Professional Documents
Culture Documents
1 s2.0 S2352710222013389 Main
1 s2.0 S2352710222013389 Main
A R T I C L E I N F O A B S T R A C T
Keywords: In this paper, a metamodel-based approach involving simulation data collection and data-driven
Building energy optimization techniques was used to forecast and optimize heating and cooling loads in three different climates
CO2 emissions in Morocco. The metamodel method is gaining popularity as it offers a better balance between
Metamodel accuracy and calculation time. In addition, a wrapper method was used as a feature selection
Feature selection method approach to find the best feature subsets in order to reduce models’ complexity. Therefore, the
Artificial neural network
performances of eleven state-of-the-art algorithms including evolutionary-based algorithms,
Support vector machine
swarm intelligence-based algorithms and human-based algorithms, coupled with a learner such as
Artificial Neural Networks (ANNs) and Support vector machine (SVM), were assessed. Hybrid
model based on League championship algorithm (LCA), Discrete symbiotic organisms search
(DSOS) algorithm, Particle swarm optimization (PSO) algorithm showed better results in terms of
accuracy and reduction of feature input parameters. Indeed, the best desired performances were
obtained with LCA-SVM for cooling load, DSOS-SVM for heating load, PSO-ANN for both heating
and cooling loads. On the other hand, to optimize the annual thermal load, the NSGA-II algorithm
was used. Results showed a reduction of 68% of the total annual load compared to the base case in
Meknes city, 73% in Ifrane and 67% in Marrakech. The solution that reflects a compromise be
tween the two objective functions (i.e., cooling load and heating load) gives better results in terms
of CO2 emissions reduction in all climates evaluated, except in the cold climate.
Nomenclature
* Corresponding author.
** Corresponding author.
E-mail addresses: nawaldine1@gmail.com (N. Abdou), elmghouchi.87@gmail.com (Y. El Mghouchi).
https://doi.org/10.1016/j.jobe.2022.105332
Received 11 May 2022; Received in revised form 24 September 2022; Accepted 25 September 2022
Available online 29 September 2022
2352-7102/© 2022 Elsevier Ltd. All rights reserved.
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
Acronyms
CL Cooling load
EER Energy Efficiency Ratio
GSA Global sensitivity analysis
HL Heating load
HVAC Heating, Ventilation and Air Conditioning
LHS Latin hypercube sampling
LSA Local sensitivity analysis
RMSE Root Mean Squared Error
WWR Windows-to-Wall Ratio
1. Introduction
The current worldwide context, characterized by a stressed climate as a result of rising energy demand and rising greenhouse gas
emissions, is forcing decision-makers in all sectors to recognise the importance of sustainable development and energy efficiency in all
countries and regions. According to the International Energy Agency (IEA), the building sector consumes more than a third of global
final energy consumption and emits more than 40% of total direct and indirect CO2 emissions.
In Morocco, due to the significant increase in the population density and lifestyle requirements, the building sector has become the
second largest energy consumer in the country, accounting for almost 33% of the total final energy consumption [1]. The Moroccan
government invites all operators and stakeholders to control energy demand in order to achieve a primary energy saving of 15% by
2030. However, to provide robust long-term solutions that not only reduce CO2 emissions but also minimise the need for
non-renewable energy sources, urgent but prudent actions such as improving the energy efficiency of buildings are needed [2]. The
common method to improve the energy performance of buildings is based on computer simulation using general purpose software such
as EnergyPlus [3], DOE-2 [4], ESP-r [5], and TRNSYS [6]. But due to the high number of alternatives and the complex interaction
between the input design variables and the output objective, intelligent optimization techniques are usually adopted to reduce the
calculation time [7]. The first approach in building optimization is to couple the above-mentioned software with an intelligent al
gorithm such as Genetic algorithm, Particle swarm optimization (PSO) or Ant Colony Optimization (ACO).
Therefore, to highlight the necessity of using various heating systems and fuels, as well as the effects of insulating thermal bridges,
Schwartz et al. [8] performed a muti-objective optimization method using a multi-objective genetic algorithm (MOGA), connected
with jEPlus, to optimize the life cycle carbon footprint and life cycle cost of a residential complex in the UK. Bamdad et al. [9] used
EnergyPlus coupled with Ant colony optimization (ACO) algorithm to optimize a commercial building in Australia. Optimization
results showed that more than 11.4% of energy savings can be achieved. Abdou et al. [10] investigated the optimal retrofit solutions of
a residential building in six Moroccan climatic zones considering energy savings, life cycle cost and thermal comfort as objective
functions using TRNSYS software coupled with MOBO (multi-objective building optimization tool). The study clearly demonstrated
the conflict between the three objective functions considered in the context of a zero-energy building target.
