Abstract: To address the issues of simple structure, delayed response, and insufficient precision in the operation of canal diversion projects, we proposed an intelligent device system aimed at multi-objective regulation including water supply, hydropower generation and ecological replenishment. A joint automatic control method was developed based on the coupling feedback between water level and discharge, integrating a multi-gate deployment with hierarchical sensing and a two-stage neural network control strategy. The system implements a closed-loop adjustment mechanism combining adaptive PID parameter tuning and dynamic correction via simulation-based prediction. A nine-day ecological replenishment experiment was conducted in a typical canal irrigation district in northern Xinjiang to validate the system′s responsiveness and stability under variable hydraulic loads. Results demonstrate that the system maintains water level fluctuations within ±0.10 m, reduces response time to 35 s, and increases energy conversion efficiency by over 30%, exhibiting strong self-learning capability and wide-range adaptability. This research provides technical support for the intelligent transformation of canal diversion systems and unattended operation of national smart water networks.