatic cou-
order estimate the influence of each ERM on the seasonal and
pling between EnergyPlus (BPS tool) and MATLABÒ (optimization
annual values of TED. A positive value of SRRC indicates that the
engine), as outlined in Fig. 2, which shows the algorithm flowchart.
ERM yields an increase of the considered output, thereby exerting
In addition, the GA pseudocode, proposed by the authors in [50], is
a negative effect, whereas the opposite occurs for a negative value
reported below:
of SRRC. This allows to identify both the best and the worst ERMs.
The best ERMs are the most energy-efficient ones that determine a
robust reduction of TEDtot. Conversely, the worst ERMs are the
energy-inefficient ones that penalize TEDtot. The detection of the
best ERMs is a critical issue because these measures should have
a definite positive effect on TEDtot, also in presence of other ERMs,
by not needing optimization processes to identify their most
energy-efficient option. Thus, this issue requires a thorough analy-
sis of SA outcomes and ERMs’ peculiarities, as carried out for the
explored case study (see Section 4.1). On the other hand, the pres-
ence of energy-inefficient ERMs could sound strange, but it can
occur because ERMs designed to reduce energy demand in the
heating season can have a negative effect during the cooling sea-
son, and vice versa, thereby causing a possible increment of TEDtot.
The detection of the best and worst ERMs is exploited in the first
optimization stage in order to properly set the design variables,
thereby minimizing the computational time required by the opti-
mization routine (see Section 2.2). Tornado plots are employed to
provide an immediate representation of SRRCs.

2.2. First optimization stage

In this phase, non-dominated (i.e., optimal) combinations of


ERMs for TED reduction are identified by running a simulation-
based bi-objective genetic algorithm (GA) that performs Pareto
optimization [41]. The two diverging objective functions to be min-
imized are TEDheat and TEDcool, which are included in the vector
F0 = [TEDheat, TEDcool], while the design variables are those associ- Fig. 2. Flowchart of the bi-objective GA that addresses the optimization of ERMs (x)
ated to the implementation of the ERMs for TED reduction. This for the minimization of thermal energy demand for space heating (TEDheat) and
stage does not consider all ERMs investigated in the preliminary space cooling (TEDcool).
F. Ascione et al. / Applied Energy 174 (2016) 37–68 43

Note that this optimization pseudocode was proposed by Ascione that minimizes TEDtot is achieved, and thus the strong interaction
et al. in [17] between primary energy systems and thermal energy demand is
not considered properly.
t = 1 (generations’ index) The non-dominated combinations of ERMs detected by the GA are
Create the initial population P(1)  {x0 i(1)}i=1,. . .,s of s individuals compared to the non-optimized combinations of the best ERMs
detected by the SA (preliminary analysis), in terms of TEDheat and
Calculate F0 (x0 (1)
i ) for i = 1, . . . , s
TEDcool. Thus, the performances of all these solutions are represented
Evaluate the rank value and the average crowding distance for
in the plane TEDheat–TEDcool in order to appreciate the impact of the
each individual of P(1)
GA in detecting optimal combinations of ERMs for TED reduction. In
DO UNTIL at least one stop criterion is satisfied
addition, for each solution, the hourly values of thermal energy
t=t+1
demand for space heating, space cooling, DHW and of electricity
Select the parents from P(t1)
demand for direct electric uses, provided by EnergyPlus, are recorded.
Generate P(t)  {x0 i(t)}i=1,. . .,s from crossover and mutation of
the parents: elite parents survive 2.3. Second optimization stage
Calculate F0 (x0 i(t)) for i = 1, . . . , s
Evaluate the rank value and the average crowding distance In this stage, the whole energy retrofit is optimized by consid-
for each individual of P(t) ering of all types of ERMs, i.e., for the reduction of TED, for the
END improvement of primary systems’ energy efficiency, for the
Return the Pareto front exploitation of renewable energy sources (RESs). Most notably, a
Pareto tri-objective optimization is performed and the cost-
optimal retrofit package is identified in presence of different eco-
The GA provides the ‘Darwininan’ evolution of a population of s indi- nomic budgets.
viduals through a series of generations. The individuals represent The three objective functions that are minimized in this stage are
packages of ERMs, encoded by the vector x0 . From generation to gen- the investment cost (IC [€]), the differences in primary energy con-
eration, the population undergoes an iterative improvement through sumption (dPEC [W h/m2 a]) and global cost (dGC [€]) compared to
the processes of crossover and mutation of the best individuals (par- the baseline. These form the vector F00 = [IC, dPEC, dGC]. IC is the
ents), detected based on the values of F0 (ranking) and on the average sum of the initial investments costs of the ERMs included in a generic
crowding distance; ce (elite count) of these best individuals survive to whole retrofit package. The primary energy consumption (PEC [W h/
each generation. The individuals that derive from crossover (chil- m2 a]) refers to energy uses for space conditioning, DHW, fans,
dren) randomly inherit the design variables (i.e., bit strings) from pumps, lighting and equipment and is calculated per unit of condi-
the combination of two parents; fc (crossover fraction) denotes the tioned area as recommended by the EPBD Recast [5]. The global cost
population fraction that originates from crossover. The remaining (GC [€]) is the lifecycle cost that takes into account investment and
individuals (mutated children) derive from the mutation of random replacement costs of ERMs, state financial incentives and operation
parents, by changing each bit with a mutation probability equal to costs associated to the mentioned energy uses. dPEC and dGC are
fm. Either the evolution stops when a maximum number of genera- preferred to the absolute values PEC and GC, as objective functions,
tions, equal to gmax, is reached or the change of Pareto front’s spread because they provide an immediate estimate of energy and eco-
between two successive generations is lower than the tolerance tol. nomic savings, and thus of retrofit effectiveness. The design vari-
In this study, most of the mentioned GA parameters take the same ables refer to the implementation of all types of ERMs. The ERMs
values employed in [17], namely: ce = 2, fc = 0.6, fm = 0.1, tol = 0.001, for TED reduction are those investigated in the previous stages. In
chosen based on authors’ expertise and previous authoritative stud- addition, suitable ERMs are identified to improve the energy effi-
ies [73,74]. On the other hand, s and gmax need to be properly set ciency of primary heating/cooling/electricity systems and to exploit
depending on the case study complexity, because they extremely RESs, based on building peculiarities and best practice. As for the pri-
affect computational times and reliability of results. Several previous mary systems, the explored ERMs concern the installation of new
studies [17,20–26] – concerning the use GAs for BPO – showed that devices for heating, cooling and DHW production, as well as of com-
reliable s values are included in the range 2–6 times the number of bined heating and power (CHP) and combined cooling, heating and
design variables, whereas reliable gmax values are included in the power (CCHP) systems. As for RESs, thermal solar collectors and pho-
range 10–100 generations. In this study, different values of gmax have tovoltaic (PV) panels are explored. Once defined all types of ERMs, a
been investigated, in order to identify the one that ensures the best discrete design variable is associated to the ii-th ERM and is encoded
trade-off between computational burden and reliability, as done in with a string of nii bits, as done in the first optimization stage. These
[23]. Conversely, s is fixed and set equal to 4, i.e., the mean value of strings of bits form the vector x00 , whose values need to be optimized
the mentioned range, because it exerts a lower influence on GA con-
in order to minimize F00 (x00 ). As aforementioned, the analysis is per-
vergence compared to gmax [23], and thus its variation in the range 2–
formed in presence of different economic budgets (B), and thus a
6 would not imply significant deviations.
constraint is introduced. Finally, the three-objective optimization
The outcome of the GA is a bi-dimensional Pareto front, which is
problem has the following mathematical formulation, where d00
the set of non-dominated solutions. These represent the optimal
denotes the number of design variables:
combinations of ERMs (for TED reduction) that minimize TEDheat
and TEDcool. Multi-objective optimization could appear pointless
min F00 ðx00 Þ ¼ ½ICðx00 Þ; dPECðx00 Þ; dGCðx00 Þ
because TEDheat and TEDcool can be summed in TEDtot, which can
be minimized by using a simpler mono-objective approach. This subject to
2 3
would be true if the ultimate goal was the minimization of TEDtot,
nd00 þ100
but this is not the case because the proposed methodology aims at x ¼ 4x001 ; . . . ; x00n1 ; . . . ; xP
00
d00
 ; . . . ; x00Pd00 5 with
nii n
the minimization of primary energy consumption and global cost. ii¼1 ii
ii¼1

In this regard, the impact and effectiveness of new energy primary  Xd00
0 0
systems, which are investigated in the second optimization stage, x0j ¼ for j ¼ 1; . . . ; n
ii¼1 ii
1
highly depend on the presence of ERMs for TED reduction. If a
mono-objective approach is employed, only the package of ERMs ICðx00 Þ 6 B
44 F. Ascione et al. / Applied Energy 174 (2016) 37–68

The problem is solved by means of a smart exhaustive sampling, period of 80 h per month is considered for maintenance, resulting
which is carried out by exploring the impact on energy consump- in an annual operation period of 7800 h. Similarly, the hourly val-
tion and global cost of all possible combinations of the proposed ues of thermal energy produced by solar collectors and electricity
ERMs addressed to primary systems and RESs’ exploitation, in produced by PV panels are calculated, depending on the specifica-
the following cases: tion of the systems, size and external climatic conditions. Then, the
thermal energy provided by CHP and CCHP systems is subtracted
 absence of ERMs for the TED reduction; from the thermal energy demand for heating, DHW and cooling,
 presence of non-optimized combinations of the best ERMs for based on hourly energy balances. The priority is given to space
TED reduction, detected by the SA in the preliminary analysis; heating and cooling, compared to DHW production, because these
 presence of the non-dominated (i.e., optimal) combinations of produce a more stable energy demand (during the day) thereby
ERMs for TED reduction, detected by the GA in the first opti- allowing a simpler and more effective plant operation. On the other
mization stage. hand, daily energy balances are used to evaluate the thermal
energy produced by solar collectors and self-used by the building,
The exhaustive sampling is defined ‘smart’ because the investi- by assuming a heat storage system with a daily capacity, typical of
gated ERMs for TED reduction have been selected properly, accord- solar thermal systems. Thermal energy produced by solar collec-
ing to the outcomes of previous stages. The result is a huge number tors is used only for DHW production, in order to avoid an exces-
of whole energy retrofit packages that allow to consider the inter- sive complexity of the plant. The thermal energy that is produced
actions among the different types of ERMs. The first multi- by CHP, CCHP or solar thermal systems and not consumed by the
objective optimization stage is fundamental to identify optimal facility goes wasted. After considering the mentioned subtractive
combinations of ERMs for TED reduction, one of which could imply terms, the hourly values of the resulting thermal energy demand
cost-optimality if combined with synergic primary systems and for space heating, cooling and DHW production are turned into
RESs. However, the cost-optimality depends on the ERMs’ cost- hourly demands of electricity and/or fuel (based on primary sys-
effectiveness, which is due not only to potential energy savings tems’ type) by exploiting the performance curves that characterize
but, obviously, also to the investment entity. Thus, the cost- the HVAC systems’ dynamic behavior. Hence, the hourly values of
optimal whole retrofit package could not include one of the non- the total electricity demand are achieved, and the electricity pro-
dominated combinations of ERMs for TED reduction. Therefore, vided by CHP, CCHP systems and PV panels is detracted based on
this stage explores also the absence of ERMs for TED reduction as hourly energy balances. The electricity converted and not con-
well as the presence of non-optimized combinations of the best sumed by the building is supplied to the grid and remunerated.
ERMs detected by the SA. Reasonably, this allows to cover the The final demands of electricity and fuels are converted into pri-
domain of energy retrofit scenarios that should include the cost- mary energy, by employing the primary energy factors established
optimal solution. by Italian current laws. The PEC is so assessed for each investigated
IC, dPEC and dGC are calculated in MATLABÒ environment, retrofit package, and the procedure is also used to calculate the PEC
without needing further EnergyPlus simulations. This represents for the baseline RB (PECRB). The objective dPEC is achieved by sub-
one of the main benefits of multi-stage approaches [4,21] com- tracting PECRB from PEC (dPEC = PEC  PECRB). Given the final
pared to standard simulation-based holistic ones [17,22], because demands of electricity and fuels, also the annual operating costs
it allows a substantial reduction of computational times. Indeed, are calculated by considering mean annual prices. This allows to
holistic approaches perform a unique optimization stage by con- assess GC for each whole retrofit package and GCRB too, by follow-
sidering all types of ERMs. Consequently, the number of the ing the guidelines of the EPBD Recast [5,6]. The objective dGC is
simulation-based optimization’s design variables increases com- achieved by subtracting GCRB from GC (dGC = GC  GCRB). The real
pared to the adopted multi-stage approach, thereby requiring a interest rate and the energy price escalation rate set equal to 3%
higher number of iterations, and thus energy simulations, to and 2%, respectively, according to cost-optimal regulations [6,75].
achieve the algorithm convergence. In addition, the explored The annual energy demand is assumed constant in the calculation
domain of retrofit scenarios cannot be excessive in order to avoid period, which is set equal to 20 years in this study, as established
a prohibitive computational burden. Conversely, the proposed for non-residential buildings [5,6] such as hospitals.
approach minimizes the running time required by the The described procedure provides the values of IC, dPEC, dGC
simulation-based optimization (first optimization stage) because for the explored whole energy retrofit packages. The outcomes
only ERMs for TED reduction are considered. The ERMs addressed are represented in a three-dimensional space, which shows a sec-
to primary energy systems and RESs’ exploitation are investigated ond three-dimensional Pareto front. Then, all points are projected
in this second optimization stage that employs MATLABÒ to calcu- onto the plane dPEC–dGC in order to achieve the cost-optimal
late IC, dPEC and dGC with a much lower computational time curve, thereby identifying the cost-optimal retrofit package in
(around 1 s for each retrofit scenario) compared to an EnergyPlus presence of a limitless economic availability. This cost-optimal
simulation (order of magnitude of minutes for complex buildings). solution occurs in correspondence of an investment equal to ICopt.
This allows to explore a huge domain of retrofit scenarios with a The same procedure is carried out in presence of different limited
reasonable computational burden. The assessment of IC is immedi- economic budgets (availabilities) lower than ICopt. Considering
ate, while the assessment of dPEC and dGC requires the hourly val- economic budgets higher than ICopt is worthless because the
ues of thermal energy demand and electricity demand for direct cost-optimal solution is, obviously, the same detected for a limit-
electric uses, provided by EnergyPlus and recorded in the previous less budget. Therefore, for each investigated economic budget,
stages. Indeed, the implementation of efficient primary energy the retrofit packages that require an investment cost higher than
and/or RES systems, investigated in this last stage, does not alter this budget are excluded. The remaining solutions are represented
such values. As done by the authors in [4,24], a code written in in the space IC–dPEC–dGC, thereby by obtaining the three-
MATLABÒ environment handles the hourly values of thermal dimensional Pareto front, and are then projected onto the plane
energy and electricity demand and calculate PEC and GC for each dPEC–dGC, in order to achieve the cost-optimal curve. Finally,
explored retrofit package. Firstly, the hourly values of thermal the cost-optimal retrofit package is identified in correspondence
and electrical energy provided by CHP or CCHP systems are calcu- of each budget. Hence, the decision maker (e.g., designer, owner,
lated, depending on the systems’ specifications and size by assum- constructor) can establish a maximum level of investment accord-
ing an operation at fixed electrical power. A typical plant shutdown ing to his needs and wills, and the methodology provides the
F. Ascione et al. / Applied Energy 174 (2016) 37–68 45