Another approach in building optimization is the use of a metamodel, also known as a surrogate model, which consists of creating a
mathematical model such as a simple or multiple linear regression model or training a machine learning model such as Artificial Neural
Networks (ANNs) or Support vector machine (SVM), to replace the original simulation-based tool. Romani et al. [11] studied heating
and cooling loads optimization using polynomial regression as metamodel and D-optimal as design of experiments method. The al
gorithm used in the optimization process was the simplex Nelder-Mead algorithm. Gengembre et al. [12] used a Kriging model, which
is a regression function corrected by a Gaussian technique, to optimize the life cycle cost of a single-zone building. The algorithm used
in the optimization process was Particle swarm optimization. Gou et al. [13] performed a multi-objective optimization using ANNs and
NSGA-II algorithm to improve the thermal comfort and building energy load. The methodology used included a sensitivity analysis
2
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
technique, followed by an optimization phase. The main challenge of the metamodel approach, which involves the use of a physical
simulation data collection and a data-driven technique, is the accuracy of prediction, which depends on the informativeness of samples
and the metamodel parameters.
Therefore, several researchers use hybrid methods, often known as improved models [14], which combine an optimization al
gorithm with a machine learning model. Many parameters can be optimized in an improved model, like as weights and biases [15],
hyperparameters [16], etc., depending on learner’s model. There is a lack of studies on energy optimization of buildings using a feature
selection method. Most existing studies using metamodel technique adopted a sequential approach with fixed input parameters as
predictor variables [17–20]. In our study, an integrated feature selection approach using wrapper method is proposed. The proposed
method uses improved machine learning models based on input feature selection process. The feature selection approach allows us to
reduce model complexity and computation time by removing unimportant input variables [21]. To assess the accuracy of the proposed
method, a case study in three different climates in Morocco was performed to predict and optimize heating and cooling loads.
The remainder of this paper is organized as follows: Section 2 and 3 describe the different methods and algorithms used as well as
the building case study and the optimization problem. Section 4 presents the results, including the selection of the best hybrid models
and optimization results. The 5th section presents the conclusions of this study and proposes perspectives for future works.
2. Methodology
The main challenge in building energy optimization is the computation time. The metamodel method is becoming more popular
because it provides a better balance between accuracy and calculation time [22]. The proposed methodology therefore aims to reduce
the computational time by using state-of-the-art methods for building energy optimization, such as Latin hypercube random sample
method and machine learning algorithms. The focus of this research is based on the use of feature selection methods in order to reduce
model’s complexity. The methodology was developed with the following steps (Fig. 1):
• Definition of the building variables: a sufficiently wide range of alternative designs should be considered such as building orien
tation, interior and external wall and roof type, glazing type, window shading and infiltration rate, etc.
• Data base generation: Latin hypercube sampling, which is a space-filling method used to generate quasi-random samples, was
selected with 1000 points as initial sample size [17].
• Data base simulation: TRNSYS was used as computer simulation software to run the building energy simulations in each climate.
• Metamodel training: before modelling, the data collected are scaled and normalised as pre-processing and then randomly divided
into a training data set (70%), a test data set (15%) and a validation data set (the remaining data) [23]. Fig. 2 shows the flowchart of
the training phase with an integrated feature selection method (i.e., wrapper method).
• Finally, based on the performance of statistical analysis, the optimized metamodel is evaluated in terms of prediction quality and
accuracy.
In the following sections, a sensitivity analysis method, as well as feature selection methods and different algorithms used to
optimize the metamodel, are described.
3
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
time, factorial factor design, or Latin hypercube sampling (LHS), are categorized based on how the design conditions are prepared [24,
25]. LHS is one of the most popular techniques for creating a constrained set of variables with a limited and representative sample of
the population. In this method, all parameters and input variables are discretized, which means that they have a limited number of
spaces. In most cases, the entire space is assigned a single value, like midway, and the value assigned represents the probability related
to that space. LHS provides good coverage of the input parameter space, since it first divides the parameter space into areas of equal
probability and then randomly selects samples within these multidimensional areas. The simplicity of this method may be an
advantage, but if the number of input variables is increased, the drawback may be a large number of sample points.
4
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
[ ( ⃒ )]
Var E Y ⃒Xj
Sj = (1)
Var(Y)
[ ( ⃒ )]
Var E Y ⃒Xj′
ST = 1 − (2)
Var(Y)
Where Y is the model output, Xj is the matrix of input factor j and Xj′ is the matrix of the remaining factors.
