Indoor Temperature and Humidity Control System Based on Transfer Deep Reinforcement Learning

When adjusting the temperature and humidity of multiple rooms simultaneously, the training process can be time-consuming. To address this issue, we propose an indoor temper-ature and humidity control system based on multi-agent transfer deep reinforcement learning. This approach combines transfer learning and multi-agent reinforcement learning to enhance the adaptability and efficiency of the indoor climate control system. Each agent controls an air inlet, achieving indoor temperature and humidity control through multi-agent cooperation. Transfer learning is employed to transfer knowledge learned from a source domain to the target domain, which can significantly reduce the amount of data required for learning and improve the control sys-tem's performance in the target domain. The results demonstrate that the multi-agent transfer deep reinforcement learning-based control system reduces adjustment time and energy consumption by 95% compared to direct control, with a 14.4% and 15.1% improvement in temperature and humidity accuracy, respectively, and a 6.2% and 3.3% improvement in uniformity. This approach holds considerable potential for enhancing the performance and energy efficiency of indoor temperature and humidity control systems in real-world applications.