MA Li-ya, LEI Xiao-hui, JIANG Yun-zhong, WANG Hao (China Institute of Water Resource and Hydropower Research, Beijing 100038, China)

Abstract: Dynamic Programming is a common method of solving multistep decision optimization problems and is widely used in reservoir optimal operation. The main disadvantage of DP is the "curse of dimensionality" when it is applied to cascade reservoirs. Successive Approximation Dynamic Programming (DPSA) can alleviate the problem by breaking the multidimensional problems into a number of single dimensional problems. Based on reservoir operation simulation model, this paper presents a long-term reservoir optimal operation model. With the upstream of the Hanjiang River cascade reservoirs as the study object, this paper takes the generated power as the objective, discretizes the storage of each reservoir and realizes the optimization by means of DPSA. It is proved that this model has significant value for reservoir operation.

Key words: cascade reservoirs, optimal operation, dynamic programming, successive approximation