5
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
1
f (z) = z
(4)
1 + e−
where Yl is the output of the neurons, wl is the weights and bl is the bias.
where φ(x) is the transformed matrix. Wand b are estimated according to an optimization problem defined as:
{ }
1 ∑n
( )
Min 2
‖W‖ + C ∗
ξi + ξi (6)
W,b,ξ∗i ,ξi 2
i=1
Subject to
⎧( )
⎪ T
⎨ ( yi − W φ(xi ) − b) ≤ ε + ξi
⎪
W T φ(xi ) + b − yi ≤ ε + ξ∗i (7)
⎪
⎪
⎩ ξ ≥ 0, ξ∗ ≥ 0, i = 1, …, n
i i
where C is a regularisation parameter, ξ∗i and ξi are the slack variables which allow the constraints flexibility.
Where yi , yi , y are the observed value, the predicted value and the average of yi . In addition, standard deviation (STD), which indicates
′
how far the RMSE differs from the average RMSE, was computed [27]. Smaller RMSE and STD indicate greater accuracy of the
predicted values while a higher value of R2 close to 1 means a perfect linear relationship between data. Moreover, in order to rank the
different models, a score-based approach was performed [42] using Eq. (10). Performance score (PS) indicates the overall score of each
model in terms of correlation and approximations. A smaller PS value denotes a better performance of the model according to the
criteria considered.
( )
PS = rank R2 + rank(RMSE) + rank(STD) (10)
3. Case study
3.1. Weather data and locations
The weather data used was generated by METEONORM software and represents a typical meteorological year. Three cities were
selected according to the type of climate, which can range from cold to semi-arid (i.e., Ifrane, Meknes and Marrakech). Meknes has a
Mediterranean climate, with moderate, rainy winters and hot, dry summers. The temperature of the warmest month varies between
6
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
Table 1
Locations and climate characteristics.
Location Climatic Climate type Minimum DBT Mean DBT Maximum DBT CDD base 24 ◦ C HDD base 18 ◦ C
(Morocco) zone (◦ C) (◦ C) (◦ C)
Fig. 3. (a): Views of the building studied, (b): Model for simulation.
Table 2
Building construction materials (Properties: Data source [11]).
Building Material (layers) Thickness (cm) Thermal Density (Kg/m3) Thermal Overall U-value (W/m2.K)
Components Conductivity, KJ/(h.m.K) Capacity, kJ/(kg. K)
30 ◦ C and 44 ◦ C, and between 0 ◦ C and 7 ◦ C for the coldest month. Ifrane has a humid and temperate climate. In winter, the rainfall is
much higher than in summer. The climate is classified as Csb on the Köppen-Geiger climate map. The average annual temperature is
15 ◦ C. The city of Marrakech, however, has a semi-arid climate. The annual average temperature is 20 ◦ C. The average annual rainfall is
281 mm, which is lower than the Mediterranean climatic zone’s average. Some climatic characteristics such as dry bulb temperature
(DBT), cooling degree days (CDD) and heating degree days (HDD) are listed in Table 1.
7
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
Table 3
Building input parameters and their range of variation.
Table 4
Different types of external window (X11).
Glazing type for windows Symbol Thickness (mm) U-value (W/m2.K) SHGC
• Total internal gains per flat each year are 2500 kWh, evenly distributed, with a ventilation rate set at 1 ACH in winter, while in
summer it is set to 4 ACH when the outdoor temperature is below the set temperature [43]. For all zones, a 0.5 ACH infiltration rate
was assumed [44].
• The building was modelled as 42 thermal zones using Sketchup. Each building zone is described by a single air node with uniform
state variables.
• In each zone, the initial air temperature and humidity were fixed at 20 ◦ C and 50%, respectively.
• The following equations Eq. (11) and Eq. (12) were used to compute the convective heat transfer coefficient for exterior surfaces
and internal surfaces, respectively, where V is the wind velocity [45], the variables C and A depend on the type of surface: floor,
ceiling, or vertical wall.
houtside = 2.8 + 3.2V (11)
( )A
hinside = C Tsurf − Tair (12)
8
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
to 8%.
CO2 emissions from residential buildings are classified into operational CO2 emissions and material production emissions, which
represent the CO2 emissions during the production and transport of building materials [52]. In this study, the effects of building retrofit
measures on CO2 emissions achieved by reducing the need for electricity and natural gas for cooling and heating respectively were
assessed [53]. Therefore, based on the habits of the Moroccan middle class, a split system residential air conditioner (EER = 2.8) and a
natural gas fired boiler (efficiency η = 0.90) were considered as cooling and heating systems respectively [10,54]. Fig. 4 summarizes
the optimization process.
