During the dehumidification process, the liquid desiccant is in contact with air through a permeable membrane that allows water vapor interaction but prevents the flow of LiCl into the air. LiCl absorbs the water vapor in the air until it reaches water vapor pressure equilibrium with the air. This process is exothermic. The desiccant is kept cool by the evaporation of water to the exhaust heat stream. Figure 34 shows the dehumidifier unit schematic. Equations for each stream are presented below. Mass transfer is driven by a vapor pressure differential between the air and desiccant solution as shown in Equation 66. The supply air side heat transfer for the dehumidifier is given by Equation 68. The heat transferred to the solution includes the sensible heat due to the temperature difference between the desiccant and air, and desiccant and water, plus the latent heat of absorption and enthalpy of dilution, given by Equation 72.Absorption process weakens the desiccant solution and reduce its ability to absorb water vapor. To desorb water vapor from LiCl, the desiccant is heated to have equilibrium water vapor pressure that is higher than that of the air. This regeneration process is the reverse of dehumidification and can use low grade heat sources. In this study, the SOFC system exhaust heat is used in this regeneration process to increase the concentration of LiCl in solution. Then, the concentrated liquid desiccant solution is stored. When moisture must be removed, the high concentration solution is used to dehumidify the outside air. Figure 36 shows the regenerator schematic.The dehumidifier’s model inlet parameters are the weather conditions, the return air condition,cannabis drying rack desiccant temperature and concentration, and cold water temperature and flow. The model output is supply air temperature, desiccant outlet temperature and exhaust air temperature.

In order to keep the humidity of the air cooling the servers below the allowable limits, the air humidity in the dehumidifier is controlled by manipulating the percentage of return air. In this model, the desiccant outlet concentration is also controlled by the desiccant flow rate. In the regenerator system the inputs of the model are weather condition, desiccant inlet temperature and concentration, and hot water temperature and flow. The outlets are air temperature, desiccant temperature, and hot water temperature. To use the desiccant for dehumidification purposes, it is required to regenerate it to a certain concentration. In this model the manipulating parameter to control the desiccant concentration is the desiccant flow rate. The validity of the model can be assessed by comparing its predicted supply conditions to the measured supply conditions. These comparisons are done for each stage independently: the first-stage dehumidifier and the second-stage Indirect Evaporative Cooler . Experimental data from DEVap prototype testing is used to verify the dehumidifier and indirect evaporative cooler in the next two section, respectively.For indirect evaporative cooler, 5 different cases from have been used to verify the model. Table 7 shows the input conditions as well as experimental outcome and model output for supply air temperature and relative humidity. Figure 38 compares the model predictions and the experiments of the relative temperature of the supply side air. For the Indirect evaporative cooler, the measured supply-side temperature change predicted by the model matches the experiments within 10% except for test number 3. In case 3 the temperature difference is higher and has low mass flow which shows the weakness of bulk model to predict the result and the need for a discretized model. Modern data centers try to use adiabatic cooling whenever the weather condition allows. However, adiabatic cooling is not possible in all locations at all times. Different types of common data center cooling systems were presented in chapter 2.

This chapter presents the method for calculating the cooling demand of data center at various locations. Also, the cooling demand for seven different data center locations that are used as case studies of this research are analyzed. In order to calculate the amount of cooling required for data center a MATLAB data center model has been developed which calculates the amount of cooling required by a data center. The inputs of the model are weather data associated with data center locations including temperature, pressure, and relative humidity. The weather data are obtained on an hourly basis from Typical Meteorological Year data from 2006 to 2016. The model takes this data and contains multiple functions that have been developed for calculating thermodynamic parameters such as saturated temperature, wet bulb, and dew point temperatures based upon the knowns weather data. In order to calculate the load, the acceptable operating conditions for servers within a data center are required. ASHRAE is the association that updates and releases an industry standard for data center operations every couple of years, based upon industry technology improvements. Table 8 shows the boundaries that define the ASHRAE recommended and allowable environmental envelope from the 2016 standard.In order to calculate the number of hours that the data centers in each location need mechanical cooling, TMY data for seven locations in the United States that are home to Microsoft data centers have been used as the input for the code. The number of hours of each cooling type that is required in each location based on both allowable and recommended envelope is shown in Figure 39. As expected, by expanding the range of temperature and humidity, the number of hours that mechanical cooling is required decreases. For data centers located in California, Seattle, and Wyoming a mechanical cooling system is barely required, while economizer and evaporative cooling will be sufficient throughout the year to keep the servers in acceptable range. However, Illinois, Iowa, Virginia and Texas require between 1000hr to 4500hr of mechanical cooling based on the location and ASHRAE requirements. Server load profiles tend to be confidential information that are rarely published.

