Effect of existing residential renovation strategies on heating and cooling load in Shanghai

. To provide a reference for the renovation of Shanghai’s existing residential districts, this study quantifies and compares the relationships between common renovation strategies, microclimate, heating and cooling loads. These common strategies include improving the greening rate (G), improving the reflectivity of pavement (P), improving the reflectivity of wall materials (W), and improving the reflectivity of roof materials or applying green roof (R). These strategies are applied to a typical model extracted from existing residential areas in Shanghai, China. ENVI-met and EnergyPlus are combined to simulate the microclimate represented by the average meteorological parameters in front of building surfaces and the building heating and cooling load on a typical meteorological day in winter and summer. The results show that applying microclimate data around target building contributes to a significant difference in air conditioning load in both summer and winter. For summer, G, W, and R reduced their total cooling load, whereas P increased this parameter. R contributed the most significant decrease in the total cooling load, followed by W, and G contributed the least. G3P1W3R3 and G1P3W1R1 were the scenarios with the lowest and highest cooling load. The total cooling load under G3P1W3R3 was 136 kWh (12.7%) less than that under G1P3W1R1. For winter, P and applying green roof (R4) reduced the heating load of the target building, whereas G, W and improving roof reflectivity (R2, R3) increased it. G1P3W1R4 and G3P1W3R3 were the scenarios with the lowest and highest heating load. The heating load under G1P3W1R4 was 145 kWh (14.5%) less than that under G3P1W3R3.


Introduction
In some of China's metropolises, such as Beijing, Shanghai, and Guangzhou, many existing residential districts were built before 2000 or even before 1990. Existing residential districts require urgent improvement owing to aging facilities and poor outdoor environments [1]. The main renovation objects include building envelopes, outdoor landscape environments, and energy systems [2]. Some renovation strategies may affect the microclimate of residential districts and, consequently, the energy consumption of residential buildings.
Microclimates are influenced by large-scale climatic conditions that are closely related to natural conditions and are difficult to modify. In addition, microclimates are affected by the heat and mass exchange between the ground, vegetation, buildings, and atmosphere on a small scale [3], which should be considered in urban planning and architectural design, including the renovation of existing residential districts. The microclimate of residential districts is related to the underlying surface, waterbody, greening, building layout, building surface, artificial heat, and many other factors [4]. For existing residential districts, some of the aforementioned factors cannot be controlled. First, waterbodies are difficult to increase owing to the high * Corresponding author: zhanghuibo@sjtu.edu.cn building density. Second, the layout of existing buildings should be maintained. Third, green walls are rarely applied because of their tedious construction. Therefore, this study selected several common renovation strategies, including improving the reflectivity of the underlying surface, increasing the vegetation, improving the reflectivity of the wall and roof, or applying a green roof.
Because these strategies significantly affect the microclimate, the meteorological boundaries of buildings are changed, which ultimately affect building energy consumption. According to Yang et al. [5], the total cooling load decreased by 10.6% after considering the full microclimatic environment, whereas the total heating load increased by 0.3%. Gobakis et al. [6] found that modifying EnergyPlus Weather File (EPW) with microclimatic conditions resulted in a difference of ±10% in the heating and cooling requirements. According to Zhang et al. [7], a higher temperature and lower wind speed in microclimate results caused a higher cooling demand and lower heating demand, with maximum difference of 16.4% and 23.4%, respectively. The difference in the energy consumption using two types of climatic data varied in different studies.
These studies demonstrate the importance of using microclimate results to simulate building energy consumption, particularly when the research object affects the microclimate. The present study aims to compare the effect of common renovation strategies on air conditioning loads by investigating a typical model extracted from existing residential districts in Shanghai. The strategies include increasing the vegetation and applying high-reflectivity materials on the ground, wall, and roof; a green roof was also included. ENVI-met and EnergyPlus were integrated to simulate the microclimate around the building envelope and the airconditioning load. This study quantified and compared the extent to which different strategies affect the microclimate around a building and the heating and cooling loads of the building in Shanghai, China.

Research framework
This study mainly focused on three main steps, as shown in Fig. 1. First, typical weather days (TMD) in July and January, which are the hottest and coldest, respectively, were selected, and a typical model of the existing residential area was extracted. Second, ENVI-met was used to simulate the microclimate of the model on the typical meteorological days under different renovation strategies after measurement and evaluation. Third, the average meteorological parameters around the building in the ENVI-met simulation results were used to generate a new weather file, and the reflectivity settings corresponding to each strategy were adjusted in EnergyPlus to simulate the air conditioning load of the target building, including cooling and heating loads.

