| Direct Expansion (DX) unitary systems are currently used in providing thermal comfort for a wide array of building types, sizes and in different climates. The design of these systems constitutes a very large impact on the energy usage and operating cost of the buildings they serve. Many of these systems currently in use today are not properly designed or sized correctly, causing excess amounts of energy wasted. The ability to predict the performance of these systems is integral to designing more energy-efficient and sustainable building systems. This thesis proposes the use of Artificial Neural Networks (ANNs) as a modeling tool for predicting the steady-state performance of DX Heat Pump systems. ANNs are computational models based on the biological central nervous system, and are trained based upon a set of given input and output data. The input variables considered are the mixed airflow rate, mixed airflow temperature, the mixed airflow moisture content and outside dry-bulb temperature. The output variables include the temperature of the supply air, and compressor power output of the system. The physical data was collected for both heating and cooling modes, from a 3-ton DX Heat Pump unit. This system was specially modified for research purposes, and is located North Carolina A&T's HVAC Laboratory in Graham Hall. The Neural Network was then trained with this given data, and its prediction results were compared to the actual data using the mean square and regression analyses. The testing results validate the Artificial Neural Network model created as an accurate tool for predicting the energy performance of the Heat Pump system. |
