Research of Transformer Protection Based on Joint Deep Learning

Abstract

As the total electricity load and the proportion of renewable energy sources continue to rise in China, the power grid is experiencing an expansion in scale and an increasing complexity in its structure. As the most important equipment in the power system, the operation status of transformers directly affects the safety and stability of the system. Once a malfunction occurs, it will bring serious economic losses and harm. This paper proposes a transformer protection scheme based on joint deep learning method. Firstly, collect signals through the circuit breakers on both sides of the transformer to complete real-time data collection. Then, a gated recurrent neural network is used to achieve short-term and ultra short-term state recognition. In addition, self supervised learning task is added for joint training. Then the transformer fault diagnosis and protection are realized. Finally, using PSCAD software to construct a typical transformer model structure and conduct simulation verification using Jupyter Lab. The results show that the protection scheme has good performance in different sampling period lengths, noise interference, and data loss situations.