Journal of the Korean Solar Energy SocietyISSN:1598-6411(Print) 2508-3562(Online)한국태양에너지학회

Utilizing the machin learning (ML) algorithm for the building cooling/heating operations requires the actual building data. The robustness of the ML-based controller depends on the amount and quality of the data. However, reserving actual building operation data is challenging because of the cost and time required. This study proposes the methodology for generating the synthetic simulation data the development of the ML (specifically, reinforcement learning) algorithm when the actual data are lacking or unavailable. As the first step toward investigating this methodology, we built two building models, an EnergyPlus simulation model and a grey-box model. The prediction performances of two models were quantified and evaluated.