Abstract: In order to improve the fast recognition ability of power system wideband oscillation, a wideband oscillation algorithm based on adaptive variational mode decomposition (AVMD) and multi-scale fuzzy entropy is designed to introduce variable prediction model. The time-frequency parameters of oscillating signals are obtained by AVMD technology, and the time-domain characteristics of each modal component are evaluated by multi scale fuzzy entropy, which completes the IMF feature vector reduction and the accurate classification of wide band oscillation data. The results show that when different levels of white noise are mixed into the signal, the classification accuracy of the model is reduced, and the average accuracy is only reduced by 4.86%. When dealing with the parameters of new experimental samples, the training model can accurately identify the type of wide-band oscillation without overfitting. Multi-scale fuzzy entropy can show the degree of self-similarity and complexity of time series at various scales. Compared with SVM and BP, VPMCD shows the optimal classification effect, with an accuracy of 99.56%, and it only takes a short time to complete online classification.