基于贝叶斯统计与MCMC思想的水库随机优化调度研究

针对马尔柯夫随机动态规划中的维数灾问题,提出一种改进马尔柯夫随机动态规划方法,基于贝叶斯统计原理,采用马尔柯夫链蒙特卡洛方法(Markov Chain Monte Carlo,MCMC)从数学角度出发,推求出一定预报级别下的实际来流概率密度函数,建立与预报级别相关的实际来流概率矩阵,在考虑预报误差发生的情况下进行不确定性优化调度,并且将该方法计算结果与有无预报时段相结合的马尔柯夫随机动态规划方法计算结果进行比较。结果表明,该方法所得到的结果比马尔柯夫随机动态规划结果更加贴近实际多年平均发电量,并且能够有效地减少计算量,缩短计算时间,从一定程度上解决了维数灾问题,本方法为不确定性优化调度提供重要理论参考。

Reservoir stochastic optimization scheduling research based on Bayesian statistics and MCMC

In order to solve the dimension disaster problem in the theory of Markov stochastic dynamic programming,the paper proposed an improved Markov stochastic dynamic programming. Based on Bayesian statistical principle, the paper used the Markov Chain Monte Carlo method to establish the actual flow probability matrix related to the forecast values according the actual inflow probability density function. The uncertainty optimal scheduling result of the method was compared to the result of Markov stochastic dynamic programming of combination of prediction and non-prediction period. The results show that the results by the improved Markov stochastic dynamic programming are more close to the actual average annual generation, and the computation complexity is reduced effectively and the computing time by the improved method is less than the time by the Markov stochastic optimization scheduling. The method solves the dimensionality problem to some extent and it can provide an important theoretical reference for the optimal scheduling of uncertainty.