Aiming at the problem of the traditional harmonic monitoring data early warning method, which only considering the distribution of data value in a short period or a single moment, rather than paying attention to the change of data trend, this paper proposes a trend warning method based on the correlation analysis of harmonic monitoring data: on the basis of analyzing the periodic regularity and similarity of the trend change of harmonic monitoring data over time during long-term operation, the change trend of real-time monitoring data and normal state data is compared. If the trend difference is large, early warning will be given. Firstly, the moving average smoothing method is used to extract the overall trend change of data to avoid the influence of short-term fluctuations and other interferences on the overall trend analysis. Then, according to the dynamic time structuring algorithm, the difference of data change law in different periods is quantified, and abnormal trend warning is given. Finally, the correctness and effectiveness of the proposed method are verified by monitoring data.