Abstract: This paper presents an improved multi-objective cuckoo search algorithm (IMOCS) to effectively solve the multi-objective joint optimization problem of cascade hydropower stations and optimize their generation benefit and capacity benefit. To avoid the slow convergence of traditional cuckoo search, this new algorithm adopts probabilities and steps of dynamic detection and is combined with a non-dominated sorting method and a crowding distance parameter as a strategy for maintaining external files. Its effectiveness was verified by using test functions. Then, we applied IMOCS to the Wu River cascade hydropower stations and obtained non-dominated solutions of uniform distribution. Finally, a subjective-objective method was used to determine the objective weights of a fuzzy decision model and choose a compromise plan, and a scheme for regulating the time variation in the reservoir stage of each station was obtained. The results show that the dispatch plan is effective and reliable in various operation conditions of different constraints and the improved algorithm is useful for optimizing joint operation of cascade hydropower stations.