Abstract: The traditional identifications for the steady-state power quality anomalies, such as harmonics anomalies, only compare the data values with the threshold without considering the trend of the data changes. This paper puts forward a method of harmonic anomaly identification that takes time series trend analysis into account in view of the periodic harmonic operating conditions and the trend change characteristics of the harmonics monitoring data. Firstly, the main trend change features are extracted by piecewise linear representation, and the trend of each segment is represented by model values. Then, the trend and numerical value anomaly indexes are determined based on the similarity of the trend sequences and the outliers' proportion respectively, and the composite anomaly indexes are obtained when the two indexes are weighted. The analysis period of sequences and the benchmark normal data are selected from the historical data. Finally, the composite anomaly index of the subsequent data to be identified is calculated to identify whether there are harmonic variation anomalies at the monitoring point. Through simulation examples and case analysis, it is proved that the proposed method is accurate, applicable and easy to implement, and can be expediently integrated into the existing power quality monitoring system.