In view of the complexity of environmental factors around nuclear power plants and the high false alarm rate of current early warning methods, an adaptive early warning method for abnormal radiation environmental monitoring data around nuclear power plants is proposed. In order to improve the validity of the data, the data are corrected according to the attribute distribution of the radiation environment monitoring data in the periphery of the nuclear power plant. The monitoring data are discretized and decomposed to obtain the first-order lag variables of the data attributes. Combined with the similarity measure between the data, the abnormal characteristics of the data are identified, and the adaptive early warning model of abnormal data is constructed, and the early warning factors output by the model are compared with the preset values, so as to realize the early warning of abnormal data. Taking a nuclear power plant operation and maintenance system as a case study, the early warning performance of the proposed method is tested, and the results show that the method can effectively realize the abnormal early warning of monitoring data, with low false alarm rate and good early warning effect.