RL vs. MPC for HVAC: Energy savings achieved, but RL faces comfort challenges

A new study published on arXiv compares the effectiveness of Reinforcement Learning (RL) and Model Predictive Control (MPC) for residential HVAC systems. Both methods demonstrated energy savings compared to traditional controls, with RL achieving slightly higher savings (20.9%) than MPC (18.1%). However, RL experienced initial difficulties with occupant comfort, leading to three reports of discomfort during its adaptation phase, while MPC maintained acceptable comfort levels. The research suggests RL requires less engineering effort for deployment but faces challenges in safe initialization and state/action space mismatches. AI

IMPACT This research highlights potential AI applications in energy efficiency for residential systems, though practical deployment challenges like occupant comfort need further attention.

RANK_REASON Academic paper comparing two control methods for HVAC systems. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv cs.LG →

• arXiv
• model predictive control
• Ozan Barış Mulayim
• reinforcement learning

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →