Multi-Scenario and Multi-Objective Optimal Operation of the Yujiang Reservoir System Using Particle Swarm Optimization Under Extreme Drought Conditions

Navigation water demand associated with the Pinglu Canal intensifies competition among water supply, ecological flow, navigation, and hydropower generation in the Yujiang River Basin during extremely dry years. A multi-objective joint-operation model was developed for six reservoirs—Baise, Wacun, Dongba, Chengbihe, Laokou, and Xijin—using water-supply deviation, ecological and navigation-flow deviation, and cascade hydropower generation as the objectives. The Muskingum routing method was embedded in each particle-fitness evaluation, and particle swarm optimization (PSO) was used to determine reservoir releases and water-supply trajectories. Three representative extreme drought years (1992, 2005, and 2011) were examined under a baseline and four optimized scenarios. Optimization scenario 1 provided the principal improvement, raising the Pinglu Canal navigation reliability from 49.04–63.84% under the baseline to 89.89–96.44%. Under optimization scenario 4, the rates reached 98.40%, 100%, and 99.70%. In the independent temporal test (WY2006–WY2010 and WY2012–WY2019), scenario 4 under the frozen PSO configuration increased mean navigation reliability relative to scenario 1 from 0.845600 to 0.961880 (Holm-adjusted p = 0.001831); mean cascade generation was numerically lower under scenario 4 (48.3445 × 108 versus 48.9319 × 108 kWh), and the one-sided test did not support an increase (Holm-adjusted p = 0.916138). These findings support coordinated allocation, adaptive intake rules, end-of-flood-season storage, and downstream re-regulation as a means of improving navigation reliability under the evaluated conditions.