Dynamic programming is a classisc method to solve reservoir optimized operation. However, with the increasing number of reservoir power stations, computation amount is increasing exponentially, resulting in a dramatic decrease in the timeliness of solving and even causing 'curse of dimensionality'. In response to this, we improved the serial recursion calculation process of dynamic programming and introduced parallel dynamic programming based on stage reconstruction. Through the proposed algorithm a multistage decision problem can be repeatedly reconstructed in a parallel environment and gradually transferred to a single stage issue. This algorithm was then applied to solve the optimized operation of cascade reservoirs in the lower reach of Yalong River in China. Analog computation was carried out to evaluate the effects of parameter control on the parallel calculation performance of the algorithm. Results indicate that the calculating efficiency, compared with serial dynamic programming, can be significantly improved without sacrificing the accuracy with parallel dynamic programming based on stage reconstruction.