Automatic identification of time-synchronization signals in intelligent substations based on improved isolated forest algorithm

To achieve automatic identification of time synchronization signal types and interface types in smart substations, this paper proposes a recognition method based on an improved isolated forest algorithm. The signal is decomposed using the wavelet lifting method to effectively eliminate common-mode noise from inter-station communications. The signal’s voltage range and voltage variation patterns are extracted as feature vectors. The K-means clustering algorithm is employed to refine the decision boundaries for multi-type samples within the isolated forest. By calculating the average path length thresholds for each signal category based on cluster centers, the method achieves automated multi-class signal recognition decisions. Experimental results demonstrate that the proposed method achieves an AUC value as high as 1 for different signal recognition targets, with overall AUC consistently exceeding 0.95. During the testing phase, full signal recognition completed in just 1.163 s, with a processing delay of 0.014 s. This approach exhibits high recognition accuracy and execution efficiency, effectively ensuring the time synchronization precision between devices.