Voltage Transformer Anomaly Detection Using LSTM Based Autoencoder
Xin Yan, Cheng He, Jing Yin, Chuanji Zhang, Hongbin Li · 2023
Potential transformers (PTs) are essential measurement equipment in power systems. Developing monitoring techniques for them is crucial to enable measurement-based applications, where the challenge lies in establishing continuous online monitoring without relying on standard PTs. Since the temporal-spatial correlation among multiple PTs in the same substation can be leveraged, detecting abnormal changes in measurement data can be used for PT monitoring. In this study, we propose a long short-term memory autoencoder (LSTM-AE) based approach for anomaly identification in multivariate time-series measurement data of PTs. Generally, the LSTM-AE is trained to learn the normal patterns exhibited by the measurement data of PTs. Then the mean square error of the reconstruction serves as a metric for detecting anomalies in a substation with multiple PTs. To evaluate the effectiveness of the proposed method, we conduct experiments using both artificial and real-world datasets. Experimental results demonstrate the success in anomaly detection, indicating the capability of our approach for PT monitoring.