Paper Conference

Proceedings of ASim Conference 2016: 3rd Asia conference of IBPSA-China, Japan, Korea


Possibility on Increasing Timesteps of Radiant Systems with Intelligent Load Predictions for Reducing Calculation Time

Woong Chung, Myoung Souk, Kwang Kim

Abstract: As the occupants in the building require better quality of the thermal environment for the higher productivity, smart buildings started to apply the predictive control for maintaining the thermal comfort. In order to reflect actual building conditions, the data-driven method is used with intelligent method, which can explain the correlation between the significant parameters and building load from the historical data. Recent researches propose smaller timestep for load prediction in order to precisely maintain the thermal comfort. However, intelligent method with smaller timesteps may consume a lot of calculation time that will cause the difficulty on applying the predictive control when calculation time is longer than the timesteps. Therefore, possibility to increase the timesteps of load predictions for reducing the calculation time should be inspected without significant changes in thermal comfort. In order to inspect the possibility to increase the timesteps without changing the thermal comfort, the characteristics of the thermal comfort should be analyzed. One of the most popular index to indicate the thermal comfort is Predicted Mean Vote (PMV). And in predicted mean vote, the radiant temperature is one of the most important factors. Since radiant temperature may be changed with different terminal systems, thermal comfort and mechanisms of radiant system and air system was analyzed to determine the appropriate timesteps for each system. Typical office building was simulated with EnergyPlus and one of the common data-driven method, artificial neural network, was selected. Variety of timesteps were used for the simulation and PMV was analyzed to select the appropriate timesteps. As a result, since radiant system may provide the better thermal comfort, the timesteps of load prediction may be longer than the timesteps of load prediction for the air system. Possibility of longer timesteps of the load prediction may open the possibility to reduce the calculation time maintaining the thermal comfort.
Keywords: Timesteps, Calculation Time, Neural Network, Predicted Mean Vote