Establishing a stage–discharge rating curve in an irrigation canal might look outdated, but it is still an important issue of agricultural water management. Therefore, in this study, a method by using machine learning (neural network) to establish a stage–discharge rating curve is proposed. The machine learning trained by the observed gage height was used to estimate velocities in an irrigation canal. The estimated velocities were used to compute the discharge. Then, the observed gage height and estimated discharge are applied to establish a simple stage–discharge rating curve. The data collected in the Wan-Dan Canal are used to evaluate the proposed method. The results showed that machine learning could effectively simulate the velocity distribution in an irrigation canal from its bottom to the water surface, as well as the flow fields at ungagged sites. Therefore, using the velocity derived from machine learning, the discharge of an irrigation canal can be accurately determined. Meanwhile, an accurate and dependable stage–discharge rating curve can be established. Our proposed method is applicable to forecast the discharges of other irrigation canals to manage agricultural water effectively.

在灌溉渠中建立水位-流量额定曲线可能看起来过时了，但这仍然是农业用水管理的一个重要问题。因此，在本研究中，提出了一种利用机器学习（神经网络）建立阶段-放电额定值曲线的方法。通过观察到的计量高度训练的机器学习被用来估计灌溉渠道中的速度。估计的速度用于计算流量。然后，应用观察到的标高高度和估计的流量来建立简单的阶段-流量额定值曲线。在万丹运河收集的数据用于评估所提出的方法。结果表明，机器学习可以有效地模拟灌溉渠道从底部到水面的速度分布，以及 ungagged 站点的流场。因此，使用机器学习得出的速度，可以准确地确定灌溉渠道的流量。同时，可以建立准确可靠的阶段-放电额定值曲线。我们提出的方法适用于预测其他灌溉渠道的流量，以有效管理农业用水。