The optimization problem was defined as:
( )/
F1 (x) × FCO2,NG F2 (x) × FCO2,EL
Min F(x) = + Anet (13)
η EER
⎧
⎪
⎪ F1 > 0 and F2 > 0
⎪
⎪ 0.3 ≤ X1 , X3 ≤ 1
⎪
⎪
⎪
⎪
⎨ 0.2 ≤ X2 , X4 ≤ 0.8
Subject to : 0.3 ≤ X5 ≤ 2 (14)
⎪
⎪
⎪
⎪ 0.5 ≤ X6 ≤ 2
⎪
⎪
⎪
⎪ 10 ≤ X7 , X8 , X9 , X10 ≤ 40
⎩
(F1 + F2 ) ≤ Rlimit .Anet
For all x = [x1 ,x2 ,…,x11 ], x ∈ X, where X is all the possible combination. F is the annual equivalent carbone dioxide emissions CO2 ,
F1 and F2 are the sub-objective function which represents the annual heating and cooling loads, respectively. FCO2,EL and FCO2,NG are the
specific emissions factors for electricity and natural gas, respectively and Anet is the net floor area. A specific emissions factors of 0.8088
kgCO2 /(kWh) [55] and 12.6 gCO2 /(MJ) [56] were used for FCO2,EL and FCO2,NG , respectively. Rlimit is the regulatory limit that defines the
threshold for annual thermal load which differs from location to another. In addition, to select the optimal solution, the weighted sum
method was used taking into account the relative importance of each sub-objective function using the following formula [10]:
( ) ( )
F1 (x) − F1 min F2 (x) − F2 min
Min [u(F1 (x), F2 (x))] = w1 + w2 (15)
F1 max − F1 min F2 max − F2 min
9
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
where w1 and w2 are the weight coefficient, Fi min and Fi max are the best and worst values of the ith sub-objective function, respectively.
10
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
Fig. 7. Performance ranking of hybrid models according to the learner and metric (R2, RMSE) in Meknes.
from location to another. Thus, it can be noted through this sensitivity analysis that some parameters such as X5, X10 and X9 can be
considered as non-influential parameters for both annual cooling and heating load. X9 is the window to wall ratio to the west, X10 is
the window to wall ratio to the north and X5 is the ground floor transmission coefficient. However, X5 can be a highly influential
parameter depending on the building typology and use [1]. In our study, X5 is not a key factor because the zones of the building in
contact with the ground are not conditioned since the ground floor is for retail use [43].
11
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
Fig. 8. Performance ranking of hybrid models according to the learner and metric (STD, PS) in Meknes.
Fig. 9. Performance ranking of hybrid models according to the location and learner.
12
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
Table 5
Input variable according to the learner and output.
Location Output (Load) Learner Best hybrid model Number of input variable Input variable
Table 6
Setting parameters of NSGA-II algorithm.
Population size 60
Generations 100
Probability of Mutation (%) 2
Crossover probability (%) 80
13
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
respectively. A, B and C represent the min (F1 ), the trade-off between F1 and F2 and the min (F2 ), respectively. Figs. 10–12 show the
results.
As part of a building optimization process, optimal parameters are determined and recommended. It is possible to observe a trend of
the considered building factors according to the lower, middle, and upper values for each factor, as well as the reduction of heating and
cooling loads. For such climatic conditions, according to the compromise solution B, the lower option is recommended for X1 and X6.
For X2, X3, X4, X7, X8, and X11, a value between the lower and average options is recommended. For X11, the average option is better.
Table 7 reports the results. In addition, to ensure results quality, a comparison between the results of the models and simulation-based
was performed and listed in Table 7. The TRNSYS simulations of the three optimal design solutions show similar results overall with
the metamodel prediction.
Besides, the comparison between the optimal solutions and the base case shows an important reduction in heating and cooling
loads. Indeed, the total annual load for the base case in kWh/m2.year is 124.25, 202.36 and 128.25 in Meknes, Ifrane and Marrakech,
respectively. For the solution B, the total annual load in kWh/m2.year is 39.26, 54.41, 42.86 in Meknes, Ifrane and Marrakech,
respectively. Therefore, a reduction of 68% of the total annual load was achieved in Meknes, 73% in Ifrane and 67% in Marrakech.
Finally, the annual CO2 emissions for the three optimal solutions and the base case are presented in Fig. 13. The lowest emissions
were obtained with solution B, except in Ifrane. Solution A, which minimises the heating load, shows better results in the cold climate
of Ifrane.
5. Conclusion
In this work, a multi-objective optimization was carried out to improve the energy efficiency of a typical building in three different
14
N. Abdou et al.
Table 7
Optimal design solutions and objective functions in each region.
