The speed governor is known as one of the most important control parts in the pumped-storage unit (PSU). When operating in transient processes, the stability of the PSU’s hydraulic and mechanical states maintains by the precise regulation of the speed governor. For the sake of improving the governing control performances, the lumped-parameter equivalent circuit model is first introduced for precise modelling of hydraulic-mechanical characteristics of PSU. Subsequently, a multi-variable generalized predictive control system based on receding horizon technique is applied at every sampling instant to optimize the control trajectory within certain finite time horizon. To validate the effectiveness of the proposed predictive control method, the simulation experiment of governing movements during turbine start-up process is conducted under a nonlinear lumped-parameter numerical simulation platform of PSU by comparative case studies, the overall control performance of the proposed control scheme is evaluated and compared with that of the traditional PID algorithm. Simulation results indicate that the proposed governing control method can achieve stronger frequency fluctuation alleviation and hydraulic oscillation damping performances than traditional PID.