Abstract: With the widespread construction of high dams and large reservoirs, together with increasingly complex operating environments, dam safety early warning systems are facing greater demands. There is an urgent need to address the limitations of traditional methods, which often lack predictive capability under complex conditions. This paper systematically reviews recent advances in intelligent dam safety early warning, with a focus on three core aspects: intelligent risk assessment, the intelligent development of early warning indicators, and risk warning models based on intelligent optimization. The study examines the evolution of risk assessment from static threshold-based judgment to multidimensional probabilistic state characterization. It also describes the shift in early warning indicator selection from manual, experience-based screening to data-driven automatic optimization. Furthermore, it highlights the critical role of intelligent optimization algorithms in addressing parameter sensitivity and local optimum problems in machine learning and deep learning models. Research indicates that integrating intelligent optimization strategies can significantly reduce reliance on empirical parameters while improving model generalizability and stability in complex nonlinear environments. This study aims to advance dam safety early warning systems toward real-time, accurate, and adaptive operation, thereby providing a solid theoretical foundation and key technical support for the development of a new generation of intelligent dam safety warning frameworks.

Key words: risk prediction, early warning indicators, early warning models, machine learning, deep learning