Abstract: With the deep integration of Internet of Things technology in ports, the fusion of multi-source heterogeneous data and collaborative operational safety have become critical challenges. This paper presents a digital twin-based IoT data fusion architecture and a dynamic safety optimization model. A four-layer system, consisting of "Perception-Edge-Platform-Application", is developed, where an improved D-S evidence theory is applied for multi-sensor data fusion. A dynamic evaluation model for operational safety index is established, incorporating equipment status, environmental factors, human factors, and process compliance. Additionally, a safety-constrained scheduling optimization algorithm is developed. The research outcomes enable effective collaboration of multi-source data and accurate assessment of safety risks, providing an innovative solution to enhance port operational efficiency and safety. The findings offer significant theoretical and practical implications for advancing smart port development.