Cascade Reservoirs Multiobjective Optimal Scheduling Based on an Improved Two-Stage Particle Swarm Optimization Algorithm

Abstract

The multiobjective scheduling of cascade reservoir systems faces challenges due to high-dimensional nonlinearity, where traditional optimization methods struggle to achieve globally balanced solutions. This study proposes a Two-Stage Multi-Objective Particle Swarm Optimization (TSMOPSO) algorithm, incorporating two innovative components to enhance optimization performance. The first component employs Piecewise mapping, adapts weights and introduces two operators to improve optimization efficiency and convergence speed. The second component features a two-stage refinement mechanism, implementing a two-level adjustment of upstream and downstream water levels based on constraint evaluations, effectively alleviating constraint limitations. A case study is conducted on cascade reservoirs system in the Jinsha River Basin of the Upper Yangtze River (JRBUY), with a multiobjective model integrating power generation, power output, and navigation demands. Numerical experiments demonstrate that TSMOPSO achieves remarkable performance under wet-year conditions: power generation of 2087.46 KW h, power output of 16,435.75 MW, and a navigation index of 3052.92 m³/s. Compared wtih other algorithms, TSMOPSO exhibits significant advantages in hypervolume (HV) indicators and solution set coverage. Pareto front analysis reveals competitive mechanisms among the three objectives. This approach provides a novel technical pathway for multiobjective optimization of complex cascade reservoir systems.