Generation and validation of a typical meteorological day (TMD)
A single day was selected as the typical meteorological day (TMD) to represent the weather condition of the entire month because CFD simulation is time consuming. The selection criteria were based on the Sandia method developed by Hall et al. [8]. The procedure adopted to select the TMD is the same as our previous work [9].
Finally the selected TMD for Shanghai are July 1 and January 24 for summer and winter. After determining the TMDs for the July and January, we compared the hourly air conditioning load on TMD and the monthly mean air conditioning load of each hour in one day to verify the accuracy of the method. The correlation between the two data groups is evident from correlation analysis, as shown in Fig. 2. In general, the air conditioning load on the TMDs determined using the above method can accurately represent that of the entire month.

Model information
and key input parameters

Model information
Existing residential districts (ERDs) tend to have various geometric features, such as building layouts, orientations, and so on. To make this study more representative, a typical model of ERDs when exploring the relationship between renovation strategies and building cooling load should be determined. This study randomly selected 50 ERDs based on a list of ERDs to be renovated in Shanghai. All the ERDs were built before 2000 (a substantial portion was built before 1990), with high density and poor envelope. Data items were collected, and the average value of each data item was obtained. The typical model of ERDs consists of 12 buildings with a determinant layout and a south-north orientation, as shown in Fig. 3. Each building is 48 m from east to west, 12 m from north to south, and 18 m in height. The building distances are 20 m for north-south and 12 m in the east-west direction. The window-towall ratio is 33% in the south and 17% in the rest.

Simulation model and setup for microclimate
ENVI-met 4.4.5 was used to simulate the microclimate of the residential district, which is a three-dimensional and holistic microclimate model that covers fluid dynamics, thermodynamics, plant physiology, and soil science [10]. Many researchers have validated the model in different climate zones [11][12][13][14]. Moreover, expanding the calculation area is crucial to consider the surroundings of the study area and improve the simulation accuracy, which also led to a higher computing load. Based on Chapter 2.3.1, a series of presimulations were carried out. These simulations focused on the number of buildings added outside the study area and the distance between the outer buildings and the boundary. Finally, the study area was determined to expand one building in four directions, with a distance of 40 m, as shown in Fig. 3.
The other input parameters are listed in Table 1. The geographical coordinates of the model are 31.4° N and 121.4° E, which is in the hot summer and cold winter climate zones. The calculation area the model was 408 m × 276 m ×54 m (x ×y ×z), and the grid resolution was 4 m ×4 m ×1.5 m (x ×y ×z). To reduce the influence of the initial conditions, we set the starting time of the calculation as 18:00 the day before the TMD, and the simulation duration was 30 h. Fully forced lateral boundary conditions were adopted, and temperature, relative humidity, wind speed, wind direction, and solar radiation were all hourly forced. When the wind speed was 0, 0.5 m/s was used instead to avoid numerical errors in the calculation. Table 1. Initial and boundary conditions used in ENVI-met.

Simulation model and setup for building energy consumption
EnergyPlus 8.9.0, a comprehensive simulation software developed for the building energy consumption dynamic simulation program was applied to build the energy consumption simulation model. The target building for the energy consumption simulation is shown in Fig. 3. Table 2 shows the construction of typical residential buildings in Shanghai. Besides, other settings in EnergyPlus are listed in Table 3. Table 2. Construction of typical residential buildings.

Scenarios' description
As mentioned in the Introduction section, the renovation objects in this study consist of vegetation, pavement, wall, and roof. Specific strategies include improving the greening rate (G), improving the reflectivity of pavement (P), improving the reflectivity of wall materials (W), and improving the reflectivity of roof materials or applying green roof (R). For G, the vegetation was set to the south and north sides of the building, as shown in Fig. 3. The area ratio of trees to grassland was 1:1, the height of trees was 10 m, and the width of the tree crown was 4 m. Furthermore, when adjusting the reflectivity of materials, other material properties should be maintained. The parameters were determined based on field research and regulations. The three greening rate parameters were 8.66%, 17.32%, and 25.98%, which correspond to the grid size of ENVI-met, as presented in Table 4. The reflectivity parameters were selected, based on the Thermal Design Code for Civil Building and Design Standard for Thermal Environment of Urban Residential Districts, as 0.15, 0.26, and 0.50 for pavement; 0.21, 0.50, and 0.70 for wall materials; and 0.14, 0.26, and 0.48 for roof materials. In addition, the emissivity of these materials to longwave radiation was set to 0.9. When applying green roof, its leaf area index (LAI) was 1, and the soil thickness was 30 cm, which also reduced the roof U-value to 1.6 W/(m 2 •K). With different parameters combined with typical models, the cases for simulation are listed in Table 5. Moreover, when exploring one specific strategy, all other strategies adopted parameter 2.

