Study of the Thermal Comfort of a Building that does not Comply with Construction Standards in Madagascar : Experimentation and Simulation with OMEdit

A very large number of buildings in developing countries are far from complying with the standards of housing. This paper presents the subjective study of the thermal comfort of a building that does not comply with construction standards or thermal regulations, located in Madagascar. Modeling was done using the Modelica tool, especially its BuildSysPro library. In order to minimize the inaccuracies, a step of an experimental adjustment of the developed numerical model was also carried out usingexperimental reference data that were obtained from the temperaturemeasurementsof the studied building elements as well as the wind speed and the received solar radiation flux. It was found that despite the obvious non-compliance with building standards and thermal regulations, the building has an acceptable thermal environment vis-à-vis its occupant.


I. INTRODUCTION
Dynamic modeling of buildings is currently gaining importance in the field of research and has also become a dominant, attractive subject treated by many authors [1][2][3][4][5][6][7][8][9][10][11].As in many other disciplines (mechanics, electronics, electricity, fluids, thermodynamics, ...), there is a choice between special simulators for each discipline or more general dynamic modeling techniques, a number of them are presented in [12].
Several simulation tools specialized in the simulation of the thermo-aerodynamic behavior of buildings have been developed.
CodyMa [13], which is a calculation code for thermal, aerodynamic and moisture simulation of multi-zone buildings, has a multi-model reception structure, allowing different levels of finesse.Several research works have been undertaken to contribute to the validation of this code; the authors are particularly interested in its global validation, including the verification phase, by inter-software comparison using the BESTEST procedure [14], the experimental validation phase.
CoDyBa is defined as a tool for calculating building heating and cooling consumption and is optimized for a limited number of thermal zones [15,16].In addition, it does not propose a model of ventilation with heat recovery [17].The modular architecture computing core uses the modal reduction of the physical problem, making it very fast Published on April 28, 2019.J. Ratsimbazafiharivola and H. T. Rakotondramiarana are with Institute for the Management of Energy (IME), University of Antananarivo, P.O.Box 566, Antananarivo 101, Madagascar.(e-mail: anpproba@ gmail.com).and evolved to Kozibu [18] which retained the CoDyBa resolution methods but was coded using a class hierarchy for representing the building.
CodyRun [19] is a computing code exclusively dedicated to the thermal behavior of dwellings.It solves a system describing thermal and aerodynamic phenomena for modeling multi-zone buildings [20,21] while integrating building thermal, natural ventilation, moisture and pollutant transfers.More precisely, the building is modeled as an assembly of components, from the elements of its walls to the air conditioning systems [21].While its user interface being designed for the Microsoft Windows environment, its user has a large choice of heat transfer models and can use multizone models [22].
EnergyPlus [23] is a freely available tool based on two simulation tools BLAST and DOE-2 [24].It allows the simulation of complex energy systems; its library of models is important but rather adapted to the simulation of existing systems in the United States.It includes the resolution with a time step being less than one hour.For modeling building in 3D easily, a plugin was developed in Google Sketchup for the geometrical seizure of the building.Its interface is not user-friendly but the Design Builder tool [25] integrates a user-friendly interface while using the robustness of the algorithms developed by EnergyPlus, which makes it relevant and effective for energy optimization of buildings without guaranteeing an excellent accuracy level of temperature in buildings in the case of free evolution.
TRNSYS [26] is a modular and highly flexible environment that includes a graphical interface and a component library to model the energy systems applied to buildings.While being able to describe the transient behavior of thermal systems, TRNSYS consists of two elements [17].On the one hand, it has a computing core that allows you working with imported data while solving systems, that is, integrating meteorological data or various time-dependent functions particularly.On the other hand, it has a large library of modules including many elements of thermal or electrical energy systems where it is possible to modify or create components, described by Fortran procedures [17], [22].It integrates the ability to create its own components and can be coupled with other tools like COMIS, CONTAM, EES, Excel, FLUENT [27], GenOpt and Matlab [28].While having functionality similar to EnergyPlus [23], it is interesting by its modular aspect which facilitates the development of specific models.Its modular structure computing core which can integrate any component from its component library or created by the user is the strength of this software.

