The application of Artificial Neural Networks (ANN) coupled with other models or optimization algorithms for runoff prediction is gradually increasing. This review systematically categorizes and summarizes four aspects: the coupling of ANN with physical models, the combination of multiple ANN models, the integration of decomposition techniques with machine learning methods, and the incorporation of ANN models with intelligent optimization algorithms， and shows the reasons for improving the prediction accuracy and the advantages of each method. Furthermore, the present research challenges and future perspectives are discussed. This study highlights the significance of employing multi-model integration and decomposition techniques in runoff prediction, which can further advance research and development in the field and provide valuable support for accurate and reliable runoff prediction and water resources management.