Model coupling
ENVI-met and EnergyPlus were integrated using weather files. The weather files for the energy consumption simulation were modified with the ENVImet output of the average meteorological parameters around the target building, including air temperature, relative humidity, and wind speed. The potential air temperature output from ENVI-met 4.4.5 must be converted to air temperature using the following function: where T is the air temperature (in K) for the energy consumption simulation, ߠ is the potential air temperature (in K) output by ENVI-met, R is the gas constant of air, and cp is the specific heat capacity at a constant pressure. Further, ܴ ܿ ⁄ =0.286 for air (meteorology).
Because some renovation strategies alter the solar gain of the building, in addition to these modifications, the change in the solar gain of the building should be considered. From the perspective of building energy consumption simulation, EnergyPlus calculates solar radiation more accurately than ENVI-met [5]. Therefore, in this study, solar radiation obtained from the ENVImet output of the building façade was not employed; instead, the settings corresponding to different strategies in EnergyPlus were adjusted to accurately calculate the solar gain.

Effect on microclimate around the target building
Since the building layout was not changed, a minimal difference was observed in the average wind speed (WSP) around the target building for different strategies; therefore, the WSP data were the results of G2P2W2R2. Moreover, in the model coupling process, considering EnergyPlus's more accurate calculation of radiation, the change of solar radiation obtained by the building was reflected in the relevant settings in EnergyPlus. As a result, this section focuses on the changes in temperature and relative humidity.
As mentioned above, this study focused on the average meteorological parameters around the building, which refers to the mean value of the grids closest to the building in ENVI-met. In the following analysis, Ta and RH represent the average temperature and relative humidity around the target building, respectively. Furthermore, taking scenario 1 as the base case, this section analysis the differences of Ta and RH in the other scenarios compared with scenario 1 in summer and winter.
3.1.1 Effect on microclimate around the target building in summer Fig. 4 shows the decrease in Ta and increase in RH in the other scenarios compared with scenario 1 for each strategy in summer. At the same time, the average values of the impact of each strategy on microclimate for 24 h as well as for air-conditioning hours are analyzed. For Ta, the influence of G was the most significant, whereas that of W was the weakest. G can reduce the average temperature of the whole day by 0.23 ℃ at most. For P and R, the effect of the latter was a little stronger.  For RH, greening-related strategies had more significant effects. G had the most noticeable influence, followed by the application of a green roof. The transpiration of vegetation added to the moisture content in the atmosphere, which eventually led to an increase in RH. On the other hand, the effects of high-reflectivity strategies on RH were weak, particularly for W. G increased the average RH of the whole day by 1.66% at most and only by 0.07% for W. The differences in Ta and RH on air conditioning hours were less significant than those throughout the day because air conditioning was turned on in residential districts at night. Previous studies [15] indicated that the effect of various strategies on microclimate weakened at night. Fig. 5 shows the decrease in Ta and increase in RH in the other scenarios compared with scenario 1 for each strategy in winter. For Ta, the influence of applying green roof (R4) was the most significant, followed by P and W, whereas that of G was the weakest. In fact, all strategies' effects in winter were unapparent, with the greatest hourly Ta reduction 0.1 ℃ and average Ta reduction for 24 hours 0.05 ℃. For greening-related strategies, the effect of G was weaker than green roof, caused by the following two reasons. Firstly, deciduous trees had low Leaf area density (LAD) in winter, while the grassland on the roof was evergreen. Secondly, the lower solar altitude in winter and the high density of ERDs contributed to less solar radiation received by the trees on the ground. Therefore, the transpiration of trees on the ground was weakened. In addition, the effect of G on Ta fluctuated more greatly than green roof (R4). G3-G1 increased hourly Ta by 0.07 ℃ at most and decreased by 0.10 ℃ at most. The decrease of Ta mainly occurred in the daytime, while the increase mainly occurred at night. The addition of trees caused the WSP to decrease, which slowed down the heat dissipation of the building at night and ultimately led to an increase in Ta.

