Model tests and CFD were carried out to find out the cause of cavitation surge in hydraulic power plants. In experiments the cavitation surge was observed at flow rate, both with and without a surge tank placed just upstream of the inlet volute. The surge frequency at smaller flow rate was much smaller than the swirl mode frequency caused by the whirl of vortex rope. An unsteady CFD was carried out with two boundary conditions: (1) the flow rate is fixed to be constant at the volute inlet, (2) the total pressure is kept constant at the volute inlet, corresponding to the experiments without/with the surge tank. The surge was observed with both boundary conditions at both higher and lower flow rates. Discussions as to the cause of the surge are made based on additional tests with an orifice at the diffuser exit, and with the diffuser replaced with a straight pipe.