cost-optimal retrofit solution in presence of this constraint. Other-  geometry/shape: geometrical and type-shape related character-
wise, if there are no financial limits, the decision maker will choose istics of the building;
the unconstrained cost-optimal solution, that corresponds to a  function: use classifications of the different parts of the build-
limitless budget. ing, occupancy and use schedules, location description and
determination of the internal heat gains;
3. Definition of a hospital reference building  system: size and type of HVAC systems, which affect thermal
zoning in conjunction with function;
The described methodology is applied to properly design the  envelope: thermo-physics properties of the building envelope
energy retrofit of a reference building (RB) that represents hospital components.
buildings built in South Italy between 1991 and 2005. This time
horizon has been chosen because the Italian law (n.10/1991 [76]) By following this subdivision, the statistical analysis of plans
in force for the regulation of new buildings’ energy design was and elevations provided by ASLNAP1 has allowed to deduce the
the same. Therefore, the Italian hospitals built in this period have geometry and thermal zoning of the RB, as depicted in Fig. 3. In
similar energy characteristics. This section illustrates the original particular, in order to assess the parameters related to geometry
approach employed to characterize the RB. and shape, listed in Table 1, a weighted average has been carried
The aim is to define a building model that represents a hospital out by considering the total floor area of each hospital as weight.
block that includes all main typical use classifications: operating Concerning the thermal zoning, clearly it depends on the per-
theatres, diagnostics, intensive care units, laboratories, generic centage subdivision of the useful floor area for use classifications,
wards, ambulatories and offices. In this regard, a theoretical RB has which highly affect hospitals’ energy performance [46]. In this
been developed, based on statistical data and authors’ experience. regard, the statistical analysis shows the contemporary presence
The employed statistical data have been provided by the ‘Azienda of all main use classifications typical of hospitals:
Sanitaria Locale of Naples’ (ASLNAP1) and refer to the hospitals of
the district of Naples, which is the first district in South Italy con-  ‘high-tech’ zones (operating theatres, diagnostics, intensive care
cerning number of hospitals: it holds 55 of the 317, around 17.4%, units and laboratories), which account for 12.4% of useful floor
of the examined building category [77]. By considering that the con- area of the explored building sample;
structive practice related to hospital buildings is quite homogeneous  generic wards, which account for 23.8%;
in South Italy, the building sample related to Naples’ hospital stock  ambulatories, which account for 23.0%;
can be reasonably assumed as representative of the whole category.  offices, which account for 36.8%, thereby representing the
According to USA benchmark models [13], in order to define a RB, the predominant one;
essential data can be organized into four main sub-sets:  common areas, which account for 4.0%.

Fig. 3. Developed hospital reference building (RB): rendering (a); thermal zoning of the five floors above the ground (b–f).
46 F. Ascione et al. / Applied Energy 174 (2016) 37–68

Table 1 Finally, the considered thermal zone types are summarized in


Characterization of geometry and shape of the developed hospital reference building. Table 2, by characterizing microclimatic control, internal heat
Parameter Value gains and DHW (district hot water) demand. In this regard, the
Building shape Rectangular set point values for indoor temperature and relative humidity as
True north – building north angle 0° wells as the ventilation needs have been set in compliance with
Length (south–north) 90.0 m Italian regulations concerning hospitals’ air-conditioning. In this
Width (east–west) 42.1 m matter, there is not a clear and direct legislative/regulatory frame-
Floor to floor height 3.96 m
Surface to volume ratio (S/V) 0.20
work in Italy, but only an old Ministerial Circular, some general
Number of floors 6 (1 buried) UNI standards and guidelines. At today, the main references are:
Gross volume 90,828 m3 Ministerial Circular LL.PP. n.13011/22.11.1974 [78], UNI 10339
Total floor area 22,711 m2 [79], UNI 8199 [80], ISPESL Guidelines [81]. On the other hand,
Useful floor area 18,926 m2
internal heat gains, DHW demand as well as the hourly schedules
Conditioned floor area 18,176 m2
Roof area 3785 m2 of building use and HVAC systems’ operation have been taken from
Window opening area 1126 m2 the database of DesignBuilder [58] for each thermal zone type.
Window fraction (window to wall ratio) As depicted in Fig. 3, the hospital RB has been subdivided into
South 26.0% the described thermal zone types, by complying with the percent-
East 24.0% age subdivision (reported in Fig. 4) related to the whole building
North 15.0% sample. For the subdivision of the five floors above the ground, a
West 23.0%
typical hospital layout has been employed deriving from the inves-
Total 23.0%
tigation of the considered sample. For instance, generic wards are
located at the last floors, whereas the operating theatre occupies
This is a preliminary percentage characterization of the useful a significant part of the first floor. The buried floor, not reported
floor area. A detailed subdivision into thermal zones requires to in Fig. 3, hosts plant rooms, and therefore it does not account for
define also the type of HVAC system installed in the indoor envi- the useful floor area. It has been modeled has a unique thermal
ronments. Four options are considered, consistently with the zone, which obviously does not present microclimatic control
examined statistical data, denoted as follows: and is characterized by an ACH equal to 1 h1 and heat internal
gains equal to 50 W/m2 [58].
 All-air: all-air systems served by air handling units. These allow Concerning the primary energy systems, the most common con-
to control temperature and relative humidity in both heating figurations within the building sample have been assumed and the
and cooling operation by varying the inlet air supply tempera- performance indicators – i.e., efficiency of boilers (g) and energy
ture via post-heating coils. The indoor air quality (IAQ) is efficiency ratio (EER) of electrical chillers – have been assessed
ensured by setting a proper inlet air flow. by means of weighted averages, considering the conditioned floor
 Heating: two-pipe fan coil systems for heating operation, which area of each hospital as weight. Most notably:
allow to control only temperature. The control type is on/off and
the IAQ is handled by natural ventilation.  the primary heating system presents a natural gas boiler, with
 Heating & cooling: four-pipe fan coil systems for both heating peak thermal power (Pth) equal to 1200 kWth and nominal
and cooling operations, which allow to control only tempera- LCV (lower calorific value) g equal to 0.90;
ture. The control type is on/off and the IAQ is handled by natural  the primary cooling system presents two identical air-cooled
ventilation. chillers, each with Pth equal to 1000 kWth and EER equal to
 No HVAC system: the air conditioning system is absent, and 2.5 at rated conditions, that is water inlet/outlet tempera-
thus neither temperature or relative humidity are controlled. tures = 12/7 °C and external air temperature equal to 35 °C;
The IAQ is handled by natural ventilation.  the primary DHW system presents a natural gas boiler, with Pth
equal to 300 kWth and nominal LCV g equal to 0.88; the system
Considering the examined building sample, the final percentage size has been set to fulfill DHW demand;
subdivision of the useful floor area, based on use classification and  electricity is taken from the grid. The base electrical load (Pel)
type of HVAC system, is summarized in Fig. 4. Hi-tech zones are for direct electric uses, which is almost always demanded dur-
integrally characterized by the presence of all-air systems because ing the year, is 200 kWel, whereas the maximum one is
a rigid microclimatic control is required, whereas for other use 800 kWel;
classifications fan coils are the most popular conditioning system.  no RES (renewable energy source) systems are installed.
All common areas do not present HVAC systems.

Fig. 4. Percentage subdivision of the useful floor area for the examined hospital building sample (ASLNAP1), based on use classification and type of HVAC system.
Table 2
Characterization of thermal zone types of the hospital reference building.

Thermal zoning parameters


Thermal zone type Microclimate control Internal gains DHW demand
Set-point temperature Set-point relative humidity Air change per hour People density Artificial lighting Equipment
High-tech (all-air)
Operating theatre 24 °C in both heating and cooling 50% in both heating and cooling 15 h1 0.125 person/m2 50 W/m2 75 W/m2 1.3 l/m2day
Diagnostics 20 °C in heating 40% in heating 6 h1 0.100 person/m2 50 W/m2 150 W/m2 0.6 l/m2day
26 °C in cooling 60% in cooling
Intensive care unit 24 °C in both heating and cooling 50% in both heating and cooling 6 h1 0.125 person/m2 50 W/m2 75 W/m2 1.3 l/m2day

F. Ascione et al. / Applied Energy 174 (2016) 37–68


Laboratory 20 °C in heating 40% in heating 6 h1 0.050 person/m2 50 W/m2 75 W/m2 1.3 l/m2day
26 °C in cooling 60% in cooling
Generic wards
Generic ward (all-air) 20 °C in heating 40% in heating 3 h1 0.175 person/m2 7.5 W/m2 13 W/m2 1.3 l/m2day
26 °C in cooling 60% in cooling
Generic ward (heating) 20 °C in both heating and cooling – 0.5 h1 0.175 person/m2 7.5 W/m2 13 W/m2 1.3 l/m2day
Generic ward (heating & cooling) 20 °C in heating – 0.5 h1 0.175 person/m2 7.5 W/m2 13 W/m2 1.3 l/m2day
26 °C in cooling
Ambulatories
Ambulatory (all-air) 20 °C in heating 40% in heating 6 h1 0.195 person/m2 17.5 W/m2 27.3 W/m2 0.17 l/m2day
26 °C in cooling 60% in cooling
Ambulatory (heating) 20 °C in both heating and cooling – 0.5 h1 0.195 person/m2 17.5 W/m2 27.3 W/m2 0.17 l/m2day
Ambulatory (heating & cooling) 20 °C in heating – 0.5 h1 0.195 person/m2 17.5 W/m2 27.3 W/m2 0.17 l/m2day
26 °C in cooling
Offices
Office (all-air) 20 °C in heating 40% in heating 3 h1 0.195 person/m2 17.5 W/m2 27.3 W/m2 0.17 l/m2day
26 °C in cooling 60% in cooling
Office (heating) 20 °C in both heating and cooling – 0.5 h1 0.195 person/m2 17.5 W/m2 27.3 W/m2 0.17 l/m2day
Office (heating & cooling) 20 °C in heating – 0.5 h1 0.195 person/m2 5.0 W/m2 27.3 W/m2 0.17 l/m2day
26 °C in cooling
Common areas (no HVAC) – – 0.5 h1 0.213 person/m2 5.0 W/m2 5.0 W/m2 –