Effect on microclimate around the target building in winter
For RH, similar to summer greening-related strategies had more significant effects. However, the impact of all strategies was weak. Applying green roof (R4) had the most noticeable influence, followed by G and P, with improving building surfaces' reflectivity the weakest. Besides, the influence of each strategy on RH was stronger than that of air conditioning time.
In general, the influence of the strategies on the microclimate around the building was more obvious in summer than winter.

Effect of different strategies on air conditioning load
This section presents the effect of different strategies on air conditioning load, including cooling and heating loads. Table 6 presents the difference in the percentage of the all-day cooling load for different strategies compared with Scenario 1, including the sensible, latent, and total cooling loads. 3-1(%) -1.9 +2.3 -3.9 -5.7

Effect of different strategies on cooling load
G, W, and R reduced the sensible cooling load of the target building, whereas P increased it. The reduction in the sensible cooling load by R is the most significant, and the percentage reduction by R3-R1 was 8.2%. Improving the roof reflectivity significantly reduces the solar gain of the roof, thereby reducing the sensible cooling load of the top floor, which accounts for most of the sensible cooling load reduction. Despite the low effect on Ta, W still cut down the sensible cooling load to a certain extent owing to the solar gain reduction of the exterior walls. The poor thermal performance of existing residential buildings (ERBs) made it easier for the solar gain of the building facades to be converted into indoor heat gain, which is the main reason for the significant effect of W and R. In spite of the smaller area and minor reflectivity variations, R decreased the cooling load more than W. This was because, when the wall reflectivity of ERBs was increased, the surrounding buildings reflected more solar radiation to the target building.
Despite having the most significant effect on Ta, G produced the slightest reduction in the sensible cooling load. This is because G affected the sensible cooling load through infiltration and heat transfer, resulting in a minimal difference in the solar gain of the building. Similarly, G still reduced the sensible cooling load by 3.3%, demonstrating the importance of using microclimate results to simulate energy consumption. P decreased Ta but increased sensible cooling load. With an increase in pavement reflectivity, more solar radiation is reflected. The high density of ERDs resulted in a higher solar gain of the exterior walls and windows, and the sensible cooling load increased by this phenomenon exceeded the load reduction produced by the decrease in Ta, which eventually resulted in an increase in the sensible cooling load by P.
The effect of all strategies on latent cooling load was weaker than that on sensible cooling load. W and R cut down the latent cooling load of the building, while G and P increased it. Plant transpiration significantly increased RH and, hence, increased the relative humidity of the indoor air through infiltration, eventually increasing the latent cooling load. However, although the green roof increased RH, it reduced the latent cooling load of the target building, which was mainly caused by the decrease of the latent cooling load of the top floor. The latent cooling load of the top floor decreased by 2.08 kWh from R1 to R4, but it was 2.05 kWh for the entire building, indicating that the total latent cooling load of the other five floors increased. For the high-reflectivity strategies, P increased the latent cooling load, while W and R decreased it.
In summary, G, W, and R decreased the total cooling load of the target building, whereas P increased it. R reduced the total cooling load the most, followed by W and G. In terms of R, increasing the roof reflectivity to 0.48 (R3) further decreased the total cooling load, compared with the green roof (R4). This demonstrates that blindly applying a green roof without analyzing specific projects is not advisable.
Based on the analysis of single strategy's effect, the lowest load scenarios under G, P, W and R were 3, 1, 3 and 3, respectively; the highest were 1, 3, 1 and 1, respectively. After combining the scenarios with the lowest and highest cooling load, the loads of the target building under G3P1W3R3 and G1P3W1R1 were compared. The total cooling load under G3P1W3R3 was 136 kWh (12.7%) less than that under G1P3W1R1. The decrease of sensible cooling load accounted for most of reduction, up to 89.1%. Table 7 presents the difference in the percentage of the all-day heating load for different strategies compared with Scenario 1. P and applying green roof (R4) reduced the heating load of the target building, whereas G, W and improving roof reflectivity (R2, R3) increased it.

