Thermal decay in underfloor air distribution (UFAD) systems: Fundamentals and influence on system performance

Underfloor air distribution (UFAD) is a mechanical ventilation strategy in which the conditioned air is primarily delivered to the zone from a pressurized plenum through floor mounted diffusers. Compared to conventional overhead (OH) mixing systems, UFAD has several potential advantages, such as improved thermal comfort and indoor air quality (IAQ), layout flexibility, reduced life cycle costs and improved energy efficiency in suitable climates. In ducted OH systems designers have reasonably accurate control of the diffuser supply temperature, while in UFAD this temperature is difficult to predict due to the heat gain of the conditioned air in the supply plenum. The increase in temperature between the air entering the plenum and air leaving through a diffuser is known as thermal decay. In this study, the detailed whole-building energy simulation program, EnergyPlus, was used to explain the fundamentals of thermal decay, to investigate its influence on energy consumption and to study the parameters that affect thermal decay. It turns out that the temperature rise is considerable (annual median=3.7 K, with 50% of the values between 2.4 and 4.7 K based on annual simulations). Compared to an idealized simulated UFAD case with no thermal decay, elevated diffuser air temperatures can lead to higher supply airflow rate and increased fan and chiller energy consumption. The thermal decay in summer is higher than in winter and it also depends on the climate. The ground floor with a slab on grade has less temperature rise compared to middle and top floors. An increase of the supply air temperature causes a decrease in thermal decay. The temperature rise is not significantly affected by the perimeter zone orientation, the internal heat gain and the window-to-wall ratio.


INTRODUCTION
Underfloor air distribution (UFAD) is a method of providing space conditioning and ventilation to offices and commercial buildings. UFAD systems use an underfloor supply plenum located between the structural concrete slab and a raised access floor system to supply conditioned air through floor diffusers directly into the occupied zone [1]. UFAD systems have several potential advantages over traditional overhead systems, such as layout flexibility, improved energy efficiency in suitable climates and reduced life cycle costs [1]. Their performance has been investigated through field study investigations [2][3][4], full and bench-scale laboratory testing [5][6][7], computational fluid dynamics (CFD) and analytical modeling [5][6][7][8], and whole-building energy simulation [5,[9][10][11][12][13][14][15]. In UFAD systems, two distinguishing characteristics combine to change the dominant heat transfer dynamics related to energy balance in the conditioned space under cooling operation. These are: (1) room air stratification, in which comfortable air temperatures are maintained in the occupied zone near the floor but warmer air exists near the ceiling; and (2) underfloor air supply plenums, through which cool supply air is distributed to floor diffusers. The underfloor plenum creates a relatively cool reservoir of air extending across the entire building

Simulation software
The energy simulation program, EnergyPlus, was used for this study [19]. EnergyPlus is a relatively new building energy simulation program, which has greater capabilities than other programs [20]. EnergyPlus was selected because its capabilities were recently upgraded to model UFAD systems, allowing it to do the following: model each underfloor plenum as a completely separate zone; perform a full heat balance on the underfloor plenum which accounts for the heat gain into the plenum from both the room above and the return plenum from the floor below [9]; calculate the surface temperatures in each time step by conducting a detailed heat balance on each surface that includes the radiant heat exchange between surfaces and internal loads [9]; take into account the thermal mass effect; and model the temperature stratification generated by the UFAD system [5]. The EnergyPlus UFAD model has not been independently verified yet. For more information about validation of the EnergyPlus program, see [21].

