Research and Verification of Predictive Control Algorithm for Open Channel Gates Based on the Integral Time-Delay Model

Under complex disturbances and backwater time-delay conditions, traditional open-channel gate water level control suffers from insufficient accuracy and slow response, readily causing water level overruns, control instability, and engineering safety risks. To overcome the limitations of conventional controllers in responding to rainfall disturbances, this study proposes a Model Predictive Control (MPC) algorithm based on the Integrator Delay (ID) model. The approach first integrates an LSTM-KAN (Kolmogorov–Arnold Network) model for accurate rainfall prediction, providing reliable inputs for disturbance feedforward. Subsequently, leveraging the SWMM simulation model and the PySWMM library, ID model parameters (backwater area and lag time) are identified in real time through impulse response testing. A state-space representation is then formulated and incorporated into the MPC rolling optimization framework, enabling precise water level forecasting over the prediction horizon. Simulation results demonstrate that the average computation time for 24-hour tests is only 240 seconds, with markedly reduced water level deviations. Experimental validation confirms superior performance under steady flow conditions (flow fluctuations < 0.007 m3/s; settling time ≈ 210 seconds) and constant water level control, achieving water level deviations < 0.05 m in known disturbance scenarios. Compared with the conventional Linear Quadratic Regulator (LQR), the proposed MPC algorithm reduces gate response time by 6.38–19.80% under the tested rainfall conditions. The proposed method establishes a complete closed-loop framework integrating rainfall prediction, water level forecasting, and combined feedforward-feedback control, offering an efficient and practical solution for open-channel gate water level management in smart water conservancy systems. It holds considerable theoretical significance and application value.