Energy Saving and Thermal Comfort Performance of Passive Retroﬁtting Measures for Traditional Rammed Earth House in Lingnan, China

: The traditional rammed earth houses sharing similar patterns in the Lingnan region, south China, and distributed in rectangular arrays, are gradually losing their vitality and becoming uninhabited under modern living conditions. This research examined a typical pattern called the “Four-point gold” house and analyzed the suitability of different retroﬁtting technologies by ﬁeld measurements and building simulation. To optimize energy consumption, indoor thermal comfort, and the corresponding economic performance of the retroﬁtting measures for the prototypical house, ﬁve measures, including wall insulation, reﬂective roof coating, carpet, sunshade


Introduction
Rammed earth refers to a compacted mixture of mud blocks that are strong, dense, and have few gaps. It is used as a construction material and has very strong plasticity [1]. Since the 1970s, different groups around the world have begun to conduct scientific research on the properties of rammed earth [2]. Compared with conventional building materials, rammed earth is low-cost, readily available, recyclable, and considered environmentally friendly and sustainable [3], especially when the material is locally available [4,5]. Rammed earth building has good thermal performance under specific climatic and technical conditions [6]. This is due to its high thermal inertia [7] and good thermal insulation performance [8]. Rammed earth buildings do not need too much external energy input throughout the year [9], and indoor thermal comfort can be achieved during most of the time of the year [3,7]. In addition, it has been shown that due to its good bioclimatic properties, an earth envelope can lead to a reduction in energy used for heating, cooling, and humidity control [10]. In general, the quality of the indoor thermal environment in summer can be acceptable [11,12]. The comfortable time can be extended when other design measures, such as insulation, double glazing, shading, and ventilation, are taken into account [11,13]. Small and large window shadings lead to lower energy loads in the torrid-arid area [14]. Moreover, the results of the research show that a combination of large design measures, such as insulation, double glazing, shading, and ventilation, are taken into account [11,13]. Small and large window shadings lead to lower energy loads in the torrid-arid area [14]. Moreover, the results of the research show that a combination of large compacted earth walls, nighttime cross ventilation, and overhangs can effectively reduce the energy demand for space cooling [15]. However, thermal comfort in winter cannot be fully achieved without heating [12,13]. Thus, it is possible to archive acceptable comfort conditions by passive means [16][17][18][19], which contributes to reducing the heating/cooling energy demand of the traditional rammed earth buildings [20][21][22].
What is more, historical and traditional rammed earth houses in Lingnan account for a large part of the existing building stock, and they are quite distributed in the region inhabited by the clans in the Chaoshan area.. There are mainly two styles of "climbing lion" and "Four-point gold", which are generally distributed in a rectangular array [23]. As shown in Figure 1, the "Climbing lion" house looks like a lion poised to climb. The gate indicates its mouth, the two small rooms are the front claws, the hall is the belly, and the two other bedrooms on the sides of the hall are the rear claws. "Four-point gold" has two extra rooms near the gate compared with the "Climbing lion" house. The shape is like the Chinese character '金' (gold) and can be seen in Figure 2. Therefore, the architectural styles in traditional villages are the same and arranged in the same direction, and "cold lanes" are formed between buildings [24,25]. Research has shown that the residents in Tulou (earth building) had better satisfaction with the indoor thermal environment and overall indoor environmental quality than residents in normal rural buildings [26]. Building integrated photovoltaic was employed to optimize the solar energy application on the rooftop of traditional Lingnan buildings [27]. However, there are few studies on the indoor thermal environment and energy-saving renovation of rammed earth houses in Lingnan, and there are no rules, norms, or standards for reference [28]. The industrialization of the construction industry (materials, systems, etc.) led to the standardization of buildings. The performance of traditional buildings can gradually deteriorate over time. As a consequence, they became more dependent on heating, ventilation, and air conditioning (HVAC) systems to cater to thermal comfort requirements, which resulted in energy use that was not a necessity in previous vernacular buildings. This situation led to the disregard for the vernacular wisdom in passive design, changing not only the way buildings and indoor spaces were designed (commonly arranged by function and thermal needs) but also the occupants' living habits [29]. Coupled with the outflow of the rural population in China due to the urbanization process, many traditional rammed earth houses are now uninhabited. Traditional buildings are gradually abandoned by the descendants of their original residents in the process of modernization. The industrialization of the construction industry (materials, systems, etc.) led to the standardization of buildings. The performance of traditional buildings can gradually deteriorate over time. As a consequence, they became more dependent on heating, ventilation, and air conditioning (HVAC) systems to cater to thermal comfort requirements, which resulted in energy use that was not a necessity in previous vernacular buildings. This situation led to the disregard for the vernacular wisdom in passive design, changing not only the way buildings and indoor spaces were designed (commonly arranged by function and thermal needs) but also the occupants' living habits [29]. Coupled with the outflow of the rural population in China due to the urbanization process, many traditional rammed earth houses are now uninhabited. Traditional buildings are gradually abandoned by the descendants of their original residents in the process of modernization.
Therefore, the investigation of Lingnan rammed earth houses is not only conducive to the protection and inheritance of historical buildings but also of great significance to the sustainable development of society. In this study, the indoor and outdoor thermal environment parameters of the traditional rammed earth house "Four-point gold" in Lingnan were tested. The building model is established using DesignBuilder and calibrated by onsite measured data. Then the thermal performance of the rammed earth house is analyzed, and the passive energy-saving measures are put forward. Finally, the corresponding optimal scheme for each optimization objective and the comprehensively optimal scheme are proposed by the orthogonal experiment method. This research helps to better understand the characteristics of rammed earth houses in the Lingnan region and provides technical support for energy saving in the residential sector. It is anticipated that the renovation measures can be applied to not only the prototypical building in this research but also to the rammed earth buildings in the region so that these traditional houses in Lingnan will have the chance to rejuvenate in terms of modern living conditions and continue to maintain vitality in the future. Therefore, the investigation of Lingnan rammed earth houses is not only conducive to the protection and inheritance of historical buildings but also of great significance to the sustainable development of society. In this study, the indoor and outdoor thermal environment parameters of the traditional rammed earth house "Four-point gold" in Lingnan were tested. The building model is established using DesignBuilder and calibrated by onsite measured data. Then the thermal performance of the rammed earth house is analyzed, and the passive energy-saving measures are put forward. Finally, the corresponding optimal scheme for each optimization objective and the comprehensively optimal scheme are proposed by the orthogonal experiment method. This research helps to better understand the characteristics of rammed earth houses in the Lingnan region and provides technical support for energy saving in the residential sector. It is anticipated that the renovation measures can be applied to not only the prototypical building in this research but also to the rammed earth buildings in the region so that these traditional houses in Lingnan will have the chance to rejuvenate in terms of modern living conditions and continue to maintain vitality in the future.

