نوع مقاله : مقاله پژوهشی
نویسندگان
1 گروه علوم و مهندسی آب، دانشکده کشاورزی، دانشگاه بوعلی سینا، همدان، ایران.
2 گروه مهندسی برق، دانشکده فنی و مهندسی، دانشگاه یزد، یزد، ایران.
3 دانشکده حکمرانی، دانشکدگان مدیریت، دانشگاه تهران، تهران، ایران.
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Optimizing water management in irrigation networks, especially in arid and semi-arid regions, is of critical importance. This research aimed to design an intelligent control system for surface water conveyance canals. To this end, an integrator-delay linear model was first developed, capable of simulating the canal's hydraulic behavior. Subsequently, an innovative hybrid control system based on the integration of a classic PID controller and a continuous reinforcement learning agent was designed and implemented. The controller gains were set to 0.209, 0.243, and 0.086 via manual tuning, and to 1.69, 0.055, and 0.086 respectively when tuned by reinforcement learning. Performance evaluation under scenarios with 10%, 20%, and 30% flow changes demonstrated that the RL-tuned controller has significantly superior stability and accuracy compared to the manually tuned controller. The most important indicator of this superiority was a five-fold reduction in the maximum depth error for the 10% input change. The proposed system represents an effective step towards intelligent water management in irrigation canals; however, its application under critical conditions requires integration with more complex non-linear models, which is suggested for future research.
کلیدواژهها [English]
)