LSTM-Based Model Predictive Control for Optimal Temperature Set-Point Planning

The growing interest in saving building energy has increasingly motivated studies on model predictive control (MPC), where system operation proceeds according to a planned strategy. Data-driven models that perform learning using past data of buildings are favorable for MPC applications owing their fast computation speed. However, it is difficult apply with insufficient data, as the prediction accuracy varies depending used learning. To address this, we propose method involves generating through detailed and utilizing long short-term memory (LSTM) network performs an model. was verified comparison reference same optimization algorithm. In objective function, which reduce electrical expenditure by optimizing indoor temperature target building, approximately 35% grid consumption reduced compared case, increasing self-consumption photovoltaic (PV) avoiding PV curtailment. Further, required time 30%, even including generation daily learning, thereby confirming feasibility employs LSTM.