During abnormal operating conditions in nuclear power plants (NPPs), a single fault may trigger a large number of sequential alarms. Conventional static alarm classification reflects alarm importance and urgency, but provides limited support for distinguishing causal priorities among alarms of the same level. This paper presents an Intelligent Alarm Management System (IAMS) and a dynamic alarm prioritization approach based on intelligent alarm analysis. Multilevel Flow Models (MFM) are used to provide functional causal knowledge. Real-time alarms are mapped to functional states, and an alarm causal matrix is updated to identify root-cause alarms, consequence alarms, and causal propagation paths. Without altering the existing Red/Yellow/Green/White classification framework, a four-field prioritization code is introduced using alarm severity, causal attribute, unexplainable range, and timestamp. In a representative Chemical and Volume Control (RCV) system scenario, two independent root-cause alarms were prioritized ahead of consequence alarms at the same severity level. Across ten typical single- and multiple-fault simulator scenarios, IAMS correctly identified the root-cause alarms in all tested cases, while alarm management time was reduced by 7%-32% compared with conventional operation. The proposed method therefore provides an interpretable and engineering-compatible means of supporting alarm response prioritization in complex NPP operating conditions.

alarm analysis; alarm ranking; Multilevel Flow Models; main control room; decision-making; nuclear power plant