The primary circuit transient of nuclear power plant is one of the important references for characterizing the fatigue of primary circuit materials. For the classification of transients, the commonly used method is to determine the category of transients according to the change in value and the operating operation of the unit. This paper forms a complete set of digital transient characteristics through the study of the thermal parameters behind various transients and the exploration of the law of transient occurrence. Combined with new computer technologies such as machine learning, the automatic processing of transient classification is realized. It provides ideas for the key links in the development of the whole-process automation platform for transient management.