Abstract: With the access of various types of intermittent energy, multivariate heterogeneous distribution networks have begun to encountered with multi-problems such as: increased uncertainty, significant increase in power disturbances, changes in the distribution of power flow characteristics, and dynamic characteristics of continuously changing operating conditions. The applicability of the traditional power flow method based on single time section is declining. At the same time, the voltage risk problem is also one of the major problems in distribution network. The technology of voltage warning, assessment and regulation is able to guarantee the voltage in distribution network and ensures its normal operation. For distribution networks with distributed generations, the coordinated control of distributed generations and the multi-level hierarchical consumption of power leads to bidirectional distribution of power flows. Which also lead to the random features and time-varying characteristics in distribution network. All of these make the traditional method for power flow calculation such as forward-returning and other methods no longer applicable. Therefore, it is necessary to further study the dynamic power flow analysis method for distribution network based on the fusion of cyber and physics, so as to further reveal the complex internal characteristics of the distribution network operation. Based on dynamic power flow model, this paper carries out distribution network risk assessment, identifies and discovers potential critical risk items that can cause adverse voltage. And this paper proposes a risk assessment model that considers both the conditional probability of risk occurrence and the severity of the consequences. When distribution network is in warning state, based on the dynamic power flow results and the simulation solution we can judge the corresponding risk indicators. The risk state types and state transition rules for the safe operation of the distribution network are established, and voltage optimization adjustment methods are further proposed to provide support for flexible control and efficient operation of distributed generation.