Enhancing the Accuracy of Water–Air Flow Prediction via Hybrid CatBoost Modelling
Annotatsiya
ABSTRACT The unclear hydraulic performance of aeration pipelines is an important factor restricting their further application in agricultural fields. Methods for predicting changes in the flow of aeration pipelines often rely on empirical models and often lack high precision. In this paper, a new method was proposed to predict the flow rate of a non‐discharge pipeline along an air‐filled path via the CatBoost machine learning algorithm. A total of 3600 sets of flow data were collected and randomly divided into 7:3 ratios for training and testing using the PLC‐based water–gas automation system. The grid search method was used for multiple cross‐validations and empirical calculations, and parameters such as n estimators, tree depth and learning rate were optimized. A prediction model was then built using these optimal parameters. The results revealed that the mean absolute error (MAE) and root mean square error (RMSE) of the CatBoost model were 86 and 71, respectively. This research provides a foundation for optimizing aeration irrigation systems, enabling precise flow control that can achieve 5%–10% water savings while maintaining crop productivity. The developed model can be integrated into automated irrigation management systems to support sustainable agricultural practices.