Abstract: With the entry of smart homes into thousands of households, intelligent control algorithms for home appliances are increasingly becoming a hot topic in academic research. An important research direction of intelligent control algorithms is the optimization of control parameters that vary in real-time with dynamic factors such as the environment. In response to this issue, this article combines research progress in related fields and proposes a real-time adaptive algorithm based on (real-time) human feedback reinforcement learning and multi-objective parameter optimization algorithms. The feasibility and effectiveness of the algorithm were verified on the experimental data of air conditioning, and it also points out an optimization direction for intelligent air conditioning control. The contribution of this algorithm lies in: 1) fitting prediction algorithm, which first generates a fitting model for environmental factors such as room type and air conditioning parameters based on cluster analysis, and then predicts changes in air conditioning status based on the current air conditioning status. 2) Feedback reinforcement learning algorithm, using air conditioning status changes and user intervention as feedback, combined with real-time state (environment), real-time optimization decision model. 3) multi-objective parameter optimization algorithm, which searches for the optimal solution of control parameters for user experience, air conditioning energy conservation, and other multi-objective objectives.

Key words: Real time, Intelligent home, Appliance control, Air condition, Parameter optimization, Human feedback, Reinforcement learning, Multi-objective optimization