Power transformer is a key equipment of power system, and transformer protection is still an important research hotspot for the safe operation of the electric power system. Kinds of transform protection algorithms have been proposed including many artificial intelligence (AI) algorithms. But the requirement of big data limits the generalization of AI algorithms. Different types of ferromagnetic material have the same shape of magnetization hysteresis which is an effective indicator for transformer protection. We propose transfer learning based equivalent magnetization hysteresis recognition algorithm for transformer protection which use little trial/real data with simulation data to solve the few-shot problem. In our experiments, we validate that the proposed algorithm can reach an improved accuracy compared with other two mentioned in the article.

equivalent magnetization hysteresis,transformer protection,transfer learning,few-shot problem