Field Survey
This study selected a house in Jieyang, Guangdong Province, located in the north of Xinzhizhai. There are 14 "climbing lion" and 13 "Four-point gold" houses in this village. The "Four-point gold" houses are the target prototype in this research. Its floor plan and locations are shown in Figure 2. The house retains the traditional Lingnan architectural styles. The archaic rammed earth technique of Chaoshan is adopted, which is made of red mud, coarse sand, and other materials. The top of the wall directly supports the wooden beams, and the rooftop is covered with tiles. The overhanging eaves in front of the hall are undertaken by stone pillars in front of the courtyard. There are walls on the periphery of the building. Behind the entrance door, there lays the front hall, and the rooms on the left and right side are the front rooms (bedrooms). Further ahead is a courtyard, and the rooms on both sides are the kitchen and utility room. On the other side of the courtyard, a multifunctional space with a large room on each side (bedroom) is designed. The large room is divided into two floors by a wooden partition, and the attics are used for storage. Influenced by the traditional concept of "withholding wealth", "Four-point gold" generally does not open windows to the outside but only openings to the inner courtyard. The house only has external windows on the east and west sides, which is limited by the strength of the rammed earth walls. The window-to-wall ratios of the east and west external walls of the bedrooms are 0.017 and 0.011, according to onsite measurements.

Onsite Test
Field measurements and investigations were conducted from 2-9 March in winter and from 12-30 July in summer in 2019. The indoor temperature, relative humidity, and black bulb temperature were tested by HOBO temperature and humidity and the global temperature meter. The detailed parameters of the instruments are shown in Table 1. To ensure the accuracy of the tested data, the room was not occupied during the measurement period. Doors and windows are closed to test the condition without natural ventilation. The thermal environment of the utility room and hall was measured and monitored, as shown in Figure 3. locations are shown in Figure 2. The house retains the traditional Lingnan architectural styles. The archaic rammed earth technique of Chaoshan is adopted, which is made of red mud, coarse sand, and other materials. The top of the wall directly supports the wooden beams, and the rooftop is covered with tiles. The overhanging eaves in front of the hall are undertaken by stone pillars in front of the courtyard. There are walls on the periphery of the building. Behind the entrance door, there lays the front hall, and the rooms on the left and right side are the front rooms (bedrooms). Further ahead is a courtyard, and the rooms on both sides are the kitchen and utility room. On the other side of the courtyard, a multifunctional space with a large room on each side (bedroom) is designed. The large room is divided into two floors by a wooden partition, and the attics are used for storage. Influenced by the traditional concept of "withholding wealth", "Four-point gold" generally does not open windows to the outside but only openings to the inner courtyard. The house only has external windows on the east and west sides, which is limited by the strength of the rammed earth walls. The window-to-wall ratios of the east and west external walls of the bedrooms are 0.017 and 0.011, according to onsite measurements.

Onsite Test
Field measurements and investigations were conducted from 2-9 March in winter and from 12-30 July in summer in 2019. The indoor temperature, relative humidity, and black bulb temperature were tested by HOBO temperature and humidity and the global temperature meter. The detailed parameters of the instruments are shown in Table 1. To ensure the accuracy of the tested data, the room was not occupied during the measurement period. Doors and windows are closed to test the condition without natural ventilation. The thermal environment of the utility room and hall was measured and monitored, as shown in Figure 3.

