The embodiment of the invention provides an anomaly detection method and device. The method comprises the following steps: acquiring the target time sequence data of a power grid system, and converting the target time sequence data to a first matrix; and inputting the first matrix to a pre-trained coding and decoding model to acquire a second matrix, calculating a first error of the second matrix with respect to the first matrix, and determining that the target time sequence data is the abnormal sequence data when the first error exceeds the preset error range. By adopting the anomaly detection method and device provided by the embodiment of the invention, the anomaly is detected by using the pre-trained coding and decoding model, and thus the manual workload in the anomaly detection of power systems can be reduced, and the efficiency of anomaly detection can be increased.