Based on the problem of abandoned water and insufficient output caused by runoff randomness, the short-term runoff inflow interval prediction method combined with nonparametric kernel density estimation (NKDE)and Monte Carlo (MC)sampling is proposed. The method analyzes the impact of hydropower station abandonment risk and insufficient output risk caused by inflow uncertainty on hydropower station output and thermal power output in the power grid, and the corresponding output adjustment strategy is given. This paper mainly carries out these studies : real-time load distribution, shortterm runoff interval prediction and real-time output adjustment of hydro-thermal power plant under uncertain incoming water:(1)Risk factors of water and electricity may be involved in the scheduling process, the process of considering runoff uncertainty on hydropower unit output, analyze the impact of hydropower abandoned water risk , estimate based on short-term nonparametric kernel density estimation (NKDE) runoff forecasting methods, given the range of the forecast runoff, reducing forecast error runoff, combine Monte Carlo (MC) sampling and propose a new real-time power control adjustment strategy.(2)Taking Qingjiang cascade power station as a case study, considering the increase of incoming water and the decrease of incoming water, the impact of water abandonment risk and insufficient output risk on short-term power generation plan is verified. The results show that the proposed hydropower and thermal power output adjustment can effectively deal with the risk of abandoned water and the risk of insufficient output. It is of positive significance to make full use of hydropower resources. It is helpful for the safe and stable operation of Hubei Power Grid.