Voltage Sag Identification Based on Deep Learning Method with Gated Recurrent Unit

The voltage sag can be regarded as a power quality issue, which is attracting more and more attention in both industry and academia due to its serious economic losses. It is therefore essential to realize sag type recognition accurately, which is essential to clarify the responsibility causing sag and helpful for economic losses reduction. Therefore, it is becoming an urgent problem to realize the accurate sag type recognition. The input for the model is the voltage waveforms collected through monitors, and the output is the corresponding sag type, where sag caused by motor starting, transformer energizing and short circuit are considered. Through the presented approach by adopting the Gated Recurrent Unit in this paper, the nonlinear correspondence between the input voltage waveforms and output type of voltage sag is established automatically. After that, with the help of Python language and TensorFlow tool, dataset is generated through Matlab/Simulink, and then, tests are carried out and the corresponding result proves that our presented method can realize sag type identification with high accuracy.