Inflow and water demand prediction plays an important role in reservoir operation rulemaking. Uncertainties in meteorological and hydrological modeling increase the complexity of inflow and water demand prediction, which are not sufficiently considered in conventional inflow and water demand prediction. Thus, this study took cascade reservoirs as the research object. These cascade reservoirs were located in the mainstream of the Yellow River Basin (YRB). A three-step framework was proposed: (1) developing a distributed Soil & Water Assessment Tool (SWAT) model to determine historical and future water availability in the reservoirs; (2) constructing the AquaCrop agricultural irrigation water demand model to predict future water demand of water users at 264 nodes (reservoirs and hydrological stations) in the YRB; (3) developing an optimal regulation model of cascade reservoirs in the YRB mainstream to quantitively determine the influence of future environmental changes on hydropower generation and water supply benefits of cascade reservoirs. The results indicate that the inflow of cascade reservoirs increased in the flood season and decreased significantly during the non-flood season compared with that in the historical period (1970–2016). However, there was a slight increase in the agricultural irrigation water demand, and the contradiction became more prominent between water resource supply and demand. During the future period (2020–2050), the guaranteed rate of hydropower generation and water supply in the YRB’s lower reaches will increase significantly. Uncertainties in climate models greatly impact hydropower generation benefits and reservoir water supply, especially in the flood season. These findings can facilitate the understanding of the complexity of hydropower management and production of cascade reservoirs under future environmental changes (inflow and demand) and encourage the use of regulation models for impact assessment studies and adaptive reservoir management to mitigate climate change.