Building Simulation Model
The indoor environment and energy-saving effect of various retrofitting measures were simulated by DesignBuilder, which uses EnergyPlus as the simulation engine and energy use analysis. According to the onsite investigation, the geometric information of the house was collected and fed into DesignBuilder. As shown in Figure 4, the building

Building Simulation Model
The indoor environment and energy-saving effect of various retrofitting measures were simulated by DesignBuilder, which uses EnergyPlus as the simulation engine and energy use analysis. According to the onsite investigation, the geometric information of the house was collected and fed into DesignBuilder. As shown in Figure 4, the building has two stories with wide eaves, which can effectively shade and ensure ventilation and daylight in the courtyard. We manually tuned the model to fit the simulation results to the onsite measurements as model calibration. The calibrated thermophysical properties of building walls, roofs, and floors are shown in Table 2. The thermal resistance of the rammed earth wall is 0.57 m 2 ·K/W, which is higher than that of the tile roof and concrete floor. This research only considers the basic structure of the building due to its simple interior furnishings. There is no larger appliance, so the model does not account for heat loads other than lamps. An occupancy intensity of two people per bedroom is assumed, and the bedrooms are not used during 9:00-22:00. Due to the aging of the house and poor airtightness of doors and windows, the air infiltration of the house is calibrated to be 3.5 h −1 . Currently, the house is not equipped with an HVAC system. We propose a simple packaged terminal heat pump system for the building to meet the requirement of thermal comfort in the simulation. The parameters of the HVAC system are listed in Table 3. The cooling temperature is 25 • C, and the heating temperature is 18 • C. Location coordinates and meteorological parameters are also input parameters.
loads other than lamps. An occupancy intensity of two people per bedroom is assumed, and the bedrooms are not used during 9:00-22:00. Due to the aging of the house and poor airtightness of doors and windows, the air infiltration of the house is calibrated to be 3.5 h-1. Currently, the house is not equipped with an HVAC system. We propose a simple packaged terminal heat pump system for the building to meet the requirement of thermal comfort in the simulation. The parameters of the HVAC system are listed in Table 3. The cooling temperature is 25 °C, and the heating temperature is 18 °C. Location coordinates and meteorological parameters are also input parameters.
Retrofitting measures could be put forward through the analysis of original building's heat transfer and indoor thermal environment. The corresponding thermophysical parameters of the measures are new inputs to simulate energy consumption and thermal comfort. The best retrofitting scheme can be found according to the characteristics of low HVAC system energy consumption, long comfortable period, and short dynamic investment payback period. Orthogonal experimental method is adopted to select suitable scheme with evenly dispersed study cases. Each average of sensitivity index (K jm ) is the average experimental index of the corresponding cases with same retrofitting measure (factor j) and level (m), which can evaluate the performance of the measures. Maximum difference (R j ) of retrofitting measures is the difference between maximum and minimum K jm . Higher R j indicates more sensitivity of the specific retrofitting measure.    Retrofitting measures could be put forward through the analysis of original building's heat transfer and indoor thermal environment. The corresponding thermophysical parameters of the measures are new inputs to simulate energy consumption and thermal comfort. The best retrofitting scheme can be found according to the characteristics of low HVAC system energy consumption, long comfortable period, and short dynamic investment payback period. Orthogonal experimental method is adopted to select suitable scheme with evenly dispersed study cases. Each average of sensitivity index (K jm ) is the average experimental index of the corresponding cases with same retrofitting measure (factor j) and level (m), which can evaluate the performance of the measures. Maximum difference (R j ) of retrofitting measures is the difference between maximum and minimum K jm . Higher R j indicates more sensitivity of the specific retrofitting measure.

Dynamic Investment Payback Period
For building energy-saving retrofitting, economic evaluation is important to the feasibility of the renovation project. The payback period, incremental cost, and net present value are the usually used indicators of economic analysis. Incremental cost refers to the direct and indirect costs caused by transformation, including materials, construction measures, machinery, and labor costs. The dynamic payback period is used for economic analysis, as shown in Equations (1) and (2). The discount rate is 8% when that is calculated [30].
where NPV refers to the net present value of year t, n refers to the calculation cycle, r refers to the discount rate, C t refers to cash flow in year t, C 0 refers to initial incremental investment cost, ∆P t refers to the dynamic investment payback period, N refers to the year of cumulative net cash flow first becoming positive, A N−1 refers to the cumulative net cash flows with the last negative item, C N refers to the cumulative net cash flow of year N.

K-Means Clustering Based Decision Making
In this study, a combined indicator of building energy consumption, payback period, and indoor thermal comfort was used to evaluate the combined performance of different energy efficiency retrofit measures. However, since energy consumption, payback period, and indoor thermal comfort have different unit systems, it is not possible to further evaluate the merits of the optimized measures. Therefore, the K-means clustering method [31] was used to make the different optimization target values dimensionless and facilitate the comparative analysis of the integrated indexes. By clustering, the experimental results of the three optimization objectives were classified into 11 categories. The degree of merit of the categories is expressed by the magnitude of their values, i.e., measures with higher value have better improvement effects. The composite index is the sum of the values of the improvement targets corresponding to the clustering categories, and the final optimization scheme can be determined by assessing the magnitude of the composite index.

