Abstract: Protection measures of a long-distance water pipeline system are usually necessary to mitigate the water hammer effect since it can cause vibration and rupture. To optimize the water hammer protection measures and improve their cost and reliability, a multi-objective optimization method is developed based on the Random Forest (RF) algorithm and Non-dominated Sorting Genetic Algorithm-II (NSGA-II). First, a hydraulic transient model based on the method of characteristics is used to obtain an expected sample set; based on this set, an RF prediction model is used to establish the relationships between the optimization variables and the optimization objectives. Then, a multi-objective optimization model is constructed and used to find the Pareto frontier solution set, by taking the highest water hammer pressure, the highest dimensionless reversal speed, and the lowest protection cost as its objective functions. This study shows our method can quickly generate optimized protection schemes that balance the requirements of different targets, thus significantly improving the design of water hammer protection for long pipeline systems.

Key words: water hammer protection, multi-objective optimization, random forest, multi-objective genetic algorithm, long-distance water transfer system