A Method for Anomaly Detection of State Information of Transformer Based on Association Rules and Elman Neural Network

In order to identify the transformer abnormal state accurately, this paper presents a method considering the correlation degree between the monitoring sequences and using the Elman neural network. Firstly, the degree of association between the state monitoring of the transformer is calculated by the association rule, and the prediction error formula is modified by the sequence correlation degree. And then use the trained Elman neural network to predict the sequence to find the abnormal points. The abnormal data is divided into abnormal sensor data and abnormal device status data. For sensor abnormalities, data cleaning is required. For device status abnormalities, we need analyze the device's abnormal type. The test results show that the correlation analysis of sequence data and the combination of Elman neural network can effectively improve the accuracy of transformer state anomaly detection.