Based on the demand for population dynamic risk assessment during urban flood and waterlogging disasters, a multi-agent model based dynamic risk assessment method for urban flood and waterlogging disasters was proposed on the basis of urban hydrologic-hydrodynamic models. The operation rules of disaster-producing environment agent, disaster-causing agent and disaster-bearing agent were set, respectively, and the communication mechanism among multiple agents was proposed to achieve the quantitative assessment of population dynamic risk during urban flooding and waterlogging disaster. A typical drainage area of the Qinghe River Basin in Haidian District of Beijing City was selected as an example, SWMM and LISFLOOD-FP models were used to simulate the flood disaster process under the scenario of design rainfall with the return period of 50 a, and based on NetLogo model, the risk assessment model of urban flood and waterlogging disaster victims was developed to assess the population dynamic risk in the study area. The results showed that under the scenario of design storm with the return period of 50 a, the flood and waterlogging risk area in the study area increased from 0.608 km² with a rainfall duration of T=4 h to 2.202 km² with T=30 h. Under conventional simulation scenarios, the low-risk population in the study area continues to decline while the high-risk population continues to rise. Ultimately, the stable proportion of low-risk population is 75.75%, while the stable proportion of high-risk population is 22.85%. With the implementation of the risk avoidance strategy, the number of low-risk population in the study area first decreased and then increased, with the smallest number of 92.30% at T = 4.5 h while the number of high-risk population first increased and then decreased, with the highest number of 7.07%, and finally the stable number of low-risk population was 100%.