Localization Method of Wide-band Oscillation Disturbance Sources Based on Compressed Sensing and Graph Convolutional Neural Networks

The wide-band oscillation caused by the grid connection of new energy seriously threatens the security of the power grid. It is particularly necessary to realize the online location of the broadband oscillation source and take timely suppression measures to ensure the safety and stability of the system. In this paper, a broadband oscillator location method combining compression sampling and graph convolution neural network is proposed. This method firstly sparsely samples the time-series oscillation signal in the substation to obtain its low dimensional observation sequence as the time sequence information of the node, and then captures the adjacency of each node in the topology structure of the master station fusion system, comprehensively considers the time-space characteristics of the system oscillation, and uses graph convolution neural network to locate the oscillation disturbance sources. Finally, the broadband oscillation sample set is used for simulation verification. The results show that the proposed method has high positioning accuracy when the measurement data contains noise, the transmission data is missing and the transmission data is biased.