Transformer Abnormal State Identification Based on TCN-Transformer Model in Multiphysics
Junjie Feng, Ruosong Shang, Ming Zhang, Guojun Jiang, Qiong Wang, Guangyong Zhang, Wei Jin
Abstract
Transformers are critical components in power systems, and their operational stability plays a decisive role in ensuring the safety and reliability of the power grid. To address the challenges of accurately assessing transformer health due to the influence of load and environmental factors during actual operation, this paper analyzes the electrical, thermal, and vibrational characteristics of transformers. A k-means++ algorithm is employed to classify operating conditions based on three key parameters: transformer load current, ambient temperature, and operating voltage. A fusion model based on the Temporal Convolutional Network-Transformer (TCN-Transformer) is proposed to identify abnormal operating states of transformers. Experiments were conducted using a 500 kV transformer as an example. The results demonstrate that the proposed TCN-Transformer model significantly outperforms comparative algorithms in terms of prediction accuracy. The model effectively captures critical information within the data, achieving superior multivariate feature sequence prediction for transformers. These findings validate the reasonableness and accuracy of the proposed method for identifying abnormal transformer operating states.