Unsupervised online poor contact detection in the secondary circuit of voltage transformers

Abstract

Voltage transformers (VTs) are fundamental and crucial instrument devices for accurate voltage measurement. Generally, measurement errors can be caused by both the VT and the secondary circuit. Most research focused on the VT but ignored the secondary circuit between it and the measurement device. Poor contact in the secondary circuit is more challenging to capture than those caused by faulty VTs. In this study, we present an unsupervised online poor contact detection method. It first exploits a time series similarity matrix derived from the measured voltage data. The physical constraints deriving from the electrical topology are utilized to eliminate the non-fault fluctuation. To ensure the versatility of the proposed method, an adaptive threshold determination strategy is designed based on knee point location. Numerical experiments compared with some state-of-the-art methods have validated the effectiveness of the proposed method. The versatility of the proposed method is further validated in three real-world cases.