Predictive HVAC Energy Modeling using a Hybrid Simulation and Neural Network Approach: A Case Study

Fast, reliable energy prediction is critical in domains such as marine, aerospace, industrial, residential, commercial, medical facilities and educational buildings, where variable occupancy, stringent environmental requirements, and mission-critical operations prevail. Motivated by these application needs, this case study analyzes an educational building in Tehran to examine how changes in classroom-occupancy parameters affect HVAC energy consumption. Using Carrier HAP, we varied the occupancy parameter and simulated two standard system configurations—the air-cooled chiller with fan-coil units (FCUs) and the split direct-expansion (DX) system—then compared their performance. The simulation outputs were subsequently used to train an artificial neural network (ANN) for rapid prediction. Higher occupancy consistently increased both cooling and heating loads; the split DX system performed better in summer, whereas the chiller-FCU configuration was more efficient in winter. The ANN achieved strong predictive accuracy on the simulation data (MAPE: ^,V% for cooling, ۱,۱% for heating).

کلیدواژه ها:

HVAC ، Artificial Neural Network (ANN) ، Carrier HAP simulation ، Energy consumption ، Energy Prediction