Intelligent Transformer Protection Method Based on Convolutional Neural Network

Transformer is an important energy hub in power system and it is an important power equipment to ensure the safe and reliable operation of power grid. At present, the identification of inrush current and internal fault is still the core problem of transformer protection. Based on the ontology structure of transformer, this paper adopts the idea of image recognition to supervise and study the equivalent magnetization curve by using Convolutional Neural Network (CNN), and proposes a transformer intelligent protection algorithm with strong recognition ability and high recognition accuracy. First, the multi-physical field simulation model of transformer is established by COMSOL software. Then, the inrush current, internal fault and external fault data are obtained, the equivalent magnetization curves are drawn and two-dimensional gray scale images are constructed as the training set and test set samples of CNN. On this basis, this paper establishes the intelligent protection method of transformer based on Convolution Neural Network. Finally, the simulation results illustrate that the intelligent protection method based on Convolutional Neural Network can identify the inrush current, internal fault and external fault of transformer accurately, and the accuracy of inrush current is up to 93.5%.