47
48 F. Ascione et al. / Applied Energy 174 (2016) 37–68

The sizes of primary heating and cooling systems have been set TEDheat;RB ¼ 51:2 kW h=m2 a; TEDcool;RB
by considering the peak thermal loads for space heating and cool- ¼ 58:2 kW h=m2 a; TEDtot;RB ¼ 109:4 kW h=m2 a ð1Þ
ing, respectively. These loads have been assessed once defined the
thermal characteristics of the building envelope. In this regard, the Thermal energy demand for cooling is higher compared to heat-
RB’s structural frame is in reinforced concrete with vertical walls in ing because of the significant entity of endogenous and solar heat
hollow bricks and horizontal envelope in mixed brick-reinforced gains for the considered building use and climatic location.
concrete. Widows are double-glazed with air filling and the frame Then, the implementation of the following nine (n) energy ret-
is in aluminum. These types of opaque and transparent envelopes rofit measures (ERMs) for the reduction of TEDheat and/or TEDcool is
are the most common within the considered building sample. The investigated:
thermal properties of the mentioned components have been set in
order to respect the Italian law n.10/1991 [76], which regulated (a) heat recovery systems;
building energy design in the considered period (1991–2005). This (b) solar shading systems;
law established that new buildings had to present an energy index, (c) energy-efficient windows;
denoted as FEN, lower than a threshold value, denoted as FENlim. In (d) variation of the roof’s solar absorptance (a);
particular, the FEN [J/m3 DD] is defined as the annual primary (e) variation of the roof’s thermal emissivity (e);
energy required for space heating per unit of conditioned volume (f) roof’s thermal insulation of thickness t;
and degree-days (DD) in order to ensure a constant indoor temper- (g) variation of the external walls’ solar absorptance (a);
ature of 20 °C with an adequate air change. The standard UNI (h) variation of the external walls’ thermal emissivity (e);
13790 [82] illustrates the procedure for FEN calculation. The FENlim (i) external walls’ thermal insulation of thickness t.
depends on sundry factors, such as climatic conditions, geometry, Hereinafter, these ERMs are delineated with the italic letters
distribution of use classifications, efficiency of primary energy sys- reported in the above list. They are detailed in Table 3, concerning
tems (that is why the systems’ efficiency has been defined previ- energy characteristics, investigated options, investment cost and
ously). It has been assessed for the investigated case study available public financial incentives. The investment costs are
according to [82] and results equal to 113 kJ/m3 DD. Thus, the ther- taken from direct quotations of suppliers for the examined build-
mal characteristics of the RB envelope have been set to achieve a ings. It is clear that most ERMs can exert opposite effects on heat-
FEN equal to FENlim, thereby implying the following values of ther- ing and cooling operations, by implying a potential decrease of
mal transmittance (U [W/m2 K]): TEDheat and, at the same time, a potential increase of TEDcool, or
vice versa. Therefore, a careful bi-objective analysis is required. A
 opaque building envelope: for the external vertical walls, variable is associated to each ERM, which is discrete for the ERMs
Uwalls = 0.72 W/m2 K; (a)–(c) because a limited number of options is explored, whereas is
for the roof, Uroof = 1.20 W/m2 K; for basement, Ubasement = 1.20 - continuous for the ERMs (d)–(i) because the options are limitless.
W/m2 K; Most notably, the ERMs (a) and (b) introduce Boolean variables
 for the windows, Uw = 0.72 W/m2 K. Furthermore, the solar heat that are equal to 0 if the measure is absent, 1 if the measure is pre-
gain coefficient (SHGC) has been set equal to 0.764, as provided sent. The ERM (c) introduces a discrete variable that is equal to 0 if
by the EnergyPlus database [36] for the considered window the measure is absent, 1 in presence of new solar control windows,
type. 2 in presence of new low-e windows, 3 in presence of new solar
4. Results and discussion control and low-e windows. The ERMs (d)–(i) introduce continuous
variables that correspond to roof’s a, e, t, and external walls’ a, e, t,
The proposed multi-stage approach for the assessment of cost- respectively, whose ranges of variability are reported in Table 3.
optimal energy retrofit is applied to the developed hospital refer- Latin hypercube sampling (LHS) is employed to sample the
ence building (RB), which represents hospital buildings built in mentioned nine variables, thereby generating the sampling set S
South Italy between 1991 and 2005. The IWEC weather data file that collects 80 (N) cases, corresponding to different combinations
[83], accredited at international level, of the Italian city of Naples of the ERMs. The resulting ratio (r) between the number N of sam-
is used in energy simulations. Naples’ weather conditions are quite pled cases and the number n of sampled variables is 8.9, which is
close to typical average conditions in South Italy. Therefore, the quite high compared to other similar studies [4,63–66] that
outcomes provided for Naples can be extended with an acceptable adopted a r value included in the range 2–5. However, this choice
approximation to a wide part of the considered category. ensures a strong accuracy and reliability, since Hopfe [63] con-
It is noted that a processor IntelÒ CoreTM i7 at 2.00 GHz speed cluded that, in most cases, 80 is the value of N that ensures reliable
was employed. results with a huge confidence. In other words, if a lower N value is
The presentation of the achieved results is organized in three set, reliable results are not guaranteed. Furthermore, the analysis
subsections, which follow the three stages of the methodology of the outcomes, achieved by employing the mentioned high r-
(see Section 2), in order to provide a clear and systematic value, allows to assess which is the minimum r-value (rmin) that
investigation. implies consistent results. This can be exploited in future applica-
tions of the methodology to other case studies (or in similar stud-
ies), thereby yielding the best compromise between computational
4.1. Preliminary analysis of energy retrofit measures for the reduction
burden and reliability.
of thermal energy demand for space conditioning
Therefore, 80 cases are generated by means of LHS. For each of
these, an EnergyPlus simulation (automatically launched by
The RB is modeled and simulated in EnergyPlus in order assess
MATLABÒ) is performed to assess TEDheat, TEDcool and TEDtot. Firstly,
the hourly values of thermal energy demand for space heating,
such outcomes are investigated in order to identify the value of rmin,
space cooling, DHW and of electricity demand for direct electric
by means of a procedure similar to the one adopted by the authors in
uses. The running time of an EnergyPlus simulation is around
[4] to detect the minimum size of the representative building sam-
180 s (3 min), by using the ‘conduction transfer function’ algorithm
ple (RBS) of a stock. In particular, the trends of mean values and
with 6 time-steps per hour. Hence, the values of thermal energy
standard deviations of TEDheat, TEDcool and TEDtot are examined in
demand for space heating (TEDheat), space cooling (TEDcool) and
function of the number of sampled cases (i.e., retrofit scenarios),
total space conditioning (TEDtot) are assessed:
F. Ascione et al. / Applied Energy 174 (2016) 37–68 49

Table 3
Characterization of the proposed ERMs for TED reduction.

Energy retrofit measures for the reduction of thermal energy demand


Energy retrofit Description Options (if Investment costa Incentivesb
measure more
configurations
are explored)
(a) Heat recovery Installation of heat recovery systems for the air handling units 80,000 € –
systems (AHUs), with a total air flow (Vair) equal to around 150,000 m3/h
and an efficiency (e) equal to 0.8
(b) Solar shading Installation of interior solar shading systems, i.e., diffusive 40 € per m2 of windows’ surface 40% of IC up to a limit of 20,000 €,
systems blinds, characterized by solar transmittance (Tsol) equal to 0.5 accorded in 5 years
and visible light transmittance (Tvis) equal to 0.5. Shading is
active if beam plus diffuse solar radiation incident on the
window exceeds 450 W/m2 [4]
(c) Energy- Installation of argon-filled double-glazed Solar control 240 € per m2 of windows’ surface 40% of IC up to a limit of 45,000 €,
efficient windows with thermal insulated PVC frames. Uw = 2.51 W/ accorded in 5 years
windows The values of overall thermal transmittance m2 K,
(Uw) and solar heat gain coefficient (SHGC) SHGC = 0.499,
are lower compared to the reference Tvis = 0.664
building’s windows. The visible light Low-e 250 € per m2 of windows’ surface
transmittance (Tvis) is always higher than Uw = 1.71 W/
0.65 in order to avoid a significant increment m2 K,
of artificial lighting demand, which is thus SHGC = 0.691,
neglected. Three options are investigated, Tvis = 0.744
characterized by different coatings. In Solar 280 € per m2 of windows’ surface
particular, solar control and low-emissive control ± Low-e
(low-e) coatings are explored Uw = 1.64 W/
m2 K,
SHGC = 0.433,
Tvis = 0.691
(d) Variation of Plastering of the exterior side of the roof with a 2 [0.10, 0.90] Since the implementations of these ERMs 40% of IC up to a limit of
the roof’s paints characterized by a new value of the are connected and the investment costs 250,000 €, accorded in 5 years
solar absorptance to solar radiation (a). Different depend on the selected options, the (state incentives for the
absorptance options are investigated cumulative investment cost is assessed improvement of the building
(a) ad hoc for each package of these ERMs envelope’s energy efficiency)
(e) Variation of Plastering of the exterior side of the roof with e 2 [0.40, 0.90] investigated by the cost-optimal analysis
the roof’s paints characterized by a new value of the (see Table 5)
thermal thermal infrared emissivity (e). Different
emissivity (e) options are investigated
(f) Roof’s Installation of a standard thermal insulant t2
thermal (polyurethane, kb = 0.028 W/m K, q = 25 kg/ [0 m, 0.10 m]
insulation of m3, c = 1340 J/kg K) on the exterior side of the
thickness t roof. Different options are considered for
insulation layer’s thickness (t)
(g) Variation of Plastering of the exterior side of the external a 2 [0.10, 0.90] Since the implementations of these ERMs
the external vertical walls with paints characterized by a are connected and the investment costs
walls’ solar new value of the absorptance to solar depend on the selected options, the
absorptance radiation (a). Different options are cumulative investment cost is assessed
(a) investigated ad hoc for each package of these ERMs
(h) Variation of Plastering of the exterior side of the external e 2 [0.40, 0.90] investigated by the cost-optimal analysis
the external vertical walls with paints characterized by a (see Table 5)
walls’ thermal new value of the thermal infrared emissivity
emissivity (e) (e). Different options are investigated
(i) External Installation of a standard thermal insulant t2
walls’ thermal (polyurethane, k = 0.028 W/m K, q = 25 kg/ [0 m, 0.10 m]
insulation of m3, c = 1340 J/kg K) on the external side of
thickness t the external vertical walls. Different options
are considered for insulation layer’s thickness
(t)
c
k [W/m K] represents thermal conductivity; q [kg/m3] represents density (volumic mass), c [J/kg K] represents specific heat.
a
Investment cost obtained from direct quotations of suppliers.
b
Financial grants provided by the Italian Government according to [84].

as represented in Fig. 5. The stabilization of the depicted functions impact of heat recovery systems and thermal insulation on TEDheat,
occurs after 35–40 cases, thereby showing that rmin is around 4, and thus also on TEDtot even if to a lesser extent. Definitely, the
and thus it is included in the mentioned range of values (2–5) used installation of heat recovery systems is the ERM most affecting
in similar studies. energy demand for space conditioning, to the extent that its pres-
The values assumed by TEDheat, TEDcool and TEDtot for the 80 ence/absence determines a discontinuity in the distributions of
sampled cases are depicted in the histograms of Fig. 6, where the TEDheat and TEDtot values. On the other hand, in some cases, the
dots refer to the values achieved for the baseline RB. Also the nor- implementation of ERMs induces an increase of TEDcool (see
mal distributions that approximate the three sets of values are rep- Fig. 6b), since thermal insulation can have a significant negative
resented. In all cases, TEDheat and TEDtot decrease compared to the impact on cooling demand because of the high solar and endoge-
baseline (see Fig. 6a and c). This is mainly due to the high positive nous heat gains. This is a reasonable outcome, consistent with
50 F. Ascione et al. / Applied Energy 174 (2016) 37–68

Fig. 5. Assessment of a representative sample of building retrofit configurations. Mean values and standard deviations, in function of the number of sampled cases, of: (a)
TEDheat; (b) TEDcool; (c) TEDtot = TEDheat + TEDcool.

Fig. 6. Impact of the proposed ERMs on: (a) TEDheat; (b) TEDcool; (c) TEDtot = TEDheat + TEDcool.

thermo-physical considerations, which occurs also for a similar tively. Definitely, sensitivity analysis (SA) and multi-objective opti-
case study, i.e., office buildings at the same climatic location [4]. mization are fundamental.
Finally, whereas heat recovery systems are clearly energy- Thus, SA is performed by assessing the standardized rank
efficient, the choice of the other ERMs is a crucial issue, because regression coefficients (SRRCs), for the nine variables (inputs) asso-
opposite effects occur in the heating and cooling seasons, respec- ciated to the mentioned ERMs in relation to TEDheat, TEDcool and
F. Ascione et al. / Applied Energy 174 (2016) 37–68 51

Fig. 7. Sensitivity analysis. Assessment of the standardized rank regression coefficients (SRRCs) of the proposed ERMs in relation to: (a) TEDheat; (b) TEDcool; (c)
TEDtot = TEDheat + TEDcool.

TEDtot (outputs), as reported in Fig. 7. A positive value of SRRC whereas the negative effect on TEDheat is much lower, so that they
means that the ERM implies an increase of the output thereby imply a clear reduction of TEDtot. Also for this ERM, one option is
exerting a negative effect, whereas the opposite occurs for a nega- examined, and thus optimization is not required. In particular, only
tive value. internal shading systems are considered in order to simplify main-
The achieved sensitivity indices confirm that most ERMs have tenance and cleaning processes, as required in hospital buildings.
opposite effects on heating and cooling demands, respectively, In addition, the values of solar transmittance (Tsol) and light trans-
except for heat recovery systems, which induce a strong reduction mittance (Tvis), which are interdependent, are set equal to 0.5, as in
of TEDheat without affecting TEDcool, and energy-efficient windows, [4], to ensure a sufficient incoming solar radiation and natural
which induce a decrease of both TEDheat and TEDcool. As antici- light. This is essential for hospital buildings and, at the same time,
pated, thermal insulation of both roof and external walls has a sig- allows to avoid a significant increase of electricity demand for arti-
nificant positive effect on heating demand (SRRC < 0), but a ficial lighting, which, therefore, is assumed constant in this study.
negative effect (SRRC > 0) on cooling demand. Thus, lower values of Tsol and/or Tvis are excluded, and higher val-
The SRRCs related to TEDtot are all negative, because the nine vari- ues are not considered because they would yield a worsening of
ables vary in the direction that implies a reduction of TEDtot, thereby solar shading benefits in the cooling season, which are predomi-
playing a positive effect. About it, concerning both roof and external nant. Finally, also this ERM is included among the best ones.
walls, the SRRCs are assessed for increasing values of thermal emis- Energy-efficient windows exert a positive effect on both TEDheat
sivity (e) and thermal insulation thickness (t), and for decreasing val- and TEDcool, and obviously also on TEDtot. However, this double
ues of solar absorptance (a). This aids to identify the worst (i.e., benefit is due to different causes. The decrease of TEDheat, which
inefficient) ERMs, which are the measures that move the mentioned is slighter, depends on the reduction of windows’ thermal trans-
variables in the opposite direction compared to the one that implies mittance (Uw), which could be negative for the cooling operation
a TEDtot reduction. On the other hand, the best ERMs are the ones by virtue of the high internal hat gains. Conversely, the decrease
that move the variables in the right direction and provide the highest of TEDcool, which is more significant, depends on the reduction of
absolute values of SRRC. However, some of these ERMs present dif- solar heat gain coefficient (SHGC) of windows, which is negative
ferent options that need to be optimized in the following stage, and/ for the heating operation. This is why different windows’ types
or exert a strong negative effect on heating or cooling demand, are investigated, characterized by different values of Uw and SHGC,
which can be amplified in presence of other measures, thereby not and, globally, the installation of these new windows yields a
ensuring a robust and definite positive effect on TEDtot. Therefore, decrease of TEDtot. Nevertheless, definitely the detection of the best
the detection of the best ERMs, whose implementation is fixed in type needs multi-objective optimization. Furthermore, the impact
the first optimization stage, as well as of the energy-inefficient of energy-efficient windows on TEDtot is comparable, in terms of
ERMs, whose implementation is excluded from optimization, need absolute value of SRRC, with other ERMs that are extremely
a detailed analysis of each measure based on the outcomes of cheaper, such as solar shading systems (see Fig. 7 and Table 3).
Fig. 7, as reported in the following lines. Thus, most likely, energy-efficient windows are not cost-
Heat recovery systems induce the highest reduction of TEDheat effective, as shown in [4] for the same climatic location and similar
and TEDtot without affecting TEDcool, and they do not need opti- building use. For these reasons, this ERM, albeit energy-efficient, is
mization because only an option, deriving from the best practice, not included among the best ones and needs a to be optimized in
is considered. Therefore, these are surely included among the best the following stage.
ERMs. Furthermore, since this ERM is highly energy-efficient and The variation of roof’s solar absorptance (a) and thermal emis-
not too expensive, most likely, it will be part of the cost-optimal sivity (e) are considered together, because these are associated to
energy retrofit package (as confirmed by the outcomes of Sec- the same action, that is the installation of a new roof’s external
tion 4.3). This makes TEDcool much higher than TEDcool, and thus plastering. Clearly, the variation of a and e have opposite effects
energy retrofit measures for TED reduction, that favor the cooling on heating and cooling demands, respectively, but the impact on
operation, are more effective. cooling demand is more significant, because the roof is more sen-
Solar shading systems have a significant positive effect on sible to summertime solar radiation, which is more perpendicular.
TEDcool, which is predominant on TEDheat as aforementioned, Therefore, the minimization of absorptance and the maximization
52 F. Ascione et al. / Applied Energy 174 (2016) 37–68