Description of office building
A three-story prototype office building with a rectangular shape (75 m x 51 m) and aspect ratio of 1.5 was chosen for this study. The floor plate size is 3,720 m 2 (total floor area is 11,200 m 2 ) and each floor is composed of 4 perimeter zones, an interior zone and a service core, which represent approximately 28%, 56% and 16% of the floor area, respectively (see Figure 2). The floor to floor height is 3.96 m and the return plenum height is 0.6 m. The raised floor height is 0.4 m. Strip windows are evenly distributed (i.e., a "ribbon" window) in the walls and the baseline window-to-wall ratio (WWR) is 40%. Different WWRs are achieved by varying the window height only. The constructions and the thermal properties of windows change based on each climate and they comply with table 5.5 of ASHRAE 90.1-2004 [22]. When doing the design day simulation, ASHRAE 0.4% summer and 99.6% winter design conditions were assumed [23].An internal or external shading system was not used in the simulations.

Internal temperature, ventilation and infiltration rate, and HVAC system
From 5:00 till19:00 the system controls the internal air temperature to a cooling and heating temperature setpoint of 23.9°C and 21.1°C, respectively. During the nighttime the system is switched off. The infiltration was assumed equal to 0.000333 m 3 /(s m 2 ) (flow per exterior surface area). The minimum outdoor air flow rate was set to be 0.762 L/(s m 2 ) (flow per gross area) and was provided from 5:00 until 19:00. Regarding the airflow outlets of each zone, the air is distributed through swirl diffusers in interior zones and linear bar grilles in the perimeter zones. Variable-speed fan coil units (FCU) supply air to the bar grilles in the perimeter zones only. The FCU shuts off when zone temperatures are in the dead-band between 21.1 and 23.9°C. In cooling mode the fan is on and the heating coil is off, and in heating mode, the fan and the heating coil are on. The building is served by a single variable-speed central station air handling unit (AHU) including an economizer, chilled water cooling coil, hot water heating coil and supply fan. The AHU fan is controlled with a static pressure reset strategy. The central plant consists of a centrifugal chiller with variable-speed pumps and a two-speed cooling tower. A gas fired boiler provides hot water to all heating coils. Table 1 shows details of system and plant inputs.

Internal heat gains and occupancy
In this paper and in Figure 3, the fraction of the design value is defined as the ratio of the actual load to the peak or maximum load. The occupants' presence in the building varied according to Table 2 and Figure 3. Three levels of internal cooling load (people, equipment and lighting) were simulated, and they are summarized in Table 2. The equipment and lighting loads follow the schedules shown in Figure 3.  Figure 3. Occupancy, lighting, equipment and HVAC schedules.

Simulated cases
The purpose of the study is to investigate the influence of thermal decay on the UFAD thermal behavior. Some important parameters affecting the thermal decay that were studied are: floor level, zone orientation, central air handler supply air temperature (SAT), climate, window-to-wall ratio, internal load, and plenum configuration. The simulated cases are listed in Table 3. Regarding the plenum configuration, the "series" option indicates that all the conditioned air leaving the air handler first enters the interior plenum and, after gaining heat due to thermal decay, the warmer air leaving the interior plenum then enters each perimeter plenum. In the "parallel" plenum configuration, the conditioned air from the AHU independently enters each plenum in parallel (equivalent to ducting supply air out to a separate (subdivided) perimeter plenum. "Ducted" option is an idealized configuration in which the conditioned air is ducted all the way to the diffusers and thus, no thermal decay exists. As shown in Table 1, AHU fan design static pressure for series and parallel configurations use 750 Pa, as opposed to 1025 Pa for the ducted configuration (equivalent to a fully ducted overhead air distribution system).
In addition, three different levels of internal load are summarized in Table 2. * Air handling unit supply air temperature (ºC) ** Window-to-wall ratio (%) *** Details summarized in Table 2 3. RESULTS AND DISCUSSION