Effect of different strategies on heating load
The reduction in the heating load by R4 was the most significant, and the percentage reduction by R4-R1 was 2.7%. R4 decreased the U-value of roof from 2.68 W/ m 2 •K to 1.6 W/ m 2 •K, thus reducing the heating load of the top floor. P made the target building receive more solar radiation during the daytime, leading to a slight reduction of the heating load.
W and improving roof reflectivity (R2, R3) reduced the solar radiation received by target building, increasing the heating load to a certain extent. Compared with other strategies, the effect of these two strategies was relatively obvious due to the poor thermal performance of the ERBs. As mentioned in Section 3.1.2, all strategies had a weak effect on microclimate in winter, and G had almost no impact on the solar radiation obtained by the target building, so the effect of G on heating load was extremely weak. Based on the above analysis, the lowest load scenarios under G, P, W and R were 1, 3, 1 and 4, respectively; the highest were 3, 1, 3 and 3, respectively. After combining the scenarios with the lowest and highest cooling load, the loads of the target building under G1P3W1R4 and G3P1W3R3 were compared. The heating load under G1P3W1R4 was 145 kWh (4.5%) less than that under G3P1W3R3. Meantime, the heating load under G1P3W1R1, with highest cooling load in summer, was 125 kWh (12.5%) less than G3P1W3R3.
Compared with the effect on the total cooling load in summer, the effect of all strategies on the heating load was reverse, except for applying green roof (R4). Considering the lower U-value, R4 reduce the airconditioning load in both winter and summer. Besides, in terms of the difference in percentage, G, P and R had stronger effect in summer, while W had stronger effect in winter.

Compared different strategies in summer and winter
To make a more intuitive comparison, Table 8 presents the difference in the air conditioning load in summer and winter for different strategies. As mentioned above, G changed Ta and RH, and thus affected the sensible cooling load through infiltration and heat transfer. In summer, G had a more significant effect on Ta and RH than in winter, contributing to a relatively obvious effect on air conditioning load in summer and a neglectable effect in winter. P decreased Ta in winter and summer and reflected more solar radiation to the target building, increasing the cooling load in the summer and decreasing the heating load in winter. Because of the stronger solar radiation and higher solar altitude in summer, P had a more obvious effect in summer. The effect caused by P on air conditioning load in summer was three times more than that in winter. W decreased Ta in winter and summer and reduced solar radiation obtained by the target building, decreasing cooling load in the summer and increasing heating load in winter. Despite stronger solar radiation in summer, higher solar altitude caused solar gain of the target building to change more obviously, thus W had a more significant effect on air conditioning load in winter. For R, improving roof reflectivity (R2, R3) was similar to W, decreasing cooling load in the summer and increasing heating load in winter, but stronger effect in summer. Applying green roof lowered the U-value of roof and decreased the air conditioning load in both summer and winter. To sum up, G and R decrease the annual air conditioning load, while P and W increase. Among them, applying the green roof reduces the annual air conditioning load most obviously.

Conclusions
In this study, 50 existing residential districts in Shanghai were selected as samples, and a typical model was extracted. Common renovation strategies on building surfaces and ground, including improving the greening rate (G), reflectivity of pavement (P), reflectivity of wall materials (W), and reflectivity of roof materials or applying green roofs (R), were applied to this model. ENVI-met and EnergyPlus were integrated to simulate the cooling and heating load of the target building under various scenarios. The relationship between air conditioning load and different renovation strategies was investigated through horizontal comparison, and the results were as follows: (1) For summer, G, W, and R reduced their total cooling load, whereas P increased this parameter. R contributed the most significant decrease in the total cooling load, followed by W, and G contributed the least. G3P1W3R3 and G1P3W1R1 were the scenarios with the lowest and highest cooling load. The total cooling load under G3P1W3R3 was 136 kWh (12.7%) less than that under G1P3W1R1.
(2) For winter, P and applying green roof (R4) reduced the heating load of the target building, whereas G, W and improving roof reflectivity (R2, R3) increased it. The reduction by R4 was the most significant. G1P3W1R4 and G3P1W3R3 were the scenarios with the lowest and highest heating load.
The heating load under G1P3W1R4 was 145 kWh (14.5%) less than that under G3P1W3R3. (3) For Shanghai, located in hot summer and cold winter zone, G and R decreased the annual air conditioning load, whereas P and W increased it. Applying green roof (R4) contributed to the greatest reduction of air conditioning load. (4) A comparison of the microclimate results and EPW indicate that the difference in WSP was the most significant, followed by RH and then Ta. At night, Ta in the microclimate results tended to be lower than that in EPW but higher in winter, leading to lower cooling load and heating load. Lower RH in microclimate results contributed to decrease of latent cooling load in summer. Lower WSP in microclimate results caused higher cooling load in summer and lower heating load in winter. (5) The specific results showed that, replacing air temperature with Ta, the difference of air conditioning load was obvious in summer and weak in winter, -4.6% in summer and -1% at most. Replacing relative humidity in EPW with RH caused a significant difference in summer, up to -10%. As to WSP, the difference of cooling load reached 5.8% and heating load -6.7%.