Study of the Thermal Comfort of a Building that does not
Comply with Construction Standards in Madagascar: Experimentation and Simulation with OMEdit Jaurès Ratsimbazafiharivola, H. T. Rakotondramiarana PHPP [29] is software that uses the Excel tool in which the data of the building is informed.It uses a static calculation method relating to a monthly energy balance based on European standards, in particular EN 13829.It allows obtaining the heating and air-conditioning needs, the power of heating and cooling, and the need in primary energy as well as a trend for the number of hours of discomfort in summer.This tool is mainly used in Germany; it is used as a basis for the award of the PassivHaus label and only concerns the realization of passive house.There are no dynamic simulations carried out and the results are derived from statistical data on buildings of the same type.
ESP-r, presented by Bartak et al. [30], is an energy modeling tool for simulating the thermal and acoustic performance of buildings, energy use and greenhouse gas emissions, all associated with environmental control systems.The studied system is thus equipped to model the flows of heat, air, moisture and consumed energy.The building is represented as a set of nodes representing rooms or connections with air treatment systems (such as heating for example) and connections between certain nodes modeling elements such as doors, windows, ventilation or infiltrations [30].More precisely, a numerical model of fluid dynamics is integrated in ESP-r.Its results are compared with experimental data obtained in a test room of which internal and external boundary conditions are controlled.Bartak et al. [30] cited experimental inaccuracies, measurement uncertainties in particular with respect to the test room volume and the non-instantaneous average speeds in the CFD (Computational Fluid Dynamics) modeling.The obtained simulation results are nevertheless in agreement with measurement results.Except for points close to the ground, the differences between simulated temperatures and measured temperatures are not greater than 0.5 °C, which corresponds to the measurement uncertainties.
The approximations made as well as the modeling and simulation methods used by some of the aforementioned tools are rarely exposed.As a result, access to more detailed information by users is extremely limited.In contrast, Modelica [31] is a general purpose object-oriented language for modeling and simulation of physical systems [32], the evolution of which is described in [33].
In this work, modeling and simulation of a building was carried out using the Modelica tools for which the BuildSysPro [34] library was used.This present study is also concerned with the analysis of the thermal comfort of a typical building located in developing countries.Modelica tools provide additional capabilities for various applications, such as the design of advanced control systems and a wide range of specialized libraries.As a result, the OMEdit (OpenModelica Editor) environment [35] was used for basic modeling.A parameter adjustment of the building model [36] was also performed in order to obtain a model with more reliability.
The types of buildings this work focuses on are buildings that do not meet building standards.Indeed, these buildings are for the majority a real case of dwelling in particular in the developing countries, with humid tropical climate.

A. The study area Located between 20° S and 47°00 E, Madagascar is almost entirely inside the tropics. It is an island in the Indian
Ocean with an area of approximately 592 000 km².It is the fourth largest island in the world and is separated from Africa by the Mozambique Channel, located about 400 km.A mountain spine located between 1200 m and 1500 m crosses the island from north to south along its entire length.This geographical relief, the maritime influence and the wind conditions are at the origin of the very varied climatic conditions encountered on this island.
The average annual temperature is between 14°C and 27.5°C.In the coastal areas, it depends on the latitude and ranges between 23°C in the South and 27°C in the North [37].The West coast is warmer than the East one (1°C to 3°C).On the uplands, the average annual temperature ranges from 14°C to 22°C.The average temperature reaches its minimum in July in the whole country; the maximum one occurs in January and February for most areas, with the exception of a few places in the highlands and the Northwestern region, where it is observed in November.
The measurement campaigns during the experiment were carried out on a building located in Antananarivo (Figure 1).

