<p>To address the planning challenges associated with structurally complex, large-scale canal networks, this study develops a multi-objective, multi-level optimization model that coordinates flow relationships across the system. The model was applied to the Baojixia Irrigation District with the primary objective of minimizing water conveyance losses and irrigation water shortages. To analyze water distribution under varying hydrological and demand conditions, two distinct algorithms were employed: a genetic algorithm (GA) with elitist preservation for water-deficit scenarios, and a hybrid particle swarm optimization (PSO) integrating simulated annealing and genetic operations for water-sufficient scenarios. This study represents one of the first attempts to apply an improved GA and PSO in parallel within a unified multi-level framework for large-scale canal networks, thereby addressing both water-scarce and water-abundant conditions. The results indicate that: (1) following optimization, flow rates in most lower-level channels exceed 71% of design flow capacity, with 58.6% of channels surpassing 80%. (2) the GA-derived scheme for water-shortage conditions requires a shorter duration, whereas the PSO scheme more closely approaches the design flow rate, thereby enhancing channel capacity utilization. (3) the conveyance loss rate is maintained between 20% and 30%, with water shortages in most channels kept below 3%. The proposed model and algorithms can be flexibly applied to other large irrigation districts featuring multi-level gravity water supply networks. The framework thus provides a robust tool for supporting scientific water allocation planning and advancing the intelligent modernization of irrigation management.</p>