Development of a Grey-box Based Building Load Model Using TRNSYS Components

Abstract

Model predictive control has the potential to reduce energy consumption during the operational phase of a building by predetermining optimal operating strategies. However, while the predictive performance of the model is important, there is very little data available from actual buildings in operation. Therefore, this study proposes a method to build a physical model that can predict indoor temperature using measured indoor thermal environment data (only partial data, mainly for air conditioning) for an actual operating building. Furthermore, a modeling methodology using TRNSYS components, which can be interlocked on the same platform as the air conditioning system is proposed. By comparing the simulated indoor temperature of the proposed model with the measured values, the prediction accuracy and temperature distribution pattern were improved compared to the original uncalibrated model. This improvement in accuracy was achieved by calibrating the model to the optimal parameters for each season through clustering. In addition, the 7.05% CVRMSE met ASHRAE Guideline 14 criteria, demonstrating its potential for future use in MPC.