Operation activity has always been the biggest threat to the safety of petrochemical enterprises. There is continuous dynamic interaction at operation area among operators/supervisors, machine and tools, hazard materials and the environment. Abnormal changes in any factor may lead to accidents. However, failure sequence of operation risk prevention measures is disordered in the evolution of operation accident. Thus, traditional static or sequential risk assessment methods are not suitable for this field, which restricts the assessment and control of operation risk. Therefore, this paper uses System Dynamics (SD) to construct a multi factor coupling feedback model for operation risk identification firstly, and describes the coupling and interaction relationship within the subsystems and between subsystems. Secondly, key indicators and corresponding monitoring techniques are determined, i.e., operators’ unsafe behaviors, gas leakage and so on. Finally, Bayesian theory (BT) is adopted to fuse multi-source information collected by the monitoring techniques, and then assessment operational risk. What’s more, a hierarchical early warning rule is built to determine the priority of risk control. This method can guide the realization of operation risk intelligent management and control.