Multiobjective Collaborative Optimization of Water Hammer Protection for Long-Distance High-Lift Water Conveyance System Based on Machine Learning

To address the challenges in pump-stoppage water hammer protection design for long-distance, high-head water conveyance systems, specifically the complexity of coupled multidevice parameters, the reliance of traditional methods on empirical trials, and the difficulty in synergistically optimizing safety and economy, this study proposes a systematic optimization framework integrating machine learning with multiobjective optimization algorithms. By comparing various machine learning models combined with Bayesian optimization, the sparrow search algorithm feed-forward backpropagation (SSA-FFBP) surrogate model was identified as having the best overall performance to replace time-consuming hydraulic transient simulations. Furthermore, the Shapley additive explanations method was introduced to reveal the dominant and coupling effects of key parameters on extreme water hammer pressures. The surrogate model was then coupled with the NSGA-II algorithm to establish a multiobjective optimization model targeting water hammer pressure and air tank volume, and the Entropy Weight-TOPSIS method was employed to select the optimal solution from the Pareto set. A case study demonstrates that the modified scheme, while ensuring pipeline pressure remains within safe thresholds, reduces the maximum pressure by 19.01%, increases the minimum pressure by 136.67%, and decreases the air tank volume by 40%, achieving a synergistic optimization of safety and economy. The framework’s generalization capability and robustness were also validated through testing on an independent project, providing an intelligent decision-making solution for complex water conveyance system protection.