Transformer State Data Cleaning Method Based on DBSCAN and ESN

To address the problem of noise in transformer status monitoring data, this paper proposes a transformer status data cleaning method based on density-based spatial clustering of applications with noise (DBSCAN) and echo state network (ESN) to improve data quality. Firstly, all historical moment data are clustered by DBSCAN to get a number of normal clusters. Secondly, for the data to be cleaned that do not belong to the normal clusters, they are distinguished as device status abnormalities and sensor abnormalities according to this abnormality detection rule. Finally, the state parametric data due to sensor anomalies are repaired by using echo state network. The results show that this method can effectively detect the noise of transformer state parameter data, and the data repair effect is good, which verifies the effectiveness of this method.