Abstract: Multi uncertainty of hydro - wind - Pv systems and its optimal solution with high dimension is a key challenge in the long - term scheduling of hydro- wind- Pv multi energy complementary systems. By employing a hydro- wind-Pv scene generation method based on Markov chain and copula function ,and utilizing a reduction technique to reduce the number of scenes ,the uncertainties of the hydro- wind- Pv system can be quantified. Taking the reduced scenes as input ,we developed a long - term two- stage stochastic optimal scheduling model that incorporates Benders decomposition algorithm and convex linearization to realize high efficient solution for high dimension problems. The model was used to simulate the scheduling process of a clean energy base in downstream of Jinsha River , which demonstrated the method /s effectiveness in enhancing adaptability to the uncertain hydro- wind- Pv systems and in improving overall benefits. In out- of - sample testing , the proposed method increased 55 . 2 million kWh power generation and decreased 169 . 4 million m3 abandoned water compared to traditional methods ,demonstrating a greater performance.