상업용 건물 내 HVAC 시스템의 온라인 지도학습 기반 최적 수요반응

Abstract

This paper proposes an online supervised-learning-based strategy for the optimal operation of a heating, ventilation, and air conditioning (HVAC) system in a commercial building. Conventionally, it requires time-consuming techniques to estimate physics-based parameters, preventing practical applications to building energy management systems (BEMSs). In this paper, an artificial neural network (ANN) is trained online with building energy data and equivalently represented using a set of linear and nonlinear equations. An optimization problem is formulated using the equation set and repeatedly solved for every day, producing data on optimal HVAC load schedules under various conditions of building thermal environments. Given the data, a BEMS updates the ANN-based model of an HVAC system and the optimization problem for optimal HVAC load scheduling. The case study results verify that the proposed strategy is effective in exploiting an HVAC system as a price-based demand response resource.