Winter
As shown in Figure 5, from 3-9 March, the overall outdoor temperature showed a downward trend. The maximum outdoor temperature was 25.13 • C, and the lowest temperature was 13.38 • C on 8 March. The average and minimum indoor temperatures are 18.7 • C and 16.3 • C, fluctuating with the outdoor temperature, and this is able to create a comfortable indoor thermal environment most of the time. The variation trend of the indoor global temperature is consistent with the indoor temperature, but that is always slightly higher than the indoor air temperature. The air temperature in the hall is almost the same as the ambient temperature, and the global temperature in the hall is also slightly higher than the hall temperature. Due to the strong solar radiation intensity on the 13th and 14th, and the semi-open structure of the hall, the global temperature was significantly higher than the air temperature, while that was still slightly higher than the air temperature in the measured room. As shown in Figure 5b, the indoor relative humidity ranges from 71.9% to 87.7%. Even in winter, the ambient relative humidity is high, with an average of 93.94%. The calibrated model can accurately simulate indoor temperature in winter. The difference between the simulated indoor temperature and the measured temperature is very small, and the average absolute error is only 0.47 • C with an average relative error of 2.51%. With the change of time, the simulated indoor relative humidity has roughly the same trend as the measured value. On the 3rd, 4th, and 9th, the model showed relatively higher accuracy. During other time periods, the absolute error of this model is relatively large, but the mean absolute error is still well controlled at 5.76%, and the mean relative error is 8.22%.
Buildings 2022, 12, x FOR PEER REVIEW 7 of 18 the same as the ambient temperature, and the global temperature in the hall is also slightly higher than the hall temperature. Due to the strong solar radiation intensity on the 13th and 14th, and the semi-open structure of the hall, the global temperature was significantly higher than the air temperature, while that was still slightly higher than the air temperature in the measured room. As shown in Figure 5b, the indoor relative humidity ranges from 71.9% to 87.7%. Even in winter, the ambient relative humidity is high, with an average of 93.94%. The calibrated model can accurately simulate indoor temperature in winter. The difference between the simulated indoor temperature and the measured temperature is very small, and the average absolute error is only 0.47 °C with an average relative error of 2.51%. With the change of time, the simulated indoor relative humidity has roughly the same trend as the measured value. On the 3rd, 4th, and 9th, the model showed relatively higher accuracy. During other time periods, the absolute error of this model is relatively large, but the mean absolute error is still well controlled at 5.76%, and the mean relative error is 8.22%.

Summer
The measured and simulated temperatures from 17-30 July are shown in Figure 6 a,c. It can be seen that the highest outdoor air temperature can reach 37.4 °C, the average value is 29.25 °C, and the daily temperature variation is large. During the day, due to the thermal resistance and thermal inertia of the envelope, the peaks of the hall temperature

Summer
The measured and simulated temperatures from 17-30 July are shown in Figure 6a,c. It can be seen that the highest outdoor air temperature can reach 37.4 • C, the average value is 29.25 • C, and the daily temperature variation is large. During the day, due to the thermal resistance and thermal inertia of the envelope, the peaks of the hall temperature and indoor temperature are lower and delayed compared with the outdoor air temperature, but the average temperatures of them are 30.39 • C and 30.36 • C, respectively. However, on 21, 23, and 30 July, the peak air temperature of the hall was slightly higher than the outdoor air temperature. This was due to the thermal inertia of the house itself not reducing the air temperature as quickly as the rain did to the outdoor air. The wind speed was about 0.7 m/s at that time, so the heat accumulated inside the hall was dissipated slowly, resulting in higher air temperature. The air temperature at night is lower than the indoor air temperature, thanks to night flushing. The indoor global temperature is slightly higher than the air temperature, and the peak values of the hall are also significantly higher than that of the indoor. As shown in Figure 6b, the city of Puning has high humidity in summer. Even on sunny days, the outdoor relative humidity can reach about 90%, and the daily humidity fluctuates greatly. Although the hall is a semi-open space, the humidity change trend is quite different from that of the outdoors. The outdoor humidity peak usually occurs at 20:00-22:00, while the maximum humidity in the hall occurs at dawn. Since the doors and windows are closed, the indoor humidity changes are relatively less affected by the outdoor humidity, and the relative humidity fluctuates between 70% and 90%. After the rain from the 21st to the 30th, the indoor relative humidity was mostly above 80%. In such high humidity, people are more likely to feel stuffy in summer. and indoor temperature are lower and delayed compared with the outdoor air temperature, but the average temperatures of them are 30.39 °C and 30.36 °C, respectively. However, on 21, 23, and 30 July, the peak air temperature of the hall was slightly higher than the outdoor air temperature. This was due to the thermal inertia of the house itself not reducing the air temperature as quickly as the rain did to the outdoor air. The wind speed was about 0.7 m/s at that time, so the heat accumulated inside the hall was dissipated slowly, resulting in higher air temperature. The air temperature at night is lower than the indoor air temperature, thanks to night flushing. The indoor global temperature is slightly higher than the air temperature, and the peak values of the hall are also significantly higher than that of the indoor. As shown in Figure 6b, the city of Puning has high humidity in summer. Even on sunny days, the outdoor relative humidity can reach about 90%, and the daily humidity fluctuates greatly. Although the hall is a semi-open space, the humidity change trend is quite different from that of the outdoors. The outdoor humidity peak usually occurs at 20:00-22:00, while the maximum humidity in the hall occurs at dawn. Since the doors and windows are closed, the indoor humidity changes are relatively less affected by the outdoor humidity, and the relative humidity fluctuates between 70% and 90%. After the rain from the 21st to the 30th, the indoor relative humidity was mostly above 80%. In such high humidity, people are more likely to feel stuffy in summer.  The weather station data of the city of Puning in 2019 was used as the weather input for building simulation. The variation trend of indoor air temperature predicted by simulation is consistent with the measured value, and it shows good prediction performance on sunny days. The measured relative humidity is close to the simulated value, as shown in Figure 6b. Overall, the manually calibrated model can accurately simulate indoor temperature and relative humidity most of the time, with an average absolute error of 0.78 • C and 5.17%, respectively. Therefore, this model is deemed suitable for the simulation of an indoor thermal environment and further calculation of energy use if an HVAC system is applied to the house.

