Abstract: Analyzing the uncertain factors and dynamic characteristics of a system, the system safety is dynamically assessed via SPA and Markov chain. A dynamic safety assessment method is proposed combining the state assessment with trend analysis, thus providing a new approach for safety assessment and verifying its feasibility by examplification. In the method the hazard grade is determined by the correlation coefficient of SPA, and a 3-variable model combining the safety state with set pair potential and pessimistic potential is developed to analyze further the development trend of the system and assess dynamically the current situation of system safety. According to the ergodicity of Markov chain, the final level of system safety can be predicted to provide a theoretic reference for decision-making in safety consideration.