Up to now， temporal information of alarms has not yet been fully utilized in existing alarm processing methods. The temporal constraint network is a kind of directed acyclic graph（DAG）， and could well represent temporal logics. Given this background， an analytic model for alarm processing with temporal information taken into account is developed. The model could not only find out the events causing the reported alarms together with the time domain concerned， but also identify abnormal and missing alarms. The concepts of the time-point and time-distance constraints in the temporal constraint network provide a path for describing the uncertainty of the event occurrences， and hence precise timing is no longer required as in the existing alarm processing models accommodating temporal information. Finally， an actual power system is served for demonstrating the feasibility and efficiency of the proposed model and method.