Water level prediction model of the upper Yangtze River by cloud computing and AutoML

ABSTRACT The change in water level in the upper reaches of the Yangtze River is of great significance to flood control and navigation. As the first water control station for the mainstream of the Yangtze River after the Minjiang, Hengjiang, Tuojiang River, and other important tributaries flow into the Yangtze River, it is imperative to forecast the water level of Zhutuo Station accurately. The present study utilizes Microsoft Azure's automated machine learning platform (AutoML) and recurrent neural network (RNN) model to predict water levels at Zhutuo Station. The AutoML approach demonstrates certain advantages over RNN methodologies in terms of operability, resource utilization, computational efficiency, and hardware configuration requirements when predicting the water level of Zhutuo. The results show that the future 1-h forecast performance is similar, the mean absolute error (MAE) and root mean square error (RMSE) of the AutoML platform are 0.0098 and 0.012, respectively, and the MAE and RMSE of the RNN model are 0.0088 and 0.011, respectively. The current study's outcomes contribute valuable insights for the real-time monitoring and predictive analytics of water levels, thereby enabling waterway managers to obtain the balance between model complexity and modeling convenience.