Heat Gain and Loss
The heat gain and loss of the envelope are simulated by feeding the typical meteorological year (TMY) data to the calibrated EnergyPlus model, assuming an HVAC system is installed and turned on. As can be seen from Figure 7, in summer, most of the heat gain of the house comes from the roof, and the maximum hourly heat gain can be reached at 4.93 kW. In winter, the heat loss of the roof is also larger than the floor and walls, with the maximum hourly heat loss in November of 3.34 kW. Affected by the ambient temperature and solar radiation, the wall is constantly being heated by solar heat gain from June to October, with a maximum hourly heat gain of 2.66 kW. The daily temperature is low in February and March, and the indoor air temperature is higher than the ambient temperature resulting in the wall being the heat-loss surface most of the time. Puning is located in hot summer and warm winter zone, and the summer is hot and long. The exterior rammed earth wall is a heated surface most of the time during the year, and the hourly mean heat gain is 0.41 kW. In addition, the heat transfer through the floor is also large, which is conducive to the passive cooling of the house during summer. From May to October, the floor always transfers heat to the ground with an average hourly heat loss of 2.25 kW. Thus, in summer, the floor is the heat loss surface, and the roof is the main heat gain surface. In winter, the roof is also the main heat gain surface, which potentially raises the indoor air temperature and creates a warm indoor environment for the occupants.
The weather station data of the city of Puning in 2019 was used as the weather i for building simulation. The variation trend of indoor air temperature predicted by s lation is consistent with the measured value, and it shows good prediction perform on sunny days. The measured relative humidity is close to the simulated value, as sh in Figure 6b. Overall, the manually calibrated model can accurately simulate indoor perature and relative humidity most of the time, with an average absolute error of 0.7 and 5.17%, respectively. Therefore, this model is deemed suitable for the simulation indoor thermal environment and further calculation of energy use if an HVAC syste applied to the house.

Heat Gain and Loss
The heat gain and loss of the envelope are simulated by feeding the typical mete logical year (TMY) data to the calibrated EnergyPlus model, assuming an HVAC sy is installed and turned on. As can be seen from Error! Reference source not found summer, most of the heat gain of the house comes from the roof, and the maximum ho heat gain can be reached at 4.93 kW. In winter, the heat loss of the roof is also larger the floor and walls, with the maximum hourly heat loss in November of 3.34 kW. Affe by the ambient temperature and solar radiation, the wall is constantly being heate solar heat gain from June to October, with a maximum hourly heat gain of 2.66 kW daily temperature is low in February and March, and the indoor air temperature is hi than the ambient temperature resulting in the wall being the heat-loss surface most o time. Puning is located in hot summer and warm winter zone, and the summer is hot long. The exterior rammed earth wall is a heated surface most of the time during the and the hourly mean heat gain is 0.41 kW. In addition, the heat transfer through the is also large, which is conducive to the passive cooling of the house during summer. F May to October, the floor always transfers heat to the ground with an average hourly loss of 2.25 kW. Thus, in summer, the floor is the heat loss surface, and the roof is the m heat gain surface. In winter, the roof is also the main heat gain surface, which potent raises the indoor air temperature and creates a warm indoor environment for the o pants.

Natural Ventilation
The hourly indoor environmental condition under natural ventilation condition simulated by DesignBuilder, and the simulated results are compared with the Chi thermal comfort standard GB/T 50785-2012 "Civil Building Thermal and Humid Env ment Evaluation Standard" [32] and the American ASHRAE 55-2017 standard [33]. uncomfortable duration of the whole indoor thermal environment is shown in Figu   Figure 7. Hourly heat transfer of house envelope.

