A Real-Time Control Strategy of Air Conditioning Systems in University Buildings Based on Computer Vision and Deep Learning
摘要
University buildings are characterized by complex, dynamic occupant behavior and high comfort demands, making traditional air conditioning systems inefficient due to slow response times. This paper proposes a pre-control strategy for air conditioning using real-time occupant distribution estimation via computer vision and deep learning. A Multi-Column Convolutional Neural Network (MCNN) model is trained to estimate crowd density and real-time occupant loads from images. Load grading and zoning regulation strategies are developed, and simulation tests assess the impact on temperature, PMV value, and cooling energy consumption. Results show that the proposed strategy enhances indoor comfort and reduces cooling energy consumption by up to 6.5%.