Abstract: It is an important technical means to address large-scale flexible resource demands reshaping the pumped storage function of conventional hydropower stations, gradually shifting their role from a power supplier to a power supplier+ battery regulator. In this regard, this paper takes the cascade hydropower-pumping-storage-wind-photovoltaic multi-energy complementary system (CHPMCS)as the research object and establishes a short-term optimal operation model with the objective of maximizing the benefits of system power generation in view of the flexible conversion of power generation-pumping and storage bidirectional operating conditions and the characteristics of complementary consumption. Secondly, considering the continuous and adjustable output of the CHPMCS, the paper proposes to transform the optimized dispatching problem into a Markov decision process, thereby transforming the multi constraint optimization problem into an unconstrained deep reinforcement learning problem. Then, to address the shortcomings of low training efficiency and susceptibility to local optima in the deep deterministic policy gradient (DDPG) algorithm, it uses an improved DDPG algorithm to solve the optimized dispatching decision process. Finally, it verifies the effectiveness of the proposed model and algorithm through numerical examples. The results show that the CHPMCS can effectively enhance its flexibility and regulatory ability through the reshaping of hydropower functions, improve the consumption capacity of new energy and the utilization rate of water resources, and improve the power generation efficiency of the system through low storage and high generation.