使用随机对偶动态规划设计考虑多重耦合不确定性及年末留蓄的水-风-光能源系统长期运行策略

Abstract Hydropower with reservoirs is increasingly important for balancing seasonal variability of growing variable renewable energy (VRE) through its reservoir regulation capability. However, the coupling of the seasonal variability and randomness of VRE with the stochastic nature of inflows makes it extremely challenging to manage long-term hydropower operations related to generation decisions within the current scheduling periods and future end-of-year carryover storage control. To address these challenges, we propose a stochastic dual dynamic programming-based framework for designing long-term hydro-wind-solar complementary operation policies. Inflow and VRE output uncertainties are captured by two different approaches: Markov chain and AutoRegressive Moving Average. These approaches enable the integration of stage-wise dependent randomness into the stochastic decision-making process. Model reconstruction techniques based on Disjunctive Programming are proposed to transform stage-wise nonlinear models into linear ones. Subsequently, Benders cuts families are constructed to constrain the feasible decision space related to hydropower operation and stochastic parameters , while managing the end-of-year carryover storage requirement. Case studies of a large-scale hydro-wind-solar energy system in China indicate that the proposed framework can derive effective complementary operation policies considering future reservoir storage management requirements under multiple coupled uncertainties. Real simulation results indicate that the framework can effectively enhance channel utilization by leveraging hydropower flexibility to support VRE integration, with the monthly average channel utilization rate exceeding 80 %. Besides, hydro-wind-solar complementary operation policies with varying end-of-year carryover storage requirements can be designed, with lower storage requirements trending to enhance hydropower output in a hydro-wind-solar complementary mode.