Natural Ventilation
The hourly indoor environmental condition under natural ventilation conditions are simulated by DesignBuilder, and the simulated results are compared with the Chinese thermal comfort standard GB/T 50785-2012 "Civil Building Thermal and Humid Environment Evaluation Standard" [32] and the American ASHRAE 55-2017 standard [33]. The uncomfortable duration of the whole indoor thermal environment is shown in Figure 8. Under the provisions of the Chinese thermal comfort standard, the uncomfortable durations under the I and II levels are 3310 h and 829 h, respectively, accounting for 37.8% and 9.5% of the whole year. Under the ASHRAE 55-2017 standard, the duration of not meeting the acceptable ranges for 90% and 80% are 4418 h and 2984 h, respectively, which is longer than the uncomfortable duration calculated by the Chinese standard. Considering the differences in climate, culture, and physiology, this study adopts the Chinese thermal comfort standard to evaluate indoor thermal comfort in Lingnan. Table 4 summarizes the uncomfortable durations for the four bedrooms under different criteria. Bedroom 1 has the best comfort, while bedroom 2 and bedroom 3 are less comfortable. On the whole, according to Chinese thermal comfort standard level II, the uncomfortable duration of each room is within 615~870 h. That is, the thermal comfort requirements are not met for about a month, and all of them appear in winter.
Under the provisions of the Chinese thermal comfort standard, the uncomfortable d tions under the I and II levels are 3310 h and 829 h, respectively, accounting for 37.8% 9.5% of the whole year. Under the ASHRAE 55-2017 standard, the duration of not me the acceptable ranges for 90% and 80% are 4418 h and 2984 h, respectively, which is lo than the uncomfortable duration calculated by the Chinese standard. Considering the ferences in climate, culture, and physiology, this study adopts the Chinese thermal fort standard to evaluate indoor thermal comfort in Lingnan. Table 4 summarizes th comfortable durations for the four bedrooms under different criteria. Bedroom 1 ha best comfort, while bedroom 2 and bedroom 3 are less comfortable. On the whole, acc ing to Chinese thermal comfort standard level II, the uncomfortable duration of each r is within 615~870 h. That is, the thermal comfort requirements are not met for abo month, and all of them appear in winter.

HVAC
To further improve indoor thermal comfort, the bedroom thermal environme simulated while the HVAC system is turned on. Indoor thermal comfort was evalu by predicted mean vote (PMV). The clothing insulation of winter and summer is 1 clo 0.5 clo, respectively, and the human metabolic index is 0.9. Figure 9 shows the chang comfort indicators in each bedroom over time. It can be seen that about 95% of the the PMV is between −0.5 and 0.5, and the corresponding predicted dissatisfaction per age (PPD) is within 10%. Indoor PMV that does not meet the comfort range appear April, May, and November. PMV is between −0.85 and 0.85 about 99% of the time, PPD is within 20%, which meets the requirements of human comfort.

HVAC
To further improve indoor thermal comfort, the bedroom thermal environment is simulated while the HVAC system is turned on. Indoor thermal comfort was evaluated by predicted mean vote (PMV). The clothing insulation of winter and summer is 1 clo and 0.5 clo, respectively, and the human metabolic index is 0.9. Figure 9 shows the changes in comfort indicators in each bedroom over time. It can be seen that about 95% of the time, the PMV is between −0.5 and 0.5, and the corresponding predicted dissatisfaction percentage (PPD) is within 10%. Indoor PMV that does not meet the comfort range appeared in April, May, and November. PMV is between −0.85 and 0.85 about 99% of the time, and PPD is within 20%, which meets the requirements of human comfort.

Energy Saving and Thermal Performance Retrofitting
From the heat gain and loss in Section 3.2, it can be known that in summer, introduced into the room mainly through the roof and exterior walls, and the ma hourly heat gain of the roof is close to 5 kW. Therefore, in summer, the optim measures can be focused on improving the heat absorption by the roof. In winter, e walls are heat-loss surfaces most of the time because the indoor air temperature is than the ambient temperature. Therefore, in winter, the optimization measures ca cused on the exterior walls. Retrofitting measures of the envelope by reducing t heat gain and enhancing the wall insulation are included. In order to ensure that th inal historical features of the "Four-point gold" are not damaged, an orthogonal ment method is used to carry out the retrofitting. Natural ventilation, enhanced flection and insulation, wall insulation, window shading, and carpet laying were s as design variables. The orthogonal experiment design table of the retrofitting is in Table 5. The detail of the design rules and data processing method of the orth experiments is shown in Appendix A.
The physical properties of materials and construction costs of each meas shown in Tables 6-8, respectively. Among them, carpets are only laid when the o temperature is below 15 °C, which is from 2 December to 4 March in the typical m logical year. In this research, carpets, polyvinyl chloride (PVC) carpets, and textile are considered. The expandable polystyrene (EPS) and extruded polystyrene (XPS) of different thicknesses are used for wall insulation with a heat conductivity coeffi 0.8 W/(m·k) and 0.034 W/(m·k), respectively. To minimize the heat gain of the ro coatings with high thermal reflectivity, solar reflectivity, and hemispheric emissivi considered. In addition, aluminum sunshades of different widths are planned to stalled above the east and west windows.