Meknes A 0.30 0.40 0.53 0.38 × 0.50 24.84 19.41 × × Glz4 5.28 6.80 42.65 42.81
B 0.31 0.34 0.64 0.42 × 0.50 15.32 19.04 × × Glz3 14.04 16.66 20.10 22.60
15
C 0.66 0.20 0.68 0.32 × 0.56 10.55 17.27 × × Glz3 33.67 35.77 12.58 15.28
Ifrane A 0.32 0.63 0.59 0.60 × 0.50 31.48 18.60 × × Glz4 11.95 17.04 47.64 46.84
B 0.30 0.42 0.60 0.42 × 0.50 16.34 21.69 × × Glz3 30.75 36.56 16.25 17.85
C 0.56 0.20 0.63 0.36 × 0.50 11.83 15.20 × × Glz3 55.16 58.96 8.37 10.43
Marrakech A 0.38 0.65 0.51 0.43 × 0.50 26.40 25.78 × × Glz3 2.22 3.41 52.35 53.19
B 0.31 0.36 0.59 0.39 × 0.50 15.11 18.87 × × Glz3 5.30 6.64 33.35 36.22
C 0.63 0.20 0.65 0.32 × 0.50 10.19 16.90 × × Glz3 15.35 17.59 24.16 28.12
climates. The selected climatic zones are Meknes (zone 3), Ifrane (zone 4) and Marrakech (zone 5), where the climate is ranging from
cold to semi-arid. A metamodel approach was used, based on simulation method and data driving techniques using wrapper method
which takes into account the performance of a learner such as Artificial Neural Networks and Support vector machine. The results
showed that the best feature subsets depend on the location (i.e., weather conditions) and target output. SVM as a learner and search
algorithms like League championship algorithm (LCA), Discrete symbiotic organisms search (DSOS) algorithm, Particle swarm opti
mization (PSO) showed better results in terms of accuracy and reduction of feature input parameters. For ANN as a learner, hybrid
models based on Particle swarm optimization (PSO), Genetic algorithm (GA), Learning automata (LA), Forest optimization algorithm
(FOA) showed better performance. Moreover, the optimization results show some recommendations. According to the compromise
solution B, a value between the lower and average options is recommended for all considered input predictors. Therefore, a reduction
of 68% of the total annual load can be achieved in Meknes, 73% in Ifrane and 67% in Marrakech.
The findings of this paper may offer helpful recommendations for passive energy efficiency measures. The suggested method has a
strong potential to deal with complex building problems, and it can be applied as a tool for decision making. The main outcomes of the
study are as follows:
• The parameters chosen for an optimization process must be carefully and thoroughly evaluated. The sensitivity analysis showed
that the window to wall ratio, north and west, is not a key element and can be kept constant in the process. Thus, when choosing a
technical solution, the decision-maker should pay particular attention to elements that have a high potential for improving energy
efficiency, such as the windows type.
• The points on the Pareto front that are commonly considered as targets must be carefully analysed in order to find the optimum
solution. As demonstrated in this study by using carbon dioxide equivalent emissions as the objective function, heating and cooling
loads as sub-objective functions, the trade-off solution between the sub-objective functions is not always the best option.
• The hybrid models based on ANN and SVM can be used to forecast and optimize building energy loads, but the need to adjust the
neural architecture for ANN, to search for optimal hyperparameters for SVM, to search for optimal input parameters, increases the
computational cost.
As future work, other models based on new generation algorithms and objective functions, which take into account the impact of
climate change on buildings and inter-buildings, will be considered to improve the presented approach.
Data availability
16
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
References
[1] B. Chegari, M. Tabaa, E. Simeu, F. Moutaouakkil, H. Medromi, Multi-objective optimization of building energy performance and indoor thermal comfort by
combining artificial neural networks and metaheuristic algorithms, Energy Build 239 (2021), 110839, https://doi.org/10.1016/j.enbuild.2021.110839.
[2] K. Bamdad, M.E. Cholette, S. Omrani, J. Bell, Future energy-optimised buildings - addressing the impact of climate change on buildings, Energy Build 231
(2021), 110610, https://doi.org/10.1016/j.enbuild.2020.110610.
[3] F. Bre, V.D. Fachinotti, A computational multi-objective optimization method to improve energy efficiency and thermal comfort in dwellings, Energy Build 154
(2017) 283–294, https://doi.org/10.1016/j.enbuild.2017.08.002.
[4] S. Asadi, S.S. Amiri, M. Mottahedi, On the development of multi-linear regression analysis to assess energy consumption in the early stages of building design,
Energy Build 85 (2014) 246–255, https://doi.org/10.1016/j.enbuild.2014.07.096.
[5] D. Katunsky, A. Korjenic, J. Katunska, M. Lopusniak, S. Korjenic, S. Doroudiani, Analysis of thermal energy demand and saving in industrial buildings: a case
study in Slovakia, Build. Environ. 67 (2013) 138–146, https://doi.org/10.1016/j.buildenv.2013.05.014.
[6] P. Penna, A. Prada, F. Cappelletti, A. Gasparella, Multi-objectives optimization of energy efficiency measures in existing buildings, Energy Build 95 (2015)
57–69, https://doi.org/10.1016/j.enbuild.2014.11.003.
