Wind power, photovoltaic and hydropower (referred to as scenery and water) complementary power generation system is an important innovative mode to improve the overall efficiency of clean energy utilization. With the vigorous development of clean energy, how to solve the problem of flexibility demand brought by the large-scale grid connection of new energy has become the core and difficult point in the research and development of multi-energy complementary system. In this paper, we firstly consider the uncertainty of scenery output and use regression analysis to predict the scenery output; secondly, we make use of the good regulation performance of hydropower units to smooth out the uncertainty of scenery output and propose a short-term scheduling model based on adaptive variational particle swarm optimization algorithm to efficiently and accurately define the output size of hydropower units. The development of short-term optimal dispatching strategy will help promote the healthy development of clean energy in China.