深度学习在数据驱动水文模型方向的应用研究是当前水文预报的研究热点。综述近期深度学习模型在水 文预报中的应用研究进展，归纳以往数据驱动模型未表现出的新特点，分析深度学习在水利工程影响下水文建模、 不确定性分析等问题中的应用潜力以及深度学习水文预报模型解释的知识发现作用。从物理机制模型中融入深 度学习和物理机制指导的深度学习两方面讨论深度学习与水文物理机制整合的研究实例。总结深度学习水文预 报研究仍面临的重要挑战，展望可能的发展趋势，以期为深度学习在水文预报中的研究发展提供有效参考。

Many studies focus on the application of deep learning in data-driven hydrological models for hydrological forecasting. The recent research progress of deep learning models on hydrological forecasting are reviewed. Deep learning shows new features unseen in non-deep learning data-driven models. It can also help to address some old challenges in hydrological modeling, such as hydrological modeling under the impact of hydraulic projects and uncertainty analysis. Deep learning can be a tool for knowledge discovery. We also describe some studies for the integration of deep learning and domain knowledge in hydrology, including incorporating deep learning in physical mechanism models and physics-guided deep learning. We also highlight the challenges in applying deep learning to hydrological forecasting and propose potential opportunities. This review can provide a useful reference for further research on deep-learning-based hydrological forecasting.