水锤计算对保障长距离输水工程管网系统安全稳定运行具有重要意义，但传统水锤数值方法存在模型复杂、计算量大的问题。为此，在自主开发的瞬态流实验平台上，通过支路快速关阀产生水锤，获取了不同流量和压力条件下的瞬态水锤压力。试验参数范围为：体积流量15~55 m3/h，压力150~450 kPa。采用集合经验模态分解（Ensemble Empirical Mode Decomposition，EEMD）方法对水锤信号进行滤波，并对水锤压力的变化规律进行了深入的研究分析。基于双向门控循环单元（Bidirectional Gate Recurrent Unit，BiGRU），建立了用于水锤压力预测的序列到序列（Sequence-to-Sequence，Seq2Seq）双向模型。结果表明，Seq2Seq双向预测模型能有效预测支路水锤，其预测数据决定系数在0.8以上，水锤特征参数预测准确率超过98%。研究成果可为水锤压力预测提供了一种新方法。

The calculation of water hammer is of great significance for ensuring the safe and stable operation of long-distance water supply network systems. However, traditional numerical methods for water hammer calculation suffer from issues of complex modeling and large computational requirements. To address this, transient flow experiments were conducted on a self-developed experimental platform, where water hammer was induced by rapidly closing a branch valve, and transient water hammer pressures were obtained under different flow rates and pressure conditions. The experimental parameters ranged from a volume flow rate of 15 to 55 m3/h and a pressure of 150 to 450 kPa. The Ensemble Empirical Mode Decomposition (EEMD) method was employed to filter the water hammer signals, and a thorough analysis of the water hammer pressure variations was conducted. Subsequently, leveraging the Bidirectional Gate Recurrent Unit (BiGRU), a Sequence-to-Sequence (Seq2Seq) bidirectional model was established for predicting water hammer pressures. The results indicated that the Seq2Seq bidirectional prediction model could effectively forecast branch water hammer, with a determination coefficient of over 0.8 for the predicted data, and an accuracy of over 98% in predicting water hammer characteristic parameters. This research outcome provides a new approach for water hammer pressure prediction.