[7] Y. Zhai, Y. Wang, Y. Huang, X. Meng, A multi-objective optimization methodology for window design considering energy consumption, thermal environment
and visual performance, Renew. Energy 134 (2019) 1190–1199, https://doi.org/10.1016/j.renene.2018.09.024.
[8] Y. Schwartz, R. Raslan, D. Mumovic, Implementing multi objective genetic algorithm for life cycle carbon footprint and life cycle cost minimisation: a building
refurbishment case study, Energy 97 (2016) 58–68, https://doi.org/10.1016/j.energy.2015.11.056.
[9] K. Bamdad, M.E. Cholette, L. Guan, J. Bell, Ant colony algorithm for building energy optimisation problems and comparison with benchmark algorithms, Energy
Build 154 (2017) 404–414, https://doi.org/10.1016/j.enbuild.2017.08.071.
[10] N. Abdou, Y. EL Mghouchi, S. Hamdaoui, N. EL Asri, M. Mouqallid, Multi-objective optimization of passive energy efficiency measures for net-zero energy
building in Morocco, Build. Environ. 204 (2021), 108141, https://doi.org/10.1016/j.buildenv.2021.108141.
[11] Z. Romani, A. Draoui, F. Allard, Metamodeling the heating and cooling energy needs and simultaneous building envelope optimization for low energy building
design in Morocco, Energy Build 102 (2015) 139–148, https://doi.org/10.1016/j.enbuild.2015.04.014.
[12] E. Gengembre, B. Ladevie, O. Fudym, A. Thuillier, A Kriging constrained efficient global optimization approach applied to low-energy building design problems,
Inverse Probl. Sci. Eng. 20 (7) (2012) 1101–1114, https://doi.org/10.1080/17415977.2012.727084.
[13] S. Gou, V.M. Nik, J.L. Scartezzini, Q. Zhao, Z. Li, Passive design optimization of newly-built residential buildings in Shanghai for improving indoor thermal
comfort while reducing building energy demand, Energy Build 169 (2018) 484–506, https://doi.org/10.1016/j.enbuild.2017.09.095.
[14] M. Bourdeau, X. qiang Zhai, E. Nefzaoui, X. Guo, P. Chatellier, Modeling and forecasting building energy consumption: a review of data-driven techniques,
Sustain. Cities Soc. 48 (2019), 101533, https://doi.org/10.1016/j.scs.2019.101533. November 2018.
[15] W. Yu, B. Li, H. Jia, M. Zhang, D. Wang, Application of multi-objective genetic algorithm to optimize energy efficiency and thermal comfort in building design,
Energy Build 88 (2015) 135–143, https://doi.org/10.1016/j.enbuild.2014.11.063.
[16] J. Zhao, X. Liu, A hybrid method of dynamic cooling and heating load forecasting for office buildings based on artificial intelligence and regression analysis,
Energy Build 174 (2018) 293–308, https://doi.org/10.1016/j.enbuild.2018.06.050.
[17] F. Bre, N. Roman, V.D. Fachinotti, An efficient metamodel-based method to carry out multi-objective building performance optimizations, Energy Build 206
(2020), 109576, https://doi.org/10.1016/j.enbuild.2019.109576.
[18] A. Prada, A. Gasparella, P. Baggio, On the performance of meta-models in building design optimization, Appl. Energy 225 (April) (2018) 814–826, https://doi.
org/10.1016/j.apenergy.2018.04.129.
[19] R. Azari, S. Garshasbi, P. Amini, H. Rashed-Ali, Y. Mohammadi, Multi-objective optimization of building envelope design for life cycle environmental
performance, Energy Build 126 (2016) 524–534, https://doi.org/10.1016/j.enbuild.2016.05.054.
[20] K. Bamdad, M.E. Cholette, J. Bell, Building energy optimization using surrogate model and active sampling, J. Build. Perform. Simul. 13 (6) (2020) 760–776,
https://doi.org/10.1080/19401493.2020.1821094.
[21] A. González-Vidal, F. Jiménez, A.F. Gómez-Skarmeta, A methodology for energy multivariate time series forecasting in smart buildings based on feature
selection, Energy Build 196 (2019) 71–82, https://doi.org/10.1016/j.enbuild.2019.05.021.
[22] Z. Romani, A. Draoui, F. Allard, Metamodeling and multicriteria analysis for sustainable and passive residential building refurbishment: a case study of French
housing stock, Build. Simul., no. May (2021), https://doi.org/10.1007/s12273-021-0806-7.
[23] M. Ilbeigi, M. Ghomeishi, A. Dehghanbanadaki, Prediction and optimization of energy consumption in an office building using artificial neural network and a
genetic algorithm, Sustain. Cities Soc. 61 (May) (2020), 102325, https://doi.org/10.1016/j.scs.2020.102325.
[24] Y. Choi, D. Song, S. Yoon, J. Koo, Comparison of factorial and Latin hypercube sampling designs for meta-models of building heating and cooling loads, Energies
14 (2) (2021), https://doi.org/10.3390/en14020512.