Table 4
Characterization of the design variables of the genetic algorithm (first optimization step).

Decision variables Range of variability Number of possible Number of bits for


discrete values variable encoding
Windows’ type (1) air-filled double glazed with aluminum frames (refer- 4 2
ence system)
(2) argon-filled double glazed with solar control coating and
thermal insulated PVC frame
(3) argon-filled double glazed with low-emissive coating
and thermal insulated PVC frame
(4) argon- filled double glazed, with solar control and low-e
coating and thermal insulated PVC frame
Thickness (t) of roof’s thermal insulation (external From 0.00 m to 0.15 m with a step of 0.01 m 16 4
layer of polyurethane)
External walls’ solar absorptance (a) From 0.1 to 0.9 with a step of 0.05 (the values 0.45 and 0.55 are 15 4
excluded)
External walls’ thermal emissivity (e) From 0.4 to 0.9 with a step of 0.05 11 4
Thickness (t) of external wall’s thermal insulation From 0.00 m to 0.15 m with a step of 0.01 m 16 4
(external layer of polyurethane)

Please note that all the solutions investigated by the genetic algorithm are characterized by the implementation of the best ERMs for TED reduction, which consist of: heat
recovery systems, solar shading systems, cool roof.

Fig. 8. Bi-objective optimization of the proposed ERMs by minimizing TEDheat and TEDcool: Pareto fronts in function of the number of generations (gmax).

of emissivity – inside the ranges reported in Table 3 – imply a sub- be classified ‘a priori’ among the best or inefficient ones, and need a
stantial and robust positive effect on TEDtot. Definitely, the instal- deeper analysis by means of multi-objective optimization. Finally,
lation of a roof ‘cool’ plastering with a = 0.1 and e = 0.9, the best ERMs for TED reduction, whose implementation is fixed
hereinafter denoted as cool roof, can be included among the best during the first optimization stage, are:
ERMs. Conversely, the installation of a dark roof coating is ineffi-
cient, and thus it is excluded in the next methodology stages.  installation of heat recovery systems (ERM (a));
Similarly, the variation of external walls’ a and e are considered  installation of solar shading systems (ERM (b));
together. Also in this case, the SRRCs show that TEDtot decreases  installation of a roof cool plastering, with a = 0.1 and e = 0.9, i.e.,
with decreasing values of a and increasing values of e. However, cool roof (ERM (d) + ERM (e)).
the opposite effects on heating and cooling demand, respectively,
are more balanced, because external walls are more sensible to One inefficient ERM has been identified, that is the installation
wintertime solar radiation, which is less perpendicular. Therefore, of a roof dark plastering, which obviously is not considered in the
the positive impact of a cool coating on TEDtot is less robust and optimization stage because the cool roof is applied to all solutions.
definite compared to the roof case, and could be reversed in pres-
ence of other ERMs. Finally, these measures cannot be classified ‘a 4.2. Bi-objective optimization of energy retrofit measures for the
priori’ as best or inefficient ones, and need a deeper analysis by reduction of thermal energy demand for space conditioning
means of multi-objective optimization.
As aforementioned, the thermal insulation of roof and external The genetic algorithm (GA) is run in order to achieve the first
walls exert a strong positive effect on TEDheat, and thus also on TEDtot (bi-dimensional) Pareto front, which collects the non-dominated
even if to a lesser extent. Nevertheless, it implies an increase of (optimal) combinations of ERMs for TED reduction, thereby mini-
TEDcool, which is predominant on TEDheat, especially in presence of mizing TEDheat and TEDcool. The best ERMs identified by the prelim-
heat recovery systems. For this reason, the positive impact of ther- inary analysis are applied to all solutions, while the configurations
mal insulation on TEDtot is not robust and definite, and the optimal of the other explored ERMs for TED reduction are optimized by
thickness (t) of insulation layers must be identified, by considering considering five design variables, handled as discrete and charac-
the synergy with other ERMs. Therefore, also these measures cannot terized in Table 4. This also reports the number of explored options
F. Ascione et al. / Applied Energy 174 (2016) 37–68 53

Table 5
Packages of ERMs for TED reduction, selected for the cost-optimal analysis (second optimization stage): non-dominated Pareto solutions (first optimization stage); non-optimized
combinations of the best ERMs; absence of ERMs (RB baseline). Please see Table 3 for a detailed description of the ERMs.
Selected packages of energy retrofit measures for the reduction of thermal energy demand

Package Characterization of the ERMs Investment cost (€) Objective functionsa


(P.)
Single ERM Package TEDheat TEDcool
(kW h/m2 a) (kW h/m2 a)

Non-dominated PARETO solutions


Points with the lowest TEDheat
P. 1 Heat recovery systems: ERM (a) Vair = 150,000 m3/h; e = 0.8 80,000 1,545,900 14.3 57.7
Solar shading systems: ERM (b) Interior shades: Tsol = 0.5; Tvis = 0.5 45,000
Energy-efficient windows: ERM (c) Low-emissive: Uw = 1.71 W/m2 K; 281,500
SHGC = 0.691
Roof’s radiative characteristics: ERMs a = 0.10; e = 0.90 (cool roof) 510,800 (135 € per m2 of
(d) + (e) roof)
Roof’s thermal insulation: ERM (f) Polyurethane: t = 0.13 m; new Uroof = 0.18 W/
m2 K
Walls’ radiative characteristics: ERMs a = 0.90; e = 0.40 628,600 (119 € per m2 of
(g) + (h) walls)
Walls’ thermal insulation: ERM (i) Polyurethane: t = 0.15 m; new Uwalls = 0.15 W/
m2 K
P. 2 Heat recovery systems: ERM (a) Vair = 150,000 m3/h; e = 0.8 80,000 1,456,100 14.4 57.5
Solar shading systems: ERM (b) Interior shades: Tsol = 0.5; Tvis = 0.5 45,000
Energy-efficient windows: ERM (c) Low-emissive: Uw = 1.71 W/m2 K; 281,500
SHGC = 0.691
Roof’s radiative characteristics: ERMs a = 0.10; e = 0.90 (cool roof) 510,800 (135 € per m2 of
(d) + (e) roof)
Roof’s thermal insulation: ERM (f) Polyurethane: t = 0.13 m; new Uroof = 0.18 W/
m2 K
Walls’ radiative characteristics: ERMs a = 0.90; e = 0.40 538,800 (102 € per m2 of
(g) + (h) walls)
Walls’ thermal insulation: ERM (i) Polyurethane: t = 0.12 m; new Uwalls = 0.18 W/
m2 K
P. 3 Heat recovery systems: ERM (a) Vair = 150,000 m3/h; e = 0.8 80,000 1,329,300 14.8 56.6
Solar shading systems: ERM (b) Interior shades: Tsol = 0.5; Tvis = 0.5 45,000
Energy-efficient windows: ERM (c) Low-emissive: Uw = 1.71 W/m2 K; 281,500
SHGC = 0.691
Roof’s radiative characteristics: ERMs a = 0.10; e = 0.90 (cool roof) 510,800 (135 € per m2 of
(d) + (e) roof)
Roof’s thermal insulation: ERM (f) Polyurethane: t = 0.13 m; new Uroof = 0.18 W/
m2 K
Walls’ radiative characteristics: ERMs a = 0.90; e = 0.40 412,000 (78 € per m2 of
(g) + (h) walls)
Walls’ thermal insulation: ERM (i) Polyurethane: t = 0.08 m; new Uwalls = 0.24 W/
m2 K
P. 4 Heat recovery systems: ERM (a) Vair = 150,000 m3/h; e = 0.8 80,000 1,265,000 14.9 56.4
Solar shading systems: ERM (b) Interior shades: Tsol = 0.5; Tvis = 0.5 45,000
Energy-efficient windows: ERM (c) Low-emissive: Uw = 1.71 W/m2 K; 281,500
SHGC = 0.691
Roof’s radiative characteristics: ERMs a = 0.10; e = 0.90 (cool roof) 446,500 (118 € per m2 of
(d) + (e) roof)
Roof’s thermal insulation: ERM (f) Polyurethane: t = 0.11 m; new Uroof = 0.21 W/
m2 K
Walls’ radiative characteristics: ERMs a = 0.90; e = 0.40 412,000 (78 € per m2 of
(g) + (h) walls)
Walls’ thermal insulation: ERM (i) Polyurethane: t = 0.08 m; new Uwalls = 0.24 W/
m2 K
P. 5 Heat recovery systems: ERM (a) Vair = 150,000 m3/h; e = 0.8 80,000 1,579,700 14.9 54.7
Solar shading systems: ERM (b) Interior shades: Tsol = 0.5; Tvis = 0.5 45,000
Energy-efficient windows: ERM (c) Low-e + solar control: Uw = 1.64 W/m2 K; 315,300
SHGC = 0.433
Roof’s radiative characteristics: ERMs a = 0.10; e = 0.90 (cool roof 510,800 (135 € per m2 of
(d) + (e) roof)
Roof’s thermal insulation: ERM (f) Polyurethane: t = 0.13 m; new U roof = 0.18 W/
m2 K
Walls’ radiative characteristics: ERMs a = 0.90; e = 0.40 628,600 (119 € per m2 of
(g) + (h) walls)
Walls’ thermal insulation: ERM (i) Polyurethane: t = 0.15 m; new Uwalls = 0.15 W/
m2 K

Intermediate points
P. 6 Heat recovery systems: ERM (a) Vair = 150,000 m3; e = 0.8 80,000 1,396,300 15.3 53.7
Solar shading systems: ERM (b) Interior shades: Tsol = 0.5; Tvis = 0.5 45,000
Energy-efficient windows: ERM (c) Low-e + solar control: Uw = 1.64 W/m2 K; 315,300
SHGC = 0.433

(continued on next page)


54 F. Ascione et al. / Applied Energy 174 (2016) 37–68

Table 5 (continued)

Selected packages of energy retrofit measures for the reduction of thermal energy demand

Package Characterization of the ERMs Investment cost (€) Objective functionsa


(P.)
Single ERM Package TEDheat TEDcool
(kW h/m2 a) (kW h/m2 a)

Roof’s radiative characteristics: ERMs a = 0.10; e = 0.90 (cool roof) 480,600 (127 € per m2 of
(d) + (e) roof)
Roof’s thermal insulation: ERM (f) Polyurethane: t = 0.12 m; new Uroof = 0.20 W/
m2 K
Walls’ radiative characteristics: ERMs a = 0.90; e = 0.40 475,400 (90 € per m2 of
(g) + (h) walls)
Walls’ thermal insulation: ERM (i) Polyurethane: t = 0.10 m; new Uwalls = 0.20 W/
m2 K
P. 7 Heat recovery systems: ERM (a) Vair = 150,000 m3/h; e = 0.8 80,000 1,132,300 16.1 52.7
Solar shading systems: ERM (b) Interior shades: Tsol = 0.5; Tvis = 0.5 45,000
Energy-efficient windows: ERM (c) Low-e + solar control: Uw = 1.64 W/m2 K; 315,300
SHGC = 0.433
Roof’s radiative characteristics: ERMs a = 0.10; e = 0.90 (cool roof) 280,000 (74 € per m2 of
(d) + (e) roof)
Roof’s thermal insulation: ERM (f) Polyurethane: t = 0.05 m; new Uroof = 0.38 W/
m2 K
Walls’ radiative characteristics: ERMs a = 0.90; e = 0.40 412,000 (78 € per m2 of
(g) + (h) walls)
Walls’ thermal insulation: ERM (i) Polyurethane: t = 0.08 m; new Uwalls = 0.24 W/
m2 K

P. 8 Heat recovery systems: ERM (a) Vair = 150,000 m3/h; e = 0.8 80,000 1,028,800 16.8 52.1
Solar shading systems: ERM (b) Interior shades: Tsol = 0.5; Tvis = 0.5 45,000
Energy-efficient windows: ERM (c) Low-e + solar control: Uw = 1.64 W/m2 K; 315,300
SHGC = 0.433
Roof’s radiative characteristics: ERMs a = 0.10; e = 0.90 (cool roof) 234,600 (62 € per m2 of
(d) + (e) roof)
Roof’s thermal insulation: ERM (f) Polyurethane: t = 0.03 m; new Uroof = 0.53 W/
m2 K
Walls’ radiative characteristics: ERMs a = 0.90; e = 0.40 353,900 (67 € per m2 of
(g) + (h) walls)
Walls’ thermal insulation: ERM (i) Polyurethane: t = 0.06 m; new Uwalls = 0.28 W/
m2 K
P. 9 Heat recovery systems: ERM (a) Vair = 150,000 m3/h; e = 0.8 80,000 982,000 17.3 51.6
Solar shading systems: ERM (b) Interior shades: Tsol = 0.5; Tvis = 0.5 45,000
Energy-efficient windows: ERM (c) Low-e + solar control: Uw = 1.64 W/m2 K; 315,300
SHGC = 0.433
Roof’s radiative characteristics: ERMs a = 0.10; e = 0.90 (cool roof) 219,500 (58 € per m2 of
(d) + (e) roof)
Roof’s thermal insulation: ERM (f) Polyurethane: t = 0.02 m; new Uroof = 0.65 W/
m2 K
Walls’ radiative characteristics: ERMs a = 0.90; e = 0.40 322,200 (61 € per m2 of
(g) + (h) walls)
Walls’ thermal insulation: ERM (i) Polyurethane: t = 0.05 m; new Uwalls = 0.32 W/
m2 K

Points with the Lowest TEDcool


P. 10 Heat recovery systems: ERM (a) Vair = 150,000 m3/h; e = 0.8 80,000 907,300 17.7 51.3
Solar shading systems: ERM (b) Interior shades: Tsol = 0.5; Tvis = 0.5 45,000
Energy-efficient windows: ERM (c) Low-e + solar control: Uw = 1.64 W/m2 K; 315,300
SHGC = 0.433
Roof’s radiative characteristics: ERMs a = 0.10; e = 0.90 (cool roof) 234,600 (62 € per m2 of
(d) + (e) roof)
Roof’s thermal insulation: ERM (f) Polyurethane: t = 0.03 m; new Uroof = 0.53W/
m2K
Walls’ radiative characteristics: ERMs a = 0.90; e = 0.40 232,400 (44 € per m2 of
(g) + (h) walls)
Walls’ thermal insulation: ERM (i) Polyurethane: t = 0.01 m; new Uwalls = 0.57 W/
m2 K
P. 11 Heat recovery systems: ERM (a) Vair = 150,000 m3/h; e = 0.8 80,000 935,100 18.3 50.5
Solar shading systems: ERM (b) Interior shades: Tsol = 0.5; Tvis = 0.5 45,000
Energy-efficient windows: ERM (c) Low-e + solar control: Uw = 1.64 W/m2 K; 315,300
SHGC = 0.433
Roof’s radiative characteristics: ERMs a = 0.10; e = 0.90 (cool roof) 204,300 (54 € per m2 of
(d) + (e) roof)
Roof’s thermal insulation: ERM (f) Polyurethane: t = 0.01 m; new Uroof = 0.84 W/
m2 K
Walls’ radiative characteristics: ERMs a = 0.90; e = 0.40 290,500 (55 € per m2 of
(g) + (h) walls)
Walls’ thermal insulation: ERM (i) Polyurethane: t = 0.04 m; new Uwalls = 0.35W/
m2 K
F. Ascione et al. / Applied Energy 174 (2016) 37–68 55

Table 5 (continued)

(continued on next page)


56 F. Ascione et al. / Applied Energy 174 (2016) 37–68

Table 5 (continued)

Legend: – ; – configuration (=ERM) non present in the RB baseline.


a
Objective functions of the first optimization stage: Thermal Energy Demand for space heating (TEDheat), Thermal Energy Demand for space cooling (TEDcool).

Fig. 9. Bi-objective optimization of the proposed ERMs by minimizing of TEDheat and TEDcool: selected Pareto front (gmax = 20). Also the performances of the non-optimized
combinations of the best ERMs are represented.

and required encoding bits for each variable. The parameters of the Pareto front provides 13 non-dominated solutions, which are
GA are set according to the values reported in Section 2.2. Most compared with the non-optimized combinations of the best ERMs
notably, the population size (s) is equal to 20 (4 times the number detected by the SA (preliminary analysis), in terms of TEDheat and
of decision variables), while three values are investigated for the TEDcool. In these combinations, the installation of heat recovery
maximum number of generations (gmax), i.e., 10, 20 and 30 gener- systems is always considered because they are definitely the most
ations, in order to find best trade-off between computational bur- efficient and effective ERM. Thus, the following packages of the
den and reliability. The Pareto fronts obtained for the three values best ERMs for TED reduction are investigated:
of gmax are depicted in Fig. 8. The satisfied stopping criterion is the
achievement of the maximum number of generations, in all cases.  heat recovery systems;
Obviously, when gmax increases, the Pareto front improves thereby  heat recovery systems + solar shading systems;
moving left. This improvement is clear from 10 to 20 generations,  heat recovery systems + cool roof;
whereas it is minimal from 20 to 30 generations, since the Pareto  heat recovery systems + solar shading systems + cool roof.
fronts achieved in the last two cases are almost overlapped. There-
fore, from 20 to 30 generations, the increase of computational These packages and the non-dominated ones provided by the
times, equal to 50%, is not motivated by a significant improvement GA are characterized in Table 5, which delineates ERMs, invest-
of reliability. Definitely, the best trade-off between computational ment cost, TEDheat and TEDcool for each package. The overall num-
burden and reliability is provided by gmax equal to 20, which can be ber of achieved packages is 17 and also the base configuration of
exploited in future applications. the RB is characterized.
Hence, the Pareto front achieved after 20 generations is selected All solutions are depicted in Fig. 9, which shows that non-
for the implementation of the second optimization stage. This optimized packages and baseline RB are, obviously, dominated
F. Ascione et al. / Applied Energy 174 (2016) 37–68 57

Table 6
Characterization of the proposed ERMs for improving the energy efficiency of the primary heating, cooling, district hot water (DHW) production system and for implementing
combined heating and power (CHP) or combined cooling, heating and power (CCHP) systems.

Energy retrofit measures addressed to the primary generation systems


Energy retrofit measure Description Options (if more Investment Incentivesa
configurations are cost (€)
explored)
Heating system
Efficient gas boiler Installation of an energy-efficient natural gas boiler, with a peak thermal power (Pth) equal to 55,500b –
1200 kWth and by a nominal LCV (lower calorific value) efficiency (g) equal to 0.95
Condensing gas boiler Installation of a condensing natural gas boiler, characterized Pth equal to 1200 kWth and g 97,900b 40% of IC up to a limit
equal to 1.05 at rated condition, that is water inlet/outlet temperatures = 35/55 °C of 26,000 €, accorded
in 5 years
Air-source heat pump Installation of two identical air-source heat pumps, characterized by Pth equal to 600 kWth 185,000b 18,300 € per year for a
and COP (coefficient of performance) equal to 3.5 at rated condition, that is water inlet/outlet total duration of
temperatures = 40/45 °C and external air temperature equal to 7 °C 5 years
Heating & cooling
Ground-source reversible heat Installation of two identical reversible ground-source heat pumps with geothermal vertical 720,000c 24,900 € per year for a
pump probes: total duration of
– Heating operation: each heat pump presents Pth equal to 600 kWth and COP equal to 4.5, 5 years
at rated condition, that is water inlet/outlet temperatures = 40/45 °C
– Cooling operation: each heat pump presents Pth equal to 1000 kWth and EER (energy effi-
ciency ratio) equal to 6.2, at rated condition, that is water inlet/outlet temperatures = 12/
7 °C
Cooling system
Air-cooled MagLev chiller Installation of two identical air-cooled chillers with magnetic levitation (MagLev) 380,000c –
compressors, each of them characterized by Pth equal to 1000 kWth and EER equal to 3.5 at
rated condition, that is water inlet/outlet temperatures = 12/7 °C and external air temperature
equal to 35 °C
Water-cooled chiller Installation of two identical water-cooled chillers equipped with cooling towers, each of them 516,000b –
characterized by Pth equal to 1000 kWth and EER equal to 5.5 at rated condition, that is water
inlet/outlet temperatures = 12/7 °C
DHW system
Efficient gas boiler Installation of an energy-efficient natural gas boiler, characterized by a peak thermal power 15,000b –
(Pth) equal to 300 kWth and by a nominal LCV (lower calorific value) efficiency (g) equal to
0.95
CHP or CCHP system
CHP: Co-generation system Installation of a co-generation system (Combined Heating and Power), PCHP
el = 200 kWel, 250,000c –
composed of a gas natural internal combustion engine and a waste heat gCHP
el = 0.390,
recovery boiler. Recovered heat is used for both space heating and DHW gth = 0.410
CHP

CHP
production, alternatively. The system is defined by the following Pel = 300 kWel, 350,000c
parameters: gCHP
el = 0.390,
– the electrical power, denoted as PCHPel gCHP
th = 0.410
– the electrical efficiency (ratio between electrical energy output and PCHP
el = 400 kWel, 440,000c
primary energy input), denoted as gCHP el gCHP
el = 0.400,
– the thermal efficiency (ratio between thermal energy output and gth = 0.400
CHP

primary energy input), denoted as gCHP th CHP


Pel = 500 kWel, 520,000c
Seven options are investigated gCHP
el = 0.400,
gth
CHP
= 0.400
PCHP
el = 600 kWel, 590,000c
gCHP
el = 0.410,
gCHP
th = 0.390
CHP
Pel = 700 kWel, 645,000c
gCHP
el = 0.410,
gCHP
th = 0.390
CHP
Pel = 800 kWel, 695,000c
gCHP
el = 0.415,
gCHP
th = 0.485

CCHP: Tri-generation system Installation of a tri-generation system (Combined Cooling Heating and PCCHP
el = 200 kWel, 280,000c –
Power), composed of a gas natural internal combustion engine, a waste gCCHP
el = 0.390,
heat recovery boiler and an air-cooled single effect absorption chiller. gCCHP
th = 0.410
Recovered heat is used for space heating, space cooling and DHW PCCHP
el = 300 kWel, 401,000c
production, alternatively. The system is defined by PCCHP
el , gCCHP
el , gCCHP
t , gCCHP
el = 0.390,
which have the same meanings used for the CHP system, and by the gCCHP
th = 0.410
EER of the absorption chiller (EERass) PCCHP
el = 400 kWel, 511,000c
Seven options are investigated, all characterized by the same value of gCCHP
el = 0.400,
the nominal EERass, which is set equal to 0.7 at rated condition, that is gCCHP
th = 0.400
CCHP
water inlet/outlet temperatures = 12/7 °C and external air temperature Pel = 500 kWel, 604,000c
equal to 35 °C gel = 0.400,
CCHP

gCCHP
th = 0.400
PCCHP
el = 600 kWel, 686,000c
gCCHP
el = 0.410,
gCCHP
th = 0.390
PCCHP
el = 700 kWel, 750,000c
gel = 0.410,
CCHP

(continued on next page)


58 F. Ascione et al. / Applied Energy 174 (2016) 37–68

Table 6 (continued)

Energy retrofit measures addressed to the primary generation systems


Energy retrofit measure Description Options (if more Investment Incentivesa
configurations are cost (€)
explored)
gCCHP
th = 0.390
PCCHP
el = 800 kWel, 808,000c
gCCHP
el = 0.415,
gCCHP
th = 0.485
a
Financial grants provided by the Italian Government according to [84].
b
Investment cost taken from the relations proposed in [4].
c
Investment cost obtained from direct quotations of suppliers.

Table 7
Characterization of the proposed ERMs for the exploitation of RESs.

