Abstract: To address the time-consuming nature of traditional hydrodynamic model combined with scheduling model, we propose a surrogate model as an approximation to enhance efficiency. The research focuses on the practical scheduling of cascade reservoirs from lower Jinsha River to the Three Gorges. By combining hydrodynamics theory with the Kriging surrogate model, we established a multi-objective dispatching model for a multi-reservoir system, and subsequently solved the model using a multi-objective evolutionary algorithm. Ultimately, we identified the multi-objective Pareto front of the model by adopting the projection pursuit method in decision-making. Research findings demonstrate that the surrogate model exhibits an average simulation error of less than 1.5%, effectively replacing the time-consuming hydrodynamic model while generating a competitive Pareto front and providing reasonable compromise solutions. These outcomes lay a theoretical foundation for advancing the comprehensive benefits of cascade reservoirs.

Key words: multi-objective optimal operation, Kriging model, hydrodynamic model, projection pursuit, cascade reservoirs