Toolset Development for Modelling Sympathetic Phenomenon and its Detection by a Neural Network

One of the most important elements of any AC power supply system is the transformer. Each stakeholder, such as Distribution Network Operators (DNOs), Transmission System Operators (TSOs), and other generators, is paying close attention to investigating possible negative impacts that could disrupt the normal distribution of electricity. One of them is short-term transient processes that occur in the network when the transformer core is turned on. These transients can cause significant distortion on the power line, which in turn can cause nuisance tripping of protection devices. For a small and isolated electrical system, this is not usually considered a major problem. However, as the electrical grid grows, this becomes more important. This is a problem that machine learning can help with. However, the problem is that there is currently no significant data on the transformer excitation phenomenon. This article gives a brief explanation of transformer excitation phenomena in an introductory section. With this information in mind, we developed a SIMULINK model and a MATLAB script to describe and automatically run a new experiment design and generate data. Finally, we use the generated data as a training dataset to fit it into a supervised convolutional neural network. The final data set that we created and used for the purposes of this work, as well as the SIMULINK model and the MATLAB script for automatically generating the experiment design, are in the public domain.