Ensemble Anomaly Data Detection Method for Oil-immersed Transformers

With the development of science and technology, there is an increasing demand for data processing in various fields. Especially in the construction of a smart grid, high-quality data on the operation of power equipment is more needed. Oil-immersed transformers as one of the important equipment in the power system, and the abnormal detection of its operating data is directly related to the accuracy of transformer fault diagnosis, which in turn may have an impact on the safety and reliability of the power supply system. This paper overviews the current research status of anomalous data detection techniques, introduces the concept of machine learning-based outlier detection methods, and uses five different methods such as statistics-based, tree-based, and classification-based to detect anomalous data in the collected dissolved gas concentration (DGC) data. Based on the detection results and the idea of the ensemble algorithm, the anomalous data in the dataset was successfully identified.