Abstract: The paper presents an improved real allocation based accelerating genetic algorithm (RAGA) and its application to optimal feeding reservoir group operation,which overcomes the shortcoming of dynamic plan in the reservoir group operation,such as "dimension difficulty" or the problems of obtaining global optimum solution.The local search is embedded into RAGA to accelerate its convergence and enhance its global optimization ability.According to the water conservancy contact among reservoirs,an algorithm frame is designed base on RAGA and an idea of water supply strategy interval is put forward.To reveal its outstanding characters,the algorithm is contrasted with optimization by reservoir turns based on the dynamical planning in the optimal operation of the two reservoirs in series.The results show that it is reasonable and feasible and its convergence is quite fast,which proves its advantages of the computing speed and global optimization.