Kummerow, AndréAndréKummerowDirbas, MohammadMohammadDirbasMonsalve, ChristianChristianMonsalveNicolai, SteffenSteffenNicolaiBretschneider, PeterPeterBretschneider2022-03-152022-03-152021https://publica.fraunhofer.de/handle/publica/41282910.1109/SEST50973.2021.9543278The automated classification of grid disturbances based on phasor measurement units (PMU) is a key application for a fast and reliable monitoring and control of future power systems. The predominant use of dynamic simulations for the training of the classification models can lead to severe misclassifications during the application phase due to measurement induced error signals. As an advancement to standard white noise approaches, an optimization-based error model is introduced for the synthesis of PMU measurement signals with specific noise characteristics. This approach allows a flexible creation of more sophisticated error signals. Extensive simulation studies are performed for a disturbance classification model based on a recurrent neural network using a large electrical transmission grid.enPhasor Measurement Unitrecurrent neural networksdisturbance classification004670conference paper