Heat transfer fundamentals of thermal decay
EnergyPlus performs a fundamental energy balance on each building surface by explicitly calculating all three heat transfer components (radiation, convection, conduction) for each hourly time-step of the simulation. Figure  4 shows a schematic diagram identifying the important horizontal building surfaces in a UFAD system. Thermal decay is caused by heat transfer into the underfloor plenum through the concrete slab (from the return plenum below) and through the raised floor panels (from the room above).  surface of the concrete slab, the results indicate that there is more conduction occurring at the top of the slab in the morning hours than in the afternoon hours due to thermal storage effects and the operating schedule of the HVAC system. When the UFAD system is turned on in the morning, heat that has been stored in the slab overnight is released into the cool plenum by convection. The time delay in heat conducting through the slab is also demonstrated by the fact that the pulse of heat entering the bottom of the slab during the late afternoon hours (peak loads) never reaches the top of the slab during normal operating hours. The storage of this afternoon heat in the slab clearly contributes to the higher conduction values at the top of the slab in the morning. The magnitude of the convective heat flux from the top of the slab into the plenum air (contributing to thermal decay) varies between about 6-8 W/m 2 over the day. The slight increase in the late afternoon is actually due to radiant heat transfer down from the underside of the floor panels.    Figure 6 illustrates the hourly variations of diffuser discharge temperature for each zone in the middle floor during the summer design day period for Cases 1, 2 and 3 (see Table 3). By comparing Cases 1 and 3, the influence of the thermal decay can be clearly observed. As shown, the series plenum configuration (Case 1) always shows higher discharge temperatures than the air handler supply air temperature (SAT) setpoint due to heat gain in the underfloor plenum. On the other hand, in Case 3 it shows a constant discharge temperature of 15.6°C, which is the same as the air handler SAT (duct heat gain is assumed to be zero for Case 3). In Case 1 it can be observed that the average temperature rise during occupied hours is no lower than 4.6 K for north, 4.8 K for east, 4.6 for south, 4.2 K for west and 2.6 K for the interior zone. The temperature rise in Case 1 is always higher in the perimeter zones than in the interior (core) zone. This is due to the "series" plenum configuration where all the conditioned air leaving the air handler first passes through the interior plenum and the warmer air leaving the interior plenum then enters each perimeter plenum. Therefore, the discharge temperatures in the perimeter zones are the result of two temperature rises: first in the interior and then in the perimeter plenum, making the diffuser discharge temperature of the perimeter always higher than that of the interior. The sharp peaks in the early morning are due to the start-up condition. During the night (the system operation is from 6 am to 10 pm), the system is off, no conditioned air is supplied into the plenum and thus the heat is accumulated in  the plenum. When the system starts to operate in the early morning, the accumulated heat is removed first, producing a high plenum temperature rise in the early warm-up period.    Figure 7 illustrates the thermal decay for the middle floor interior and west perimeter zones. For the west perimeter zone, Case 1 (series plenum) shows higher thermal decay than Case 2 (parallel plenum) due to the fact that the plenum air gains heat as it passes through the interior plenum zone on its way out to the perimeter in the "series" plenum configuration, as discussed earlier. On the other hand, for the interior zone, Case 2 shows higher thermal decay then Case 1 due to the lower airflow rate passing through the interior supply plenum in parallel plenum configuration. Unlike the series plenum configuration, the supply airflows for the perimeter zones do not pass through the interior plenum in Case 2. In Case 3, thermal decay is assumed to be zero for all zones. Figure 8 presents the hourly supply airflows of the west and interior zones in Cases 1, 2 and 3 during the design day period. For the west perimeter zone, Case 1 (with thermal decay) shows higher airflow rates than Case 3 (no thermal decay), as expected, because for Case 1 the larger supply airflow rate results from the increased diffuser discharge temperature. The increase in airflow during the afternoon is due to the solar radiation reaching the west zone. However, in the interior zone, the supply airflow turns out to be higher in Case 3 without thermal decay than in Case 1 with thermal decay. This is due to the fact that there is no conditioned air supplied into the plenum in Case 3 and thus the heat is continuously accumulated in the plenum, increasing the surface temperatures of the adjacent raised floor and the concrete slab. Therefore, as can be seen in Figure 9 for the interior zone, the convective heat gain to the room from both raised floor and suspended ceiling in Case 3 (without thermal decay) is much higher than those in Case 1 (with thermal decay), therefore raising the supply airflow rate despite the low diffuser temperature in Case 3. Although Case 3 does not represent a system configuration that occurs in practice, the results demonstrate that EnergyPlus is properly accounting for the heat transfer processes.    Figure 10 shows a comparison of room cooling load and supply air to room temperature difference (DeltaT rm ) for the West perimeter zone for Cases 1 and 3. While the cooling load is greater for Case 3 as previously discussed, the DeltaT rm is proportionally much greater causing the room airflow to be less for Case 3. This is reflected at the system level as shown in Figure 11 where air handler (AHU) airflow and supply return temperature difference (DeltaT sys ) are shown. However, there is only a small difference in the product of these two factors (i.e., system demand) for the two cases. On an annual basis (not shown) Case 1 demand is about 10% greater than Case 3, apparently due to heat transfer and thermal storage differences. Figure 6 through Figure 11 show that the thermal decay can have a large impact on the supply air flow rate and temperature. These affect fan and chiller energy consumption, indicating that the thermal decay should be carefully considered to assess UFAD performance. Although not reported here, the cooling load split between the room, supply plenum and return plenum was analyzed. Results similar to those reported in [14] were obtained.