[25] I. Jaffal, C. Inard, C. Ghiaus, Fast method to predict building heating demand based on the design of experiments, Energy Build 41 (6) (2009) 669–677, https://
doi.org/10.1016/j.enbuild.2009.01.006.
[26] T. Wei, A review of sensitivity analysis methods in building energy analysis, Renew. Sustain. Energy Rev. 20 (2013) 411–419, https://doi.org/10.1016/j.
rser.2012.12.014.
[27] Y. Masoudi-Sobhanzadeh, H. Motieghader, A. Masoudi-Nejad, FeatureSelect: a software for feature selection based on machine learning approaches, BMC
Bioinformatics 20 (1) (2019) 1–17, https://doi.org/10.1186/s12859-019-2754-0.
[28] A. Husseinzadeh Kashan, League Championship Algorithm (LCA): an algorithm for global optimization inspired by sport championships, Appl. Soft Comput. J.
16 (2014) 171–200, https://doi.org/10.1016/j.asoc.2013.12.005.
[29] Y. Masoudi-Sobhanzadeh, H. Motieghader, World Competitive Contests (WCC) algorithm: a novel intelligent optimization algorithm for biological and non-
biological problems, Informatics Med. Unlocked 3 (2016) 15–28, https://doi.org/10.1016/j.imu.2016.06.002.
[30] E. Atashpaz-Gargari, C. Lucas, Imperialist competitive algorithm: an algorithm for optimization inspired by imperialistic competition, in: 2007 IEEE Congr. Evol.
Comput. CEC 2007, 2007, pp. 4661–4667, https://doi.org/10.1109/CEC.2007.4425083.
[31] V.K. Patel, V.J. Savsani, Heat transfer search (HTS): a novel optimization algorithm, Inf. Sci. 324 (2015) 217–246, https://doi.org/10.1016/j.ins.2015.06.044.
[32] A.E.S. Ezugwu, A.O. Adewumi, Discrete symbiotic organisms search algorithm for travelling salesman problem, Expert Syst. Appl. 87 (2017) 70–78, https://doi.
org/10.1016/j.eswa.2017.06.007.
[33] M.Y. Cheng, D. Prayogo, Symbiotic Organisms Search: a new metaheuristic optimization algorithm, Comput. Struct. 139 (2014) 98–112, https://doi.org/
10.1016/j.compstruc.2014.03.007.
[34] R. Rajabioun, Cuckoo optimization algorithm, Appl. Soft Comput. J. 11 (8) (2011) 5508–5518, https://doi.org/10.1016/j.asoc.2011.05.008.
[35] M. Ghaemi, M.R. Feizi-Derakhshi, Forest optimization algorithm, Expert Syst. Appl. 41 (15) (2014) 6676–6687, https://doi.org/10.1016/j.eswa.2014.05.009.
[36] J.H. Holland, Genetic algorithms, Sci. Am. 267 (1) (1992) 66–73.
[37] R. Eberhart, J. Kennedy, A new optimizer using particle swarm theory, Proc. Int. Symp. Micro Mach. Hum. Sci. (1995) 39–43.
[38] M. Dorigo, M. Birattari, T. Stutzle, Ant colony optimisation, IEEE Comput Intell 1 (4) (2006) 28–39.
[39] M.R. Meybodi, H. Beigy, New learning automata based algorithms for adaptation of backpropagation algorithm parameters, Int. J. Neural Syst. 12 (1) (2002)
45–67.
[40] C. Lu, S. Li, Z. Lu, Building energy prediction using artificial neural networks: a literature survey, Energy Build 262 (2022), 111718, https://doi.org/10.1016/j.
enbuild.2021.111718.
[41] C. Cortes, V. Vapnik, Support-vector networks, Machine Learning 20 (3) (1995) 273–297.
[42] W. Gao, J. Alsarraf, H. Moayedi, A. Shahsavar, H. Nguyen, Comprehensive preference learning and feature validity for designing energy-efficient residential
buildings using machine learning paradigms, Appl. Soft Comput. J. 84 (2019), 105748, https://doi.org/10.1016/j.asoc.2019.105748.
17
N. Abdou et al. Journal of Building Engineering 61 (2022) 105332
[43] F. Sick, S. Schade, A. Mourtada, D. Uh, M. Grausam, Dynamic building simulations for the establishment of a Moroccan thermal regulation for buildings,
J. Green Build. 9 (1) (2014) 145–165, https://doi.org/10.3992/1943-4618-9.1.145.
[44] H. Mastouri, B. Benhamou, H. Hamdi, E. Mouyal, Thermal performance assessment of passive techniques integrated into a residential building in semi-arid
climate, Energy Build 143 (2017) 1–16, https://doi.org/10.1016/j.enbuild.2017.03.022.
