The new generation of smart substations uses the acquisition and execution unit for unified sampling. Subsequent equipment will rely on the data from this device. Therefore, the abnormal data generated during the sampling process of the device will affect the normal operation of multiple systems such as substation protection and measurement and control. How to efficiently identify these abnormal data is a crucial part of power system sampling and the basis for the safety and stability of smart substations. Traditional anomaly data detection methods are mainly designed to deal with occasional anomalous sampling points in low sampling rate scenarios. As the sampling rate of intelligent substations increases and the electromagnetic interference problem intensifies, it has become a common phenomenon that multiple sampling points are abnormal at the same time during sampling. This makes the original recognition algorithm less accurate, and the unrecognized anomalous data may cause the accuracy of the subsequent measurement and control devices to be reduced or even the protection devices to be inadvertently activated. Given the shortcomings of traditional detection methods, a real-time algorithm for identifying abnormal data of AC sampling based on density-based spatial clustering of applications with noise (DBSCAN) algorithm is proposed. The algorithm makes use of the spatial density difference between abnormal and normal data to effectively distinguish the abnormal sampling points with lower density, so as to realize the identification of abnormal sampling data in intelligent substations. Compared with the traditional method, the method proposed has more accurate results in sampling abnormal data identification.