Abstract: With the rapid development of the national economy, the size of the distribution network is growing and the connection is more and more complex. The demand for the reliability of the distribution system is becoming higher and higher by the power supply enterprises and the power users. The distribution network is at the end of the power system and directly supplies electricity to the power users. According to statistics, about 80% of the users' power failure is caused by the failure of distribution system. According to the results of risk assessment of distribution network, it is particularly important to screen out the most likely failures risk, prevent and control them, so as to improve the reliability of power supply. The amount of information obtained by the risk assessment calculation is very huge. For risk online prevention and control, it is impossible to carry out risk online prevention and control analysis of all the evaluation results. It needs to select the most serious risk information from the mass result data, then trigger the risk online prevention and control measures to reduce the risk level to a certain degree. Previous ranking and screening of risk assessment results are based on the performance index method, which belongs to the deterministic research category. In this paper, , the fault risk is sorted and screened to determine the highest level of risk based on the uncertainty theory. For the pre ranking fault risk, before the fault can be prevented in advance, such as the distribution network reconfiguration, the isolation of isolated island, DG rescheduling and other measures to minimize the loss. Finally, the reliability of the algorithm is verified by actual example analysis.