B. Climate data
Climatic classification is extremely useful for designing buildings with thermal comfort [38].
Thus, daily outdoor data (temperature, precipitation, wind speed, relative humidity, dew point temperature, etc.) during the year 2018 were collected from many weather stations.The different data were measured at a height ranging from 3mto 10m and at a frequency of 60 minutes.
Indeed, these data will be taken as parameters in the building model to be implemented under OMEdit [35].
In Figure 2 are shown some samples of meteorological data of Antananarivo during the year 2018.
Figure 2.a shows the hourly change of the outside temperature as well as the dew point temperature.

C. Experimental Protocols
In order to carry out the measurements, it is essential to install in this experimental system the devices listed below.
Figure 3 shows a simplified schematic representation of the methods for measuring the respective temperatures of the walls and the indoor environment of the studied building.Digital temperature sensors in the form of a probe were used to measure surface temperatures.These sensors were plugged into an Arduino Uno board that was connected to a computer.Two libraries named One-wire.hand DallasTemperature.hwere used to ensure the communication between the probes and the arduino.
In addition, since several temperature sensors have been used, a test plate and male-male wires shown in Figure 4 were used for circuit mounting.
While being equipped with an anemometer, a thermometer and a hygrometer, the Testo 410-2 which is powered by two alkaline batteries (Zn-Mn) of 1.5 Volts allowed us performing manual measurement of the internal humidity of the cell.
An SP-Lite (Silicon Pyranometer) pyrometer from Kypp&Zonen allows the measurement of the global solar radiation arriving around the studied cell.It is connected by means of a USB 2.0 A to B cable for the printer, to a USBdevice from National Instrument.The NI USB-6008 is a terminal block that has eight unbalanced analog input (AI) channels, two analog output (AO) channels, twelve digital input / output channels, a 32-bit counter, and full speed USB interface.
For the data storage as well as the data processing, two computers were used: one is a desktop computer used to store in the Excel the solar irradiation data coming from the Pyranometer; the other is a laptop, which is connected with the Arduino, allowing acquiring the temperatures collected by the sensors.
The instruments used in the experiments are shown in Figure 4.

D. Implementation of the model 1) Model development
The Modelica model development for buildings that are subject to measurement campaigns is based on the use of the BuildSysPro [34] library in OMEdit [35].The digital model of the building envelope is shown in Figure 5.

2) Thermal comfort model
Adaptive comfort is a recent model of thermal comfort that determines comfort temperatures in an environment where there are temperature changes.
Several research works agree on a formula of type involving the outside temperature: According to De Dear et al. [39], one can estimate as being equal to 0.31 whereasbeing equal to 17.8.
As for the operative temperature [40], it corresponds to the temperature that is felt by the occupant and takes into account the air temperature in the zone of occupancy and the radiation effects.
Mathematically, it is given by: in which, Tmp denotes the weighted average temperature of the room surfaces while Ta is the air temperature.Equation ( 2) can be used for air velocities below 0.2 m.s -1 (valid hypothesis for the inside of a building).

III. DESCRIPTION OF THE STUDIED BUILDING
The building subject of our experiment is located in Antananarivo (Latitude: 47.51667 -Longitude: -18.91667) -Madagascar.The latter has a humid tropical climate and is influenced by several types of wind [38].Tropical cyclones and their consequences represent a great threat every year [41,42].

A. Architectural Properties
The architectural properties of the studied building as well as the thermophysical properties of the building materials used are grouped together in Tables 1 and 2.

B. Thermophysical properties
Table 2 presents some thermophysical properties of the materials used for the construction of the studied building.

