Dams located in heavy snowfall regions are required to restore their water utilization capacity during snow melting period because the inflow volume into reservoir decreases during the snowfall period. In this study, a dam operation model using deep reinforcement learning was trained by inputting snowmelt and temperature data of the catchment area to achieve both control of snowmelt floods and operation to restore the water utilization capacity of dams in heavy snowfall areas. From the results, this model can be adopted to both flood control and restoreing water utilization capacity during snowmelt period. We also obtain the results that this model can control discharge rate according to mass of snowmelt or weather conditions.