The contradiction between the ever-increasing water demand of Danfeng County and the limited water supply capacity of the reservoirs is becoming more and more prominent. To fully tap the project’s water supply potential and solve the problems of excessive water abandonment and low water supply guarantee rate in the operation of the existing projects, an optimal operation model is established with the objective of minimum abandoned water, and the improved particle swarm optimization is used to find the joint optimal water supply solution for Qingfeng Reservoir and Longtan Reservoir. Two utilizable capacities of 2.0×106 m3 and 2.2×106 m3 of Qingfeng Reservoir are simulated with a month as the calculation period, under the condition of giving priority to the ecological flow, evapotranspiration and leakage loss of the river. The results show that the objective function tends to stabilize when the number of iterations reaches 36 times. Compared with the conventional scheduling, the average annual water abandonments of the two utilizable capacities are significantly reduced, and the water supply guarantee rate is improved to some extent. It shows that the improved particle swarm optimization based on linear decreasing inertia weight is superior to the other calculation methods and it is applicable to the joint optimization of water supply of Qingfeng and Longtan reservoirs.