Conventional intelligent optimization algorithms are inefficient or even impossible in finding feasible solutions to the cascade reservoir optimal operation calculations with small time step and a large number of calculation periods. On this basis, an improved NSGA-Ⅱ algorithm is proposed to solve the problem. Based on the framework of NSGA-Ⅱ algorithm, the improved algorithm introduces reference goal value, potential goal value, offset degree and mutation guiding operator to optimize the population evolution process and to enhance the quality of populations in the evolution process, making the solution set as close as possible to the true Pareto-optimal front. The verification results from a multi-objective optimal operation case study of cascade reservoirs in the Jinxi River Basin of Fujian Province show that the improved NSGA-Ⅱ algorithm has higher computational efficiency and better optimization results compared to other algorithms, demonstrating good practicality.