Research on Comfort Control of Smart Home Environment Based on Deep Reinforcement Learning

The comfort of the indoor environment and the energy consumption of smart devices are key issues in smart home research. The comfort of the indoor environment directly affects the physical and mental health of people working and living in it, while the control of energy consumption helps to realize low-carbon environmental protection and cost savings. Currently, the mainstream programs of “whole-house intelligence” in the market are based on logic control and scenario control. These programs are often simple and effective but cannot effectively balance indoor comfort and power consumption. Domestic and international scholars' research on indoor comfort control focuses more on modeling by understanding the user's habit of using the equipment and the condition of the equipment. However, users do not know how to operate or make mistakes in operation will cause obstacles to model learning, and this control method lacks objective and effective control standards. Therefore, in order to solve the above problems, this paper proposes an autonomous smart device control method based on the deep reinforcement learning method, taking the illumination and temperature environments in the indoor environment as the research objects. The main contributions are as follows: 1) By using the DQN method to solve the device control problem, the energy consumption is significantly reduced while ensuring comfort compared to PID and MPC schemes. 2) The environment is feature extracted to express the indoor comfort level more accurately. At the same time, the state space dimension of the DQN method is reduced, and the performance of the DQN method is improved. 3) The indoor temperature comfort is calculated using the BP neural network, and the feature extraction ability of temperature comfort is improved under the condition of limited parameters.