The Research Of Qingjiang Cascade Reservoir's Optimal Scheduling Rules Of Power Generation Based On Stochastic Dynamic Programming

With the rapid development of economy and promoting the process of well-off society in our country, power demand is increasing year by year. Hydropower resource that is rich was scientificlly developed and reasonable used. Twelve hydropower bases which have been built or are being built are into a situation of multi-reservoir jiont operation gradually. In the thesis, in the basis of the analysis of runoff regular pattern of reservoir, exceedence probability curve of runoff was drew, the stochastic dynamic programming model of Hydropower generation scheduling was established and solved, then, Qingjiang cascade hydropower generation scheduling rules was developed. The main research results as follows:(1) The low of runoff running into reservoirs was analyzed. By the Weibull empirical probability principle, exceedence probability curve of runoff was drew, and the runoff in the valley in each time bucket was showed.(2) Matrix of transition probabilities based on artificial runoff series were established. Due to limitations of historical runoff data, the vast data in the matrix of transition probabilities established by traditional method were zero, which indicated that accidental factors appeared frequently. For his problem, the principle of stochastic hydrology was used in this paper, runoff data were multiplied, and artificial runoff transition probability matrix was established based on the revised data.(3) The research on correcting method of Artificial runoff series. The correction method of traditional artificial runoff has some limitations, which makes artificial runoff series has some difficulties and bias in the process of correction. A new direct correction method of artificial runoff was used in this article to correct Artificial runoff series and got a very good test results.(4) Scheduling rules of reservoir generation was made. The scheduling rules of Qingjianggeheyan's reservoir generation was made by using results of the stochastic dynamic model. The results calculated by the rules and the result of dynamic programming were compared with, and shew that the scheduling rules had reliability and was reasonable.Finally, conclusions were maked. Because of the limit of time, it existed shortcomings, and was necessary for further research.