The invention relates to an anomaly detection method and system of terminal access data. The method includes steps: abstracting a related attribute of original message data to a recognizable characteristic vector, and analyzing the characteristic vector; performing reduplication removal processing on the analyzed message characteristic data, and performing quantification processing on characteristic data with a large value range; detecting normal data and abnormal data in the terminal access data, and generating a corresponding model file; and performing online classification and detection onthe message data through loading of the model file, and generating a detection result. The false alarm rate and the missing alarm rate are low, and reliable operation of a power consumption information acquisition system can be guaranteed.