A Multi-Agent Reinforcement Learning Approach to real-time Demand Response in Cruise Ship Cabins

This paper explores a multi-agent reinforcement learning approach for real-time control of HVAC systems during demand response events. The heating, ventilation and air conditioning (HVAC) systems are major energy consumers on cruise ships. At the same time, the large time constants of HVAC systems make them a strong candidate for demand response, as energy consumption can be adjusted with delayed effects on occupant comfort. However, this inherent slowness also makes control a complex and challenging task. Additional factors, such as solar heat load, significantly affect cabin temperature, thus complicating the control strategy. Since actions in chillers and air handling units influence fan coil units, optimization must be done holistically.