[45] S. Soutullo, E. Ricardo, S.J. Cristina, F.J. Antonio, H. Rosario, Energy balances of four office buildings in different locations in Spain, Proc. IBPSA-Canada’s
Bienn. Conf. (2010).
[46] T. Niemelä, R. Kosonen, J. Jokisalo, Energy performance and environmental impact analysis of cost-optimal renovation solutions of large panel apartment
buildings in Finland, Sustain. Cities Soc. 32 (2017) 9–30, https://doi.org/10.1016/j.scs.2017.02.017.
[47] P. Murray, J. Marquant, M. Niffeler, G. Mavromatidis, K. Orehounig, Optimal transformation strategies for buildings, neighbourhoods and districts to reach CO2
emission reduction targets, Energy Build 207 (2020), 109569, https://doi.org/10.1016/j.enbuild.2019.109569.
[48] S.M. Garriga, M. Dabbagh, M. Krarti, Optimal Carbon-Neutral Retrofit of Residential Communities in Barcelona, Spain, vol. 208, Energy Build., 2020, https://
doi.org/10.1016/j.enbuild.2019.109651.
[49] J. Kneifel, Life-cycle carbon and cost analysis of energy efficiency measures in new commercial buildings, Energy Build 42 (3) (2010) 333–340, https://doi.org/
10.1016/j.enbuild.2009.09.011.
[50] Y. Huang, J.L. Niu, T.M. Chung, Energy and carbon emission payback analysis for energy-efficient retrofitting in buildings - overhang shading option, Energy
Build 44 (1) (2012) 94–103, https://doi.org/10.1016/j.enbuild.2011.10.027.
[51] U.Y.A. Tettey, A. Dodoo, L. Gustavsson, Effects of different insulation materials on primary energy and CO 2 emission of a multi-storey residential building,
Energy Build 82 (2014) 369–377, https://doi.org/10.1016/j.enbuild.2014.07.009.
[52] K. Li, L. Pan, W. Xue, H. Jiang, H. Mao, Multi-objective optimization for energy performance improvement of residential buildings: a comparative study,
Energies 10 (2) (2017), https://doi.org/10.3390/en10020245.
[53] Y. Lou, Y. Yang, Y. Ye, W. Zuo, J. Wang, The effect of building retrofit measures on CO2 emission reduction-A case study with U.S. medium office buildings,
Energy Build 253 (2021), 111514, https://doi.org/10.1016/j.enbuild.2021.111514.
[54] M. Krarti, P. Ihm, Evaluation of net-zero energy residential buildings in the MENA region, Sustain. Cities Soc. 22 (2016) 116–125, https://doi.org/10.1016/j.
scs.2016.02.007.
[55] S. Hamdaoui, M. Mahdaoui, A. Allouhi, R. El Alaiji, T. Kousksou, A. El Bouardi, Energy demand and environmental impact of various construction scenarios of
an office building in Morocco, J. Clean. Prod. 188 (2018) 113–124, https://doi.org/10.1016/j.jclepro.2018.03.298.
[56] S. Timmerberg, A. Sanna, M. Kaltschmitt, M. Finkbeiner, Renewable electricity targets in selected MENA countries - assessment of available resources,
generation costs and GHG emissions, Energy Reports 5 (2019) 1470–1487, https://doi.org/10.1016/j.egyr.2019.10.003.
[57] A.T. Nguyen, S. Reiter, A performance comparison of sensitivity analysis methods for building energy models, Build. Simul. 8 (6) (2015) 651–664, https://doi.
org/10.1007/s12273-015-0245-4.
[58] L. Zhu, et al., Uncertainty and sensitivity analysis of cooling and heating loads for building energy planning, J. Build. Eng. 45 (2022), 103440, https://doi.org/
10.1016/j.jobe.2021.103440. October 2021.
[59] V. Zeferina, F.R. Wood, R. Edwards, W. Tian, Sensitivity analysis of cooling demand applied to a large office building, Energy Build 235 (2021), 110703, https://
doi.org/10.1016/j.enbuild.2020.110703.
[60] N. Abdou, Y. El Mghouchi, S. Hamdaoui, M. Mhamed, Optimal building envelope design and renewable energy systems size for net-zero energy building in
tetouan (Morocco), in: 2021 9th International Renewable and Sustainable Energy Conference (IRSEC), IEEE, 2021, pp. 1–6, https://doi.org/10.1109/
IRSEC53969.2021.9741188.
[61] B. Manrique Delgado, et al., Lifecycle cost and CO2 emissions of residential heat and electricity prosumers in Finland and The Netherlands, Energy Convers.
Manag. 160 (2018) 495–508, https://doi.org/10.1016/j.enconman.2018.01.069, 2018.
18