However, the profile is roughly constant and utilization changes between 60% to 80%. For the current work and data center simulation results, either NREL published server load profiles as shown in Figure 40 are used by scaling it to the size of the targeted data center, or it is assumed that the utilization is constant at 70% throughout the entire operating period. The following data center simulation results for each location are based upon the assumption of a 50MW designed data center that follows the load demand of. The designed temperature difference of air entering and leaving the servers is 15℃. The number of cooling hours and cooling device correspondent to that is based on ASHRAE recommended envelope. The mechanical cooling system in these results is assumed air cooled chiller. Figure 41 to Figure 44 show the results for California and Texas which are the two ends of the spectrum with California being the location with the lowest overall energy use and Texas the highest. Figure 41 shows the TMY dry bulb and wet bult temperature for California and Texas which are the parameters that determine what type of cooling is required for the data center. California and Texas average dry bulb temperature are 14℃ and 20.2℃ and wet bulb temperature are 11.6℃ and 15℃, respectively. California has the least variation in temperature throughout the year while Texas temperature changes more than 40℃ during the year which has a significant impact on change in the cooling required for Texas.PUE is the ratio of total energy used by facility to energy used by the serves. This parameter shows how effectively a data center uses energy. As the number gets closer to 1 it means that the facility becomes very efficient, with most of the energy being directly converted in the servers for the computational demands of the data center. The following two graphs shows the PUE for the entire year. The spikes that bring PUE up to the 1.4- 1.5 range are because of energy being consumed by mechanical coolers. The average PUE for California and Texas for the whole year as simulated with the current model are 1.16 and 1.32, respectively. The results for the TMY data, Power usage breakdown,vertical grow system percentage of energy usage, and PUE for the other 5 locations are presented in APPENDIX A. Figure 45 shows energy use for each location for a 50MW designed data center following Figure 40 load demand. The air temperature difference is 15℃ and ASHRAE 2016 standards is followed for temperature and humidity limits. Table 12 shows the average PUE for all the locations with California having the lowest PUE at 1.167 and Texas the highest at 1.315.The type of cooling system, designed temperature difference, and changing allowable range has a significant effect on the amount of energy that data center consumes. For example, as the technology is rising IT manufacturers are pushing the boundaries on safe temperature that IT equipment’s can tolerate. Figure 46 and Table 13 show the energy used and percentage consumed by different part of data center for various combinations. Water cooled system use less energy than air cooled system. Increasing the temperature difference for the air entering and leaving the server room means less flow of air is required, which leads to less energy required for cooling the air. As the IT technology progresses, the IT equipment can tolerate higher temperature which leads to higher range of acceptable temperature and humidity. This means wider range of outside temperature is acceptable for cooling the server, leading to lower energy usage. In this chapter, a data center cooling model has been developed to calculate the amount of cooling required by a data center.

The model takes the weather data for each location and acceptable range of temperature and humidity for data center to calculate the load. In addition, the cooling demand for California, Seattle, Wyoming, Illinois, Iowa, Virginia, and Texas have been calculated and analyzes. Texas had the highest cooling demand with a PUE of 1.315 and California had the lowest with a PUE of 1.167. Energy usage of data center based on different types of cooling device and at different designed temperature difference has been compared. Results showed that higher temperature difference and water-cooled system lead to less energy consumed by the cooling system. In this section, the possibility of using a highly efficient, zero emission SOFC system to produce electricity and cooling in various amounts to meet electricity and cooling demands of a data center is investigated. In this configuration each fuel cell powers one server rack and heat from each individual SOFC system in used in a small-scale LDD to produce cooling for one server. Figure 47 shows the integrated system configuration for rack level power and cooling. For this analysis, each server rack power is considered nominally 12kW. We assume that a fuel cell equivalent to eight 1.5kW BlueGEN SOFC systems is used to meet the server electrical demand . The exhaust of the SOFC is used to regenerate LiCl liquid desiccant to provide 1400CFM cold and dehumidified air for each server rack. Figure 48 shows the integrated SOFC-LDD system. The SOFC exhaust gas produces hot water that will supply the heat demand for regenerating the liquid desiccant. The regeneration process occurs within a heat and mass exchanger where the vapor desorbs from the desiccant and is carried away by an air stream due to desiccant solution that has higher water vapor pressure than the air. The high concentration LiCl is stored in a tank. When air conditioning is required, the high concentration LiCl is used to dehumidify the air. As mentioned before the ASHRAE recommended suitable range of temperature and humidity for all environmental classes inside the data centers is 18˚C to 28˚C dry bulb temperature and 9˚C to15˚C dew point and 60% RH . Figure 49 shows the model results of the first test on a psychrometric chart. The green line labeled ‘LiCl – 35%’ shows the humidity ratio of the air in equilibrium with the liquid desiccant at a mass fraction of 0.35 LiCl and at each of the temperatures considered. Line black shows the first stage dehumidification process for supply air and then the cooling air process. The dehumidification process is internally cooled by evaporation of water to exhaust air to keep the desiccant cooler and increase its dehumidification potential. Then, the supply air is cooled by indirect evaporative cooling at constant humidity ratio. Horizontal black line shows the second stage process for supply air, which goes through indirect evaporative cooling at constant humidity ratio.