To optimize the hazardous materials (hazmat) transportation paths in dynamic road networks and make a decision on path selection for carriers, the stochastic time dependent (STD) characteristics of transport networks were analyzed, the travel time and the number of people affected around each link were fitted in piece-wise continuous time interval. The 0-1 integer programming model was formulated by taking the arrival time window as a constraint, and taking the stochastic attribute values of travel time and transport risk as optimization criteria. A two-stage multi-dimensional label correcting algorithm was designed with First-In-First-Out property of the STD road networks, and the non-dominated paths were generated at different departure time to meet the time window constraint with predetermined confidence level. The criterion-weight and threshold-dominate methods were proposed to balance computational efficiency and accuracy. Research results indicate that the travel time and transport risk of hazmat in STD road networks are related to the setting of arrival time window and the selection of departure time. The non-dominated path depends on the departure time and the choice of carrier's preference, and the amount of non-dominated paths depends on the dominated threshold value. Different types of carriers can choose the optimal combination of departure time and transportation path according to the confidence level of arriving on time.