Abstract: To cope with voltage fluctuations and exceeding the limits caused by the strong randomness of the output power of a high proportion of distributed generations (DGs) and reduce system operating losses, a centralized and local hierarchically-coordinated optimal operation model of hybrid AC/DC distribution networks is proposed. By fully tapping into the capacity and power/voltage regulation capabilities of grid-integrated photovoltaic (PV) inverters and DC networking voltage source converter (VSC) in the system, the system's optimal operation is achieved to minimize the operating active power losses and bus voltage deviations. The proposed model simultaneously optimizes the set points of the initial reactive of the PV inverters participating in centralized regulation, as well as the preset droop functions of the PV inverters and VSCs in the local control stage, to achieve collaborative regulation between the centralized and local control layers under the random changes of PV and load power. An uncertainty probability distribution describes the typical scenarios of PV and load power, then the proposed stochastic optimization model is transformed and solved by combining the second-order cone convex relaxation and linearization. Finally, the effectiveness and superiority of the proposed hierarchically-coordinated optimal operation model are verified through a test study of an improved IEEE 33-bus hybrid AC/DC distribution network.