Energy Saving and Thermal Performance Retrofitting
From the heat gain and loss in Section 3.2, it can be known that in summer, heat is introduced into the room mainly through the roof and exterior walls, and the maximum hourly heat gain of the roof is close to 5 kW. Therefore, in summer, the optimization measures can be focused on improving the heat absorption by the roof. In winter, exterior walls are heat-loss surfaces most of the time because the indoor air temperature is higher than the ambient temperature. Therefore, in winter, the optimization measures can be focused on the exterior walls. Retrofitting measures of the envelope by reducing the roof heat gain and enhancing the wall insulation are included. In order to ensure that the original historical features of the "Four-point gold" are not damaged, an orthogonal experiment method is used to carry out the retrofitting. Natural ventilation, enhanced roof reflection and insulation, wall insulation, window shading, and carpet laying were selected as design variables. The orthogonal experiment design table of the retrofitting is shown in Table 5. The detail of the design rules and data processing method of the orthogonal experiments is shown in Appendix A. The physical properties of materials and construction costs of each measure are shown in Tables 6-8, respectively. Among them, carpets are only laid when the outdoor temper-ature is below 15 • C, which is from 2 December to 4 March in the typical meteorological year. In this research, carpets, polyvinyl chloride (PVC) carpets, and textile carpets are considered. The expandable polystyrene (EPS) and extruded polystyrene (XPS) boards of different thicknesses are used for wall insulation with a heat conductivity coefficient of 0.8 W/(m·k) and 0.034 W/(m·k), respectively. To minimize the heat gain of the roof, five coatings with high thermal reflectivity, solar reflectivity, and hemispheric emissivity were considered. In addition, aluminum sunshades of different widths are planned to be installed above the east and west windows. When the indoor temperature is 2 • C higher than the outdoor temperature, natural ventilation is used for cooling. The HVAC system is activated only when natural ventilation is insufficient to maintain comfort. The operating hours coincide with the rest period of the occupants, from 22:00 to 9:00 the next morning. In addition, the compressor was placed at the window on the second floor in conjunction with the bedroom layout of a traditional rammed house and its storage space on the second floor, which allows for more comfortable living conditions without destroying the historic character. Last but not least, the whole house is retrofitted, but the HVAC system was only used in the four bedrooms.

Energy Saving
The annual HVAC energy consumption before the retrofitting of the rammed earth house was calculated to be 4695.62 kWh, which was used as the benchmark for calculating energy savings. After taking retrofitting measures, the total energy consumption was significantly reduced. The sensitivity analysis of each energy-saving retrofitting measure is shown in Table 9. The maximum average energy saving in the orthogonal experiment is 2192.27 kWh/a. The maximum difference for energy saving of wall insulation, heatreflective and thermal insulation coating, sunshade width, natural ventilation, and carpet is 1294.96 kWh/a, 197.49 kWh/a, 159.96 kWh/a, 91.51 kWh/and 79.53 kWh/a, respectively. Therefore, the impact of each retrofitting measure on energy savings can be ranked as follows: wall insulation, heat-reflective and thermal insulation coating, sunshade width, natural ventilation, and carpet. Among wall insulations, the energy-saving effect of level 4 (XPS 50 mm) is higher than other wall insulation materials. For heat-reflective and thermal insulation coatings and sunshade width, the energy-saving effect of level 5 coating material (RLHY-A05) and level 5 sunshade width (0.5 m) is better than other measures in each group. For natural ventilation and carpets, turning off the window (door) and PVC carpet have better performance in energy saving compared with other measures. To sum up, the combined design of 50 mm XPS panel, RLHY-A05 coating, 0.5 m sunshade width, turning off the window (door), and PVC carpet can be considered as the optimal scheme for energy saving.

Economic Performance
The study has shown that the total cost of retrofitting measures is approximately 1.2-1.5 times the cost of energy-saving materials [34]. Therefore, the total cost of each retrofitting measure can be calculated based on the unit price and area of each material shown in Tables 6-8. In this research, the dynamic investment payback period was selected as the criteria to evaluate the economic performance of each retrofitting measure. The sensitivity analysis of the dynamic investment payback period for each retrofitting measure is shown in Table 10. Natural ventilation can be achieved by manually opening windows and doors without investment costs. Therefore, there is no need to discuss the economic performance of natural ventilation. The minimum average dynamic investment payback period in the orthogonal experiment is 9.17 years. When EPS materials are used, the dynamic investment payback period cannot be calculated. The maximum differences for a dynamic investment payback period of sunshade, carpet, and coating are 2.99, 2.86, and 2.83, respectively. Therefore, the impact of each retrofitting measure on the dynamic investment payback period can be ranked in order as follows: wall insulation, sunshade width, carpet, and heat-reflective and thermal insulation coating. For wall insulation, the economic performance of level 2 (XPS 30 mm) is better than other wall insulation materials. Among all heat-reflective and thermal insulation coatings, the water-based coating has the best economic performance. For carpets, the economic performance of the carpet is the best. For sunshade width, the 0.5 m sunshade width has the best performance in economic performance. In general, the combined design of 30 mm XPS panel, water-based coating, carpet, and 0.5 m sunshade width can achieve optimal economic performance.

