Optimized Control of Thermally Activated Building System in Office Buildings

Abstract

Building environments have been controlled by using heating and cooling systems to maintain comfortable interior temperatures. However, a recent popular trend in building heating and cooling system involves utilizing a radiant system to achieve energy savings and comfortable temperatures. One type of radiant system uses the concrete structure of a building as the heating and cooling system by embedding pipes inside it. This system has many advantages, such as reduced installation time, reduced building material requirements, and non-necessity of increasing the height of the building. Because the system uses the entire concrete structure of a building, individual control may be difficult to perform. Thus, the system was designed to remove the basic heating/cooling load in a building. Although many studies propose utilizing different parameters to determine the basic heating/cooling load of a building, the most common method used to remove the basic load involves keeping the radiant surface as the room setpoint temperature. When the surface temperature of the Thermally Activated Building System (TABS) is kept at the setpoint temperature of the room, the system will remove load based on how the air temperature of the room changes from the load. This concept is called the self-regulation effect, and was useful for removing load without any feedback. Because only partial heating and cooling load can be removed by TABS via the self-regulation effect, the remainder of the load was removed by an air-based heating and cooling system. In a well-designed building with a small amount of heating and cooling load, the self-regulation effect can be very effective. However, the current method used for self-regulation was applied by supplying water at the room setpoint temperature. In addition, the load able to be removed by TABS is limited in a building with a significant amount of heating and cooling load. Thus, the objective of this study is to identify the thermal mechanism of TABS and increase its utilization by adjusting the supply water temperature depending on the load. To increase the utilization of radiant systems, the current TABS control method should be carefully observed to determine how it can be improved. Because the self-regulation effect was applied by supplying water at the room setpoint temperature, the core layer where pipes are embedded will be close to the room setpoint temperature and the surface temperature of TABS cannot be kept at the room setpoint temperature from consistent effect from the load. Therefore, the heat exchange between the surface of TABS and room air temperature is lower than expected and utilization of TABS is decreased. Moreover, the air system removes the remainder of the load and keeps the room air temperature at the setpoint temperature. The amount of load removed by TABS when following the self-regulation concept decreases as the setpoint temperature is consistently met. Consequently, the active use of TABS should be executed by targeting the specific basic load. For the active utilization of TABS, its thermal mechanism should be analyzed and the target load should be identified. The thermal mechanism of TABS demonstrates a significant amount of time delay due to the large heat capacity of concrete, which was one of the advantages of TABS. However, time-delay of TABS also means that the system needs a large amount of time to supply the heat into the room. With the various changes on supply water temperature and load, the TABS should use at least daily control instead of hourly control. Once the control timestep was chosen, the target basic load can be chosen by using the minimum load achieved over a 24-hour load period. Thus, a load prediction should be performed to determine the basic load. The minimum load over a 24-hour period was identified by studying load patterns based on historical data. In this study, two load prediction methods were utilized to select the supply water temperature for TABS. One popular method involved using the outdoor air temperature to calculate the temperature of the supply water. The heating and cooling curves are derived using a resistance-capacitance (RC) network. The second method is an intelligence-based method, and uses an artificial neural network (ANN) to recognize the load pattern. In the process of learning the load patterns, the input parameters of the ANN were selected by an analysis that divided the building load into external load, solar load, and internal load. The input parameters should be obtained prior to the prediction, and are defined as follows. The input parameters that consider the external load, solar load, and internal load were outdoor air temperature, cloud coverage based on weather forecasts, and type of day. Using these three input parameters, the accuracy of the load prediction became reliable after approximately one month of pattern learning. Because accuracy was poor during the ANN learning period, this study proposes the use of load prediction based on outdoor air temperature during the ANN learning period, and then use the ANN prediction after it has been validated. Based on our understanding of the thermal mechanism of TABS and load prediction methods, the goal of removing the minimum load of a building over a 24-hour period (using TABS) was executed using a co-simulation involving EnergyPlus and MATLAB. EnergyPlus was used to realize the actual building environment, and MATLAB calculated the supply water temperature and performed load predictions with the information obtained from EnergyPlus. The Building Controls Virtual Test Bed (BCVTB) was used as middleware connecting the two simulations to exchange information at each timestep. TThrough the co-simulation, the thermal output of TABS with different control strategies was compared to verify the utilization of TABS. Among control strategies, predictive control with ANN demonstrated the greatest thermal output of TABS. The validation of the method was executed by applying different weather conditions and showed equivalent results.