Compression and Restoration Identification Method of Wide-band Oscillation Data Based on Autoencoder

With the continuous development of power electronic power systems, the dynamic characteristics of power systems become much more complex, which may cause wide-band oscillation problems. However, due to the constraints of Nyquist sampling theorem and communication bandwidth, it is difficult for the existing dynamic monitoring devices to transmit and monitor wide-band oscillation signals. In this paper, a method for wide-band oscillation data compression and restoration identification based on autoencoder is proposed. Firstly, the encoder is used at the sub-station to reduce the dimension of the measurement data. On the one hand, it is convenient to carry out the wide-band oscillation signal transmission with the existing bandwidth. On the other hand, it also effectively reduces the redundancy of the oscillation data. On this basis, the master station uses the decoder to restore and identify the uploaded compressed data, which can not only verify the effectiveness and accuracy of the proposed method, but also enable the master station to conduct a more comprehensive monitoring and analysis of wide-band oscillation. Finally, case study is carried out in the four-machine and two-area system with a wind farm. The results show that the method can be applied to the data transmission of subsynchronous, supersynchronous and even medium/high frequency oscillation, and has high restoration accuracy and good anti-noise performance.