Example of raw train operational data.

This paper focuses on optimizing the management of delayed trains in operational scenarios by scientifically categorizing train delay levels. It employs static and dynamic models grounded in real-world train delay data from high-speed railways. This classification aids dispatchers in swiftly identifying and predicting delay extents, thus enhancing mitigation strategies’ efficiency. Key indicators, encompassing initial delay duration, station impacts, average station delay, delayed trains’ cascading effects, and average delay per affected train, inform the classification. Applying the K-means clustering algorithm to standardized delay indicators yields an optimized categorization of delayed trains into four levels, reflecting varying risk levels. This static classification offers a comprehensive overview of delay dynamics. Furthermore, utilizing Markov chains, the study delves into sequential dynamic analyses, accounting for China’s railway context and specifically addressing fluctuations during the Spring Festival travel rush. This research, combining static and dynamic approaches, provides valuable insights for bolstering railway operational efficiency and resilience amidst diverse delay scenarios.