This study proposes a new computational model in which HVAC&R (Heating, Ventilation, Air Conditioning, and Refrigeration) systems are automatically simulated using only data. Unlike the existing simulation method, it is possible to understand the power and performance of the device at each time by predicting the parameters based on the physics equation with the machine learning model and substituting the values back into the physics equation. We developed a flow model that predicts the load to be handled by each device and equipment models that calculate the device's performance. The model's performance was excellent, and the highest accuracy was obtained by sequentially inputting the input variables ranked using the Spearman correlation coefficient for both the flow rate and each equipment model. In the case of pumps, all five pumps satisfied CV(RMSE) of 30% in Cases 07 to 09, to which the Spearman correlation coefficient was applied. The CV(RMSE), due to the outlet temperature of the turbo chiller, heat exchanger, and heat storage tank, satisfies ±30%. This method overcomes the limitation of not reflecting the physical characteristics of devices in data-based modeling (black-box) of HVAC&R systems. Also, this reflects the current system status from data, and it could solve the limitations of the existing simulation method (white-box) that relied on design documents. As a result, it is possible to obtain reliable calculation results by reducing the modeler's subjective intervention. Because it is modeled on device units, it can also be used for fault detection and optimal operation control calculations inside the system.
!!!All Science Journal Classification (ASJC) codes