Research on Intelligent Air Conditioning Optimization Control Algorithms Based on Neural Networks and Heuristic Algorithms

This paper designs a home-based air conditioning control system utilizing neural networks and hybrid heuristic algorithms. The system employs long short-term memory neural networks to predict outdoor temperature and uses the equivalent thermal parameter model and simplified thermal model to predict indoor temperature changes, followed by the calculation of PMV values using a fully connected neural network. An innovative hybrid heuristic algorithm called genetic algorithm-simulated annealing with greedy criterion (GGASA) is proposed and applied to generate optimal air conditioning control strategies. The system integrates various modules and transforms the daily operation strategy of air conditioning into a constrained optimization problem, striving to lower energy consumption while maintaining thermal comfort for occupants. The lightweight design of the system offers the potential for deployment in future smart homes and edge computing devices. Experimental results show that the individual modules of the system can achieve their intended functions well, and the generated control strategies can save approximately 25% of energy consumption compared to traditional ON/OFF control strategies.