Energy retrofit measures for the exploitation of renewable energy sources


Energy retrofit measure Description Options for Options for type Investment costa (€/ Incentives
size m2)
Solar thermal collectors Installation of solar thermal collectors on the cov varies in Flat-plate, not 240 € per m2 of 55 € per m2 of
building roof, south-oriented and with 34° the range 10– selective surface collectors’ surface collectors’ surface,
tilt angle. The size is expressed by the 60% with a Flat-pate selective 360 € per m2 of up to a limit of
percentage of the covered roof area, denoted step equal to surface collectors’ surface 55,000 €, accorded in
with cov 10% 5 yearsb
Six possible sizes and two options for the
solar collector type are investigated
Solar photovoltaic (PV) panels Installation of solar photovoltaic (PV) panels cov varies in Poly-crystalline 250 € per m2 of 50% of IC up to a
on the building roof, south-oriented and the range 10– Silicon panels’ surface limit of 96,000 €,
with 34° tilt angle. The size is expressed by 60% with a mono-crystalline 430 € per m2 of accorded in
the aforementioned parameter cov step equal to silicon panels’ surface 10 yearsc
Six possible sizes and two options for the PV 10%
panel type are investigated
a
Investment cost taken from the relations proposed in [24].
b
Financial grants provided by the Italian Government according to [84].
c
Financial grants provided by the Italian Government according to [85].

solutions. Hereinafter, in order to aid a proper interpretation of the (not available in the mentioned studies) from direct quotations
main results, the non-dominated solutions provided by the GA are of suppliers.
distinguished in three groups: The systems’ size is set by considering the thermal and electri-
cal loads of the baseline. In presence of ERMs for TED reduction,
 solutions that favor heating operation, thereby characterized by thermal peak loads will decrease, but this is not considered in sys-
the lowest values of TEDheat and the highest values of TEDcool; tems’ sizing in order to ensure a secure and safe operation also in
 solutions that favor cooling operation, thereby characterized by presence of breakdowns or rigid climatic conditions, as fundamen-
the lowest values of TEDcool and the highest values of TEDheat; tal for hospitals. Thus, the thermal power (Pth) of primary heating
 intermediate solutions. systems is set equal to the peak thermal load for space heating,
that is 1200 kWth. Concerning the implementation of heat pumps,
Fig. 9 shows that non-dominated solutions yield a clear two identical systems are considered, with Pth equal to 600 kWth
decrease of TEDheat and/or TEDcool compared to the other packages, each, in order to have access to Italian public financial incentives
but, definitely, they are not necessarily included in the whole cost- [84] that are provided only for heat pumps with Pth lower than
optimal retrofit. The second optimization stage is required. This is 1000 kWth, as well as to improve energy performance at part loads.
applied to all achieved 18 solutions for building thermal behavior On the other hand, the peak thermal load for space cooling is equal
(see Table 5), which consist of 17 packages of ERMs for TED reduc- to 2000 kWth. Therefore, the proposed primary cooling systems
tion (13 non-dominated solutions + 4 non-optimized combinations consist of two identical chillers, with Pth equal to 1000 kWth each,
of the best ERMs) and baseline RB. which are preferred to a bigger one in order to ensure a secure and
efficient operation also at low loads. As for combined heating and
4.3. Tri-objective optimization of the whole energy retrofit and power systems (CHP) and combined cooling, heating and power
assessment of cost-optimality systems (CCHP), different options are investigated for the size.
The minimum electrical power (Pel) is set equal to 200 kWel
Proper ERMs addressed to improving the energy efficiency of because this value is the base electrical load for direct electric uses,
primary heating/cooling/electricity systems (see Table 6) and to and, furthermore, smaller CHP and CCHP systems are scarcely
exploiting RESs (see Table 7) are identified based on building pecu- available on the market and present low energy efficiency. The
liarities and best practice. Tables 6 and 7 show these ERMs by maximum value of Pel for these systems is set equal to 800 kWel
reporting energy characteristics, investigated options, investment because it represents the peak electrical load for direct electric
costs, available public financial incentives. Some investment costs uses. Considering higher values of Pel is useless because these are
are taken from previous scientific studies (notably, [4,24]), others not cost-effective [54,55]. Finally, as RES systems, the installation
F. Ascione et al. / Applied Energy 174 (2016) 37–68 59

Fig. 10. Tri-objective optimization of the whole energy retrofit by minimizing investment cost (IC), primary energy consumption (PEC) and global cost (GC) (a), and
assessment of cost-optimality in presence of a limitless economic budget (b).

of solar thermal collectors and photovoltaic (PV) panels on the GC assessment, prices of electricity and natural gas are set equal to
building roof is proposed. Globally, these can cover a maximum 0.18 €/kW hel and 0.62 €/N m3, respectively, taken from the web-
area equal to 60% of roof surface (3785 m2), because the remaining site of the Italian Authority for Energy [87], in correspondence of
surface is reserved to devices and air handling units. These are the amount of electricity and natural gas demands of the RB. Con-
characterized by 34° tilt angle and orientation to South, in order versely, if CHP or CCHP systems are implemented, the demand of
to ensure the highest yearly energy conversion potential [86], natural gas increases and consequently the unitary price
and are properly placed in order to prevent mutual shading. decreases; thus, the price for natural gas consumed by these sys-
All possible combinations of the ERMs addressed to primary tems id equal to 0.43 €/N m3 [87]. Eventually, the electricity pro-
energy systems and RESs are considered, by excluding the presence duced by the facility, through CHP, CCHP or PV panels, and sold
of incompatible measures. Exhaustive sampling is employed by to the grid is remunerated at the price of 0.08 €/kW hel, which is
applying all these combinations to the 18 screened configurations approximately the average value of the Italian cost for pure electric
of building thermal behavior reported in Table 5. energy in the last two quarters of 2015 [87]. In absence of ERMs,
The total number of explored whole energy retrofit packages is the RB is characterized by the following values of PEC and GC:
596,700. For each package, investment cost (IC), difference in
primary energy consumption (dPEC) and in global cost (dGC) com- PECRB ¼ 558:8 kW h=m2 a; GCRB ¼ 11; 982; 610 € ð2Þ
pared to the baseline are assessed in MATLABÒ environment. As for
60 F. Ascione et al. / Applied Energy 174 (2016) 37–68

Fig. 11. Tri-objective optimization of the whole energy retrofit by minimizing investment cost (IC), primary energy consumption (PEC) and global cost (GC) (a), and
assessment of cost-optimality in presence of an available economic budget equal to 1200 k€ (b).

The computational time required to assess IC, dPEC, dGC for all limitless economic availability. The achieved cost-optimal solution
596,700 packages is around 273,900 s (76.1 h). Otherwise, it is requires an investment equal to about 1236 k€, denoted as ICopt,
prohibitive (order of magnitude of months) if standard which is interestingly around 10% of GCRB. Hence, twelve limited
simulation-based approaches, which run a BPS tool simulation economic budgets, lower than ICopt, are considered, from 1200 k€
for each package, are adopted to explore a so huge number of to 100 k€ with a step of 100 k€. For each budget, the packages that
solutions. Finally, the benefit produced by the proposed methodol- require a higher investment are excluded. The remaining solutions
ogy is substantial. are represented in the space IC–dPEC–dGC and then projected onto
The achieved values of IC, dPEC, dGC for all packages of ERMs the plane dPEC–dGC as done in presence of a limitless budget. For
are represented in a three-dimensional space, as reported in demonstrative purposes, Figs. 11–14 show the achieved three-
Fig. 10a, which shows a second Pareto front (three-dimensional) dimensional (figures a) and bi-dimensional (figures b) graphs in
that collects the non-dominated solutions concerning investment correspondence of the budgets 1200 k€, 900 k€, 600 k€ and
cost, primary energy consumption and global cost. Then, all points 300 k€, respectively.
are projected onto the plane dPEC–dGC as reported in Fig. 10b, Figs. 10–14 show that the most cost-effective packages of ERMs
which shows the cost-optimal curve and allows to identify the for TED reduction – namely the ones that imply the highest GC
cost-optimal whole energy retrofit package in presence of a savings (i.e., absolute values of dGC) – are the non-optimized
F. Ascione et al. / Applied Energy 174 (2016) 37–68 61

Fig. 12. Tri-objective optimization of the whole energy retrofit by minimizing investment cost (IC), primary energy consumption (PEC) and global cost (GC) (a), and
assessment of cost-optimality in presence of an available economic budget equal to 900 k€ (b).

combinations of the best ERMs. This means that the energy bene- occurs because, as aforementioned, TEDcool is predominant on
fits produced by the non-dominated Pareto combinations do not TEDheat, especially in presence of heat recovery systems.
justify the increment of investment costs. This is mainly due to Finally, the cost-optimal whole energy retrofit package is iden-
the explored building type, which is characterized by high endoge- tified in correspondence of each of the twelve budgets. Table 8
nous heat gains, so that the ERMs for TED reduction, not included details these cost-optimal solutions, by reporting for each of them:
among the best ones, exercise comparable opposite effects on heat-
ing and cooling operation. Therefore, also the optimal combina-  composition, in terms of ERMs addressed to TED reduction, pri-
tions of these ERMs provided by the GA are less cost-effective mary systems’ efficiency improvement and RESs’ exploitation,
compared to the packages that include only the best ERMs. Defi- respectively;
nitely, this outcome is valid for the explored case study, but in  values assumed by the three objective functions IC, dPEC and
future applications to other building categories (for instance, resi- dGC.
dential buildings) the solutions provided by the GA can be surely
part of cost-optimal retrofit packages. It is noted that, among the The unconstrained cost-optimal retrofit package in presence of
non-dominated packages of ERMs for TED reduction, the ones that a limitless economic availability, and thus the actual cost-optimal
favor the cooling operation imply the highest GC saving. This solution, includes the following ERMs:
62 F. Ascione et al. / Applied Energy 174 (2016) 37–68

Fig. 13. Tri-objective optimization of the whole energy retrofit by minimizing investment cost (IC), primary energy consumption (PEC) and global cost (GC) (a), and
assessment of cost-optimality in presence of an available economic budget equal to 600 k€ (b).

 combination of the best ERMs for TED reduction (package 17 in ICopt ¼ 1236 k€; dPECopt ¼ 67:9 kW h=m2 a; dGCopt ¼ 2932 k€
Table 5), which consist of heat recovery systems, solar shading ð3Þ
systems, cool roof;
 energy-efficient natural gas boiler for DHW production; The outcomes concerning the ERMs’ mix are consistent. As for
 (CCHP) combined cooling, heating and power (i.e., tri- ERMs addressed to TED reduction, the non-optimized combination
generation) system, characterized by an electrical power equal of the best measures is implemented because it is more cost-
to 600 kWel, lower than the electrical peak load for direct elec- effective compared to the non-dominated combinations, as clear
tric uses (800 kWel): the ratio, known as peak factor, is equal to in Figs. 10–14 and argued previously. As for ERMs addressed to pri-
75%; mary systems and RESs, the installation of a CCHP system makes
 not selective solar collectors that cover the 30% of roof surface; ineffective the replacement of the existing boiler and chillers
 mono-crystalline PV panels that that cover the 30% of roof (which remain the same) with more efficient devices, because
surface. these would not be exploited. On the other hand, a new efficient
boiler and solar collectors for DHW production are effective
This cost-optimal solution presents the following values of the because the heat recovery of CCHP is mainly employed for space
three objective functions (clearly, negative values for dPEC and heating and cooling. Furthermore, the electrical power of CCHP
dGC indicate energy and economic savings, respectively): system (600 kWel) is lower than the maximum admitted one
F. Ascione et al. / Applied Energy 174 (2016) 37–68 63

Fig. 14. Tri-objective optimization of the whole energy retrofit by minimizing investment cost (IC), primary energy consumption (PEC) and global cost (GC) (a), and
assessment of cost-optimality in presence of an available economic budget equal to 300 k€ (b).

(800 kWel) in order to minimize the amount of produced electrical complies with this limit. Figs. 10–14 and Table 8 show that when
energy which is not consumed, by taking into account also the the budget decreases:
presence of PV panels. Indeed, an excessive conversion of electric-
ity is not cost-effective because not consumed energy is sold to the  the number of admitted solutions decreases, obviously;
grid at a quite low price. Definitely, all ERMs are synergic and this  the cost-optimal solution moves up (see figures b) because the
second optimization allows to set the optimal configuration of potential GC saving (absolute value of dGC) decreases when the
each measure by considering the interaction with the other ones. constraint gets stricter;
If the decision maker has a limitless economic availability, he  the potential PEC saving (absolute value of dPEC) decreases in
will opt for the described unconstrained cost-optimal solution, most cases, but sometimes it increases because a reduction of
which is definitely the most cost-effective. However, quite com- GC saving, which is related to both energy and economic
monly, there is a limit to the investment that the decision maker aspects, does not necessarily imply a reduction of PEC saving.
does not want (or cannot) exceed. In this instance, he can establish This is mainly due the size of CHP and CCHP power systems,
a maximum budget according to his needs and wills, and select which are highly cost-effective because of the low price of nat-
from Table 8 the constrained cost-optimal retrofit package that ural gas that they consume, without representing the most
64 F. Ascione et al. / Applied Energy 174 (2016) 37–68

Table 8
Cost-optimal whole energy retrofit packages in correspondence of difference available economic budgets: characterization of the ERMs and values assumed by the objective
functions.

Legend: ; – configuration (= ERM) non present in the RB baseline.


a
Objective functions of the second optimization stage: Investment cost (IC), difference in primary energy consumption compared to RB baseline (dPEC), difference in global
cost compared to RB baseline (dGC). Obviously, negative values occur for dPEC and dGC, representing energy and economic savings, respectively.
b
Please note that in the column related to IC, also the ratio between IC and the useful floor area (18,926 m2) is reported for indicative purposes.