Parallel configuration
In addition to the series plenum configuration, another possible option is the parallel configuration where the conditioned air from the AHU independently enters each plenum in parallel. Since the plenum configuration can have significant impact on the thermal decay, it is worth looking into this configuration in detail. As opposed to the series configuration where AHU supply air first enters the interior plenum and then the perimeter plenums in series, the parallel configuration can reduce the perimeter diffuser discharge air temperature. The diffuser temperature difference between the series and parallel configurations are illustrated in Figure 6. As shown, the parallel configuration (Case 2) has lower perimeter and higher interior discharge temperatures compared to the series configuration (Case 1). The lower perimeter discharge temperature is due to the fact that there is only one temperature rise in the parallel configuration as opposed to the series configuration where there are two temperature rises, first in the interior and then in the perimeter plenum. The zone supply airflows for each zone in Cases 1 (series) and 2 (parallel) are also presented in Figure 8. As shown, the parallel configuration has lower perimeter but higher interior supply airflows compared to the series configuration. The lower temperature rise in the perimeter plenum ( Figure 6) reduces the perimeter supply airflow, while the higher plenum temperature rise in the interior plenum increases the interior supply airflow.

Influence of thermal decay on energy consumption
In this section is reported a simple evaluation of the impact of plenum configuration on energy consumption. The annual HVAC primary "source" energy usage of the three different plenum configurations under Baltimore climatic conditions is summarized in Figure 12. The secondary to primary conversion factors of 3.167 and 1.084 are assumed for the electricity and natural gas, respectively [19]. Comparing Case 1 to Case 3, the chiller and fan energy use are increased by the effects associated with thermal decay by 19% and 8.5%, respectively. As expected, Case 3 without any thermal decay shows the lowest energy consumption compared to Case 1 and Case 2. The annual total source HVAC energy of Case 3 is 12.7% less compared to baseline Case 1 with the series configuration. Comparing Case 1 to Case 2, the parallel plenum configuration (Case 2) has less thermal decay which reduces the chiller and fan energy in much the same way as Case 3 does. However, the heating energy for Case 2 is increased (due to increased reheat in the perimeter) compared to Case 1, and thus there is a tradeoff between cooling and heating energy between these two different configurations, making the overall HVAC energy difference between Case 1 and Case 2 only 2.2%. From this simple analysis, it appears that reducing the thermal decay may lead to energy savings.