IV. SIMULATION RESULTS
For the simulation, the same boundary conditions during the experiment were adopted to the models in BuildSysPro [34] under OMEdit [35].The thermophysical properties of various building materials were introduced under OMedit during the modeling of the studied building.
Figure 7 shows the simulation results of the studied building as modeled via BuildSysPro [34] in OpenModélica [35] while taking as parameters the year 2018 meteorological data.More precisely, Figures 7.a   The temperatures of the studied building elements obtained by simulation for the period between 1stand 5th December 2018 are extracted from the results presented in Figure 7 for comparison with those obtained experimentally during the same period and at the same location, as presented in Figure 8. Experimental temperature values of some elements of the studied building, measured from 1st to 5th December 2018, are presented in Figure 9.The temperature changes of the ceiling (red dotted line) exceeds by some degrees that of the other elements especially around noon.The floor temperature change remains the lowest, on average, among all the building components.
Figure 10 shows the change of the local wind speed whose direction has a North-North-East trend (≈22.5 °), and the outdoor ambient temperature as well as the solar radiation, measured at the studied building site in Antananarivo from 1st to 5th December 2018.The effects of the solar radiation can easily be seen from the appearance of the temperature measurement results presented in Figures 8  and 9.The contribution of the solar radiation (Fig. 10) to the indoor air temperature (Fig. 9) can be observed.Indeed, the indoor air shows more amplitude variations and a higher average value when the solar radiation is present.
Figure 11 shows the comparison of the temperature values of the different building elements obtained by simulation and experimentation.It follows from curve comparison results that the obtained simulation and experimental curves are in agreement moderately.
Indeed, the values of the coefficient of determination R 2 related to this theroretical versus experimental result comparison are greater than 0.75 such that it is respectively equal to 0.8204 for the indoor air (Fig. 11.a), to 0.7743 for the floor (Fig. 11.b), to 0.8868 for the ceiling, and to 0.8104 for the facades.
Although, for each studied building element, the model used under OMEdit may be considered acceptable, it is preferable to carry out a method of model readjustment as previously stated.The origin of the differences is assumed to come from the modeling hypothesis applied to the Modelica tool, the lack of accuracy of some simulation parameters as well as the uncertainties of the experimental measurements.As a result, deviations from Modelica results compared to the experimental ones range from 0 to 9.964% (see Table 3).Figure 13 shows the comparison of the temperature values of the different building elements obtained by simulation and experimentation after having carried out the parameter adjustment method.More precisely, the recalibration consisted in the local correction of the numerical model by looking at some previously identified parameters that contain dominant modeling defects.
The adjusted parameters are the airflow rate in the model's ventilation module and the sky vault temperature.
Indeed, the volume of air entering and leaving the room strongly influences the value of the indoor air temperature and that of the various elements of the studied building.Besides, the values of the sky vault temperature were derived from an estimate, from a computation and not from an accurate experimental measurement.
It was found that the different curves that were respectively obtained by simulation and experimentation deviate very little from each other.Their evolutions in time agree and it can then be asserted that these curves are in perfect agreement.More precisely, the values of the coefficient of determination R 2 are close to 1 while being respectively equal to 0.9896 for the indoor air temperature (Fig. 13.a), to 0.9829 for the floor temperature (Fig. 13.b), to 0.9942 for the ceiling temperature (Fig. 13.c), and to 0.9896 for the facade temperature (Fig. 13.d).
Therefore, for each building element, the model used under OMEdit perfectly predicts the thermal behavior of the building under study.While being the temperature felt by the occupant of the building, the operative temperature is calculated using equation (2).In fact, the average wind speed inside the building does not exceed 0.2 m.s -1 .Figure 14 compares the operative temperatures respectively obtained from experimental measurements and from simulations.Figure 15 confronts the operative temperature obtained using the adjusted model to the comfort temperature from the work of De Dear et al [39].This confrontation allows evaluating the comfort of the building occupant during both diurnal and nocturnal periods.
Indeed, in this Figure 15, we notice that, generally, the deviation of the temperature felt by the occupant from the comfort temperature digs a little in diurnal period.While in nocturnal period, the difference between the aforementioned two temperatures decreases slightly.
Thermal comfort is therefore better approached in the daytime than at night.This can be explained by the fact that during the diurnal period, the air renewal by natural ventilation induces a change of the air volume of the studied building.By cons, all openings are kept closed during nocturnal period, the air change rate is then limited, and the indoor environment is thus stable.V. CONCLUSION This work tackled the thermal comfort study of a building that does not respect building standards and thermal regulations.These types of building are quite common in developing countries, particularly in Madagascar, and constitute a real case of housing for the majority of the population.A comparison between simulation results and those obtained experimentally from the studied building was presented.The first result comes from the implementation of a model using an object-oriented modeling tool based on the Modelica language.For the experiment, a measurement campaign was carried out in Antananarivo Madagascar.The results of the simulation are also presented for a full year.Comparison simulation results vs experimental ones was made in accordance with the dates and the duration of the experimental campaign.
After comparison of the temperature values of the studied building various elements, obtained by simulation under OMEdit and by experimentation, a step of resetting the numerical model by adjusting its parameters was performed.The adjusted model was used to study and subjectively evaluate the comfort of the building occupants.Indeed, according to the work of De Dear et al [39], there is a correlation between the average indoor temperature over a period and the comfort temperature, which is probably the feedback result of the occupant behavior in relation to the indoor temperature.
During the thermal comfort study in the experimental building, a fit between the recalibrated model and the experimental data was found.It can be concluded that the model developed and recaled under OMEdit allowed the analysis of thermal comfort.It can also be noted that, despite the obvious non-compliance with building standards and thermal regulations, the building has an acceptable thermal environment for its occupants, whether by day or by night.