energy-efficient ERM. For instance, from budget 600 k€ to bud- conjunction with the cool roof, since these measures are highly
get 500 k€, the reduction of CHP system size and the introduc- synergic. Indeed, heat recovery systems imply a drastic reduction
tion of ERMs that are more energy-efficient, i.e., heat recovery of TEDheat, whereas the cool roof is the ERM that induce the highest
systems, cool roof, PV panels, imply a decrease of GC saving decrease of TEDcool, which gets predominant in presence of heat
and an increase of PEC saving. A similar conclusion is valid for recovery systems.
budgets from 300 k€ to 200 k€. As for ERMs addressed to primary systems and RESs, CHP or
CCHP systems are always present when the budget is adequate,
Also the outcomes concerning the cost-optimal solutions for the because they represent the most cost-effective ERM. Most notably,
different budgets are consistent. As for ERMs addressed to TED from budget 1200 k€ to budget 800 k€, a tri-generation system
reduction, the combination of the three best ones is present only with electrical power equal to 600 kWel is always implemented
in correspondence of a limitless budget. Conversely, for budgets in conjunction with not selective solar collectors. The combination
lower than ICopt, in different cases, ERMs for TED reduction are between these two ERMs is particularly synergic because in sum-
absent because the priority is given to improving the energy effi- mertime the heat recovery of CCHP is employed for space cooling
ciency of primary systems. In other cases, when the budget is suf- while solar thermal is used to satisfy DHW demand, which is sub-
ficient, heat recovery systems are present, sometimes in stantial in hospital buildings. However, lower budgets hamper the
F. Ascione et al. / Applied Energy 174 (2016) 37–68 65

simultaneous implementation of these two quite expensive ERMs, thermal comfort, but the pay back periods are close to 20 years and
and therefore a mere co-generation system, which in summertime therefore this ERM cannot be included in the cost-optimal solution,
exploits heat recovery for DHW production, is preferred. From which is assessed over a calculation period of 20 years.
700 k€ to 600 k€, the implemented co-generation system presents The mentioned studies and further ones in matter of hospitals’
an electrical power equal to 600 kWel, which definitely represents energy retrofit provided interesting outcomes, but they focused on
the optimal size for both CHP and CCHP systems. Lower budgets do specific and limited ERMs by considering a restricted domain of pos-
not allow the choice of this optimal size, thereby imposing the sible retrofit scenario. Therefore, the optimization of energy retrofit
implementation of smaller systems. Finally, the budgets of 200 k€ was partial and not definite. Conversely, the proposed methodology
to 100 k€ do not allow the presence CHP and CCHP systems allows to investigate several ERMs related to the different levers
because too expensive. In correspondence of these lowest budgets, affecting hospitals’ energy performance, namely the characteristics
the most cost-effective ERMs are the heat recovery systems and of building thermal envelope and energy systems, as well as the
the new efficient boiler for the production of DHW. The mere exploitation of RESs. This is an original aspect and a significant
replacement of primary heating and cooling systems is never car- improvement compared to the previous studies, because a huge
ried out, since: from budget 1200 k€ to budget 300 k€, new effi- domain of retrofit scenarios is explored and this ensures the robust
cient devices would not exploited by virtue of the presence of assessment of the actual cost-optimal energy retrofit solutions.
CHP or CCHP systems; for budgets 200 k€ and 100 k€, other ERMs In this regard, the cost-optimal solutions have been found with
are more cost-effective. Eventually, concerning RES technologies, reference to the developed RB, which represent 317 hospital build-
the mix of solar collectors and PV panels is consistent with the ings built in South Italy in the period 1991–2005. A RB has been
other implemented ERMs. Not selective solar collectors are more investigated because the achieved outcomes about the cost-
cost-effective than selective ones. On the other hand, mono- optimality are valid for most members (i.e., buildings) of the repre-
crystalline PV panels are more cost-effective than poly-crystalline sented building category with a good approximation, as stated by
ones, but these latter are preferred in some cases in order to com- the guidelines of the EPBD recast [5,6]. Therefore, the cost-
ply with the economic availability. optimal solutions characterized in Table 8 for different budgets
Finally, the decision maker can choose any of the cost-optimal can be applied to all hospital buildings (real applications) of the
solutions of Table 8 according to his needs and wills, thereby considered category. Clearly, the optimal sizes of CHP and CCHP
implying a double benefit: systems are different for each case study because the thermal
and electrical loads vary. They can be derived from the peak factors
 an economic saving for the private, in terms of reduction of life- reported in Table 8. Thus for each hospital of the category, the deci-
cycle costs compatible with an established maximum level of sion maker can choose one of the identified cost-optimal retrofit
investment; solutions in function of the economic budget. This is another inter-
 an energy and environmental benefit for the collectivity, esting and original aspect of the methodology, since it allows to
because all solutions imply a significant reduction of energy take into account the different economical availabilities that sub-
consumption, and thus polluting emissions, compared to the sist in the several case studies. Definitely, this approach can imply
baseline. a discrepancy with the actual cost-optimal solutions for category’s
members that are quite different from the RB in terms of geometry,
In this regard, the cost-optimal solutions yield a reduction of characteristics of the building thermal envelope and energy sys-
equivalent carbon dioxide (CO2-eq) emissions from 180 t/year, in tems as well as climatic location. However, these cases are few
presence the budget of 100 k€, to 1260 t/year, in presence of a lim- because the investigated category is rather restricted and the RB
itless economic availability (actual cost-optimal solution). This last has been defined ‘ad hoc’, by developing a reliable theoretical
value is impressive, since it is close to the annual CO2-eq emissions model, to represent a wide part of the category. Nevertheless,
of 1000 new average cars. The emissions have been assessed by when a very high level of accuracy is required, the developed
using the conversion factors provided in [88] for the EU members, methodology can be applied directly to the examined case study
and Italy in particular: notably 0.708 tCO2-eq/MW hel for electric- thereby finding the specific cost-optimal solutions for that build-
ity 0.237 tCO2-eq/MW hpr for natural gas. ing. In this regard, future studies will focus on this issue by assess-
ing the discrepancy in the cost-optimal solutions found for the RB
4.4. Originality of results and real applications and for different category’s members, in order to evaluate the error
that is committed if the RB is considered representative of the
Previous scientific studies provided some conclusions similar to whole category.
those found here concerning the impact of specific ERMs on hospi- All told, the cost-optimal solutions reported in Table 8 can be
tal buildings’ energy performance. For instance, also Ruan et al. extended to the 317 hospital buildings of the explored category.
[54] and Gimelli and Muccillo [55] showed the high cost- Most notably, the unconstrained cost-optimal retrofit package
effectiveness of CHP and CCHP systems, due to a stable thermal yields a substantial environmental benefit. Indeed, if the value of
demand and a favorable heat-to-power ratio for hospitals. How- RB’s greenhouse emissions is assumed as average within the cate-
ever, such systems must be carefully designed, especially as gory, the implementation of this package to all 317 buildings pro-
regards the size, in order to maximize the economic benefits. In duces a potential reduction of CO2-eq emissions around 0.40 Mt/
this regard, the optimal peak factor is rather lower than 1 [54,55] year, which is close to 0.10% of total annual CO2-eq emissions from
as also occurs for the cost-optimal solutions found in this study. all polluting sources in Italy [89]. Hence, the application of the pro-
Furthermore, as previously argued, these cost-optimal solutions posed methodology to several RBs, and thus building categories,
do not include the thermal insulation of the building envelope. would hugely reduce the environmental impact of the building
Indeed, this ERM is not cost-effective for hospital buildings because sector at large-scale. Definitely, this would be a small but crucial
of the high ventilation needs and internal heat gains, which imply step toward sustainability.
the risk of indoor overheating. The same conclusion was found by
Buonomano et al. [40], who showed that the roof insulation returns 5. Conclusions
only marginal reductions of space heating and cooling demands.
On the other hand, Ascione et al. [46] demonstrated that the envel- The paper proposes a new methodology to detect robust cost-
ope thermal insulation yields a significant improvement of indoor optimal energy retrofit solutions for buildings, in line with the
66 F. Ascione et al. / Applied Energy 174 (2016) 37–68

EPBD Recast, with an acceptable computational burden. It is based  mono-crystalline photovoltaic (PV) panels that cover the 30% of
on simulation-based multi-stage and multi-objective optimization, roof surface.
performed by coupling EnergyPlus and MATLABÒ. In detail, the
methodology framework presents the following three main stages: This solution implies a saving of primary energy consumption
around 67.9 kW h/m2 a (12.2%) and of global cost around 2932 k€
I. Preliminary phase: the energy performance of the base build- (24.5%), as well as a reduction of CO2-eq emissions equal to
ing configuration is investigated. Then, Latin hypercube 1260 t/year. The required investment cost (ICopt) is about
sampling (LHS) and sensitivity analysis (SA) are employed 1236 k€, which is around 10% of the global cost related to the base-
in order to identify the best (i.e. the most energy-efficient) line RB. The cost-optimal solutions have been identified also in
and the worst (i.e., energy-inefficient) energy retrofit mea- presence of twelve limited economic budgets, lower than ICopt,
sures (ERMs) addressed to the reduction of thermal energy from 1200 k€ to 100 k€ with a step of 100 k€. Definitely, when
demand for space conditioning (TED). the budget decreases, also the potential economic savings during
II. First optimization stage: a genetic algorithm (GA) is run in building lifecycle decrease. Globally, the outcomes show that the
order to identify non-dominated (optimal) combinations of following ERMs, reported in order of priority, are the most cost-
ERMs for TED reduction, by minimizing thermal energy effective ones:
demand for space heating and space cooling, respectively.
Thus, a bi-objective Pareto optimization is performed.  CCHP systems in presence of higher budgets, and CHP systems
III. Second optimization stage: ERMs addressed to improving the in presence of lower ones;
energy efficiency of primary energy systems and to exploit-  not selective solar collectors for DHW production, especially in
ing renewable energy sources (RESs) are introduced. Then, presence of CCHP systems;
by means of smart exhaustive sampling, all combinations  mono-crystalline PV panels, in presence of higher budgets, and
of these ERMs are applied to the baseline, as well as to the poly-crystalline PV panels in presence of lower ones;
non-dominated combinations (provided by GA) and the best  heat recovery systems, in conjunction with cool roof if the bud-
non-optimized combinations (provided by SA) of the ERMs get is sufficient;
for TED reduction. With reference to all achieved whole  efficient natural gas boiler for DHW, especially in presence of
energy retrofit packages, investment cost, primary energy CCHP systems or low budgets.
consumption saving and global cost saving are assessed.
Finally, for each building of the represented category, the deci-
Thus, a three-objective Pareto optimization is performed.
sion maker will choose the cost-optimal energy retrofit solution
Finally, the cost-optimal retrofit packages are identified, in
that complies with his possibility of investment. In any case,
presence of a limitless economic availability and of different
energy consumption, and thus polluting emissions, of the building
budgets, respectively.
category would be drastically reduced, thereby ensuring a huge
environmental benefit in name of sustainability.
Compared to standard simulation-based approaches for cost-
optimal analysis, the methodology ensures more robust and reli-
able solutions with a lower computational burden. The main rea- Acknowledgements
sons are:
The authors would like to thank the ‘Azienda Sanitaria Locale of
 a huge number of retrofit scenario can be investigated with a Naples’ (ASLNAP1), for the deep collaboration and for having pro-
reasonable running time, since EnergyPlus simulations are run vided the data exploited to develop the hospital reference building.
only in the first two stages to assess the thermal performance
of ERMs for TED reduction. Instead, the implementation of pri-
References
mary energy and RES systems is handled in MATLABÒ in the
second optimization stage. Therefore, the number of EnergyPlus [1] Odyssee. Energy efficiency trends in buildings in the EU. Lessons from the
simulations is much lower compared to the number of explored ODYSSEE MURE project Available at:Available from: <http://www.odyssee-
indicators.org>2012.
retrofit scenarios;
[2] Khatib H. IEA world energy outlook 2011: a comment. Energy Policy
 parameters and design variables of the GA in the first optimiza- 2012;48:737–43.
tion stage, which is the most time consuming one, are properly [3] Caputo P, Pasetti G. Overcoming the inertia of building energy retrofit at
set according to the outcomes of LHS and SA in order to mini- municipal level: the Italian challenge. Sustain Cities Soc 2015;15:120–34.
[4] Mauro GM, Hamdy M, Vanoli GP, Bianco N, Hensen JLM. A new methodology
mize the number of required EnergyPlus simulations. for investigating the cost-optimality of energy retrofitting a building category.
Energy Build 2015;107:456–78.
All told, the methodology is applied a reference building (RB) [5] EU Commission and Parliament. Directive 2010/31/EU of the European
Parliament and of the Council of 19 May 2010 on the energy performance of
that represents hospital buildings built in South Italy between buildings (EPBD Recast).
1991 and 2005. The RB has been developed by employing an orig- [6] EU Commission. Commission Delegated Regulation (EU) No 244/2012 of 16
inal procedure. In detail, 596,700 energy retrofit packages have January 2012 supplementing Directive 2010/31/EU of the European Parliament
and of the Council on the energy performance of buildings.
been investigated and the detected cost-optimal solution in pres- [7] Pisello AL, Goretti M, Cotana F. A method for assessing buildings’ energy
ence of a limitless economic availability includes the following efficiency by dynamic simulation and experimental activity. Appl Energy
ERMs: 2012;97:419–29.
[8] Corgnati SP, Fabrizio E, Filippi M, Monetti V. Reference buildings for cost
optimal analysis: method of definition and application. Appl Energy
 heat recovery systems, solar shading systems and cool roof, as 2013;102:983–93.
ERMs for TED reduction; [9] de Vasconcelos AB, Pinheiro MD, Manso A, Cabaço A. A Portuguese approach to
define reference buildings for cost-optimal methodologies. Appl Energy
 energy-efficient natural gas boiler for district hot water (DHW)
2015;140:316–28.
production; [10] Building Performance Institute Europe (BPIE). Principles for Nearly Zero
 combined cooling, heating and power (CCHP) system, with elec- Energy Buildings. Paving the way for effective implementation of policy
trical power equal to 600 kWel; requirements. November 2011 report.
[11] Typology approach for building stock energy assessment. Main results of the
 not selective solar collectors that cover the 30% of roof surface, TABULA project. Final project report, Institut Wohnen und Umwelt GmbH,
for DHW production; Darmstadt (Germany). Available at: <www.building-typology.eu>.
F. Ascione et al. / Applied Energy 174 (2016) 37–68 67

