A Calibration Method based on Heteroscedastic Evolutionary Bayesian Optimization for Simulation System of HVAC

In recent years, increasing the energy efficiency of buildings has become one of the objectives of facility managers. Advanced control methods can be used to improve the efficiency of Heating, Ventilation, and Air Conditioning (HVAC). However, advanced control methods cannot be effectively applied due to inconvenient commissioning. Therefore, simulating and modeling the HVAC system becomes particularly important for verifying various control algorithms. Modeling and calibrating the HVAC simulation system becomes crucial for the application of various control strategies based on actual operational data, while current calibration methods are usually based on experience. In this research, a method for the calibration of an HVAC model based on Heteroscedastic Evolutionary Bayesian Optimization (HEBO) is presented. Firstly, the HVAC model is built using the Modelica language in the Dymola environment. Secondly, real operation and weather data are fed to the HVAC model to predict the energy consumption of various equipment. Thirdly, the Root Mean Squared Error (RMSE) between predicted and real values is minimized by HEBO so that the optimal parameters of the equipment can be obtained. Finally, a case study is conducted to verify the effectiveness of the proposed method. The results show that the CV(RMSE) of calibrated models meets the guideline requirements of ASHRAE, IPMVP, and FEMP.