用“降维方法”进行梯级水电站优化调度

The Multistep Dispatching Optimization of Hydropower Stations in Cascade

本文应用马尔可夫决策规划以及“分解和多级最优化”的降维方法，为东北电网的某大型水电站梯级系统制定了长期最优运行策略。求解的目标是在满足发电保证率的前提下，梯级系统年发电量的期望值最大。与时历法相比，按该策略运行时梯级系统的年效益可提高３．５％，且发电保证率亦有提高。本文还对径流描述的现行作法进行了一些分析，提出在水文计算中，Ｃｓ／Ｃｖ的取值应以经验频率点据与Ｐ－Ⅲ型分布曲线吻合最好为依据，并采用了计算机适配理论频率曲线的方法；找出了定长期运行策略时，梯级之间水量联系的随机描述方法。本项成果已交生产单位使用。

In this paper the Marcov decision programing and the method of reducing dimensions based on decomposition and multistep optimization are used to seek the long-term optimization policy for large hydropower stations in cascade in the Northeast China Power System. The objective function is to maximize the average expected annual power output of cascade system with the prerequisite to the reliability requirement of electric generation. Compared with time-history method the policy can increase the yearly power benefit of the cascade system by 3. 5 % and the reliability requirement of electric generation. This paper also analyzes the general method of describing stream flow and puts for-ward that in hydrologic calculation the selected value Cs/Cv depends on optimum fitting experience frequency points in P- Ⅲ distribution curve and electronic computer is used to fit theory distribution curve. In this paper a stochastic descriptive method of water value contact between cascade hydroelectic stations is given when the long-term optimizing policy is being found. The achieved results of the study have been put to use in production units.