The problems of “curse of dimensionality” will be resulted from the application of stochastic dynamic programming(SDP) to the optimization operation of a multi reservoir system and it is difficult to apply SDP to multi annual regulating reservoirs. A multistage decomposition combination optimization method(MDCOM) is presented to the optimal operation of multi reservoir system, which contains a SDP operation measure with even release flows for the multi annual regulating reservoirs. The method decomposes the optimal operation problems of multi reservoir system into the optimization operation of each single reservoir with SDP in the first stage. An improved genetic algorithm simulated annealing hybrid algorithm is proposed to combine and optimize the results of the each single reservoir in the second stage. In a case study of long middle term optimal operation on generation electricity of a 3 reservoir system in series in Wujiang River, Guizhou Province, a mean annual generation electricity of 9.472 billion kW·h for the multi reservoir system is achieved, which is bigger than that from GA, SA or SDP applied respectively. The MDCOM also runs faster than the SDP. The results obtained from these applications have proved that the MDCOM is a new promising search algorithm for multi reservoir optimization problems.
