Abnormal Detection Method of Sensor Data Based on Informer

The online monitoring data of transformer oil temperature plays an important role in judging the reliability of transformer operation and monitoring the internal insulation state of transformer. However, due to the complex environment where the oil temperature sensor is located, failures are very easy to occur, so the monitoring data may appear abnormally due to sensor failure. The resulting abnormal data affects the judgment of the state of the transformer. Therefore, this paper first analyzes the abnormal types of sensor data, and then proposes an abnormal detection approach for transformer oil temperature sensor data based on Informer. The proposed approach can identify various abnormalities in long-term oil temperature monitoring data, and experiments based on multiple transformer oil temperature monitoring data verified the effectiveness and high accuracy of the proposed method.