随机动态规划在三峡梯级长期发电优化调度中的应用

对于时间序列，如果认为某时段值仅与前一时段值直接相关，则认为该时间序列具有马尔柯夫特性，径流序列近似具有这种特性。通过余留效益函数对三峡梯级建立了长期发电优化调度的随机动态规划模型。基于大量模拟的随机径流序列样本，通过简单获得径流状态概率转移矩阵，它比仅基于历史径流资料所得结果要精确得多。最后通过对模拟径流序列样本的模拟操作，对所得优化策略进行详细评估。

Long-term generation optimization of Three-Gorge cascade using stochastic dynamic programming

For a time series,when the value of any period has only direct relationship with that of the previous period,it is said that the time series show Markovian characteristic,which the runoff series show approximately.With the benefit to go function being its recursive stage objective,the stochastic dynamic programming model for long term generation optimization is established for Three Gorge cascade.Based upon the vast synthetically generated runoff samples,the runoff probability transition matrices is derived by simple statistics,which are more precise than those obtained only based on historical runoff series.Finally,the obtained optimal policy can be evaluated in detail by operation simulations with the simulated runoff samples as input.