Fig. 1 .
Fig. 1.Map and location of the study area: Antananarivo Madagascar.

Figure 2 .
b presents the hourly change of the relative humidity of the air at 2m above the ground surface.Figures 2.c and 2.d respectively depict the hourly changes of the air relative humidity at 2m above the ground surface as well as the wind speed during the year 2018.

Fig. 3 .
Fig.3.Simplified schematic representation of the configuration and experimental layout adopted for temperature measurement.

Fig. 4 .
Fig.4.Instruments used in the experimental protocol

Figure 6
Figure 6 depicts the graphic aspect under SweetHome3d (Fig. 6.a) as well as the orientation of the studied building (Fig. 6.b).
and 7.b respectively present the temperature prediction of the indoor air and the floor of the studied building.
Figure 7.c shows comparison between the roof and the ceiling predicted temperatures while Figure 7.d compares temperature curve of the facade and that of the window.It was found that climatic parameters inputted in arguments are very influential on the trend and the value of the obtained results.

Fig. 7 .
Fig.7.Evolution of the temperatures of the studied building components during a simulation year under OMEdit (Antananarivo 2018).

Fig. 8 .
Fig. 8. Temperature change of the studied building components (interior facade, ceiling, floor, indoor air) under OMEdit for a simulation related to5 day data (from 1st to 5th December 2018).

Fig. 9 .Fig. 10 .
Fig. 9. Temperature changes of the studied building elements (black dotted line for indoor air, blue dotted line for walls, red dotted line for the ceiling, green dotted line for the floor, and gray dashedline for the outdoor air) as measured from 1st to 5th December 2018.

Fig. 11 .
Fig. 11.Comparison of temperature curves of the studied building elements obtained by simulation under OMEdit and by experimentation from 1st to 05th December 2018 in Antananarivo.

Figure 12
Figure 12 compares the indoor air temperature curves obtained by experimental measurement (black solid line) and simulation (dotted line), respectively.

Fig. 12 .
Fig. 12. Visual identification of the difference between the measured points and the corresponding predicted curve for the indoor temperature case.

Fig. 13 .
Fig. 13.Values of the coefficient of determination R 2 of the studied building element temperatures after readjustment the numerical model.

Fig. 14 .
Fig. 14.Comparison of operative temperatures respectively obtained by experimental measurements and simulation: original model (black dotted line) vs calibrated model (red dotted line)

Fig. 15 .
Fig. 15.Comparison of the operative temperature obtained using the adjusted model (black solid line) and the comfort temperature from the work of De Dear et al [39] (black dotted line)

TABLE 1 :
GEOMETRIC CHARACTERISTICS OF THE BUILDING