[12] Spiekman M, Panek A, Erhorn H, Erhorn-Kluttig H, Guyot G, Carrié R, [41] Saidur R, Hasanuzzaman M, Yogeswaran S, Mohammed HA, Hossain MS. An
et al. The final recommendations of the ASIEPI project: how to make end-use energy analysis in a Malaysian public hospital. Energy 2010;25
EPB-regulations more effective? Summary report 2010. Available at: (12):4780–5.
<http://www.asiepi.eu>. [42] Čongradac V, Prebiračević B, Jorgovanović N, Stanišić D. Assessing the energy
[13] Deru M, Field K, Studer D, Benne K, Griffith B, Torcellini P, et al. U.S. consumption for heating and cooling in hospitals. Energy Build
department of energy commercial reference building models of the national 2012;48:146–54.
building stock. NREL – National Renewable Energy Laboratory; 2011. Available [43] Bagnasco A, Fresi F, Saviozzi M, Silvestro F, Vinci A. Electrical consumption
at:Available from: <http://www.nrel.gov/docs/fy11osti/46861.pdf>. forecasting in hospital facilities: an application case. Energy Build 2015;103
[14] Evins R. A review of computational optimisation methods applied to (15):261–70.
sustainable building design. Renew Sustain Energy Rev 2013;22:230–45. [44] Saidur R, Hasanuzzaman M, Mahlia TMI, Rahim NA, Mohammed HA. Chillers
[15] Nguyen AT, Reiter S, Rigo P. A review on simulation-based optimization energy consumption, energy savings and emission analysis in an institutional
methods applied to building performance analysis. Appl Energy buildings. Energy 2011;36(8):5233–8.
2014;113:1043–58. [45] Čongradac V, Prebiračević B, Petrovački N. Methods for assessing energy
[16] Conn AR, Scheinberg K, Vicente LN. Introduction to derivative-free savings in hospitals using various control techniques. Energy Build
optimization, vol. 8. Siam; 2009. 2014;69:85–92.
[17] Ascione F, Bianco N, De Stasio C, Mauro GM, Vanoli GP. A new methodology for [46] Ascione F, Bianco N, De Masi RF, Vanoli GP. Rehabilitation of the building
cost-optimal analysis by means of the multi-objective optimization of building envelope of hospitals: achievable energy savings and microclimatic control on
energy performance. Energy Build 2015;88:78–90. varying the HVAC systems in Mediterranean climates. Energy Build
[18] Wright JA, Loosemore HA, Farmani R. Optimization of building thermal design 2013;60:125–38.
and control by multi-criterion genetic algorithm. Energy Build 2002;34 [47] Van Hoof J, Kort HSM, Duijnstee MSH, Rutten PGS, Hensen JLM. The indoor
(9):959–72. environment and the integrated design of homes for older people with
[19] Magnier L, Haghighat F. Multiobjective optimization of building design using dementia. Build Environ 2010;45(5):1244–61.
TRNSYS simulations, genetic algorithm, and artificial neural network. Build [48] Van Hoof J, Kort HSM, Hensen JLM, Duijnstee MSH, Rutten PGS. Thermal
Environ 2010;45(3):739–46. comfort and the integrated design of homes for older people with dementia.
[20] Hamdy M, Hasan A, Siren K. Applying a multi-objective optimization approach Build Environ 2010;45(2):358–70.
for design of low-emission cost-effective dwellings. Build Environ 2011;46 [49] Bujack J. Heat consumption for preparing domestic hot water in hospitals.
(1):109–23. Energy Build 2010;42:1047–55.
[21] Hamdy M, Hasan A, Siren K. A multi-stage optimization method for cost- [50] Salata F, Golasi I, di Salvatore M, de Lieto Vollaro A. Energy and reliability
optimal and nearly-zero-energy building solutions in line with the EPBD- optimization of a system that combines daylighting and artificial sources. A
recast 2010. Energy Build 2013;56:189–203. case study carried out in academic buildings. Appl Energy 2016;169:250–66.
[22] Ascione F, Bianco N, De Masi RF, Mauro GM, Vanoli GP. Design of the building [51] Ma Z, Wang S. Supervisory and optimal control of central chiller plants using
envelope: a novel multi-objective approach for the optimization of energy simplified adaptive models and genetic algorithm. Appl Energy 2011;88
performance and thermal comfort. Sustainability 2015;7(8):10809–36. (1):198–211.
[23] Ascione F, Bianco N, De Stasio C, Mauro GM, Vanoli GP. Simulation- [52] Arteconi A, Hewitt NJ, Polonara F. State of the art of thermal storage for
based model predictive control by the multi-objective optimization of demand-side management. Appl Energy 2012;93:371–89.
building energy performance and thermal comfort. Energy Build [53] Arteconi A, Brandoni C, Polonara F. Distributed generation and trigeneration:
2016;111:131–44. energy saving opportunities in Italian supermarket sector. Appl Therm Eng
[24] Ascione F, Bianco N, De Masi RF, De Stasio C, Mauro GM, Vanoli GP. 2009;28(8–9):1735–43.
Multi-objective optimization of the renewable energy mix for a building. [54] Ruan Y, Liu Q, Zhou W, Firestone R, Gao W, Watanabe T. Optimal option of
Appl Therm Eng 2015. http://dx.doi.org/10.1016/j. distributed generation technologies for various commercial buildings. Appl
applthermaleng.2015.12.073 (in press). Energy 2009;86(9):1641–53.
[25] Chantrelle FP, Lahmidi H, Keilholz W, Mankibi ME, Michel P. Development of a [55] Gimelli A, Muccillo M. Optimization criteria for cogeneration systems: multi-
multicriteria tool for optimizing the renovation of buildings. Appl Energy objective approach and application in an hospital facility. Appl Energy
2011;88(4):1386–94. 2013;104:910–23.
[26] Echenagucia TM, Capozzoli A, Cascone Y, Sassone M. The early design stage of a [56] Costa A, Fichera A. A mixed-integer linear programming (MILP) model for the
building envelope: multi-objective search through heating, cooling and evaluation of CHP system in the context of hospital structures. Appl Therm Eng
lighting energy performance analysis. Appl Energy 2015;154:577–91. 2014;71(2):921–9.
[27] Diakaki C, Grigoroudis E, Kabelis N, Kolokotsa D, Kalaitzakis K, Stavrakakis G. A [57] Zheng CY, Wu JY, Zhai XQ. A novel operation strategy for CCHP systems based
multi-objective decision model for the improvement of energy efficiency in on minimum distance. Appl Energy 2014;128(1):325–35.
buildings. Energy 2010;35(12):5483–96. [58] DesignBuilder Software, V 3.2.0.067. DesignBuilder Software Ltd.
[28] Hammache A, Benali M, Aubé F. Multiobjective self-adaptive algorithm for Gloucestershire (UK); 2013. <www.designbuilder.co.uk>.
highly constrained problems: novel method and applications. Appl Energy [59] Oh SMYJ, Kim Y, Park CS, Kim I. Process-driven BIM-based optimal design
2010;87:2467–78. using integration of EnergyPlus, genetic algorithm, and Pareto optimality. In:
[29] Sanaye S, Dehghandokht M. Modeling and multi-objective optimization of Proceedings of the IBPSA building simulation 2011 conference Sydney
parallel flow condenser using evolutionary algorithm. Appl Energy 2011;88 (Australia), November 2011. p. 894–901.
(5):1568–77. [60] Kim DW, Park CS. Manual vs. optimal control of exterior and interior blind
[30] D’Errico G, Cerri T, Pertusi G. Multi-objective optimization of internal systems. In: Proceedings 11th international IBPSA conference Glasgow
combustion engine by means of 1D fluid-dynamic models. Appl Energy (Scotland), July 2009. p. 1663–70.
2011;88(3):767–77. [61] Helton JC, Johnson JD, Sallaberry C, Storlie CB. Survey of sampling-based
[31] Fesanghary M, Asadi S, Geem ZW. Design of low-emission and energy-efficient methods for uncertainty and sensitivity analysis. Reliab Eng Syst Saf
residential buildings using a multi-objective optimization algorithm. Build 2006;91:1175–209.
Environ 2012;49:245–50. [62] Jin R, Chen W, Simpson TW. Comparative studies of metamodelling techniques
[32] Karmellos M, Kiprakis A, Mavrotas G. A multi-objective approach for optimal under multiple modelling criteria. Struct Multidiscip Optim 2001;23(1):1–13.
prioritization of energy efficiency measures in buildings: model, software and [63] Hopfe CJ. Uncertainty and sensitivity analysis in building performance
case studies. Appl Energy 2015;139:131–50. simulation for decision support and design optimization [Ph.D.
[33] Delgarm N, Sajadi B, Kowsary F, Delgarm S. Multi-objective optimization of the thesis]. Eindhoven University; 2009.
building energy performance: a simulation-based approach by means of [64] Hopfe CJ, Hensen JLM. Uncertainty analysis in building performance
particle swarm optimization (PSO). Appl Energy 2016;170:293–303. simulation for design support. Energy Build 2011;43(10):2798–805.
[34] Trnsys A. Transient system simulation program. University of Wisconsin; [65] de Wit S, Augenbroe G. Analysis of uncertainty in building design evaluations
2000. and its implications. Energy Build 2002;34:951–8.
[35] IDA-ICE. IDA indoor climate and energy. <http://www.equa.se/ice/intro.html>. [66] Eisenhower B, O’Neill Z, Narayanan S, Fonoberov VA, Mezić I. A methodology
[36] US Department of Energy. Energy Efficiency and Renewable Energy Office, for meta-model based optimization in building energy models. Energy Build
Building Technology Program. EnergyPlus 8.0.0; 2013. Available at: <http:// 2012;47:292–301.
apps1.eere.energy.gov/buildings/energyplus/>. [67] Tian W. A review of sensitivity analysis methods in building energy analysis.
[37] MATLABÒ – MATrixLABoratory (2010) – 7.10.0. User’s guide. MathWorks; Renew Sustain Energy Rev 2013;20:411–9.
2010. [68] Saltelli A, Ratto S, Andres T, Campolongo F, Cariboni J, Gatelli D, et al. Global
[38] Hong T, Koo C, Kim J, Lee M, Jeong K. A review on sustainable construction sensitivity analysis. Wiley; 2008.
management strategies for monitoring, diagnosing, and retrofitting the [69] de Wilde P, Tian W. Identification of key factors for uncertainty in the
building’s dynamic energy performance: focused on the operation and prediction of the thermal performance of an office building under climate
maintenance phase. Appl Energy 2015;155:671–707. change. Build Simul 2009;2:157–74.
[39] Alexis GK, Liakos P. A case study of a cogeneration system for a hospital in [70] de Wilde P, Tian W. Predicting the performance of an office under climate
Greece. Economic and environmental impacts. Appl Therm Eng 2013;54 change: a study of metrics, sensitivity and zonal resolution. Energy Build
(2):488–96. 2010;42:1674–84.
[40] Buonomano A, Calise F, Ferruzzi G, Palombo A. Dynamic energy performance [71] Yildiz Y, Korkmaz K, Özbalta TG, Arsan ZD. An approach for developing
analysis: case study for energy efficiency retrofits of hospital buildings. Energy sensitive design parameter guidelines to reduce the energy requirements of
2014;78:555–72. low-rise apartment buildings. Appl Energy 2012;93:337–47.
68 F. Ascione et al. / Applied Energy 174 (2016) 37–68

[72] Deb K. Multi-objective optimization using evolutionary algorithms, vol. [81] ISPESL, Istituto Superiore per la Prevenzione e la Sicurezza dei Lavoro. Linee
16. Chichester: John Wiley & Sons; 2001. Guida per la definizione degli standard di sicurezza e di igiene ambientale dei
[73] Li YF, Xie M, Goh TN. A study of project selection and feature weighting for reparti operatori. Consiglio Superiore di Sanità, 26/07/2002 [in Italian].
analogy based software cost estimation. J Syst Softw 2009;82(2):241–52. [82] UNI – Italian Organization for Standardization. UNI EN ISO 13790: Prestazione
[74] Li YF, Ng SH, Xie M, Goh TN. A systematic comparison of metamodeling Energetica degli Edifici, Calcolo del Fabbisogno di Energia per il Riscaldamento
techniques for simulation optimization in decision support systems. Appl Soft e il Raffrescamento [in Italian].
Comput 2010;10(4):1257–73. [83] Available at: <http://apps1.eere.energy.gov/buildings/energyplus/
[75] European Commission. Energy directorate general, meeting document for the weatherdata_about.cfm/>.
expert workshop on the comparative framework methodology for cost optimal [84] Italian Government Decree. Decreto Ministeriale 28 Dicembre 2012.
minimum energy performance requirements. In: Preparation of a delegated Incentivazione della produzione di energia termica da fonti rinnovabili ed
act in accordance with Art 290 TFEU 6 May 2011 in Brussels; 2011. interventi di efficienza energetica di piccole dimensioni (Conto termico) [in
[76] Italian Parliament Law. Legge 9 Gennaio 1991 n. 10, G.U. 16/1/1991. Norme Italian].
per l’attuazione del Piano energetico nazionale in materia di uso razionale [85] Italian Government Law. Legge 23 Dicembre 2014 n. 190, G.U. 29/12/2014
dell’energia, di risparmio energetico e di sviluppo delle fonti rinnovabili di (Legge di Stabilità 2015) [in Italian].
energia [in Italian]. [86] PV-GIS Software. European Commission, Joint Research Centre, Institute for
[77] ISTAT (national Institute of STAtistics). Data available at: <http://dati.istat.it/ Environment and Sustainability, Renewable Energies Unit. Available at:
Index.aspx?DataSetCode=DCIS_OSPEDSSN>. <http://re.jrc.cec.eu.int/pvgis/pv/>.
[78] Italian Government Ministerial Circular. Circolare Ministeriale LL.PP. n.13011/ [87] Italian Authority for Energy site. Autorità per l’Energia Elettrica e il Gas (AEEG).
22.11.1974.Requisiti fisico-tecnici per le costruzioni edilizie ospedaliere. Available at: <http://www.autorita.energia.it/> [in Italian].
Proprietà termiche, igrometriche, di ventilazione e di illuminazione [in Italian]. [88] Technical annex to the SEAP template instructions document: THE EMISSION
[79] UNI – Italian Organization for Standardization. Norma UNI 10339: Impianti FACTORS. Available at: <http://www.eumayors.eu/IMG/pdf/technical_annex_
areaulici ai fini di benessere. Generalità, classificazione e requisiti. Regole per en.pdf/>.
la richiesta di offerta, l’offerta, l’ordine e la fornitura [in Italian]. [89] ISPRA. Italian Greenhouse Gas Inventory 1990–2012. National Inventory
[80] UNI – Italian Organization for Standardization. Norma UNI 8199: Misura in Report 2014. Available at: <http://www.isprambiente.gov.it/files/
opera e valutazione del rumore prodotto negli impianti di riscaldamento, pubblicazioni/rapporti/Rapporto_198_2014.pdf/>.
condizionamento e ventilazione [in Italian].

You might also like