Transformer Anomaly Detection Method Based on MDS and LOF Algorithm

Aiming at the problems of poor real-time performance and low efficiency of transformer anomaly detection method, a transformer anomaly detection model is constructed by combining multidimensional scaling (MDS) algorithm and local outlier factor (LOF) algorithm. MDS algorithm is used to reduce the feature dimension of high-dimensional space-time state monitoring matrix constructed by transformer parameter data, which reduces the redundancy of features and shows them in the form of coordinates in two-dimensional space. Then LOF algorithm is used to calculate the local outlier factor values of all sample points, and the local outlier factor values are compared with the threshold value, so as to screen out the abnormal sample points. The actual operation data of 10kV transformer in a substation are simulated and tested. The results show that the transformer anomaly detection method based on MDS and LOF algorithm is effective.