Abstract: With the rapid advancement of building energy management and intelligent technologies, AI-powered image recognition has emerged as a key focus in optimizing energy efficiency through smart power distribution systems. This study proposes an AI-driven approach for intelligent power distribution optimization in buildings. The system utilizes high-resolution cameras and infrared thermal imagers for data acquisition, combines AI algorithms for image preprocessing and feature extraction, and integrates visual information with electrical sensor data to enhance equipment condition assessment and anomaly detection. Through a collaborative decision-making mechanism that balances energy conservation and safety objectives, real-time optimization of equipment operations is achieved. Experimental results demonstrate that the proposed method outperforms traditional control approaches in energy savings, power quality, response speed, and system stability. It achieves an average energy reduction rate of 11.3%, reduces the voltage fluctuation coefficient to 2.1%, and maintains a system availability rate of