Flexible on-request (arranged) water delivery is applicable in existing canals that operate manually. A challenge facing this method is the diversity and multiplicity of requests across the network, making planning for water distribution very difficult and time-consuming. Therefore, we need a suitable tool to derive the proper operation of structures quickly to improve network performance compared to rotational delivery. This study uses intelligent optimization methods to prepare on-request operational instructions of irrigation canals under unsteady flow conditions. A meta-model was developed using artificial neural networks (ANNs) for the first reach of the Aghili canal for simulating unsteady flow in canals to speed up the process. The meta-model was combined with a genetic algorithm (GA) to determine optimal operational instructions. The error rate of the ANN model was smaller than 2.5%, indicating excellent performance of the developed model. Two scenarios of 3- and 6-h delivery were defined to compare the optimal operation of the ANN-GA model with conventional operation. For both scenarios, the percentage of improvement was remarkable, and on average, it was above 50%. The utility of the developed model for shorter deliveries is significant. Moreover, the model can be generalized to other reaches.

中文翻译：

---

考虑非定常流的混合遗传算法和人工神经网络按需交付的最优操作指令

灵活的按需（安排）供水适用于手动操作的现有运河。这种方法面临的一个挑战是整个网络的请求的多样性和多样性，使得水分配规划非常困难和耗时。因此，我们需要一个合适的工具来快速推导出结构的正确操作，从而与旋转交付相比提高网络性能。本研究采用智能优化方法，根据请求编制非稳态流动条件下灌溉渠的操作指令。使用人工神经网络 (ANN) 为 Aghili 运河的第一段开发了一个元模型，用于模拟运河中的不稳定流动以加速该过程。元模型与遗传算法 (GA) 相结合，以确定最佳操作指令。人工神经网络模型的错误率小于 2.5%，表明所开发模型的性能优良。定义了 3 小时和 6 小时交付的两种情况，以比较 ANN-GA 模型的最佳操作与传统操作。在这两种情况下，改善的百分比都非常显着，平均在 50% 以上。所开发模型对于更短交货期的效用是显着的。此外，该模型可以推广到其他领域。所开发模型对于更短交货期的效用是显着的。此外，该模型可以推广到其他领域。所开发模型对于更短交货期的效用是显着的。此外，该模型可以推广到其他领域。