In order to reveal the hydraulic characteristics of centrifugal pumps in the transient shutdown process，a shutdown experiment has been carried out on a low specific speed open impeller centrifugal pump under six non-rated conditions, thus obtaining the real-time evolution characteristics of the external characteristic parameters such as rotational speed, inlet and outlet pressures, head, flow rate, and shaft power over time. Meanwhile, a fitting model is established for the shutdown conditions based on the BP neural network model providing a simulation test platform for the protection of the pump station under unexpected conditions such as pump station power failure or shutdown. Based on the simulation analysis it is found that the impeller speed shows a linear and rapid decline in the initial stage of shutdown, while the flow rate decreases slowly due to the inertia and impeller effects; with an increase in the opening of outlet valves，the time required for the flow rate to decline is prolonged as well, with the shaft power showing a fluctuating and rapidly declining trend in the initial stage of shutdown. The results show that the proposed model accurately presents the hydraulic performance in the transient pump shutdown process.