Because the short-time optimized dispatcher question has some characteristics such as high dimension, non-protruding, dispersed, non-linear etc., so need one algorithm with high calculation precise and convergence speed to solve this question. This paper put forward one improve self-adaptation niche genetic algorithm, its characteristic mainly is crowding radius (Euclidean distance) can self-adaptation change with the increase of evolution, on the premise of well keeping the variation of solution, to improve the speed of convergence and precision of solution, then used this algorithm to solve the optimal dispatcher's question of cascade hydropower station, through the calculation of example, can find out, the improve self-adaptation niche genetic algorithm has greatly improve the speed of convergence and precision of solution.