A Generic Data Generation Framework for Short Circuit Detection Training of Neural Networks

Artificial Neural Networks excel in pattern recognition tasks and can therefore be used to identify short circuits in power systems. As short circuits do not frequently occur in real world grids, simulation data must be used to train neural networks. This paper describes a framework that can be used to automatically create large datasets for the training process of Neural Networks. Tests show that the framework is fast, adaptable, and works out of the box for most grid models. Additionally, a short circuit detection training is presented, that proves the applicability for Neural Network training purposes.