Abstract:

As the sole movable component in power transformers, the on-load tap changer (OLTC) plays a crucial role in security of power system. However, owing to frequent operation and harsh working conditions, the OLTC has become one of the most vulnerable parts of power transformers. Based on the existing fault diagnosis methods, a method analyzing time-domain feature parameters of OLTC vibration signals based on chaotic features is proposed. Starting from the chaotic features of OLTC vibration signals, the signals are reconstructed in a high-dimensional space and their geometric characteristics are extracted. Several chaotic features, including the correlation dimension, largest Lyapunov exponent, and Kolmogorov entropy, are calculated to construct a multidimensional feature space. Finally, K-means clustering is applied to explore the fault diagnosis in OLTCs. The results show that the method can effectively distinguish among different types of OLTC faults. However, when the fault severity is taken into account, the diagnostic effect needs to be further improved.

Key words: fault diagnosis, on-load tap changer (OLTC), chaotic feature, vibration signal, k-means clustering