Abstract: Large-scale inter-basin water transfer projects typically manifest the characteristics of multiple water sources, multiple objectives, multiple users, and multiple dimensions, whereas their different water sources, objectives, and water users are often interconnected yet mutually constraining, leading to new challenges in modeling their joint optimal scheduling and solving the problem. This study constructs a new model for water allocation optimization under the reservoirs' dynamic regulation rules, based on a genetic algorithm (GA) framework, and applied in a case study of the Jiaodong Water Transfer Project. We use this model to determine the system's water flow behaviors through accurate water balance simulations, and then design a fitness function via a comprehensive consideration of the constraints and multi-objective requirements of the project. Finally, its updated solutions are adjusted using a reverse correction strategy. To enhance algorithm efficiency, we have developed three key techniques: adopting a phase identification and division method to reduce decision variable dimensionality; determining the optimization adjustment directions based on reservoir regulation rules; and developing an intelligent mechanism for infeasible solution handling. Application of the new model to the transfer project shows a significant improvement on scheduling performance. Compared with the annual operational plan, a full scheduling period scheme raises the minimum water supply reliability at the diversion outlets by 22%, achieving superior water supply outcomes with reduced diversion flow. In the remaining period scheme, the Jihongtan Reservoir achieves a water supply reliability of 74.17%, with a water shortage of 102.2267 million m3 and a water diversion index utilization rate of 98.43%, meeting all the project requirements. This study helps consider and address optimal operation challenges in complex inter-basin water transfer systems.

Key words: water diversion project, optimal operation, dynamic regulation, genetic algorithm, multi-objective