A novel optimal water distribution model for multi-level canal system based on terminal demand-driven methodology

In light of the investigation of the poor applicability of existing canal water distribution models across different irrigation districts and scales, this study proposes a Demand-driven, Cascaded-Layer Recursive Optimization Model (DCL-ROM). The model takes the discharge at field outlets as the sole decision variable. By physically and recursively back-calculating the flow processes of upstream canals level by level, it reduces the optimization dimensionality of the multi-level system. With the objective of minimizing both the total number of gate operations and water conveyance losses, the model incorporates rotational grouping constraints, gate operation rules, and canal conveyance capacity limits. It is solved using a hybrid strategy that integrates a genetic algorithm, dynamic programming, and exhaustive search. Taking the Ganfu Plain Irrigation District as a case study, with comparable numbers of gate operations, the model reduced the water conveyance loss rate of the West Main Canal System from 50.27% to 44.57%, and that of the East Main Canal System from 47.04% to 29.64%. Further analysis reveals that the marginal water-saving benefit of gate operations on the main canal follows the law of diminishing marginal returns, which offers valuable insights for decision-making in multi-objective optimal water allocation. The results demonstrate that the proposed model integrates structural generality with engineering practicality, serving as an actionable decision-making tool for the efficient, collaborative water distribution of complex multi-level canal systems.