Sensitivity analysis: Parameters affecting thermal decay
The impact of variations in key parameters on the plenum temperature rise is shown as box-plots in Figure 13, Figure 14 and Figure 15 derived from annual simulations of the model described in Section 2.2. These parameters include the floor level, zone orientation, central air handler SAT, climate, window-to-wall ratio, internal load and plenum configuration. A box-plot is a way of graphically summarizing a data distribution. In a box-plot the dark horizontal line in the box shows the median value. The bottom and top of the box show the 25 th and 75 th percentiles, respectively. The horizontal line joined to the box by the dashed line shows either the maximum or 1.5 times the interquartile range of the data, whichever is smaller. Points beyond those lines are outliers. The interquartile range is the difference between the 25 th and 75 th percentiles [24]. In Figure 13, Figure  14 and Figure 15 each zone of each floor contributed one data point. The relevance of each point has not been weighted by the area represented by the zone. The aim of this paper was not to develop a predictive model of the thermal decay (in this case weighting by floor area would be needed) but to investigate the impact of thermal decay on UFAD system behavior and to study which are the parameters that affect the most thermal decay.
For all the simulated cases, the median temperature rise is 3.7 K during the occupied hours with 50% of the values between 2.4 and 4.7 K. In Figure 13b the box-plots of the temperature rise for the three different plenum configurations are presented. The negative thermal decay indicates that there is a heat loss instead of heat gain, but it represents a very small frequency as shown in Figure 13a. The results indicate that the series configuration shows the highest temperature rise as expected. In the series configuration, the conditioned air should travel through both interior and perimeter plenums before entering the perimeter terminal units, while it needs to pass through only the perimeter plenums in the parallel configuration. Although the temperature rise in the interior zone is lower in series configuration compared to parallel configuration, the higher temperature rise in the perimeter zones in the series configuration turns out to increase the overall temperature rise as shown in the figure. The ducted options do not show any temperature rises due to the modeling assumption of no duct heat gain. By comparing (a) and (b) graphs of Figure 14 showing the results of each zone, the parallel configuration (Case 2) in Figure 14 shows higher temperature rise in the interior, but shows lower temperature rise in the perimeter zones, eventually decreasing the overall temperature rise compared to series configuration (Case 1). The temperature rise is not significantly affected by the perimeter zone orientations. The median was around 3-4 K for all the four perimeter zones, while the distribution slightly varies among each zone.
. Lee  The thermal decay for different floor levels (ground, middle and top floors) is illustrated in the first plot of Figure  14. The results indicate that the ground floor has less temperature rise compared to middle and top floors. This is due to the fact that the ground floor slab is in direct contact with the relatively cooler soil, while the bottom surfaces of middle and top floor slabs are facing higher temperature return plenums of the floor below, increasing the heat gain of the plenum air from the concrete slab. The second plot of Figure 14 presents the results for each month. The median value increases in the summer seasons and decreases in the winter seasons due to the higher heat gains in the summer compared to winter. Although there are some hours when the temperature rise goes below zero in the winter, i.e., the plenum air is delivered into the room at a lower temperature than when it is supplied into the plenum, the median value in the winter is positive, indicating that temperature rise takes place even in the cold winter period. In addition, the number of hours when thermal decay is negative is very small, as indicated in Figure 13a. The third plot of Figure 14 illustrates the sensitivity to the central air handler SAT. As the SAT increases, the thermal decay decreases mainly due to the increased supply airflow with higher SAT and the smaller temperature difference between the plenum air and the heat transfer surfaces (slab and floor panels). The median value for SAT 13.9°C is 4.6 K as opposed to 2.6 K in SAT 17.2°C. Given the similar cooling load of the room, the supply airflow should increase as the SAT gets higher, which, in turn, reduces the temperature rise in the plenum. The annual fan energy consumption directly connected to the supply airflow explains this discussion. The fan energy of SAT 13.9°C case is 13.9 kWh/m 2 , while it is 16.1 kWh/m 2 for SAT 17.2°C case, representing a 16% increase in average airflow for SAT 17.2°C. The last plot of Figure 14 shows the climate dependency of the thermal decay.
As expected, the hot Miami climate (MFL) causes the greatest heat gain in the plenum compared to other mild and cold climates. The mean value of Miami is 4.7 K, while Baltimore, Minneapolis and San Francisco were 3.6