Wide-band Oscillation Disturbance Source Location Based on Compressed Sensing and CNN-LSTM

Due to the strong time-varying and non-linear interactions between power electronic devices and the power grid, challenges arising from wide-band oscillation risks have been increasing under the backdrop of the "dual-high" power system development, making the precise location of oscillation disturbance sources difficult. This article introduces a method for the precise location of wide-area system oscillation disturbance sources, which is based on compressed sensing and a fusion of the CNN-LSTM algorithm. Firstly, we discussed the role and application conditions of compressed sensing and generated a measurement matrix synchronized with the oscillation signal, achieving high-rate compression of the oscillation signal. Subsequently, this paper constructed a deep learning model using CNN-LSTM, taking sparse sampling signals from various substations as inputs to achieve wide-band oscillation location. Finally, we applied this method to a disturbance source location task in an IEEE-39 bus system that includes a wind farm. The results indicate that this approach can effectively bypass the constraints of the Nyquist sampling theorem and achieve highly accurate disturbance source location under low computational requirements.