Indoor Thermal Environment
The annual indoor unacceptable thermal comfort hours of the rammed earth house are selected as the criteria to evaluate the indoor thermal environment improvement. The sensitivity analysis of the annual indoor unacceptable thermal comfort hours for each retrofitting measure is shown in Table 11. The maximum average annual indoor unacceptable thermal comfort hours in the orthogonal experiment is 1766 h. The maximum difference for the annual indoor unacceptable thermal comfort hours of wall insulation, heatreflective and thermal insulation coating, sunshade width, carpet, and natural ventilation is 276.60 h, 253.00 h, 116.80 h, 85.50 h, and 5.37 h, respectively. Therefore, the impact of each retrofitting measure on the annual indoor unacceptable thermal comfort hours can be ranked in positive order as follows: wall insulation, heat-reflective, and thermal insulation coating, sunshade width, carpet, and natural ventilation. Among wall insulations, the annual indoor unacceptable thermal comfort hours of level 3 (EPS 100 mm) are higher than other wall insulation materials. For all heat-reflective and thermal insulation coatings, ZS-221 white coating results in the largest increase in the annual indoor thermal comfort period. For carpets, the annual indoor unacceptable thermal comfort hours of the textile carpet are the longest. For the sunshade width, the 0.3 m sunshade width has the best performance for the annual indoor unacceptable thermal comfort hours. For natural ventilation, opening windows (doors) can have the best performance in shrinking the annual indoor unacceptable thermal comfort hours. In general, the combined design of 100 mm EPS panel, ZS-221 white coating, textile carpet, 0.3 m sunshade width, and opening windows (doors) can achieve the optimal scheme for the indoor thermal environment.

Optimized Scheme
The above retrofitting schemes focus on a single optimization objective without considering the combined impact of various measures on building energy consumption, economic performance, and the indoor thermal environment. However, various retrofitting measures will be applied simultaneously for a single building. Therefore, it is necessary to propose a comprehensively optimized scheme which can be used in the actual retrofitting. The K-means clustering algorithm is used to find the comprehensively optimized scheme. According to the results of the dynamic investment payback period, different indicators of twenty-five cases in Appendix A are divided into eleven grades, and the number 11 represents the best grade. The comprehensive indicator is the sum of grade values for energy saving, economic performance, and indoor thermal environment. The sensitivity analysis of the comprehensively optimized scheme is shown in Table 12. The maximum difference for the performance of the comprehensively optimized scheme of wall insulation, heatreflective and thermal insulation coating, sunshade width, carpet, and natural ventilation is 10.00, 3.60, 1.80, 1.50, and 1.13, respectively. Therefore, the impact of each retrofitting measure on the performance of the comprehensively optimized scheme can be ranked in positive order as follows: wall insulation, heat-reflective and thermal insulation coating, sunshade width, carpet, and natural ventilation. Among them, 30 mm XPS, ZS-221 white coating, carpet, 0.5 m sunshade width, and turning off windows (doors) can be regarded as the combined design to achieve the comprehensively optimized scheme. In summary, improving wall insulation is the most effective way for energy saving and indoor thermal comfort mitigation, but EPS is too expensive for the dynamic investment payback period. Heat-reflective and thermal insulation coating and sunshades with low initial investments also have good performance in reducing the heat gains from solar radiation, resulting in lower energy consumption and longer comfort period. Carpet is only used in winter, so the annual benefits can be relatively insignificant. Moreover, the enhancement of natural ventilation has not yielded desired effect because the transition season in this climate is short. Based on the comprehensive index, 30 mm XPS wall insulation, ZS-221 heat-reflective and thermal insulation coating, sunshade width of 0.5 m, carpet, and natural ventilation are a recommended retrofitting scheme.

Conclusions
This study analyzed the thermal performance and environment of traditional rammed earth houses in Lingnan, south China. Energy-saving retrofitting measures are proposed to improve the indoor thermal environment by taking the building conditions and local customs into account. It not only retains the historical features and cultural value of traditional rammed earth houses but also improves thermal comfort and achieves energy saving.
The heat transfer of rammed earth houses behaves differently by seasons. In summer, most of the heat gain comes from the roof and exterior walls. During winter, the heat loss is mainly due to the exterior walls. Hence, the retrofitting measures mainly focus on reducing the heat gain of the roof and enhancing the insulation of the wall.
Five energy-saving retrofitting measures, comprised of natural ventilation, enhanced roof reflection and insulation, wall insulation, window shading, and carpet laying, are proposed. The highest performance of retrofitting scheme on energy saving, dynamic investment payback period, and the annual indoor thermal comfort increase table are 2192.27 kWh/a, 9.17 years, and 1766 h, respectively. The research puts forward a comprehensively optimized scheme, which consists of 30 mm XPS 30 mm, ZS-221 white coating, carpet, 0.5 m sunshade width, and turning off windows (doors). The optimized scheme can be practically used in renovation, which balances energy saving, economic performance, and indoor thermal environment.

Limitations
This work takes the "four-point gold" house as a sample to analyze its thermal performance and search for a suitable retrofitting scheme. The conclusions and the application of retrofitting measures are limited to this specific type of traditional rammed earth residential buildings in the Lingnan region, and it is also one of the most popular types. We will study other types of traditional rammed earth buildings in the future and achieve more universal and comprehensive results for the traditional residential houses in the region.